Data Driven Computing and Machine Learning in Engineering

DACOMA-19
DATA DRIVEN COMPUTING AND MACHINE LEARNING

IN ENGINEERING 2019 (DACOMA-19)
Advances in Computational Mechanics
Volume 4, 2019
ISSN: 1940-5820 (printed)
ISSN: 1940-5839 (online abstract)
Conference Proceedings of
Data Driven Computing
and Machine Learning
in Engineering
Organizer
HORIZON 2020
BESTOFRAC
Editor
Xiaoying Zhuang
Preface
In this age of big data, machine learning techniques have been successfully applied in image processing,
genomics, financial problems and even medical diagnosis. The emerging application of machine learning
and big data analysis has fundamentally influenced and changed our way of how we think, plan, solve and
analyze in engineering. Nevertheless, we are faced with many issues and unsolved problems when applying
data drive computing and machine learning in engineering analysis.
I have conceived the organization of this conference about two years ago with Timon Rabczuk and Hehua
Zhu, the presidents of the conference. Now with the financial support of HORIZON-2020-RISE
BESTOFRAC project, the International Conference on Data Driven Computing and Machine Learning in
Engineering (DACOMA-19) conference is made possible to take place in Shanghai in September 2019.
The conference is co-organized by Tongji University, Leibniz University Hanover, Chinese Society of
Computational Mechanics, International Chinese Society of Computational Mechanics, The German
Association of Computational Mechanics. This proceeding collects over 100 abstracts from participants of
DACOMA-19 presenting research and application in big data technology, data driving computing and
artificial intelligence in engineering as well as promoting interdisciplinary topics.
I hope readers will enjoy reading this proceeding and be inspired from this book in their research in this
area. I would like to acknowledge Qimin Wang, Minjing Cai, Bin Li and Wei Peng for their help in
collecting and organization of DACOMA.
Xiaoying Zhuang
in Shanghai August 2019
Data Driven Computing and Machine Learning in Engineering
Menu
Machine Learning Based Solution Of Partial Differential Equations In Computational Mechanics:

Concepts, Implementation And Applications
Timon Rabczuk and Xiaoying Zhuang 1
Machine Learning for Seawall Structures Under Flooding and Erosion
Hung Nguyen-Xuan 3
Data-driven Construction and Maintenance of Urban Tunnels
Hongwei Huang 5
Solving Direct And Inverse Computational Engineering Problem Through A Deep Learning Based
Data-Driven Method
Zhanli Liu, Xiang Li, Haolong Chen, Chengcheng Luo and Zhuo Zhuang 7
An Investigation Of Deep Learning In Predicting Shield Tunnel Excavation Parameters
Xiaojun Li, Dong Han and Gang Li 8
The Application Of A Deep Energy Method In The Bending And Eigenvalue Analysis Of Kirchhoff
Plate
Xiaoying Zhuang and Hongwei Guo 9
Isogeometric analysis of a phase field model for simulating electromechanical behavior and
morphological evolution of vesicles
Mohammed Ashour and Navid Valizadeh 11
Exploration signals from higher dimensions base on machine learning
Cong Cao 12
Volumetric parametrization using convolution neural networks
C.L. Chan and C. Anitescu 13
Boundary Shape Identification of the Heat Conduction Problem Based on Machine Learning
Haolong Chen, Xiang Li, Chengcheng Luo and Zhanli Liu 14
Free vibration analysis of Timoshenko nanobeam using differential transform method and general
nonlocal theory
Shirko Faroughi, Ahad Mahmodpour and Tiomn Rabczuk 15
DACOMA-2019
Data-Driven Simulation of History-Dependent Materials Using Structured Data Sets and

Transition Rules
Klaus Hackl and Kerem Ciftci 17
Artificial neural networks for solving coupled PDEs in flexoelectricity
H. D. Huynh, H. Nguyen-Xuan and X. Zhuang 18
Nonlinear mechanical behavior and failure mechanism of three-dimensional complex carbon
nanotube networks
Jiachao Ji and Junhua Zhao 20
Designing phononic crystal with anticipated band gap through a deep learning based data-driven
method
Xiang Li and Zhanli Liu 21
A comparative study of machine learning methods applied to the singular integrals in the boundary
element method
Yuan Li and Wen-tao Mao 22
Data driven thermo-mechanical calculation during additive manufacturing process
Constant PRASSETTE, Guanghua WEI, Michel BELLET and Yancheng ZHANG 24
Residual-based Goal-oriented A Posteriori Error Estimation and Adaptive Mesh Refinement in
Thermoelastic Problems
Ehsan Rabizadeh and Amir Saboor Bagherzadeh 26
Theoretical and experimental study of long-term creep in multilayered composites
Roham Rafiee and Amin Ghorbanhosseini 28
Deep learning Approach to Analysis and Prediction of Infrastructure Disaster
Manouchehr Shokri 30
Inverse design of phononic Topological insulators using Machine learning
S.S.Nanthakumar and Xiaoying Zhuang 32
A Deep Learning Method for A Fully Coupled Thermomechanical Model
Nam Vu-Bac, Xiaoying Zhuang and Timon Rabczuk 34
A Date Driven Framework to Design Seismic Metamaterial
Qimin Wang, Xiaoying Zhuang and Yanyu Chen 36
Simulation of nonlinear propagating waves in hysteresis media via higher-order central-WENO
high resolution schemes
Hassan Yousefi 38
Numerical Simulation of Stability Analysis of Jointed Rock Slope Based on SPH Method
Yongzheng Zhang and Xiaoying Zhuang 39
Effect of size of microcapsules on the self-healing concrete

Yu Zheng and Shuai Zhou 40
Application of a deep learning method for solving PDEs for compressed air energy storage
Shuwei Zhou 41
Machine Learning based Crack Growth Prediction: Application to a Helicopter Component Digital
Twin
Xuan Zhou and Leiting Dong 43
Numerical implementation of data-driven computational mechanics
Ying Zhou and Yuantong Gu 45
A Real-Time Structural Topology Optimization Method using Conditional Generative Adversarial
Networks
Xuefeng Zhu, Xiaochen Liu, Ping Hu and Zheng-Dong Ma 46
Analysis of discontinuous evolutions in phase-field fracture modelling
Patricio Rodriguez and Xiaoying Zhuang 48
Robust topology optimization using an improved adaptive Gaussian process
Somdatta Goswami and Souvik Chakraborty 50
An Intelligent Detection Model Based on Fully Convolutional Neural Network for Pavement Crack
Duo Ma, Hongyuan Fang, Binghan Xue and Fuming Wang 52
Bayesian solution for inverse elasticity problems with hybrid uncertainties using Polynomial Chaos
Surrogate dictionaries
Tittu Varghese Mathew, S. Natarajan and S.P.A. Bordas 54
Adaptive grid refinement in a 3D peridynamic model
Arman Shojaei 55
Feasibility of Machine Learning Methods for the prediction of Roadheader Performance in
Underground Excavation Project-A Case Study
Hadi Fattahi 57
Applying Adaptive Neuro Fuzzy Inference System to Predict Surface Settlement Caused By
Mechanized Tunneling-A Case Study
Hadi Fattahi 59
Classification of surrounding rock of highway tunnel based on migration learning
Xue-feng Chen 61
Automatic Lithology Identification Based on Machine Learning: A Case Study of TZ4 Well
Siyao Guo, Zhihui Ye, Han Wang, Dong Chen and Dandan Zhu 63
DACOMA-2019
Automatic Identification of Gas Hydrate Formation using Machine Learning Algorithms

Yuqiang Ning, Zhihui Ye, Han Wang, Dong Chen, Shouding Li and Dandan Zhu 65
Automatic Damage Analysis of Cracked Concrete Specimens Using Deep Learning-based Crack
Detection techniques
Pan Sun 67
A New Method of Tunnel Leakage Water Identification Based on Terrestrial Laser Scanning
Keshen Zhang, Hehua Zhu and Wei Wu 68
Dynamic Classification of Surrounding Rock Mass in Tunneling Boring Process Based on Big
Operational Data
Mengqi Zhu and Hehua Zhu 70
Prediction of Strength of Foamed Concrete under Cyclic loading Using Artificial Neural Network
Guanbao Ye, Jiangting Liu, Zhen Zhang, Fengrui Rao and Jiaqi Chen 72
A coarse-grained model for carbon nanotube/polymer nanocomposites
Behrouz Arash and Timon Rabczuk 74
Patricio Rodriguez and Xiaoying Zhuang 75
A finite element framework for vibration problem of hybrid nanocomposite beams reinforced with
graphene platelet and carbon fiber
Farzad Ebrahimi, Ali Dabbagh and Timon Rabczuk 77
Modeling effect of thin film’s surface roughness on nanoindentation test using random forest
Mohammad Javad Rabiei Faradonbeh, Saleh Akbarzadeh and Mohammad Silani 79
Assessing the spalling risk of fire-loaded tunnel linings by multifield model and machine learning
strategy
Zhiran. Gao and Yiming Zhang 80
Computational Modeling for Effect of Crack Healing Pattern in Self-Healing Concrete
John Hanna 81
A data-driven solution for the governing equation of flexoelectricity
Bo He and Xiaoying Zhuang 83
Synthesis and study of copper ferrite-copper oxide nanocomposites
Kambiz Hedayati and Tavus Hosseinabad 85
Comprehensive Study of Crack Propagation of different Percentage of Doping of Boron Atoms on
the Mechanical Properties of Polycrystalline Graphene
Mohammadreza Izadifar, Soheil Gohari, Colin Burvill and Timon Rabczuk 87
Phase field modelling of microstructural length scale effects on stressed grain growth in
polycrystalline copper thin films
Mostafa Jamshidian and Timon Rabczuk 89
Parametric Investigation of Large Stretchability of Graphene Nanoribbon Springs
Brahmanandam Javvaji, Bohayra Mortazavi and Xiaoying Zhuang 91
Multiscale computation based on FNN and RNN
Bin Li and Xiaoying Zhuang 93
A Manifold Learning Approach for Multiscale Phase Field Evolution for Fracture
Yangyuanchen Liu, Kexin Weng and Yongxing Shen 95
Numerical simulations on the material point method of impact damage problems
Jing-xin Ma 97
Numerical Modeling of Crack Propagation in the Presence of Inclusion Using Extended Finite
Element Method
Luthfi M. Mauludin 99
Active Machine Learning to Design Novel Two-dimensional Carbon Nitrides
Bohayra Mortazavi and Xiaoying Zhuang 100
Prediction of bending angle of laser bent tailor welded blanks using artificial neural network
(ANN)
Mohammad Ali Mousavi, Mehdi Safari and Mohammad Gohari 102
Multi-objective optimization of perforated variable-stiffness plates using NSGA-II
Saeid Nikbakht 103
Investigation of nonlinear self-healing theory in continuum damage-healing mechanics
Chahmi Oucif and George Z Voyiadjis 105
Application of an adaptive neuro fuzzy inference system for modelling of temperature and force in
robotic bone drilling process
Vahid Tahmasbi, Amir Hossein Rabiei and Mehdi Safari 106
Modelling of creep age forming of aluminum 7075 tailor machined blanks using adaptive neuro
fuzzy inference system (ANFIS)
Mehdi Safari, Vahid tahmasbi and Amir Hossein Rabiee 107
Isogeometric Analysis for Phase-Field Models of High-Order Surface PDEs with Application to
biomembrane modeling
Navid Valizadeh 108
DACOMA-2019
Influence of Thermostatting on Nonequilibrium Molecular Dynamics Simulations of Heat

Transport in Solids
Zhen Li, Shiyun Xiong, Charles Sievers, Yue Hu, Zheyong Fan, Ning Wei,1 Hua Bao, Shunda Chen,
Davide Donadio and Tapio Ala-Nissila 109
Crack Propagation with Different Radius Local Random Damage based on peridynamic theory
Jinhai Zhao and Meng Jia 111
Topology optimization of piezoelectric materials
Chuong Nguyen and Xiaoying Zhuang 113
A prediction model of debris flow susceptibility in Southwest China based on a quantitative data
processing method
Feng Ji and Zili Dai 114
Prediction of shield moving performance during tunnelling based on optimized LSTM neural
network
Song-Shun Lin, Shui-Long Shen, Annan Zhou 116
Data-driven approach to Inverse Problems using Deep Learning Networks
Ngoc-Hien Nguyen, Vinh Phu Nguyen, Thi-Hong-Hieu Le, and Yan Ding 118
Data-driven methods for quantifying the effects of geological variables on performance degradation
of metro tunnel in operation
Xin Wang and Hehua Zhu 119
Study on K-means Evaluation Method for Service Performance of Highway Tunnels
Chong-bang Xu, Fa-you Deng, Xiao Xu and Gang Liu 121
A LSTM framework in predicting changes in digital parameters of granite compositions under
uniaxial compression load
Yalei Wang, Jinming Xu, and Xuejie Tao 123
A newly-developed alloy gas atomization process and the powder product used in selective laser
melting process
Pengfei Yan and Biao Yan 125
Prediction of Undrained Shear Strength of Soft Clays Using LightGBM Method Based on Bayesian
Optimization
Wengang Zhang, Chongzhi Wu, Runhong Zhang and Yongqin Li 126
A Review in Recent Development of 5 Axis CNC Milling Machine Tool Operations
Mohsen Soori and Mohammed Asmael 128
Mineral Segmentation from Polarized Light Microscopy Images of Rock Slices via Deep Neural
Networks
Qimin Wang, Yanlu Ding and Xiaoying Zhuang 130
A Novel Method for Health Assessment of Subway Shield Tunnel Structure Based on Knowledge
Graph
Yadong Xue, Hongfei Zhang and Peizhe Shi 132
An adaptive collocated method for PHT splines with optimal selection of collocation points
Yue Jia, Cosmin Anitescu, Yongjie Jessica Zhang and Timon Rabczuk 134
Optimal Placement of Active Members in Adaptive Cable-Strut tensile Structure Using an Optimal
Combination Algorithm
Sha Li and Nan Xiao 135
Two Applications of Isogeometric Boundary Element Method for Axisymmetric Singular
Helmholtz Acoustic Problems
Ahmed Mostafa Shaaban 137
Parameter optimization in penalty-based implicit discontinuous approach for efficient rock failure
analysis
Fei Zheng and Xiaoying Zhuang 138
Application of Machine Learning to the Prediction of Displacement Capacities of Non-ductile RC
Beam-Columns
Reza Allahvirdizadeh, Mohammad Reza Azadi Kakavand, De-Cheng Feng and
Ertugrul Taciroglu 140
Damage Evolution and Energy Absorption Modeling of Laminated GFRP Plate with Weave
Pattern Variation under High Velocity Impact
Ilyani Akmar Abu Bakar, Muhammad Fakhrullah Ramli, Zarina Itam and
Mohd Khairul Kamarudin 142
An experimental study on natural frequencies measurement in steel plates/beams using PAK-
Mobile MKII system
Soheil Gohari, Mohammadreza Izadifar, Timon Rabczuk, Colin Burvill, Saeed Mouloodi and
Naeim Akbari Shahkhosravi 144
Assessment of pipe-pile integrity using transverse velocity signal -- Numerical simulation
Zhitang Lu, Dongjia Liu, Zili Dai, Yang Guo and Zhaibang Ke 145
Experimental and numerical analyses on mixing uniformity of water and saline in pipe flow
Bin Sun, Yuanbo Lu, Hongyuan Fang, Yisheng Zhang and Yihua Chang 147
On the Generation of Design Response Spectrum for Muscat Region in Oman
Mohamed Alaa Tabnaj, Himanshu Gaur and Ram Kishore Manchiryal 148
Effects of Radial Charge Parameters of Liner on the Velocity of Shaped Charge Jet
Mengwen Xu 150
DACOMA-2019
Application of Multi Objective Bees Algorithm to Access the Optimal Generalization of Nonlinear
Full Vehicle Model
Reza Khademi Zahedi, Abbas Moradi and Afshin Ghanbarzadeh 151
Machine learning in the engineering crack identification problems
Chao Zhang and Cuixia Wang 153
Road surface damage detection using neural network
Sara Changizi and Jafar Amani-Dashlejeh 154
Modeling the stability of a soil-rock-mixture slope based on the digital image technology and
numerical manifold method
Tao Chen, Yongtao Yang and Hong Zheng 155
DEM study of modeling method and thermal cracking of sandstone
Zhang Wengang, Wang Zhenyu, Wu Chongzhi and Gao Xuecheng 157
Hip Fracture Risk Assessment based on Different Failure Criteria using QCT-based Finite Element
Modeling
Hossein Bisheh and Yunhua Luo 158
Parametric Study of Hip Fracture Risk using QCT-based Finite Element Model
Hossein Bisheh and Yunhua Luo 160
Phase field fracture in elasto-plastic solids: a length-scale insensitive model for constitutive
behaviour of quasi-brittle materials
Jianguang Fang 161
Research on the Design Construction Building Information Exchange (DCBIE) based on BIM
Juan Ju 162
Optimization and Studying the Effect of Machining Parameters in the Cortical Bone Turning by
Sensitivity Analysis
Vahid Tahmasbi, Mehdi Safari and Jalal Joudaki 163
A machine learning-based continuum damage model for predicting complex crack propagation
Tran Quoc Thai and Xiaoying Zhuang 164
Methods Based on Artificial Neural Networks for the Solution of Partial Differential Equations on
Domains with Complex Geometries
Cosmin Anitescu, Elena Atroshchenko and Somdatta Goswami 166
Research on Application of Machine learning and Optimization Algorithm in Transparent Soil: A
Review
Bo Wang and Zhengwei Zhu 167
Thermal conductivity identification of nonlinear functionally graded materials via a machine

learning strategy
Zhuo-jia Fu, Wen-zhi Xu and Qiang Xi 169
An Alternative Optimization Algorithm to Analysis of Structure by Hybrid Artificial Neural
Network-Isogeometric Finite Element Program
D. Nguyen Kien, Xiaoying Zhuang and Timon Rabczuk 170
Deep learning algorithm for solving PDEs for fluid flow field
Juan Lin and Shuwei Zhou 172
DML based artificial neural network approach to solve transient wave equation using the
Collocation method
Abhishek Mishra, Cosmin Anitescu and Pattabhi R. Budarapu 173
Phononic crystal band-gap detection by Machine learning technique
Binh Nguyen and Xiaoying Zhuang 175
Extended isogeometric-meshfree collocation methods without branch enrichment for cracks analysis
under contact loading
Nhon Nguyen-Thanh, Weidong Li and Kun Zhou 177
Numerical simulation of Von-Karman plate problem by nonlocal operator method
Huilong Ren 179
Data Driven solution for 1D Cahn-Hilliard Equation
Jorge López-Zermeño, Somdatta Goswami and Navid Valizadeh 180
Probabilistic characterization of rock mass deformation modulus from P-wave velocity based on
information criteria
Xianbin Huang, Hehua Zhu, Keshen Zhang, Wei Wu, Lianyang Zhang 182
Stochastic multiscale modeling of thermal properties of Polymeric clay nanocomposites and global
sensitive analysis
Bokai Liu 184
Uncertainty Analysis of Engineering Structures by Time Domain Spectral Element Based Modeling
Shuvajit Mukherjee, Srinivasan Gopalakrishnan and Ranjan Ganguli 185
Probabilistic stability analysis of earth dam slope using extreme gradient boosting method
Lin Wang and Wengang Zhang 187
Probabilistic Assessment of Serviceability Limit State of Diaphragm Walls for Braced Excavation
in Clays using Point Estimate Method
Wengang Zhang, Runhong Zhang, Chongzhi WU and Anthony T.C. Goh 188
Data Driven Computing and Machine Learning in Engineering 1
Machine learning based solution of Partial Differential Equations in

Computational Mechanics: concepts, implementation and applications
Timon Rabczuk1,* and Xiaoying Zhuang 2,3
1 Institute of Structural Mechanics, Faculty of Civil Engineering, Bauhaus University Weimar, Weimar 99425, Germany
2 Department of Geotechnical Engineering, Tongji University, Shanghai 200092, China
3 Institute of Continuum Mechanics, Leibniz University Hanover, 30167 Hannover, Germany

*Corresponding Author: Timon Rabczuk. Email: timon.rabczuk@uni-weimar.de.
Abstract: Partial Differential Equations (PDE) are fundamental to model different phenomena in science
and engineering mathematically. Solving them is a crucial step towards a precise knowledge of the
behaviour of natural and engineered systems. In general, in order to solve PDEs that represent real
systems to an acceptable degree, analytical methods are usually not enough. One has to resort to
discretization methods. For engineering problems, probably the best known option is the Finite Element
Method (FEM). The fundamental idea is to approximate the solution of the PDE by means of functions
specifically built to have some desirable properties. In this contribution, we explore Deep Neural
Networks (DNNs) as an option for approximation. They have shown impressive results in areas such as
visual recognition. DNNs are regarded here as function approximation machines. There is great flexibility
to define their structure and important advances in the architecture and the efficiency of the algorithms to
implement them make DNNs a very interesting alternative to approximate the solution of a PDE. We
concentrate in applications that have an interest for Computational Mechanics. Most contributions that
have decided to explore this possibility have adopted a collocation strategy. In this contribution, we
concentrate in mechanical problems and analyze the energetic format of the PDE. The energy of a
mechanical system seems to be the natural loss function for a machine learning method to approach a
mechanical problem. As proofs of concept, we deal with several problems and explore the capabilities of
the method for applications in engineering.
DOI: 10.32604/acm.2019.00001 www.techscience.com/acm

2 DACOMA-2019
Figure 1. A comparison of displacement magnitude at the CDEF plane between reference solution and
deep energy method of a twisted Neo-Hookean 3D Cuboid.
Keywords: Deep energy method; Energy approach; Physics informed; Machine learning; Partial differential
equations; Hyperelastic model;
Acknowledgement
The second author thankfully acknowledges the Sofja Kovalevskaja Programme from Alexander von Humboldt
Foundation.
References
1. M. Raissi. (2018). Deep hidden physics models: Deep learning of nonlinear partial differentialequations, The
Journal of Machine Learning Research 19: 932-955.
2. T. Kirchdoerfer, M. Orti (2016)., Data-driven computational mechanics, Computer Methods in Applied
Mechanics and Engineering 304: 81-101.
3. H. Guo, X. Zhuang, and T. Rabczuk. (2019)A deep collocation method for the bending analysis of Kirchhoff
plate. Computers, Materials and Continua, 59:433-456.
4. C. Anitescu, E. Atroshchenko, N. Alajlan, and T. Rabczuk. (2019)Articial neural network methods for the
solution of second order boundary value problems. Computers, Materials and Continua, 59:345-359.
Machine Learning for Seawall Structures Under Flooding and Erosion
Hung Nguyen-Xuan1,*
1CIRTECH Institute, Ho Chi Minh City University of Technology (HUTECH), Ho Chi Minh City, Viet Nam
*Corresponding Author: Hung Nguyen-Xuan. Email: ngx.hung@hutech.edu.vn.
Abstract: Climate change has been changing in a harmful way to human life for recent years. Especially, the global
average surface temperature has been highly increasing with a disturbing trend. It causes negative effects on our
planet such as frequently occurring flood and hurricane with very high intensity in some area, but others had a long
period of drought. Additionally, the most dangerous effect of global warming is the high rate of melting a glacier
and sea ice at both the Earth’s North and South poles leading to the sea level risen dramatically threatening to many
countries located near coastlines. There are many fatal effects of increasing sea level. For example, the vanishing of
various lands on the world map or a lot of infrastructures close to shorelines have been destroying because of the
erosion. Therefore, researching and developing advanced methods for the resistance of flood and erosion is an
urgent task. Especially, the monitoring heath structure field can help to control the safety of a complex structure
to estimate any severe collapses. In this study, a hollow concrete block (HCB) used to prevent coastal erosion has
been designed [1, 2]. Furthermore, machine learning has been applied to this design to predict an existed damage
inside the model with a variety of fiber concrete material properties considered. Due to the lack of experiment data,
the numerical simulations were employed to generate many datasets from the HCB model (see Figure 1).
Figure 1. Framework of CNN for predicting damage of a hollow seawall block.

By changing many cases of boundary conditions and load directions on the HCB model to cover all situations can
exist in a real seawall structure. Each dataset was obtained from two HCB models, one of them was acquired from
an undamaged structure, while the other was obtained from a damaged one by applying a static load enough to make
a random failure inside it. A harmonic load was applied to these models. Then a load point displacement with
respect to time was collected as an output signal. This signal was switched to the frequency domain by employing
Fast Fourier Transform [3]. Prior to putting the treated signal into Convolution Neural Networks (CNNs) [4], the
DOI : 10.32604/acm.2019.00003 ww.techscience.com/acm

4 DACOMA-2019
signal was normalized and extracted a two-dimensional matrix which contains values one and zeros where the value
one presents a curved of the output signal. The dataset then was trained by CNNs to classify which signal is come
from a damaged or undamaged model. The result shows that this method can predict accurately an existed failure
inside the HCB model. By applying this method to the real seawall structures, we can predict which HCB should be
fixed or replaced before all structure may be collapsed as shown in Figure 1.
Keywords: seawall structure, fibre reinforced concrete, damage modelling, vibration-based damage detection,
machine learning, convolution neural network.
References
1. H. Nguyen-Xuan, GR. Liu, S. Bordas, S. Natarajan, T. Rabczuk. (2013). An adaptive singular ES-FEM for
mechanics problems with singular field of arbitrary order, Computer Methods in Applied Mechanics and
Engineering, 253: 252-273.
2. Bao-Loi Dang, Hau Nguyen-Ngoc, H. Nguyen-Xuan, Magd Abdel-Wahab, Numerical investigation of novel
prefabricated hollow concrete blocks for stepped-type seawall structures, Engineering Structures, 198, 109558,
2019.
3. E. O. Brigham, R. I. Morrow, "The Fast Fourier Transform", IEEE Spectrum, vol. 4, 63-70, December 1967.
4. Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton. ImageNet classification with deep convolutional neural
networks. Advances in neural information processing systems, DOI: 10.1145/3065386, 2012.
Data-driven Construction and Maintenance of Urban Tunnels

Hongwei Huang1,*
1Department of Geotechnical Engineering, Tongji University, Shanghai, 200092, China.
*
Corresponding Author: Hongwei Huang. Email: huanghw@tongji.edu.cn.
Abstract: With the development of structure health monitoring (SHM) and geotechnical testing techniques, huge
amounts of data are collected from the design, construction and maintenance period of urban tunnels. To face the
challenge of data explosion and data effective utilization, quite a number of data analysis methods from the grey
system, time series to modern machine learning based algorithms have been paid much attentions to. 3 major cases
are taken to illustrate the application effect of machine learning approach in the lifetime of urban tunnels:
(1) Shanghai clay database is compiled and learned for characterization of spatial variability and transformation
uncertainty. It could help for the reliability-based design of tunnel structures.
Bayesian Framework
soil parameters Non-normal Yi to δij between Correlation matrix C
to Yi normal Xi Xi、Xj
Logarithm Xi =Φ-1 [Fi (Yi )]

Bootstrap Bootstrap
transformation Fit Johnson Distributions
random field discretion

Yi Xi n=18
n=2229
δ12
δ56
(a) (b)
Fig. 1 Machine learning based estimation of soil properties: a) soil compressive modulus; b) multi-variate
distribution of soil parameters.
(2) ML algorithms (ANN, SVP, GBR) are employed to predict the shield machine posture during the construction
period. Based on the trained prediction models, construction parameters are optimized to make a better control of
tunnel settlement.
DOI : 10.32604/acm.2019.00005 www.techscience.com/acm

6 DACOMA-2019
(3) Multi sensor data are fused using the extended Kalman filter (EKF) method for a better understanding of
unknown parameters. Then an integral safety assessment is conducted based on the predicted parameters.
The study shows ML has a great prospect and can solve complicated issues in Geostructural system field.
8.00
CONVERGECNE (MM)
7.00
6.00
Assembling error &
5.00 Ground disturbance
structural quality etc.
4.00
3.00
(a) Leakage (b) Crack
Convergence with J1
2.00
Convergence with J2
1.00 JOINT OPENING (MM) Tunnel Deformation Leakage & crack
0.00
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70
σv=γH Monitoring Data Inspection Data

(c) Spalling (d) Offset
Tunnel defect diagnosis

Data Preprocessing
Structural deformation Assembly error & Inspecting
distribution material degradation data
Data Fusion
Multi-sensor Fusion
Assembly error
Process
Distribution of
Leakage of joint
joint opening
（Continuous BN） Rubber quality
Distribution of max Inspecting crack

crack width
Diagnosis of Defects Existing crack
(Discrete BN) Distribution of

internal force
Concrete spalling
Concrete strength
degradation
Reduced Bayesian network
Cause of defects
Fig. 3 Framework of dynamic tunnel structual safety assessment based on Bayesian network
Keywords: Structure Health Monitoring, Data driven construction and maintenance; Data fusion process; Machine
Learning.
Reference
1. Chen, J.-J., Zhang, W. & Wang, J.-H. (2017). Data Fusion Analysis Method for Assessment on Safety
Monitoring Results of Deep Excavations. Journal of Aerospace Engineering, 30, B4015005.
2. Ching, J., Phoon, K.K., and Chen, C.H. (2014). Modeling CPTU parameters of clays as a multivariate normal
distribution. Canadian Geotechnical Journal, 51(1): 77-91.
3. Kulhawy, F.H. and Mayne, P.W. (1990). Manual on estimating soil properties for foundation design. Report
EL-6800, Electric Power Research Institute, Palo Alto, California.
4. Wang, F., Huang, H. W., He, B., Wu, Y., Shao, H. & Wu, H. M. (2016). Wireless sensing on shield tunnels in
Shanghai. Proceedings of the International Conference on Smart Infrastructure and Construction, 191-196.
5. Huang, H.W., Xiao, L., Zhang, D.M. and Zhang, J. (2017). Influence of spatial variability of soil Young’s
modulus on tunnel convergence in soft soils, Engineering Geology, 228:457-370.
Solving direct and inverse computational engineering problem through a deep

learning based data-driven method
Zhanli Liu*, Xiang Li, Haolong Chen, Chengcheng Luo and Zhuo Zhuang
Applied Mechanics Lab., Dept. of Engineering Mechanics, School of Aerospace Engineering,

Tsinghua University, Beijing 100084 China.
*Corresponding Author: Zhanli Liu. Email: liuzhanli@tsinghua.edu.cn.
Abstract: With the steady development of computer science, machine learning and data science have made
significant progress in recent decades. These techniques generally rely on a substantial amount of data samples to
extract the abstract mapping hidden within the data. Hence, these technologies have gradually attracted the attention
of researchers in the field of computational mechanics and computational engineering. This paper aims to interpret
several forms of applications that integrate machine learning and data science with computational mechanics and
computational engineering problems. In the first application, the core algorithm of the convolutional neural network
is implemented to solve the linear elastic finite element problem. A standard finite element equation is transformed
into an optimization problem in this method. The method is verified by a plane strain linear elastic finite element
problem. The method demonstrates promising accuracy by comparing the results obtained by traditional finite
element solver. In the second application, a method is proposed to establish the implicit mapping between the
effective mechanical property and the mesoscale structure of heterogeneous materials. Shale is employed in this
paper as an example to illustrate the method. At the mesoscale, a shale sample is a complex heterogeneous
composite that consists of multiple mineral constituents. The mechanical properties of each mineral constituent vary
significantly, and mineral constituents are distributed in an utterly random manner within shale samples. Large
quantities of shale samples are generated based on mesoscale scanning electron microscopy images using a
stochastic reconstruction algorithm. Image processing techniques are employed to transform the shale sample
images to finite element models. Finite element analysis is utilized to evaluate the effective mechanical properties of
the shale samples. A convolutional neural network is trained based on the images of stochastic shale samples and
their effective moduli. The trained network is validated to be able to predict the effective moduli of real shale
samples accurately and efficiently. Not limited to shale, the proposed method can be further extended to predict
effective mechanical properties of various heterogeneous materials. In the third application, assisted by image-based
finite element analysis and deep learning, a data-driven approach is proposed for designing phononic crystals. An
auto-encoder is trained to extract the topological features from sample images. Finite element analysis is employed
to study the band gaps of samples. A multi-layer perceptron is trained to establish the inherent relation between band
gaps and topological features. The trained models are ultimately employed to design phononic crystals with
anticipated band gaps. Not limited to this material, the proposed method could be further extended to design various
structured mechanical materials with specific functionalities.
Keywords: Computational engineering, computational mechanics, deep learning, finite element analysis, inverse
problem.
Acknowledgement
This work is supported by the National Natural Science Foundation of China, under Grant No. 11722218, No.
11532008, Tsinghua University Initiative Scientific Research Program.
References
1. Li, X., Liu, Z., Cui, S., Luo, C., Li, C., & Zhuang, Z. (2019). Predicting the effective mechanical property of
heterogeneous materials by image based modeling and deep learning. Computer Methods in Applied
Mechanics and Engineering, 347: 735-753.
2. Li X., Yan Z, Liu Z. (2019). The combination and application of machine learning and computational
mechanics (in Chinese). Chin Sci Bull, 64: 635-648.

An investigation of deep learning in predicting shield tunnel excavation

parameters
Xiaojun Li*, Dong Han and Gang Li
Department of Geotechnical Engineering, College of Civil Engineering, Tongji University, Shanghai 200092 China.
*Corresponding Author: Xiaojun Li. Email: lixiaojun@tongji.edu.cn
Abstract: Shield tunneling construction method in a popular technology in building urban infrastructures. For
practitioners, optimizing shield tunnel driving parameters and reducing ground settlement induced by construction is
a major objective, especially in highly populated urban areas. Traditionally, calibrating shield tunnel driving
parameters is a try-and-error process with high cost, and it depends largely on past experiences. In this study, we
investigated the adoption of deep learning model, LSTM, in predicting shield tunnel construction parameters. The
investigation data is collected from the construction of a shield tunnel in shanghai urban areas, and totally 1,387
rings of construction parameters are recorded. The construction parameter is first treated as one-dimensional time
series data, and LSTM is used to predict each construction parameter separately. The LSTM prediction is compared
with autoregressive integrate moving average model (ARIMA). Results show the maximum errors of LSTM and
ARIMA prediction are 12% and 23%, respectively. Then, for construction under normal circumstances, the
curvature, slope and depth of the tunnel are taken as input of the deep learning model to predict the driving
parameters. For construction which crosses existing buildings, foundation types, building structure types, current
building conditions, and ground conditions are taken as additional input to predict the driving parameters. Results
show that the maximum prediction errors for normal circumstances and crossing situation are 9% and 12%,
respectively. This study shows that deep learning is a very promising method in predicting shield tunnel
construction parameters. The predicting accuracy could be very high even in very complicated situations such as
crossing existing buildings.
Keywords: Shield tunnel, construction parameters, deep learning, LSTM.
Acknowledgement
This work is supported by Shanghai Municipal Science and Technology Commission with Grant No. 14231200604.
References
1. Li X, Zhang C, Yuan D. (2013). An in-tunnel jacking above tunnel protection methodology for excavating a
tunnel under a tunnel in service. Tunnelling and Underground Space Technology, 34, 22-37.
2. Hochreiter S, Schmidhuber J. (1997). Long short-term memory. Neural Computation, 9(8), 1735-1780.
3. Sang C, Di Pierro M. (2019). Improving trading technical analysis with TensorFlow Long Short-Term Memory
(LSTM) Neural Network. The Journal of Finance and Data Science, 5(1), 1-11.

The application of a deep energy method in the bending and eigenvalue

analysis of Kirchhoff plate
Xiaoying Zhuang1,2,* and Hongwei Guo1
1 Institute of Continuum Mechanics, Leibniz Universität Hannover, Appelstraße 11, Hannover, 30157, Germany.
2 Department of Geotechnical Engineering, Tongji University, Siping Road 1239, Shanghai, 200092, China
*Corresponding Author: Xiaoying Zhuang. Email: zhuang@ikm.uni-hannover.de.
Abstract: In this article, a deep energy method is proposed for the bending and eigenvalue analysis of Kirchhoff
plate. As is known to all, deep learning has recently become a central and popular branch in research and
applications, such as image processing, object detection, finance and even engineering [1], due to its potential in
nonlinear processing of data with multiple hidden layers in supervised or unsupervised learning. As for the
application of deep learning in civil engineering [2], there are three major approaches including the solving of PDEs
with deep learning [3, 4], the surrogate of FEM by deep learning [5], and the studying of data-driven problems in
engineering such as liquefaction with deep learning [6-8]. And in this article, we further broaden the scope of
solving PDEs with deep learning. Instead of focusing on a dataset of targeted solutions, we directly deploy a deep
feedforward neural network to discover those physical patterns involved in a certain mechanical phenomenon.
Virtually, many engineering problems and natural science are governed by partial differential equations, revolving
engineering mechanics [9-13], rock and soil engineering [14-22] et al.. Many classic analytical and numerical
methods have been proposed and discussed to solve those problems, especially after the blossoming of computer
technology, such as separation of variables, lie group method, semi-analytical methods, numerical methods, et al.
After the birth of artificial neural network method, it was also applied in solving those PDEs, but did not show
excellent performance until recently. With improved algorithms in computer science, advanced computer hardware,
and more accessible data from Internet, deep learning, which is based on a deep neural network set, has shown
favorable features that can be applied in solving partial differential equation. Moreover, our research group has
observed that the deep neural networks can be further applied to unveil those physical patterns laying behind some
physical problems by placing a deep neural network prior on field variables. A neural network severs as universal
function approximators here [23], including computational graphs and backpropagation algorithms involved in deep
learning. Unlike previous applications in directly dealing with governing equations, the deep energy method, which
combines a deep learning technique with the energy minimum principle, has been established in a unified
framework for solving this higher-order problem. The whole procedure begins with generating randomly distributed
collocation points in the physical domain, as a result of which this method turns out to be truly “Meshfree”. Then a
loss function is built with the aim that the total potential energy is minimized at those collocation points. The L-
BFGS optimizer is adopted to backpropagate the loss function and train the whole model. It needs to be emphasized
here that, to better reveal its physical patterns, an autoencoder is added to the deep feedforward neural network and a
premium layer and neuron configuration is discussed in this article. As for the activation function choice, different
schemes indeed have an impact on the accuracy and efficiency of this deep energy method. Accordingly, a tailored
activation function based on analytical solution is adopted here and compared with classical tanh activation function,
which is proven to be a better choice. In this way, the constructed deep neural network can approximate deflection in
Kirchhoff bending problems, where the C1 continuity requirement has posed significant difficulties in traditional
mesh-based methods. However, this can be solved in the framework of DEM, where the deep neural network can be
deployed to approximate the transversal deflection. The DEM can be also suitable for the eigenvalue analysis of
structures, where a different formulation of loss function is adopted. Once the neural network is trained, it can be
efficient and accurate implemented to predicted deflection and stress in the whole physical domain. Numerical
examples have demonstrated that the proposed DEM can be suitable for the bending and eigenvalue analysis of
Kirchhoff plate, which ushers in a new scheme for the mechanical analysis of structural components in engineering.
Keywords: Deep learning, Deep energy method, Autoencoder layer, Activation function, Kirchhoff plate.

10 DACOMA-2019
References
1. Vargas, R., Mosavi, A., Ruiz R. (2017). Deep learning: a review. Advances in Intelligent Systems and Computing.
2. Lee, J.H., Shin J., and Realff M.J. (2018), Machine learning: Overview of the recent progresses and
implications for the process systems engineering field. Computers & Chemical Engineering, 114, 111-121.
3. Anitescu, C., et al. (2019). Artificial neural network methods for the solution of second order boundary value
problems. Computers, Materials & Continua, 59(1), 345-359.
4. Guo, H., Zhuang X., Rabczuk T. (2019). A Deep Collocation Method for the Bending Analysis of Kirchhoff
Plate. Computers, Materials & Continua, 59(2), 433-456.
5. Liang, L., et al. (2018). A deep learning approach to estimate stress distribution: a fast and accurate surrogate
of finite-element analysis. Journal of The Royal Society Interface, 15(138), 20170844.
6. Moayedi, H., et al. (2018). A systematic review and meta-analysis of artificial neural network application in
geotechnical engineering: theory and applications. Neural Computing and Applications, 1-24.
7. Ismail, A. (2018). ANN-based empirical modelling of pile behaviour under static compressive loading.
Frontiers of Structural and Civil Engineering, 12(4), 594-608.
8. Zakian, P. (2017). An efficient stochastic dynamic analysis of soil media using radial basis function artificial
neural network. Frontiers of Structural and Civil Engineering, 11(4), 470-479.
9. Ghosh, P., Rajesh S., Chand J.S. (2017). Linear and nonlinear elastic analysis of closely spaced strip
foundations using Pasternak model. Frontiers of Structural and Civil Engineering, 11(2), 228-243.
10. Yin, R., et al. (2018), Damage on lining concrete in highway tunnels under combined sulfate and chloride
attack. Frontiers of Structural and Civil Engineering, 12(3), 331-340.
11. Bera, A.K., et al. (2018), A multi-attribute decision making approach of mix design based on experimental soil
characterization. Frontiers of Structural and Civil Engineering, 12(3), 361-371.
12. Memarzadeh, R., Barani G., Ghaeini-Hessaroeyeh M. (2018). Application of a weakly compressible smoothed
particle hydrodynamics multi-phase model to non-cohesive embankment breaching due to flow overtopping.
13. Martínez-Galván, S. A., Romo M. P. (2018), Assessment of an alternative to deep foundations in compressible
clays: the structural cell foundation. Frontiers of Structural and Civil Engineering, 12(1), 67-80.
14. Wang, G., et al. (2017), Numerical study on the seismic response of the underground subway station-surrounding
soil mass-ground adjacent building system. Frontiers of Structural and Civil Engineering, 11(4), 424-435.
15. Zhang, Z., Wu J., Huang X. (2017). Application of a vertex chain operation algorithm on topological analysis of
three-dimensional fractured rock masses. Frontiers of Structural and Civil Engineering, 11(2), 187-208.
16. Han, J., Guo J. (2017). Geosynthetics used to stabilize vegetated surfaces for environmental sustainability in
civil engineering. Frontiers of Structural and Civil Engineering, 11(1), 56-65.
17. Yao, H., et al. (2017). Construction risks of Huaying mount tunnel and countermeasures. Frontiers of
Structural and Civil Engineering, 11(3), 279-285.
18. Wu, J. J., Wang L. J., Cheng Q. G. (2017). Soil arching effect of Lattice-Shaped Diaphragm Wall as bridge
foundation. Frontiers of Structural and Civil Engineering, 11(4), 446-454.
19. Ahn, I. S., Cheng L. (2017), Seismic analysis of semi-gravity RC cantilever retaining wall with TDA backfill.
20. Sun, W., Wang L., Wang Y. (2017). Mechanical properties of rock materials with related to mineralogical
characteristics and grain size through experimental investigation: a comprehensive review. Frontiers of
Structural and Civil Engineering, 11(3), 322-328.
21. Far, H., Flint D. (2017). Significance of using isolated footing technique for residential construction on
expansive soils. Frontiers of Structural and Civil Engineering, 11(1), 123-129.
22. Xu, G., et al. (2018). Shanghai center project excavation induced ground surface movements and deformations.
23. Hornik, K., Approximation capabilities of multilayer feedforward networks. Neural networks, 4(2), 251-257.
Isogeometric analysis of a phase field model for simulating electromechanical

behavior and morphological evolution of vesicles
Mohammed Ashour1,* and Navid Valizadeh1
1Bauhaus University of Weimar, Marienstraße 15, 99423 Weimar, Germany.
*Corresponding Author: Mohammed Ashour. Email: mohammed.ashour@uni-weimar.de
Abstract: In this paper, we develop a phase field model for a vesicle in an electrical field. We aim to obtain the
morphological evolution of the system by solving it using Isogeometric analysis. Both mechanical and electrical
aspects have been considered, such as elastic bending energy, flexoelectricity, dielectricity, osmotic pressure and
surface tension. An energy functional represents the system is reformulated using a phase field variable, which
produces a fourth-order PDE. This PDE is highly nonlinear with moving boundaries. Solving fourth-order PDEs
without the need to introduce additional degrees of freedom or using mixed-formulation necessitate basis functions
that are piecewise smooth and globally C1 continuous. Isogeometric analysis possesses the capability of introducing
smooth C1 globally-continuous basis functions using B-splines. The primal variation of the energy functional is
obtained from which a weak form of two primary variables is formulated. The model has axisymmetric
configurations for which the morphological evolution of the vesicle is obtained. The results demonstrate the
robustness of the phase-field model in conjunction with IGA to solve fourth-order PDEs without the need to
introduce additional mathematical complexities. The effect of the mechanical-electrical coupling, electrolytes within
the vesicle and outside, the flexoelectricity and the dielectricity are examined in details.
Keywords: Isogeometric Analysis, Phase-Field, Helmholtz Free Energy, Mechanical-Electrical coupling, Vesicles,
Cell membrane.
Acknowledgement
The First author has a DAAD scholarship, which provides him with financial support during his PhD studies in
Germany. This support is greatly acknowledge.
References
1. Du, Qiang, Chun Liu, and Xiaoqiang Wang. "A phase field approach in the numerical study of the elastic
bending energy for vesicle membranes." Journal of Computational Physics 198, no. 2 (2004): 450-468.
2. Elliott, Charles M., and Björn Stinner. "A surface phase field model for two-phase biological membranes."
SIAM Journal on Applied Mathematics 70, no. 8 (2010): 2904-2928.
3. Gao, Ling-Tian, Feng Xi-Qiao , Ya-Jun Yin, and Huajian Gao. "An electromechanical liquid crystal model of
vesicles." Journal of the Mechanics and Physics of Solids 56, no. 9 (2008): 2844-2862.
4. Gao, Ling-Tian, Xi-Qiao Feng, and Huajian Gao. "A phase field method for simulating morphological
evolution of vesicles in electric fields." Journal of Computational Physics 228 (2008): 4162-4181.
5. Gómez, H., V. M. Calo, Y. Bazilevs, and T. J. Hughes. "Isogeometric analysis of the Cahn–Hilliard phase-field
model." Computer methods in applied mechanics and engineering 197, no. 49-50 (2008): 4333-4352.
6. Valizadeh, Navid, and Timon Rabczuk. "Isogeometric analysis for phase-field models of geometric PDEs and
high-order PDEs on stationary and evolving surfaces." Computer Methods in Applied Mechanics and
Engineering 351 (2019): 599-642.

Exploration signals from higher dimensions base on machine learning

Cong Cao1
1Bauhaus University Weimar, Geschwister-Scholl-Straße 8, 99423 Weimar, Germany.
Abstract:New technologies have enabled measurements on additional levels of the gene expression processes.The
use of high-throughput technologies for finding reliable combinations poses new challenges [1]. First, in contrast to
standard blood values,patient data based on high-throughput technologies are not commonly available in hospitals.
This means, they have to be generated separately. Secondly, although the prices for generating data from high-
throughput technologies are constantly decreasing, it is still quite expensive and time-consuming to carry out these
experiments. This implies high costs for such data. The advantage and the curse at the same time is that such
experiments easily yield thousands or even far beyond 10,000 possible candidates for signal. The sample size in
expensive pilot studies is usually very limited, sometimes less than 20. From a machine learning or classification
point of view one then tries to derive a classifier from a data set with more than 1000 candidates and perhaps only
around 20 or 30 instances. Finding random associations and overfitting is therefore difficult to avoid. An essential
step for the construction of a classifier is feature selection (selecting a subset from the candidates) and sometimes
feature extraction (defining new scores based on the given candidates) [2,3]. A large number of methods for feature
selection and extraction has been developed within machine learning. However, these techniques are usually not
tailored to the typical setting of signals, where the number of features is usually very large and the number of cases
(donors, mice…) is often very limited. The study aims at adjusting state-of-the-art and developing new feature
selection and extraction methods.
Keywords: Machine learning, High-throughput technologies, Gene, Feature selection.
References
1. Bolon-Canedo V,Sanchez-Marono N,Alonso-Betanzos A,et al.A review of microarray datasets and applied
feature selection methods[J].Information Sciences An International Journal,2014,282(5):111-135.
2. Golub T R,Slonim D K,Tamayo P,et al .Molecular classification of cancer:class discovery and class prediction
by gene expression monitoring[J].Science,1999,286(5439):531-537.
3. Callow M J,Dudoit S,Gong E L,et al.Microarray expression profiling identifies genes with altered expression
in HDL-deficient mice[J].Genome Research,2000,10(12):2022-2029.

Volumetric parametrization using convolution neural networks

C.L. Chan1,* and C. Anitescu1
1Bauhaus University Weimar, Weimar, Germany.
*Corresponding Author: C. L. Chan. Email: chiu.ling.chan@uni-weimar.de.
Abstract: Currently, the usage of convolution neural networks (CNNs) for image processing has gained
wide-spread attention. The advantage of CNN is that it is capable to handle all the details, including the fine details
of given images [1]. Here, we present an approach for using a CNN for converting a given voxel-based images to
analysis suitable volumetric meshes. There is rich a literature on approaches for volumetric meshes generation. The
results are, however, highly dependent on the expertise of user. Therefore, it would be appealing to use machine
learning for volumetric mesh generation. For a given voxelized image, we propose to use the B-Spline filters as the
convolution layer of CNN for geometry features identification. The advantage of using B-Spline filters is that the
obtained volumetric parametrization is described in a standard CAD format and can be directly used for analysis
using an isoparametric approach [2]. We validate the creation of volumetric meshes using simple models such as
square and sphere, and further use it for robust learning of other shapes. We show the efficiency of our approach
through the implementation of volumetric parametrization and analysis on benchmark and real-world problems.
Keywords: Convolution neural networks, Volumetric parametrization, B-Spline filters.
Acknowledgement
The second author would like to acknowledge the support of the German Research Foundation (DFG project
number 392023639).
References
1. Sinha, J. Bai, and K. Ramani Biodiesel. (2016). Deep learning 3D Shape Surfaces Using Geometry Images.
Computer Vision – ECCV 2016, 2016: 223-240.
2. C.L. Chan, C. Anitescu, and T. Rabczuk. (2017). Volumetric parametrization from a level set boundary
representation with PHT-splines. Computer Aided Design, 2017: 29-41.

Boundary Shape Identification of the Heat Conduction Problem Based on

Machine Learning
Haolong Chen and Xiang Li and Chengcheng Luo and Zhanli Liu*
School of Aerospace Engineering, Tsinghua University, 100084, Beijing, China.

*Corresponding Author: Zhanli Liu. Email: liuzhanli@tsinghua.edu.cn.
Abstract: The development of machine learning has been highly motivated with the advancement of computer
science [1,2]. In this paper, machine learning is proposed to solve the boundary shape identification of the heat
conduction problem by the measured temperature on the surface. The heat conduction problem with irregular
boundary shape and constant thermal conductivity is transformed to the regular boundary shape with variable
thermal conductivity by the modified one-dimensional correction method [3]. Then the problem is transformed to
identify the thermal conductivity. Compared to the traditional methods, the proposed method can avoid remeshing
the domain. The relationship between the measurement temperature and the thermal conductivity is established
through the finite element calculations and machine learning. The thermal conductivity of the heat conduction
problem is successfully identified by a convolutional neural network based on the measured temperature. After that
the unknown boundary shape is evaluated. The influences of measurement point numbers and measurement errors
upon the inverse results are also investigated [4,5]. Numerical examples show that the proposed method can
reconstitute the unknown boundary geometry shapes effectively. With the increase of measurement point numbers
and the decrease of measurement errors, the inverse results are more accurate.
Keywords: Inverse problem, Identification of the boundary geometry shape, Finite element method, Deep learning,
Modified one-dimensional correction method.
References
1. Liang, Y., Lee, H., Lim, S., Lin, W., Lee, K., & Wu, C. (2002). Proper orthogonal decomposition and its
applications-part I: Theory. Journal of Sound and Vibration, 252: 527-544.
3. Fan, C., Zhang, M., Hu, S., Yang, L., & Sun, F. (2012). Identification of plate surface geometry: a numerical
and experimental study. Numerical Heat Transfer, Part B: Fundamentals, 61: 52-70.
4. Chen H., Yu B., Zhou, H., & Meng, Z. (2017). Comparison of CS, CGM and CS-CGM for prediction of pipe’s
inner surface in FGMs. CMC-Computers Materials & Continua, 53: 271-290.
5. Chen, H., Yu, B., Zhou, H., & Meng, Z. (2018). Identification of transient boundary conditions with improved
cuckoo search algorithm and polynomial approximation. Engineering Analysis with Boundary Elements, 95:
124-141.

Free vibration analysis of Timoshenko nanobeam using differential transform

method and general nonlocal theory
Shirko Faroughi1,*, Ahad Mahmodpour1and Tiomn Rabczuk2
1Faculty of Mechanical Engineering, Urmia University of Technology, Band Street, Urmia, Iran.
2 Bauhaus University Weimar, Geschwister-Scholl-Straße 8, 99423 City and Weimar, Germany.
*Corresponding Author: Shirko Faroughi. Email: Sh.farughi@uut.ac.ir.
Abstract: Most scientific problems in science and engineering fields can be defined by the partial differential
equations. A diversity of analytical and numerical approaches has been established to achieve precise estimated
solutions for the problems in the literature. The Differential Transform Method (DTM) is one of the semi-analytical
techniques to many problems. This approach is expedient to attain the solutions of linear and nonlinear differential
equations. There is no requirement to linearization or discretization. Large computational work and round-off errors
are evaded. The exact solution cannot be found or is hard to find for problems of complex nature. In such cases,
approximation approach is used. But, the exact solution can be attained by employing DTM, even for the complex
problems.
Application of nanomaterials have been extended and induce interest among the scientific communities in the area
of physics, engineering and nanotechnology. Nanomaterials play very important roles in numerous
nanoelectromechanical systems. Among these, beam kind structures show a vital role in the field of nanotechnology.
Beam type structures are generally exploited in mechanical, aerospace and civil engineering. Continuum mechanics
are classified into classical and nonclassical continuum mechanics. Wang and Hu [1] stated that the classical theory
are not able to expect the decrease in phase velocities of wave propagation in a carbon nanotube when the
wavenumber is so large. Hence several nonclassical continuum theories such as strain gradient theory, couple stress
theory, micropolar theory and nonlocal elasticity theory have been neestablished to integrate size effect by
introducing an intrinsic length scale. Among these theories, nonlocal elasticity theory suggested by Eringen [2] has
been extensively used in the vibration and static response of nanobeams. Reddy [3] examined analytical solutions
for numerous beam theories like Euler–Bernoulli, Timoshenko, Reddy and Levinson in case of simply supported
boundary condition. Shaat and Abdelkefi [4] showed that Eringen’s nonlocal theory is useless to fit the dispersion
curves of some materials. In order to remove this problem of Eringen’s nonlocal theory, a general form of the
nonlocal theory was suggested [4]. In this work we use this theory for modeling of nonlocal Timoshenko beam. This
general nonlocal theory is recognized with considering different attenuation functions for the existing material
moduli.
This work serves a semi analytical-numerical approach called DTM is used to study static and free vibration of
nanobeams corresponding on nonlocal-Timoshenko beam theory. The governing equations of motion a Timoshenko
beam is transformed to a set of simpler algebraic revolving equations by means of DTM. Then solving the
transformed equations and employing a procedure of inverse transformation to attain precise mode frequency and
displacement
Keywords: Differential Transform Method, Nonlocal Timoshenko beam, Free vibration analysis, General nonlocal
theory.
References
1. Wang, L. & Hu, H. (2005). Flexural wave propagation in single-walled carbon nanotubes. Physical Review B,
71(19): 195412.
2. Eringen, A.C. (1972). Nonlocal polar elastic continua. International Journal Engineering Science, 10(1): 1-16.

16 DACOMA-2019
3. Reddy, J. N. (2007). Nonlocal theories for bending, buckling and vibration of beams. International Journal
Engineering Science, 45 (2-8): 288-307.
4. Shaat, M., & Abdelkefi, A. (2017). New insights on the applicability of Eringen’s nonlocal theory.
International Journal of Mechanical Sciences, 121: 67-75.
Data-Driven Simulation of History-Dependent Materials Using Structured

Data Sets and Transition Rules
Klaus Hackl1,* and Kerem Ciftci1
1Ruhr-Universität Bochum, Universitätsstr. 150, 44801 Bochum, Germany.
*Corresponding Author: Klaus Hackl. Email: Klaus.hackl@rub.de.
Abstract: Data-driven computational mechanics replaces phenomenological constitutive functions by data sets in
stress-strain space. The distance of modeling values, e.g. stresses and strains in Gauss-points of a finite elements
calculation, from the data set is then minimized with respect to an appropriate metric, subject to equilibrium and
compatibility constraints, see [1]. While this approach works well for nonlinear elasticity, there are problems when
dealing with history-dependent materials, because the same point in stress-strain space might correspond to different
material behavior. In [2], this issue is treated by including local histories into the data set. There however, it is still
necessary to include models for the evolution of internal variables. Thus, we speak about a mixed formulation
consisting of a combination of classical and data-driven modeling.
In the approach presented here, we solve this issue by augmenting the data set by directions in the tangent space of
the specific points in stress-strain space. Moreover, we divide the data set into subsets corresponding to different
material behavior, e.g. elastic versus inelastic or loading versus unloading. Based on this subdivision, we introduce
transition rules mapping the modeling points to the various subsets. We demonstrate the operability of the proposed
approach by applying it to a model of small strain elastoplasticity with isotropic hardening.
Keywords: Data-driven, History-dependence, Structured data sets.
References
1. Kirchdoerfer, T., Ortiz, M. (2016). Data-driven computational mechanics. Computer Methods in Applied
2. Eggersmann, R., Kirchdoerfer, T., Reese, S., Stainier, L., Ortiz, M. (2019). Model-Free Data-Driven
inelasticity. Computer Methods in Applied Mechanics and Engineering, 350: 81-99.

Artificial neural networks for solving coupled PDEs in flexoelectricity

H. D. Huynh1, H. Nguyen-Xuan2 and X. Zhuang1,*
1 Institute of Continuum Mechanics, Leibniz Universität Hannover, Appelstraße 11, 30167 Hannover, Germany.
2CIRTECH Institute, Ho Chi Minh City University of Technology (HUTECH), Ho Chi Minh City, Viet Nam.
*
Corresponding Author: Xiaoying Zhuang. Email: zhuang@ikm.uni-hannover.de.
Abstract: In this study, a neural network is introduced to approximate solutions for a coupled model of flexoelectric
materials. Flexoelectricity is an electromechanical model of piezoelectric structures [1, 2] associated with strain
gradients typically involved with the effect of micro-structures, reported in Refs. [3, 4, 5, 6, 7, 8]. Therefore,
computation of such models does not only solve the coupled problem of mechanical and electrical behaviors but also
deals with high-order differentiation. The concept of using neural networks for integrated models of flexoelectricity
is based on the algorithm of meshless method which the computational domain is discretized by collocation points
instead of using meshes. The energy function is accordingly constructed from mathematical definition of
equilibrium equations as well as kinematic equations. The deep neural network is applied to minimize the total
energy function to find the approximation solution. This technique, discussed in [9, 10, 11] is also called by Deep
Energy Method (DEM). The DEM has provided remarkable advantages of modeling coupled models with high
computational accuracy and efficiency in comparison with conventional finite element approaches.
The implementation of DEM is based on the built-in library of Tensorflow. A 2D flexoelectric cantilever beam is
taken into the modeling of electromechanical behaviors, subjected by mechanical and electrical loading. The energy
function of the flexoelectric model is reported in Refs. [3, 12]. A neural network is established with two inputs
represented for coordinates of collocation points and three outputs represented for mechanical deformations in x and
y-direction and the electric potential φ . The architecture of the neural network is also illustrated in Figure. 1. In this
deep neural network, the hyperbolic tangent function (Tanh) is chosen to be the activation function for hidden layers.
Figure 1: The architecture of the neural network for flexoelectric model

19 DACOMA-2019
The process of seeking solution is to learn a network, or in other words to learn iteratively by updating parameters of
the neural network θ such that the sum of the energy function and the squared error of boundary conditions is
minimized. Minimizing the energy function is involved in derivatives of outputs with respect to their inputs, and
therefore the gradient of the neural network’s parameters is also defined by using gradient techniques. The LBFGS
is applied for training purpose in this study.
Keywords: Neural networks, Deep energy, Finite elements, Partial differential equations, Collocation method.
Acknowledgement
The support provided by RISE-project BESTOFRAC (734370) - H2020 is gratefully acknowledged.
References
1. S. S. Nanthakumar, T. Lahmer, T. Rabczuk. (2013). Detection of flaws in piezoelectric structures using
extended FEM. International Journal for Numerical Methods in Engineering 96(6):373-389.
2. S. S. Nanthakumar, T. Lahmer, X. Zhuang, G. Zi, T. Rabczuk. (2016). Detection of material interfaces using a
regularized level set method in piezoelectric structures. Inverse Problems in Science and Engineering, 24(1):
153-176.
3. B. H. Nguyen, X. Zhuang, T. Rabczuk. (2018). Numerical model for the characterization of Maxwell-Wagner
relaxation in piezoelectric and flexoelectric composite material. Computers & Structures, 208: 75-91.
4. B. H. Nguyen, S. S. Nanthakumar, X. Zhuang, et. al. (2018). Dynamic flexoelectric effect on piezoelectric
nanostructures. European Journal of Mechanics - A/Solids. 71: 404-409.
5. S. S. Nanthakumar, X. Zhuang, H. S.Park, T. Rabczuk. (2017). Topology optimization of flexoelectric
structures. Journal of the Mechanics & Physics of Solids, 105: 217-234.
6. H. Ghasemi, H. S. Park, N. Alajlan, T. Rabczuk. (2018). A computational framework for design and
optimization of flexoelectric materials. International Journal of Computational Methods, 15(3), 1850097.
7. H. Ghasemi, H. S. Park, T. Rabczuk. (2018). A multi-material level set-based topology optimization of
flexoelectric composites. Computer Methods in Applied Mechanics and Engineering. 332: 47-62.
8. H. Ghasemi, H. S. Park, T. Rabczuk. (2017) A level-set based IGA formulation for topology optimization of
flexoelectric materials. Computer Methods in Applied Mechanics and Engineering. 313: 239-258.
9. Khader M. Hamdia, Hamid Ghasemi, Xiaoying Zhuang, Naif Alajlan, Timon Rabczuk. (2019). Computational
Machine Learning Representation for the Flexoelectricity Effect in Truncated Pyramid Structures. Computers,
Materials & Continua, 59(1): 79-87.
10. C. Anitescu, E. Atroshchenko, N. Alajlan, T. Rabczuk. (2019). Artificial Neural Network Methods for the
Solution of Second Order Boundary Value Problems. Computers, Materials & Continua, 59(1): 345-359.
11. Justin Sirignano, Konstantinos Spiliopoulos. (2018). DGM: A deep learning algorithm for solving partial
differential equations. Journal of Computational Physics, 375, 2018, 1339-1364.
12. Amir Abdollahi, Christian Peco, Daniel Millán, Marino Arroyo, Irene Arias. (2014). Computational evaluation
of the flexoelectric effect in dielectric solids. Journal of Applied Physics, 116, 093502.
Nonlinear mechanical behavior and failure mechanism of three-dimensional

complex carbon nanotube networks
Jiachao Ji1 and Junhua Zhao1,2,*
1
Jiangsu Key Laboratory of Advanced Food Manufacturing Equipment and Technology, Jiangnan University, 214122, Wuxi,
China.
2
Institute of Mechanics and Advanced Materials, School of Mechanical Engineering, Jiangnan University, 214122 Wuxi, China.
*
Corresponding Author: Junhua Zhao. Email: junhua.zhao@163.com
Abstract: The coarse-grained (CG) potentials of single-walled carbon nanotubes (SWCNTs) were developed in our
previous work [1]. However, these CG potentials can only be used to exactly obtain the linear mechanical properties
of the carbon nanotube networks (CNNs) without crosslinked CNTs but not the nonlinear mechanical properties of
CNNs with crosslinked CNTs, because the quadratic CG stretching and bending potentials are only considered in the
CG bonded potentials, where the full-atom harmonic potentials of CNTs are only considered. To design and
assemble CNT-based flexible nanodevices in practical applications, the nonlinear mechanical behavior and failure
mechanism of CNNs are extremely important and necessary. In this study [2], the explicit expressions of the
chirality-dependent high-order nonlinear CG stretching and bending potentials in the CG bonded potentials are
systematically established based on the full-atom Morse potential and the Reactive Empirical Bond-Order (REBO)
interatomic potential of second generation, respectively. In particular, the CG non-bonded potentials are improved
by using the 18-24 Lennard-Jones (LJ) potentials. By comparison with the full-atom molecular dynamics (MD)
simulations and our analytical results, the present nonlinear CG potentials have high accuracy. The obtained
nonlinear CG potentials can be used to efficiently characterize the nonlinear mechanical behaviors of CNNs with
crosslinked CNTs under different loading conditions at minor computational cost.
Keywords: Nonlinear, Coarse-grained potentials, REBO potential, Carbon nanotube networks.
Acknowledgement
We gratefully acknowledge support from the National Natural Science Foundation of China (Grant nos. 11572140
and 11302084), the Natural Science Foundation of Jiangsu Province (Grant no. BK20180031), the 111 project
(Grant no. B18027) and the Thousand Youth Talents Plan.
References
1. Zhao, J., Jiang, J., Wang, L., Guo, W., & Rabczuk, T. (2014). Coarse-grained potentials of single-walled
carbon nanotubes. J. Mech. Phys. Solids, 71: 197-218.
2. Ji, J., Guo, W., & Zhao, J. (2019). Novel nonlinear coarse-grained potentials of carbon nanotubes. J. Mech.
Phys. Solids, 128: 79-104.

Designing phononic crystal with anticipated band gap through a deep

learning based data-driven method
Xiang Li and Zhanli Liu*
Applied Mechanics Lab., Dept. of Engineering Mechanics, School of Aerospace, Tsinghua University, Beijing 100084 China.
*
Corresponding Author: Zhanli Liu. Email: liuzhanli@tsinghua.edu.cn.
Abstract: Phononic crystal is a type of artificial heterogeneous material constituted by a periodic repetition of cells.
This characteristic of phononic crystals provides a possible solution to the accurate manipulation of acoustic and
elastic waves. For this reason, phononic crystal is of application potentials in vibration and noise reduction, filtering,
acoustic lens, acoustic imaging, acoustic stealth, etc. It is thus of significance in the fields of information,
communication, medical, and military applications. To design phononic crystal with anticipated manipulation
characteristic has become a research hotspot in recent years. Though, to accurately manipulate acoustic and
mechanical wave is still a major challenge for existing designing approaches. Hence, assisted by image-based finite
element analysis and deep learning, a data-driven approach is proposed in this study for designing phononic crystals.
In this approach, a generative deep learning model is trained to establish the inherent relation between frequency
bands and phononic crystal topologies; a stochastic modeling method is investigated to generate stochastic samples
with novel topologies. Image-based finite element analysis is developed to study the frequency bands of stochastic
samples. The topologies and frequency bands of the stochastic samples are employed to expand the knowledge
database of the deep learning model. The model is ultimately employed to design novel phononic crystals with
designated frequency band features. Not limited to phononic crystals, the proposed method can be further extended
to design various structured mechanical materials with specific functionalities.
Keywords: Material design; phononic crystal; band gap; deep learning; finite element analysis.
Acknowledgement
This work is supported by the Science Challenge Project, China, No. TZ2018001, National Natural Science
Foundation of China, under Grant No. 11722218, 11302115 and 11532008, Tsinghua University Initiative Scientific
Research Program, China.
References
1. I. Goodfellow, Y. Bengio, A. Courville, Y. Bengio, Deep learning, MIT press Cambridge2016.
2. Y. Pennec, B. Djafari-Rouhani, Fundamental properties of phononic crystal, Phononic Crystals, Springer2016,
pp. 23-50.
3. S. Halkjær, O. Sigmund, J.S.J.Z.f.K.-C.M. Jensen, Inverse design of phononic crystals by topology
optimization, 220(9-10) (2005) 895-905.
4. H.-W. Dong, X.-X. Su, Y.-S. Wang, C.J.S. Zhang, M. Optimization, Topological optimization of two-
dimensional phononic crystals based on the finite element method and genetic algorithm, 50(4) (2014) 593-604.

A comparative study of machine learning methods applied to the singular

integrals in the boundary element method
Yuan Li1,2,* and Wen-tao Mao1,2
1Henan Normal University, 46# East of Construction Road, 453007 Xin Xiang, China.
2 Engineering Lab of Intelligence Business & Internet of Things, 46# East of Construction Road, 453007 Xin Xiang, China.
*Corresponding Author: Yuan Li. Email: liyuan@htu.edu.cn
Abstract: Accurate and efficient evaluation of singular integrals and nearly singular integrals is of crucial
importance for successful implementation of the boundary element method (BEM) [1]. The singular integrals arise
when the source point is located on the element of integration. In this case, the value of the integrand is infinite
when the source point and the field point coincide. Some methods, such as coordinate transformation [2], have to be
adopted to remove singularities. The nearly singular integrals arise when the source point is close to but not on the
element of integration. They are difficult to be evaluated using the traditional Gaussian quadrature rules because the
value of the integrand varies dramatically as the source point approaches the element. Element subdivision [3] is one
of the most widely used methods for evaluating nearly singular integral. Compared with regular integrals, these two
kinds of integrals require a large number of Gauss points to obtain accurate integral results. The computational time
is considerable when there are many singular integrals and nearly singular integrals in the computational model.
This paper describes a method to evaluate the singular integrals and nearly singular integrals by using machine
learning methods. In the data preparation phase, a large number of integration samples are collected in the local
coordinate system. The element shapes and the source point positions of these samples are varies. It should be
demonstrated that the original integrand can’t be used directly for computing the training samples, because the value
of the fundamental solutions in the integrand is related to the coordinates of the points in the global coordinate
system. However, the samples can only be calculated in the hypothetical local coordinate system in the data
preparation phase. In this paper, the formulas of the fundamental solutions are divided into several sub-formulas.
The integral value of each sub-formula can be converted from the local coordinate system to the global coordinate
system, as long as the coordinate transformation matrix is obtained. In the training phase, the coordinates of the
element nodes and the source points are treated as input variables. The element integral matrices computed are
treated as output variables. Some classical machine learning methods [4] for regression problems including Extreme
Learning Machine (ELM) [5], Gaussian Process Regression(GPR), K-Nearest Neighbor (KNN), Support Vector
Regression(SVR) and Regression Tree(RT) are adopted to training the samples and their performances are
compared. In the application phase, the predictive model produced in the training phase is implemented on the
numerical quadrature process of the boundary element analysis code. For nearly singular integrals and singular
integrals, the element integral results are obtained directly by the predictive model. Traditional Gaussian quadrature
process is omitted. For the regular integrals, accurate results can be obtained using traditional Gaussian quadrature.
The fundamental solutions of elastostatics are taken as example to derive the sub-formulas and compute the integral
samples. Linear quadrilateral elements are used as integral element. Other types of element can be trained using the
same method. The results of one of the test sets are shown in Fig. 1. The Fig. 1(a) and Fig. 1(b) represent the test
results of singular integrals and nearly singular integrals respectively. The mean relative errors(MRE) of the results
of the machine learning methods are listed in the Table 1. It can be observed that the predictive results of all of the
methods are in good agreement with the test results. For the singular integral, ELM and GPR are more outstanding
than other methods. For the nearly singular integral, SVR and ELM are better than other methods.

23 DACOMA-2019
(a) results of the singular integral (b) results of the nearly singular integral
Figure 1. Comparison of the results of some classical machine learning methods
Table 1. The mean relative errors of the results of the machine learning methods
ELM GPR KNN SVR RT
MRE of the
singular 4.4569e-5 3.5646e-5 0.0052 0.0036 0.0106
integral
MRE of the
nearly singular 0.0040 0.0066 0.0081 0.0019 0.0398
integral
Keywords: Boundary element method, Singular element integral, Nearly singular element integral, Machine
learning.
Acknowledgement
This work was supported by the National Natural Science Foundation of China(No. 11702087, U1704158).
References
1. Banerjee, P., & Butterfield, R. (1981). Boundary element methods in engineering science. London: McGraw-
Hill.
2. Zhang, J., Lu, C., Zhang, X., Xie, G., Dong, Y., & Li, Y. (2015). An adaptive element subdivision method for
evaluation of weakly singular integrals in 3D BEM. Engineering Analysis with Boundary Elements, 51: 213-
219.
3. Zhang, J., Wang, P., Lu, C., & Dong, Y. (2017). A spherical element subdivision method for the numerical
evaluation of nearly singular integrals in 3D BEM. Engineering Computations, 34(6): 2074-2087.
4. Oishia, A., & Yagawab, G. (2017). Computational mechanics enhanced by deep learning. Computer Methods
in Applied Mechanics and Engineering, 327: 327-351.
5. Mao, W., He, J., Li, Y. & Yan, J. (2017). Bearing fault diagnosis with auto-encoder extreme learning machine:
A comparative study. Proceedings of the Institution of Mechanical Engineers Part C-Journal of Mechanical
Engineering Science, 231(8): 1560-1578.
Data driven thermo-mechanical calculation during additive manufacturing

process
Constant Prassette1,2, Guanghua Wei1, Michel Bellet2 and Yancheng Zhang2,*
1Shanghai Jiao Tong University, ParisTech Elite Institute of Technology, Shanghai, China.
2MINES Paristech, PSL Research University, CEMEF-Centre de Mise en Forme des Matériaux, CNRS UMR 7635, CS 10207, 1
Rue Claude Daunesse, 06904 Sophia Antipolis cedex, France.
*Corresponding Author: Yancheng Zhang. Email: yancheng.zhang@mines-paristech.fr.
Abstract: Additive manufacturing (AM), commonly known as 3D printing has developed rapidly in recent years.
This technology opened up a range of new perspectives on complex geometries which until then were considered
laborious. Its contribution to lightweighting is a profound asset for aerospace industry where AM gained significant
popularity. However, thermally induced residual deformations and residual stresses in AM parts must be effectively
predicted to guarantee wide industrial adoption of AM technology. In this work, a numerical modelling based on the
finite element and level-set methods is implemented to study direct material deposition (DMD) of metal alloys.
Metal deposition is followed up by updating the level-set layer by layer. Nevertheless, conventional thermo-
mechanical calculation for AM simulation is time consuming, especially for large parts. In order to increase the
computational efficiency and maintain the precision, the data driven thermo-mechanical model is proposed. In the
model based on the inherent strain theory, the plastic strain rate is extracted from first layer depositions and is
considered as inherent strain rate, thus non-linear calculation becomes linear problem. The model validation is
carried out with a single wall produced by depositing multiple single tracks on top of each other, and its application
is successfully extended to a rectangular wall contour geometry. The simplified method reduces noteworthy the
computational cost and it is in good agreement with the conventional calculation for predicting residual distortions.
Keywords: Data driven thermo-mechanical calculation, Inherent strain rate method, Direct material deposition.
Aknowledgement
Heartfelt thanks go to Prof. Yancheng ZHANG for his patience, motivation, enthusiasm, immense knowledge and
invaluable help.
References
1. Yancheng Zhang, Gildas Guillemot, Marc Bernacki and Michel Bellet. (2018). Macroscopic thermal finite
element modeling of additive metal manufacturing by selective laser melting process, Computer Methods in
Applied Mechanics and Engineering, 331: 514-535.
2. Xuan Liang et al. (2018). A modified method for estimating inherent strains from detailed process simulation
for fast residual distortion prediction of single-walled structures fabricated by directed energy deposition,
Additive Manufacturing, 23: 471– 486.
3. Sonja Jonsson and Sebastian Krappedal. (2018). Evaluation of residual stresses and distortions in additively
manufactured components, School of Industrial Engineering and Management, KTH Royal Institute of
Technology.
4. Joel KEUMO TEMATIO. (2018). Méthodes de réduction de complexité des simulations thermomécaniques du
procédé de fabrication additive, CEMEF, Mines ParisTech.
5. Qingcheng Yang et al.(2016). Finite element modeling and validation of thermomechanical behavior of Ti-
6Al-4V in directed energy deposition additive manufacturing, Additive Manufacturing, 12: 169-177.
6. J.C Heigel, P. Michaleris, E.W. Reutzel. (2015). Thermo-mechanical model development and validation of
directed energy deposition additive manufacturing of Ti-6Al-4V, Additive Manufacturing 5,.

25 DACOMA-2019
7. Ninshu Ma et al. (2016). Inherent strain method for residual stress measurement and welding distortion
prediction, in ASME 2016 35th International Conference on Ocean Offshore and Arctic Engineering,
OMAE2016-54184.
8. Biegler M., Graf B., Rethmeier M., In-situ distortions in LMD additive manufacturing walls can be measured
with digital image correlation and predicted using numerical simulations, Additive Manufacturing 20 (2018)
101–110.
9. Bey Vrancken, Lore Thijs, Jean-Pierre Kruth, and Jan Van Humbeeck, Heat treatment of ti6al4v produced by
selective laser melting: Microstructure and mechanical properties, Journal of Alloys and Compounds, 541:
177–185, 2012.
10. Kaufui V. Wong and Aldo Hernandez, A review of additive manufacturing, ISRN Mechanical Engineering,
2012, 2012.
11. J. Ding et al. (2014), A computationally efficient finite element model of wire and arc additive manufacture, Int.
J. Adv. Manuf. Technol., 70: 227-236.
12. N. Hopkinson, R.Hague, P. Dickens, Rapid manufacturing: An Industrial Revolution for the Digital Age, John
Wiley & Sons, 2006.
13. J.P. Kruthn M. Leu, T. Nakagawa, Progress in additive manufacturing and rapid prototyping, CIRP Ann.
Manuf. Technol., 47: 525-540, 1998.
14. Michel Rappaz, Michel Deville, Michel Bellet and Ray Snyder, Numerical Modeling in Materials Science and
Engineering, Springer, 2010.
15. M. Bellet, J.-L. Chenot, L. Fourment, E. Massoni, P. Montmitonnet, Séminaire de Plasticité, Ecole des Mines
de Paris, 1994.
16. Youssef Mesri, Hugues Digonnet and Thierry Coupez, Advanced Parallel Computing in Material Forming
with CIMLib, Revue européenne de mécanique numérique, 7: 669-694, 2009.
17. Jean Salençon, Mécanique des Milieux Continus, vol.1, 978-2-7302-1245-8, 2005.
18. J. Lemaitre, J.-L. Chaboche, Mécanique des Matériaux Solides, Dudod, 2004.
19. P. Michaleris, Modeling metal deposition in heat transfer analysis of additive manufacturing processes, Finite
Element Analysis Description, 86: 51-60, 2014.
20. M. Megahed, H.-W. Mindt, N. N’Dri, H. Duan, O. Desmaison, Metal additive manufacturing process and
residual stress modeling, IMMI, 2016.
21. J.C. Steuben, A.P. Iliopoulos, J.G. Michopoulos, Discrete element modeling of particle-based additive
manufacturing processes, Comput. Methods Appl. Mech. Engin., 2016.
Residual-based Goal-oriented A Posteriori Error Estimation and

Adaptive Mesh Refinement in Thermoelastic Problems
Ehsan Rabizadeh1,* and Amir Saboor Bagherzadeh2
1Institute
of Structural Mechanics (ISM), Bauhaus-Universität Weimar, Marienstraße 15, D-99423 Weimar, Germany.
2 Department of Applied Mathematics, Faculty of Mathematical Sciences, Ferdowsi University of Mashhad, Mashhad, Iran.
*Corresponding Author: Ehsan Rabizadeh. Email: ehsan.rabizadeh@uni-weimar.de.
Abstract: In recent years, substantial attention has been devoted to thermoelastic multifield problems and their
numerical analysis. Thermoelasticity is one of the important categories of multifield problems which deals with the
effect of mechanical and thermal disturbances on an elastic body. In other words, thermoelasticity encompasses the
phenomena that describe the elastic and thermal behavior of solids and their interactions under thermo-mechanical
loadings. Since providing an analytical solution for general coupled thermoelasticity problems is mathematically
complicated, the development of alternative numerical solution techniques seems essential [1-5].
Due to the nature of numerical analysis methods, presence of error in results is inevitable, therefore in any numerical
simulation, the main concern is the accuracy of the approximation. There are different error estimation (EE)
methods to assess the overall quality of numerical approximation. In many real-life numerical simulations, not only
the overall error, but also the local error or error in a particular quantity of interest is of main interest. The error
estimation techniques which are developed to evaluate the error in the quantity of interest are known as “goal-
oriented” error estimation (GOEE) methods [6-9].
In this research, the goal-oriented a posteriori error estimation in thermoelasticity problems is investigated.
Generally, the a posteriori error estimation techniques can be categorized into two major branches of recovery-
based [10-12] and residual-based [13-15] error estimators. This article focuses on the application of the residual-
based error estimator in thermoelasticity. Moreover, in order to reduce the error in the quantity of interest efficiently
and optimally in two-/three-dimensional (2D/3D) thermo-mechanical multifield problems, goal-oriented adaptive
mesh refinement is performed.
To investigate the goal-oriented error estimation in thermoelastic problems, a new dual weighted residual (DWR)
[16,17] method for a posteriori error estimation and mesh adaptation is developed and implemented in
thermoelasticity. The developed method, which relies on the duality principles, consists of an adjoint problem
solution. The mesh adaptivity procedure based on the DWR method is performed by adaptive local
h-refinement/coarsening with allowed hanging nodes. According to the proposed DWR method, the error
contribution of each element is evaluated. In the refinement process, the contribution of each element to the goal
error is considered as the mesh refinement criterion.
In order to substantiate the accuracy and performance of the developed methods, they are applied to several
numerical problems with analytical solutions and the estimated errors are compared with the exact errors. Here, 2D
and 3D problems under thermo-mechanical loadings are considered as benchmark problems. In these examples,
several singular functions are considered as the quantity of interest. To show the accuracy of the developed
estimator, in all examples, the goal error effectivity index as a standard measure of the quality of an estimator is
calculated. In all examples, after few initial coarse meshes, the goal error effectivity index lies in the acceptable
range and by further refinements approaches to the ideal value of 1.
Moreover, in order to demonstrate the efficiency of the proposed method and show the optimal behavior of the
employed refinement method, the results of different conventional error estimators and refinement techniques (e.g.,
global uniform refinement, Kelly, and weighted Kelly techniques) are used for comparison. This comparison shows
that among the selected error estimation and refinement techniques, the employed DWR method is the most
optimum and efficient one and the global uniform refinement has the least convergence rate. On average, in the
studied examples, the DWR, Kelly, and W-Kelly methods show, respectively, 0.68%, 0.42%, and 0.26% higher
convergence rates compared to the global uniform refinement. This efficiency enhancement leads to the
considerable computational time saving.
Keywords: Thermoelasticity, Finite element method (FEM), Residual-based error estimation, Goal-oriented error
estimation (GOEE), Mesh adaptivity.

27 DACOMA-2019
Acknowledgement
The first author would like gratefully acknowledge for the financial support of this work which was provided by the
Deutscher Akademischer Austauschdienst (DAAD).
References
1. Sternberg, E. & Chakravorty, J. (1959). On Inertia Effects in a Transient Thermoelastic Problem. Journal of
Applied Mechanics, 26: 503–509.
2. Chen, J. & Dargush, G. F. (1995) Boundary element method for dynamic poroelastic and thermoelastic
analyses. International Journal of Solids and Structures, 32: 2257–2278.
3. Hosseini-Tehrani, P. & Eslami, M.-R. (2003). Boundary element analysis of finite domains under thermal and
mechanical shock with the Lord-Shulman theory. The Journal of Strain Analysis for Engineering Design, 38:
53–64.
4. Hosseini-Tehrani, P. & Eslami, M.-R. (200). BEM analysis of thermal and mechanical shock in a two-
dimensional finite domain considering coupled thermoelasticity. Engineering Analysis with Boundary
Elements, 24: 249–257.
5. Hosseini-Tehrani, P. & Eslami, M.-R. (2000). BEM analysis of thermal and mechanical shock in a two-
dimensional finite domain considering coupled thermoelasticity. Engineering Analysis with Boundary
Elements, 24: 249–257.
6. Prudhomme, S. & Oden, J.T. (1999). On goal-oriented error estimation for elliptic problems: application to the
control of pointwise errors. Computer Methods in Applied Mechanics and Engineering, 176, 313–331.
7. Prudhomme, S. & Oden, J.T. (2001). Goal-oriented error estimation and adaptivity for the finite element
method. Computers and Mathematics with Applications, 41: 735–756.
8. Rabizadeh, E., Saboor Bagherzadeh, A., & Rabczuk, T. (2015). Adaptive thermos-mechanical finite element
formulation based on goal-oriented error estimation. Computational Materials Science, 102: 27–44.
9. Rabizadeh, E., Saboor Bagherzadeh, A., & Rabczuk, T. (2016). Goal-oriented error estimation and adaptive
mesh refinement in dynamic coupled thermoelasticity. Computers and Structures, 173: 187–211.
10. Zienkiewicz, O. C. and Zhu, J. Z. (1987). A simple error estimator and adaptive procedure for practical
engineering analysis. International Journal for Numerical Methods in Engineering, 24: 337–357.
11. Zienkiewicz, O. C. and Zhu, J. Z. (1992). The superconvergent patch recovery and a posteriori error estimates.
Part 1: The recovery technique. International Journal for Numerical Methods in Engineering, 33: 1331–1364.
12. Zienkiewicz, O. C. and Zhu, J. Z. (1992). The superconvergent patch recovery and a posteriori error estimates.
Part 2: Error estimates and adaptivity. International Journal for Numerical Methods in Engineering, 33: 1365–
1382.
13. Babuška, I. & Rheinboldt, W.C. (1978). A-posteriori error estimates for the finite element method.
International Journal for Numerical Methods in Engineering, 12: 1597–1615.
14. Ainsworth, M. & Oden, I.T. (1992). A procedure for a posteriori error estimation for h-p finite element
methods. Computer Methods in Applied Mechanics and Engineering, 101, 73–96.
15. Oden, J.T., Demkowicz, L., Rachowicz, W., & Westerman, T.A. (1989). Toward a universal hp-adaptive finite
element strategy. Part 2: A posteriori error estimation. Computer Methods in Applied Mechanics and
Engineering, 77: 113–180.
16. Becker, R. & Rannacher, R. (1996). A feed-back approach to error control in finite element methods: basic
analysis and examples. East-West Journal of Numerical Mathematics, 4: 237–264.
17. Becker, R. & Rannacher, R. (2001). An optimal control approach to a posteriori error estimation in finite
element methods. Acta Numerica, 10: 1–102.
Theoretical and experimental study of long-term creep in multilayered

composites
Roham Rafiee1,* and Amin Ghorbanhosseini2
1Composites Research Laboratory, Faculty of New Sciences & Technologies, University of Tehran, Tehran 1439957131 Iran.
*Corresponding Author: Roham Rafiee. Email: roham.rafiee@ut.ac.ir.
Abstract: In most industries, composite materials have dominated the market compared with traditional materials
and structures, due to their unique properties. Composite structures, in addition to their proper weights, have a good
corrosion resistance and have a relatively high life span. Due to their long operational time, composite structures are
usually subjected to long-term loading and experience creep phenomenon. The occurrence of creep leads to
reduction in mechanical properties and makes the composite structure more susceptible to failure. Therefore,
studying the behavior of these structures in long-term periods requires the analysis of the creep phenomenon as one
of the most important factors of their life-time and durability.
There have been very few studies conducted to investigate the long-term creep behavior of cylindrical composite [1-
10]. Most studies in this field are experimental and theoretical which have been done on a limited basis. Prediction
for any composite lamination requires repeated testing, and long-term behavior cannot be modeled without costly
and time-consuming experiments.
The main objective of this article is developing an integrated modeling for evaluating long-term creep phenomenon
in cylindrical composite structures under constant loading for a period of one year. In this model, the creep caused
by constant load on the composite structure is simulated and investigated using the long-term data of pure resin.
This data can be obtained from the Time Temperature Superposition [11]. This model can be used as a suitable tool
for predicting long-term creep behavior of composite structures and reduce the costs of long-term test programs. The
model is validated and improved on the basis of an empirical study. Experimental program consists of suitable
equipment for monitoring and collecting long-term creep data on a cylindrical composite structure subjected to
compressive transverse loading as a case study. Also, Theoretical model for simulating long-term creep is acquired
by numerical methods based on linear viscoelasticity [12-13] and with implementing UMAT subroutine into
ABAQUS.
Keywords: Composite, Creep, Cylindrical composite structures, FEM, UMAT.
References
1. Farshad, M., & Necola, A. (2004). Effect of aqueous environment on the long-term behavior of glass fiber-
reinforced plastic pipes. Polymer testing, 23(2): 163-167.
2. Farshad, M., & Necola, A. (2004). Strain corrosion of glass fiber-reinforced plastics pipes. Polymer testing,
23(5): 517-521.
3. Yao, J., & Ziegmann, G. (2006). Equivalence of moisture and temperature in accelerated test method and its
application in prediction of long-term properties of glass-fiber reinforced epoxy pipe specimen. Polymer Testing,
25(2): 149-157.
4. Guedes, R. M., Sá, A., & Faria, H. (2007). Influence of moisture absorption on creep of GRP composite pipes.
Polymer testing, 26(5): 595-605.
5. Faria, H., & Guedes, R. M. (2010). Long-term behavior of GFRP pipes: reducing the prediction test duration.
Polymer Testing, 29(3): 337-345.

29 DACOMA-2019
6. Guedes, R. M. (2010). Nonlinear viscoelastic analysis of thick-walled cylindrical composite pipes. International
Journal of Mechanical Sciences, 52(8): 1064-1073.
7. Guedes, R. M., Sá, A., & Faria, H. (2010). On the prediction of long‐term creep‐failure of GRP pipes in
aqueous environment. Polymer Composites, 31(6): 1047-1055.
8. Yoon, S. H., & Oh, J. O. (2015). Prediction of long-term performance for GRP pipes under sustained internal
pressure. Composite Structures, 134: 185-189.
9. Rafiee, R., & Mazhari, B. (2016). Simulation of the long-term hydrostatic tests on glass fiber reinforced plastic
pipes. Composite Structures, 136: 56-63.
10. Rafiee, R., & Mazhari, B. (2016). Evaluating long-term performance of Glass Fiber Reinforced Plastic pipes
subjected to internal pressure. Construction and Building Materials, 122: 694-701.
11. Ward, I. M., & Sweeney, J. (2012). Mechanical properties of solid polymers. John Wiley & Sons.
12. Boltzmann, L. (1878). Zur theorie der elastischen nachwirkung. Annalen der Physik, 241(11): 430-432.
13. Findley, W. N. (1960). Mechanism and mechanics of creep of plastics. SPE journal, 16(1): 57-65.
Deep learning Approach to Analysis and Prediction of Infrastructure Disaster

Manouchehr Shokri1,*
1Bauhaus University Weimar, Department of civil engineering/Institud fur Structure Mechanik (ISM) Marienstrasse 15, 99423,
Germany.
*Corresponding Author: Manouchehr Shokri. Email: manouchehr.shokri@uni-weimar.de.
Abstract: Natural disasters have capable of disastrous consequences impacts on the functionality of infrastructure
systems and able intense systemic and socio-economic losses. According to budget restrictions, it is crucial to
optimize decisions regarding mitigation, preparedness, response, and recovery practices for these systems. This
requires accurate and efficient means to evaluate the infrastructure system reliability. Machine learning is a hot
subject in the recent decade, and the influence of Artificial Neural Network (ANN) is especially notable, which is
the most extensively used models of machine learning in assessment of infrastructure. This study provides damage
detection assessment of seismic performance of reinforced concrete (RC) bridges, by using machine learning
methods. This study developed a deep learning (DL) model for estimating of seismic vibration on railway RC
bridges. A multi-layered perceptron (MLP) with a back-propagation (BP) algorithm neural network was
implemented to predict the seismic performances of the designated bridges. ANN models were developed, trained
and tested in a based MATLAB and Python program. A training set and a validation set of bridges were produced
from dynamic response of different RC bridges. Finite Element Analysis (FEA) was used to generate training and
testing set of ANN models. A training and a validation data set of RC bridges will produced from FEA analysis
results of the dynamic response of RC bridges by shifting parameters (the input parameters of ANN) including
accelerations and axle loads ,concrete compressive strength, reinforcement ratio, Size of column, column shape,
width of the slab, effective depth of tension reinforcement, slab shape Peak acceleration, Shear wall, Story height,
max width of bay in X, Y direction and etc., Under the near-fault earthquakes. On the other hand, Output will be
dimensional dynamic response in terms of the roof displacement, base shear forces and base bending moments,
accurately when compared with the results of conventional methods like FEA[1][2]. The method performs on the
collected feature measurements on a railway RC bridge during dynamic response of bridges structures, which for
this study were gathered in a numerical experiment using a three dimensional finite element model. Thus, the next
step consists in the design and unsupervised training of Artificial Neural Networks that used as mentioned input data
[3]. The results demonstration that the proposed method is robust and capable of capturing the physical complexities
for the dynamic damage detection force on RC bridge prediction task. Therefore results will compared with
analytical and exact deflection[4]. The results showed that an appropriately trained neural network could
consistently predict permanent earthquake-induced seismic deformation of the RC bridges. Bridges fragility analysis
to calculate failure probability was another achieved that were created by using nonlinear analysis (NA) and
artificial neural networks (ANNs). Nonlinear response history analyses were achieved, in order to calculate the
seismic performances of the bridges. To this end, 306 bridge-earthquake cases were considered. The MLP neural
networks considered herein consist of an input layers vectors, hidden layers and an output vector. In order to train
ANNs, 70% of the numerical results were selected, and the remained 30% were employed for testing the reliability
and validation of ANNs. Numerous structures of MLP neural networks were analyzed in order to obtain efficient
and effective neural networks. After achieving the best structure of neural network, it was used for generating new
data. A total number of 600 new bridge-earthquake cases were generated based on neural simulation. Finally,
probabilistic seismic safety analyses were conducted. Therefore bridges fragility analysis were developed using
numerical results, neural predictions and the combination of numerical and neural data. Consequences of this study
demonstrate that ANNs are suitable tools for predicting damage detection of seismic performances of RC bridges. It
was also shown that efficiency stresses of the reinforcements is one of the important sources of uncertainty in
fragility analysis of RC bridges [5]. It is evident from this evaluation that ANNs have been successfully applied to
many Infrastructure engineering areas like prediction, risk analysis, decision-making, resources optimization,
classification, and selection etc. It was demonstrated that the neural network based approach is highly successful to
determine response of bridges and buildings subjected to Seismic evaluation. Based on the results of case studies, it
is evident that ANNs perform better than as similarity to the conventional methods[6].

31 DACOMA-2019
Keywords: Seismic evaluation, Dynamic analysis, RC bridge, Artificial Neural Network, FEM, Deep Learning,
Damage detection
Acknowledgement
Acknowledgements and Reference heading should be left justified, bold, with the first letter capitalized but have no
numbers. Text below continues as normal.
References
1. A. Mortezaei and K. Mortezaei. (2012). Neural Network Evaluation of FRP Strengthened RC Buildings
Subjected to Near-Fault Ground Motions having Fling Step. Int. J. Civil, Environ. Struct. Constr. Archit.
Eng., vol. 6, no. 2, pp. 78–82.
2. S. G. Joshi, S. N. Londhe, and N. Kwatra. (2014). Application of artificial neural networks for dynamic
analysis of building frames. Comput. Concr., vol. 13, no. 6, pp. 765–780.
3. A. C. Neves, I. Gonz, and J. Leander. (2018). Experimental Vibration Analysis for Civil Structures, vol. 5,
no. October.
4. M. Saleh. (2015). Prediction of rigid pavement responses under axle loads using artificial neural network.
Int. J. Pavement Res. Technol. vol. 8, no. 1, pp. 10–16.
5. M. Hassan, E. A. Ahmed, and B. Benmokrane. (2012). Punching Shear Strength Prediction of Frp-
Reinforced Concrete Flat Slabs. 6th Int. Conf. Adv. Compos. Mater. Bridg. Struct., no. May, pp. 22–25.
6. S. J. S. Hakim and H. Abdul Razak. (2013). Structural damage detection of steel bridge girder using
artificial neural networks and finite element models. Steel Compos. Struct., vol. 14, no. 4, pp. 367–377.
Inverse design of phononic Topological insulators using Machine learning
S. S. Nanthakumar1 and Xiaoying Zhuang1,2,*

1
Institute of Continuum Mechanics, Leibniz Universität Hannover, Applestr. 11, 30167 Hannover, Germany.
2 College of Civil Engineering, Tongji University, 1239 Siping Road, 200092 Shanghai, China.
Abstract: Topological insulators (TIs) are a new state of matter the behavior of which depends on its topology
rather than geometry. The TIs act as insulator in the bulk and allows wave propagation along the surface. These
surface states are topologically protected which makes them immune to backscattering and remain robust even in the
presence of defects. Quantum Hall effect (QHE) is a phenomenon in which the presence of magnetic field enable
electrons to travel along the edge of a semi-conductor under external voltage. Quantum Spin Hall effect (QSHE)
[1,2,3] is a phenomenon discovered later, in which the edge protected electron movement is achieved even in the
absence of an external magnetic field. The edge state due to QSHE is helical while due to QHE is chiral. The helical
edge state leads to propagation of opposite spin states, so an electron can backscatter if the spin is flipped.
The acoustic and elastic analogue of QSHE has been achieved, while majority of them are in discrete lattice. Though
there are works available on inverse design of phononic metamaterial (MM) for band gaps, there have been only
very few works on inverse design of phononic TI for topological band gaps [4,5]. A band gap prevents wave
propagation in the bulk while topological band gap prevents wave propagation in the bulk but allows waves along
the edge. The inverse design procedures adopted utilize a level set based optimization method or SIMP. The
emergence of QSHE in a phononic metamaterial necessitates the unit cell topology to offer a double Dirac cone in
its band structure. In [4], objective function based on unit cell band structure is formulated to determine the
topologies of different topological phase.
While in [5], the objective function is formulated directly based on the amplitude of wave intensity such that the
intensity is zero in the bulk and maximum in the interface. As an alternative, it is possible to construct an objective
function based on the supercell made of two different unit cells. The band structure determined for the supercell
need to have interface modes lying in a bulk band gap. The topology can be represented by pseudo density of
elements or using level set function. The Convolution Neural Network (CNN) can be trained with supercell topology
as input and band structure determined by finite element method as output. The frame work for designing a
phononic TI based on QSHE is shown in Figure 1. The results are obtained for solving forward problem by FEM
and inverse design of supercell using Genetic algorithms such that the band structure has intersecting interface
modes in a bulk band gap. Future work includes training the CNN to solve both the forward problem and inverse
design of phononic topological insulators.
Keywords: QSHE, Topological insulators, Phononic MMs, Topology optimization, CNN

33 DACOMA-2019
Figure 1 : The Framework for designing of a phononic topological insulator based on QSHE.
References
1. Bernevig, B. A., Hughes, T. L., & Zhang, S. C. (2006). Quantum spin Hall effect and topological phase
transition in HgTe quantum wells. Science, 314(5806): 1757-1761.
2. Qi, X. L., & Zhang, S. C. (2010). The quantum spin Hall effect and topological insulators. arXiv preprint
arXiv:1001.1602.
3. Kane, C. L., & Mele, E. J. (2005). Quantum spin Hall effect in graphene. Physical review letters, 95(22):
226801.
4. Nanthakumar, S. S., Zhuang, X., Park, H. S., Nguyen, C., Chen, Y., & Rabczuk, T. (2019). Inverse design of
quantum spin hall-based phononic topological insulators. Journal of the Mechanics and Physics of Solids, 125:
550-571.
5. Christiansen, R. E., Wang, F., & Sigmund, O. (2019). Topological Insulators by Topology Optimization.
Physical Review Letters, 122(23): 234502.
A Deep Learning Method for a Fully Coupled Thermomechanical Model

Nam Vu-Bac1 and Xiaoying Zhuang1,* and Timon Rabczuk2,*
1Clusterof Excellence PhoenixD (Photonics, Optics, and Engineering-Innovation Across Disciplines), Hannover, Germany.
2
Bauhaus University Weimar, Geschwister-Scholl-Straße 8, 99423 Weimar, Germany.
*
Corresponding Author: Xiaoying Zhuang. Email: zhuang@ikm.uni-hannover.de; Timon Rabczuk. Email: timon.rabczuk@uni-
weimar.de.
Abstract: This study presents a robust method used to solve fully coupled thermomechanical model. The proposed
method is based on computational graphs and backward stochastic differential equations to solve coupled partial
differential equations (PDEs). In this way, solution of a coupled system between a transient heat conduction and
linear elastic mechanical model thus can be obtained. To validate the method numerical results using Pytorch are
implemented. The numerical solutions are compared with exact solution and good agreement is obtained illustrating
robust and accuracy of the method. The method provides a great tool that opens a new door for solving complex
multiphysics problems in industrial application.
At first the computational domain is filled by randomly generated collocation points. Kinematical quantities are
computed that are inserted into the constitutive equations. A loss function including the governing partial differential
equations and boundary conditions is then constructed and minimized at the collocation points via a combination of
optimizers adopted in the back propagation process. By doing so, a set of hyperparameters can be obtained [1-3].
Figure 1 shows an example of a deep neural network including 2 hidden layers [4].
Figure 1: A schematic diagram of a deep neural network with two hidden layers [4]
Let consider the strong forms of the coupled partial differential equations [5,6]
Balance of momentum

35 DACOMA-2019
where σ and b is the respective Cauchy stress and the body force; ū , t̄ and denotes the Dirichlet and Neumann
boundaries, respectively.
Balance for the temperature
where θ and q is the respective temperature and heat flux; θ , q̄ and denotes the Dirichlet and Neumann boundaries,
respectively.
Keywords: Deep energy method (DEM), Thermomechanical model, Artificial neural network (ANN).
Acknowledgement
We are grateful to the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)
for supporting this research under Germany’s Excellence Strategy within the Cluster of Excellence PhoenixD (EXC
2122, Project ID 390833453).
References
1. Anitescu, C., Atroshchenko, E., Alajlan, N., & Rabczuk, T. (2019). Artificial Neural Network Methods for the
heterogeneous materials by image based modeling and deep learning. Computer Methods in Applied Mechanics
and Engineering, 347: 735-753.
3. Guo, H., Zhuang, X., & Rabczuk, T. (2019). A Deep Collocation Method for the Bending Analysis of Kirchhoff
Plate. Computers, Materials & Continua, 59(2): 433-456.
4. Sheehan S, Song YS (2016) Deep Learning for Population Genetic Inference. PLOS Computational Biology
12(3): e1004845.
5. Vu-Bac, N., Duong, T. X., Lahmer, T., Zhuang, X., Sauer, R. A., Park, H. S., & Rabczuk, T. (2018). A
NURBS-based inverse analysis for reconstruction of nonlinear deformations of thin shell structures. Computer
Methods in Applied Mechanics and Engineering, 331: 427-455.
6. Vu-Bac, N., Duong, T. X., Lahmer, T., Areias, P., Sauer, R. A., Park, H. S., & Rabczuk, T. (2019). A NURBS-
based inverse analysis of thermal expansion induced morphing of thin shells. Computer Methods in Applied
A Date Driven Framework to Design Seismic Metamaterial
Qimin Wang1 , Xiaoying Zhuang1,* and Yanyu Chen2

1 Departmentof Geotechnical Engineering, Tongji University, Shanghai 200092, China.
2 Department of Mechanical Engineering, University of Louisville, Louisville, KY40292, USA.
*Corresponding Author: Xiaoying Zhuang. Email: xiaoyingzhuang@tongji.edu.cn.
Abstract: Earthquakes are one of the most harmful natural disasters to humans. When an earthquake occurs, the
source area is undergoing rapid rupture and movement, which constitutes an oscillation source. The oscillation
propagates through the Earth's medium, forming elastic waves in the continuum, namely seismic waves. Waves can
be elastic, acoustic or electromagnetic, but there is one phenomenon in the same: for certain frequency range,
periodic materials or structures, i.e., the so-called metamaterial, can be constructed to hinder wave propagation of
the corresponding frequencies. This phenomenon is the bandgap. To actively prevent structures from potential
seismic damage, design of seismic metamaterial has become an aroused research topic [1]. A seismic metamaterial
is a periodic structure made of different elastic materials, which creates the exact bandgap, mainly for controlling the
propagation of low-frequency seismic surface waves. Although the basic principle of elastic wave propagation has
been established by Brillouin [2], recent research in this field has focused on theory [3, 4] and experimental proofs
[5, 6] of the bandgap. Nevertheless, the design process of seismic metamaterials so far has been quite limited,
mainly by trial-and-error, leaving multitudinous potential optimal design space unexplored.
Machine learning has been applied as an alternative to discover optimal materials such as mechanical metamaterials
with unprecedented properties, such as the accelerated searching of graphene origami [7] and the self-learning-based
hierarchical composite design [8, 9]. The numerical simulation of phononic crystals is a complex eigenvalue
problem. However, due to structural periodicity, the wave vector can be reduced to the Irreducible Brillouin zone
(IBZ) of the solution dimension. With COMSOL and COMSOL Live Link for MATLAB, we simulated the bandgap
of two-dimensional simple periodic structures using the parameter sweeping of IBZ boundaries, and proved that
simple periodic structures can be effective as the prospective seismic material. With Python and PyTorch, we
realized the data driven framework to build a surrogate model for the eigenvalue analysis and the bandgap prediction,
as shown in Figure 1. Under the same training conditions, we compared the performance of the linear regression, the
shallow feedforward NN (neural network) and the deep CNN (Convolutional neural network).
Figure 1: Our data driven framework for seismic metamaterial design.

37 DACOMA-2019
Results show when taking the MSE (Mean square error) as the model accuracy criterion, the MSE of the NN could
be merely 10% of that of the linear regression. Moreover, after introducing the convolutional layers into the neural
network, the MSE of CNN is only 50% of that of NN. This finding proves the intrinsic difference among the classic
data driven method like linear regression and the neural network, as well as the function of convolutional layers to
learn deep features. Our study also shows that in terms of optimizers, the Stochastic Gradient Descent performs
better under relatively the smaller batch size (256) and the smaller learning rate (0.02), whereas Adam instead
performs better under the larger batch size (1480) and the larger learning rate (0.001), which might be helpful for
similar data driven problems.
Keywords: Seismic Metamaterial, Bandgap, Data Driven, Neural Network, Convolutional Neural Network.
Acknowledgement
The Authors thankfully acknowledge the funding from the Tongji University Peak Discipline Program of
Interdisciplinary Project (2019010106).
References
1. Chen Y., Feng Q., Scarpa F., et al. (2019). Harnessing multi-layered soil to design seismic metamaterials with
ultralow frequency band gaps. Materials & Design, 107813.
2. Brillouin L. (1953). Wave propagation in periodic structures: electric filters and crystal lattices, Dover.
3. Geng Q., Zhu S., & Chong K.P. (2018). Issues in design of one-dimensional metamaterials for seismic
protection. SOIL DYNAMICS AND EARTHQUAKE ENGINEERING SOUTHAMPTON, 107(1): 264–278.
4. Nanthakumar S., Zhuang X., Park H. S., et al. (2019). Inverse design of quantum spin hall-based phononic
topological insulators. Journal of the Mechanics and Physics of Solids, 125: 550–571.
5. Brûlé S., Javelaud E., Enoch S., et al. (2014). Experiments on seismic metamaterials: molding surface waves.
Physical review letters, 112(13): 133901.
6. Shi C., Sun H., Hu X., et al. (2016). Acoustic metamaterial panels based on multi frequency vibration
absorbers. Sensors and Smart Structures Technologies for Civil, Mechanical, and Aerospace Systems 2016,
980345.
7. Hanakata P.Z., Cubuk E.D., Campbell D.K., et al. (2018). Accelerated search and design of stretchable
graphene kirigami using machine learning. Physical review letters, 121 (25): 255304.
8. Gu G.X., Chen C.T., Buehler M.J. (2018). De novo composite design based on machine learning algorithm.
Extreme Mechanics Letters, 18: 19–28.
9. Gu G.X., Chen C.T., Richmond D.J., et al. (2018). Bioinspired hierarchical composite design using machine
learning: simulation, additive manufacturing, and experiment. Materials Horizons, 5(5): 939–945.
Simulation of nonlinear propagating waves in hysteresis media via higher-

order central-WENO high resolution schemes
Hassan Yousefi1,*
1School of Civil Engineering, College of Engineering, University of Tehran, Tehran, Iran.
*Corresponding Author: Hassan Yousefi. Email: hyosefi@ut.ac.ir.
Abstract: In this study, higher-order (third and fourth-order) central-WENO high resolution schemes [1-3] are
integrated with different hysteresis models to simulate non-linear wave propagation problems in hysteresis media. In
this regard, both discontinuous solutions and nonlinear waves due hysteresis media can be simulated properly. The
discontinuous solutions can be developed either to abrupt changing in media properties (e.g., density) or nonlinear
responses. For this end, wave equations will be re-expressed as first-order hyperbolic partial differential equations,
the so called strain-velocity or stress-velocity representation. The importance of using the higher-order schemes is
capturing high frequency waves with small wave numbers which is challenging problem in numerical simulations
due to numerical dissipation of different schemes [4] (for instance, the second order Kurganov-Tadmor (KT) central
high resolution method, and common finite element and finite difference methods). Media with the elasto-plastic [5]
and Duhem [6] hysteresis models were integrated with second order central schemes. Here, the higher-order central-
WENO schemes are incorporated with hysteresis models especially those developed for soil-dynamic problems
(such as, the (modified) hyperbolic hysteresis model). This is important for different projects; for example: effects of
soil-nonlinearity on high-speed railway tracks and the site-effects for important projects located near earthquake
sources which emitting wide-range of frequencies (even up to 30 Hz). Finally, it should be mentioned that the
common equivalent linear method, used in different commercial codes, can considerably underestimate the
responses in the high frequency range.
Keywords: Central-WENO, nonlinear wave, hysteresis media, higher-order schemes.
Acknowledgement
Acknowledgements and Reference heading should be left justified, bold, with the first letter capitalized but have no
numbers. Text below continues as normal.
References
1. Yousefi, H., Mohammadi, S., & Rabczuk, T. (2019). Multiscale Polynomial-Based High-Order Central High
Resolution Schemes. Journal of Scientific Computing, 80(1): 555-613
2. Levy, D., Puppo, G., & Russo, G. (1999). Central WENO Schemes for Hyperbolic Systems of Conservation
Laws. ESAIM Mathematical Modelling and Numerical Analysis, 33(3): 547-571
3. Levy, D., Puppo, G., & Russo, G. (2000). Compact Central WENO Schemes for Multidimensional
Conservation Laws. SIAM Journal on Scientific Computing, 22(2): 656–672
4. Yousefi, H., Ghorashi, S. S., & Rabczuk, T. (2016). Directly Simulation of Second Order Hyperbolic Systems in
Second Order Form via the Regularization Concept. Communications in Computational Physics, 20(1): 86-135
5. Giese, G., & Fey, M. (2002). A Genuinely Multidimensional High-Resolution Scheme for the Elastic-Plastic
Wave Equation. Journal of Computational Physics, 181(1): 338-353.
6. Meurer, T., Qu, J., & Jacobs, L. J. (2002). Wave Propagation in Nonlinear and Hysteretic Media-A Numerical
Study. International Journal of Solids and Structures, 39(21-22): 5585-5614.

Numerical Simulation of Stability Analysis of Jointed Rock Slope Based on

SPH Method
Yongzheng Zhang1 and Xiaoying Zhuang2,*
2Tongji University, 1239 Siping Road, 200092 Shanghai, China.
Abstract: In recent years, there are more and more large-scale jointed slope projects. the stability of slopes is
becoming more and more complicated and important. The anisotropic characteristic of rock mass complicates the
problems of rock mechanics. The traditional numerical simulation analysis methods are difficult to analyze the
stability coefficient and potential slip surface of the slope, and it is also difficult to analyze the large deformation
process of the slope after instability, thus, it brings hidden dangers to the safety of the slope engineering. Based on
the smooth particle dynamics (SPH) method described by Lagrange, the SPH program for simulating the stability of
jointed slopes and the large deformation of landslides after instability is compiled in Fortran language, and the
elastic-plastic constitutive equation of rock mass is introduced. The numerical simulation of the stability of jointed
slopes is carried out based on the elastoplastic constitutive model and the Mohr-coulomb yield criterion. This study
makes up for the shortcomings of current numerical simulation methods, and has theoretical and practical
significance for revealing the failure mechanism of jointed slope and optimum design of slope engineering.
Keywords: Smoothed Particle Hydrodynamics (SPH), Rock slope, Joints and fracture, Stability analysis, Numerical
Simulation.
Acknowledgement
This work was funded by China Scholarship Council The support are gratefully acknowledged.
References
1. Lucy L B.Numerical approach to testing the fission hypothesis. (1977). Astronomical Journal, 82(12): 1013-
1024．
2. Bui HH，Fukagawa Ｒ，Sako K, et al. Slope stability analysis and discontinuous slope failure simulation by
elasto-plastic smoothed particle hydrodynamics (SPH). (2011). Geotechnique, 61(7): 565-574．
3. MonaghanJ J． An introduction to SPH. (1989). Computer Physics Communications, 48: 89-96．
4. Barton N, Choubey V. The shear strength of rock joints in theory and practice. (1977). Rock Mechanics, 10(1-
2): 1-54.
5. NonoyamaH, Moriguchi S, Sawada K, et al. (2015). Slope stability analysis using smoothed particle
hydrodynamics (SPH) method. Solis and Foundations, 55(2): 458-470.

Effect of size of microcapsules on the self-healing concrete

Yu Zheng1 and Shuai Zhou1,*
1
College of Materials Science and Engineering, Chongqing University, Chongqing 400045, China
*
Corresponding Author: Shuai Zhou. Email: china_zhoushuai@163.com.
Abstract: The microcapsule-enabled self-healing materials are appealing since they can heal the cracks
automatically. Although much research has been carried out, the influence of the radius of microcapsules on the self-
healing effect is still not well investigated. This research uses the two-dimensional discrete element method (DEM)
to investigate the cracking behavior of specimens containing one microcapsule and one microcrack. The influence of
the size of microcapsules is considered. The potential healing time and the influence of the initial damage are
studied. A large amount of simulation have been conducted and analyzed. The results indicate that the crack
coalescence type between the pre-existing crack and the hole is influenced by the size of holes. The elastic modulus,
the compressive strength and the coalescence stress decrease with the rise of the radius of holes. The initial damage
in experiments should be greater than 95% of the compressive strength to enhance the self-healing effect. The larger
microcapsules require less initial damage. A new type of displacement field near the crack and a new type of
coalescence crack are observed. The research enhances the understanding of the influence of sizes of holes on the
cracking behavior of concrete containing a circular hole and a pre-existing crack.
Keywords: Microcapsule, size, cracking processes, DEM.
References
1. Amiri, F., Milan, D., Shen, Y., Rabczuk, T., & Arroyo, M. (2014). Phase-field modeling of fracture in linear
thin shells. Theoretical and Applied Fracture Mechanics, 69: 102-109.
2. Amiri, F., Anitescu, C., Arroyo, M., Bordas, S., & Rabczuk, T. (2014). XLME interpolants, a seamless bridge
between XFEM and enriched meshless methods. Computational Mechanics, 53(1): 45-57.
3. Zhuang, X. Y., Cai, Y. C., & Augarde, C. (2014): A meshless sub-region radial point interpolation method for
accurate calculation of crack tip fields. Theoretical and Applied Fracture Mechanics, 69(1): 118-125.
4. Zhuang, X. Y., Chun, J. W., & Zhu, H. H. (2014): A comparative study on unfilled and filled crack
propagation for rock-like brittle material. Theoretical and Applied Fracture Mechanics, 72(1): 110-120.
5. Zhuang, X. Y., Zhu, H. H., & Augarde, C. (2014): An improved meshless Shepard and least square method
possessing the delta property and requiring no singular weight function. Computational Mechanics, 53(2): 343-
357.
6. Zhuang, X. Y., & Zhou, S. (2018): Molecular dynamics study of an amorphous polyethylene/silica interface
with shear tests. Materials, 11(6): 929.
7. Zhuang, X. Y., & Zhou, S. (2019): The prediction of self-healing capacity of bacteria-based concrete using
machine learning approaches. Computers Materials & Continua, 59(1): 57-77.

Application of a deep learning method for solving PDEs for compressed air
energy storage
Shuwei Zhou1,*
1Tongji University, Siping Road 1239, 200092 Shanghai, China.
*Corresponding Author: Shuwei Zhou. Email: zhoushuwei1016@126.com.
Abstract: Compressed air energy storage (CAES) is an energy-storage and power generation technology that has
numerous potential applications. Compared with other energy-storage patterns, such as pumped hydroelectric
storage (PHS), CAES has lower capital investment and maintenance costs [1, 2]. A CAES plant utilizes compressed
air as an energy storage and power generation carrier. When the electricity supply exceeds demand, surplus
electrical power is used to drive air compressors and store compressed air into underground caverns. Meanwhile,
during peak hours, the stored compressed air is expelled and mixed with gas fuels. The mixed gas is then fired and
driven to facilitate gas turbine expansion, which in turn runs the electrical power generators.
An underground reservoir is the main component of a CAES plant, and excavations of new hard rock caverns could
provide additional possible site selections although no existing commercial CAES plant is using this kind of caverns.
So researchers have explored possibilities of air storage in hard rock caverns. In unlined rock caverns, the air is
normally prevented from leaking by natural groundwater or artificial water curtain. Thus the stored air pressure is
balanced by the ambient hydrostatic water pressure within the host rock. However, these groundwater control
methods are unsuitable for regions where water is scare. Thus, an alternative method, CAES in lined rock caverns, is
studied in the current study and was found to have has a better applicability. Two types of the underground cavern
for CAES are shown in Fig. 1.
Charge Discharge Charge Discharge
Rock Rock
Compressed Compressed
air air
T, p
T, p
(a) Tunnel-shaped (b) Vertical-jar-shaped

Figure 1: Two types of underground caverns for CAES
It is a great challenge to predict air temperature and pressure in an underground cavern for compressed-air energy
storage (CAES) owing to huge operational cycles and multiple field interaction [3, 4]. The deep-learning approach is
proved an effective tool for solving this [5]. We applied a deep-learning method to solve the corresponding equation
set after the strong form is established, while the solution is approximated with a deep neural network [6, 7].
Necessary train processes are required to adjust the differential operator, initial condition, and boundary conditions.
Meshfree implementation is used and we also constitute a simple analytical solution for the purpose of comparison.
The neural network is trained on batches of randomly sampled time and space points, while the deep learning
method provides the general solution to the mass conservation equation, energy conservation equation, and air state
equation. The used method has a similar spirit to Galerkin methods, with the solution approximated by a neural

42 DACOMA-2019
network instead of a linear combination of basis functions. The compassion of the results solved by the deep
learning method and the analytical solution shows a significant agreement.
Keywords: Deep learning, PDEs, Compressed air energy storage, Neural network, Air state.
Acknowledgement
The financial support provided by the National Natural Science Foundation of China (No. 51278378) and the
National High Technology Research and Development Program of China (863 Program, No. SS2012AA052501) is
gratefully acknowledged.
References
1. Zhou S, Xia C, Du S, et al. (2015). An analytical solution for mechanical responses induced by temperature
and air pressure in a lined rock cavern for underground compressed air energy storage. Rock Mechanics and
Rock Engineering, 48(2): 749-770.
2. Zhou S, Xia C, Zhao H, et al. (2017). Numerical simulation for the coupled thermo-mechanical performance of
a lined rock cavern for underground compressed air energy storage[J]. Journal of Geophysics and Engineering,
14(6): 1382-1398.
3. Zhou, S., Zhuang, X., & Rabczuk, T. (2019). Phase-field modeling of fluid-driven dynamic cracking in porous
media. Computer Methods in Applied Mechanics and Engineering, 350: 169-198.
4. Zhou, S., Zhuang, X., & Rabczuk, T. (2018). A phase-field modeling approach of fracture propagation in
poroelastic media. Engineering Geology, 240: 189-203.
5. Sirignano, J., & Spiliopoulos, K. (2018). DGM: A deep learning algorithm for solving partial differential
equations. Journal of Computational Physics, 375: 1339-1364.
Solution of Second Order Boundary Value Problems. Computers, Materials & Continua, 59, 345-359.
7. Guo, H., Zhuang, X., & Rabczuk, T. (2019). Artificial Neural Network Methods for the Solution of Second
Order Boundary Value Problems. Computers, Materials & Continua, 59: 433-456.
Machine Learning based Crack Growth Prediction: Application to a

Helicopter Component Digital Twin
Xuan Zhou1 and Leiting Dong1,*
1Beihang University, No. 37 Xueyuan Road, 100191 Beijing, China.
*Corresponding Author: Leiting Dong. Email: ltdong@buaa.edu.cn.
Abstract: Digital Twin is an integrated multiphysics, multiscale, probabilistic simulation of an as-built vehicle or
system that uses the best available physical models, sensor updates, fleet history, etc., to mirror the life of its
corresponding physical twin [1]. In the digital twin application, the load data from onboard sensors is used to predict
the crack growth and remaining life in real time. However, traditional fracture mechanics simulation methods are
time-consuming. In this article, we propose a machine learning based approach, combining with high performance
fracture mechanics simulation, to perform real-time fatigue crack growth prediction for structures with complex
geometry. High-fidelity simulations are carried out, by using the SGBEM-FEM alternating method [2] to generate
the training data. The mapping between various loads /crack geometry and stress intensity factor (SIF) of crack
fronts is established as a surrogate model by machine learning method such as gaussian progress regression (GPR)
and support vector regression (SVR). With the real-time load and current crack geometry as input, the surrogate
model can quickly export the SIF to the fatigue crack growth model to analyze the crack growth and update the
crack geometry step by step. Here an example of a flawed helicopter lift frame [3] is presented to demonstrate the
analysis procedure and capabilities of the method. The crack configuration is a small corner defect at the edge of a
large central hole in a flanged plate made of 7010 aluminum alloy, and the component is subjected to the ASTERIX
spectrum load. To reduce the variables representing the complex crack shapes, the principle component analysis
(PCA) were used to parameterize the crack fronts. The generalized Frost–Dugdale law [4] and NASGRO equation
[5] were used to predict the fatigue crack growth. The crack growth history and remaining life obtained by our
approach coincide well with simulation and experiment results presented in other papers [6-8]. Results show that the
proposed approach can accurately predict the evolution of the crack growth over entire load histories and
substantially reduce the computing time, which can be used for real-time health monitoring in digital twin
applications.
Keywords: Fatigue crack growth, Surrogate model, Machine learning, SGBEM-FEAM, Digital twin.
References
1. Glaessgen, E., & Stargel, D. (2012, April). The digital twin paradigm for future NASA and US Air Force
vehicles. In 53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference
20th AIAA/ASME/AHS Adaptive Structures Conference 14th AIAA (p. 1818).
2. Han, Z. D., & Atluri, S. N. (2002). SGBEM (for cracked local subdomain)-FEM (for uncracked global
structure) alternating method for analyzing 3D surface cracks and their fatigue-growth. CMES- Computer
Modeling in Engineering and Sciences, 3(6): 699-716.
3. Irving, P. E., Lin, J., & Bristow, J. W. (2003). Damage tolerance in helicopters. In Report on the round robin
challenge, presented at the American Helicopter Society 59th annual forum, Phoenix, AZ, USA.
4. Barter, S., Molent, L., Goldsmith, N., & Jones, R. (2005). An experimental evaluation of fatigue crack growth.
Engineering failure analysis, 12(1): 99-128.
5. Mettu, S. R., Shivakumar, V., Beek, J. M., Yeh, F., Williams, L. C., Forman, R. G., ... & Newman Jr, J. C.
(1999). NASGRO 3.0: A software for analyzing aging aircraft.
6. Vaughan, R. E., & Chang, J. H. (2004). Life predictions for high cycle dynamic components using damage
tolerance. Journal of the American Helicopter Society, 49(4): 493-500.
7. Newman Jr, J. C., Irving, P. E., Lin, J., & Le, D. D. (2006). Crack growth predictions in a complex helicopter
component under spectrum loading. Fatigue & fracture of engineering materials & structures, 29(11): 949-958.

44 DACOMA-2019
8. Tiong, U. H., & Jones, R. (2009). Damage tolerance analysis of a helicopter component. International Journal
of Fatigue, 31(6): 1046-1053.
Numerical implementation of data-driven computational mechanics

Ying Zhou1,2 and Yuantong Gu1,*
1School of Chemistry, Physics and Mechanical Engineering, Queensland University of Technology, 4001 Brisbane QLD,
Australia.
2 State IJR Center of Aerospace Design and Additive Manufacturing, Northwestern Polytechnical University, 710072 Xi’an
Shaanxi, China.
*Corresponding Author: Yuantong Gu. Email: yuantong.gu@qut.edu.au.
Abstract: Data-driven computational mechanics emerges as a new methodology in computational mechanics, which
directly uses raw material data instead of empirical material models in calculation [1,2]. The solution of data-driven
solver is to find the optimal material states (ε, σ) minimizing the distances between the material data sets and
equilibrium & compatibility constraints. The distance is defined in the metric of strain energy of a pseudo linear-
elastic material. In this paper, we thoroughly investigate the effects of the pseudo linear-elastic material on data-
driven computing solver. The data-driven solver is decomposed into two sub-problems. The inner sub-problem finds
the optimal material states (ε, σ) in the equilibrium & compatibility constraints closest to prescribed material state in
the material data set. This process is independent of the pseudo linear-elastic material. The outer sub-problem finds
the nearest neighbor of the calculated material states (ε, σ) in the material data sets. The nearest neighbor search
space can be regarded as a super-ellipsoid centered at (ε, σ). Some issues in numerical implementation are discussed
and several key parameters are investigated. It is found that the construction of super-elliptical search space along
the direction of the equilibrium & compatibility constraints will improve the efficiency and accuracy of the
iteratively data-driven computing process.
Keywords: Data-driven computational mechanics, Material data set, Nearest neighbor search.
References
1. Kirchdoerfer, T., Ortiz M. (2016). Data-driven computational mechanics. Computer Methods in Applied
2. Conti, S., Müller S., Ortiz M. (2018). Data-driven problems in elasticity. Archive for Rational Mechanics and
Analysis, 229: 79-123.

A Real-Time Structural Topology Optimization Method using Conditional

Generative Adversarial Networks
Xuefeng Zhu1,*, Xiaochen Liu1, Ping Hu1 and Zheng-Dong Ma2
1School of Automotive Engineering, Dalian University of Technology, 116024 Dalian, China.
2Department of Mechanical Engineering, Michigan University, 48130 Ann Arbor, USA.
*Corresponding Author: Xuefeng Zhu. Email: xuefeng@dlut.edu.cn.
Abstract: Traditional topology optimization methods such SIMP method [1] or level set method rely on time-
consuming iterative algorithms, and with the increase of the degrees of freedom of the background grid, the
computation time increases exponentially, that is, the so-called “dimension curse” problem. To solve this problem,
this paper proposes a real-time topology optimization method based on Conditional Generative Adversarial
Networks (CGAN) [2], which can almost real-time predict accurate optimization results according to input
displacement and load boundary conditions. First, we establish a data set using the simulation data generated by
SIMP method, and then train the dataset by using Conditional Generative Adversarial Networks with Gradient
Penalty (CGAN-GP) to generate optimized structures (see Fig. 1). To improve quality of optimized structures
generated by CGAN-GP, we use Pix2pix GAN to enhance the sharpness of these topologies, and get the structure
with high clarity. Comparing the traditional topology optimization, numerical examples illustrate that computational
time can be reduced dramatically while the accuracy rate is up to 85%.
CWGAN-GP
Real/SIMP
Figure 1: Comparison of optimized structures obtained by the presented method and traditional SIMP algorithm.
Keywords: Real-Time Structural Topology Optimization, Conditional Generative Adversarial Networks, Deep
learning, Artificial intelligence.
Acknowledgement
This work was funded by the National Natural Science Foundation of China (No.11872015, No. 51775025) and
Fundamental Research Funds for the Central Universities (No. DUT18JC47). These supports are gratefully
acknowledged.

47 DACOMA-2019
References
1. Tcherniak, D. (2002). Topology optimization of resonating structures using SIMP method. International
Journal for Numerical Methods in Engineering, 54(11): 1605-1622.
2. Mirza, M., & Osindero, S. (2014). Conditional generative adversarial nets. Computer Science, 2672-2680.

Patricio Rodriguez1 and Xiaoying Zhuang1,2,*
1TDepartment of Geotechnical Engineering, Tongji University, Shanghai, China.
2
Institute for continuum mechanics, Leibniz University, AppelStraße 11, 30167 Hannover, Germany.
*
Abstract: Phase field modelling of fracture has revitalized the study of failure phenomena in the scientific
community during the last two decades. Multiple applications in multi-field frameworks [1-3] have been attempted
to exploit the advantages of the approach. Complex crack behaviors like nucleation, merging and branching, which
used to rely on intricate numerical techniques, are naturally handled by an additional partial differential equation.
From a theoretical point of view, a phase-field crack evolution is dictated by three energetical principles [4] namely,
energy conservation, irreversibility and global stability. While irreversibility and energy conservation are grounded
on thermodynamic principles and are widely accepted, the third one is still under debate. As pointed out by [5], a
global stability criterion results in unphysical results in nonconvex variational problems since a jumps over energy
barriers would be necessary in order to reach a global minimum. In terms of fracture modelling, numerical results
show cracks propagating instantaneously through the domain when a continuously evolving crack is expected. A
local stability criterion would be more suitable, however as shown by [5], even in this scenario the formulation
allows jumps during crack growth, breaking the causality principle [5]. Modifications to the theoretical framework
have been proposed by [5] and [6] pointing out the necessity of capturing rate dependent effects during fast crack
propagation to obtain physically meaningful solutions. However the implementation of a numerical scheme capable
of representing viscous and inertial effects in a rate independent framework has not been tackled yet. Here, an
algorithm based on a Machine Learning algorithm is proposed in order to study the space of admissible solutions
during the optimization process.
Machine learning has been successfully applied in other contexts to define the optimization domain in problems
with non-convex objective functions in material science [7]. In [8] a classification of learning algorithm was used to
define an optimization domain based on data from experiments and numerical simulations. In [9] an adaptive
algorithm is defined in order to modify the definition of the space of interest during the optimization process.
Following the same line of idea, a multilayered feed-forward neural network is used to explore the effect of inertial
effects on crack propagation during fast transitions. The finite element method is used to discretize the partial
differential equations resulting from the minimization of the total energy functional. Crack propagation in the quasi-
static and dynamic cases is considered. The Alternating Minimization Algorithm (AMA) is adopted here to solve
the quasi-static problem and implicit generalized alpha integration method for the dynamic case. By feeding pairs of
simulations of the quasi-static and dynamic crack propagation examples a multilayer neural network is trained to
calculate a relaxation ratio of a sequence of solutions of the evolution problem. Afterwards, the relaxation ratio is
used along with the quasi-static scheme to obtain improved solutions that better reflect the rate dependent
phenomena that are expected during jumps in the system evolution.
The results show that the total energy of the system is not conserved during fast crack growth, however the energy
imbalance obtained with the calculated relaxation ratio agrees well, within a certain range of crack propagation
velocities, with the energy associated to inertial effects in the dynamic case. For crack tip velocities beyond this
threshold the quasi-static solution becomes inadequate to model crack propagation since phenomena like branching
and crack propagation promoted by wave reflection take on a dominating roll during the evolution.
Keywords: Phase-fields, variational formulation, fracture, rate independent.
Acknowledgement
The author would like to acknowledge the financial support from the Sofja Kovalevskaja Prize of the Alexander von
Humboldt Foundation (Germany).

49 DACOMA-2019
References
1. Wilson, Z. A., Borden, M. J., & Landis, C. M. (2013). A phase-field model for fracture in piezoelectric
ceramics. International Journal of Fracture, 183(2): 135-153.
2. Wilson, Z. A., & Landis, C. M. (2016). Phase-field modeling of hydraulic fracture. Journal of the Mechanics
and Physics of Solids, 96: 264-290.
3. Borden, M. J., Verhoosel, C. V., Scott, M. A., Hughes, T. J., & Landis, C. M. (2012). A phase-field description
of dynamic brittle fracture. Computer Methods in Applied Mechanics and Engineering, 217: 77-95.
4. Mielke, A., & Roubícek, T. (2015). Rate-independent systems. Theory and Application (in preparation).
5. Alessi, R. (2016). Energetic formulation for rate-independent processes: remarks on discontinuous evolutions
with a simple example. Acta Mechanica, 227(10): 2805-2829.
6. Larsen, C. J. (2010). Epsilon‐stable quasi‐static brittle fracture evolution. Communications on Pure and
Applied Mathematics: A Journal Issued by the Courant Institute of Mathematical Sciences, 63(5): 630-654.
7. Liu, Y., Zhao, T., Ju, W., & Shi, S. (2017). Materials discovery and design using machine learning. Journal of
Materiomics, 3(3): 159-177.
8. Balachandran, P. V. (2019). Machine learning guided design of functional materials with targeted
properties. Computational Materials Science, 164: 82-90.
9. Gopakumar, A. M., Balachandran, P. V., Xue, D., Gubernatis, J. E., & Lookman, T. (2018). Multi-objective
optimization for materials discovery via adaptive design. Scientific reports, 8(1): 3738.
Robust topology optimization using an improved adaptive Gaussian process

Somdatta Goswami1,* and Souvik Chakraborty2
1Instituteof Structural Mechanics, Bauhaus University Weimar, 99423-Weimar, Germany.
2
Center for Informatics and Computational Science (CICS), University of Notre Dame, IN, USA.
*
Corresponding Author: Somdatta Goswami. Email: somdatta.goswami@uni-weimar.de.
Abstract: Topology optimization is a mathematical tool that optimizes the geometry of an object, within a given
space and for a given set of loads and constraints. The goal remains to maximize the performance of a part of the
system by reducing its mass and the cost of manufacturing. Out of various optimization procedures, topology
optimization yields the optimum configuration. Over the years, the concept has gained huge popularity among both
researchers and practicing engineers. Popular topology optimization methods are based on density, level set,
topological derivative, phase field, evolutionary approaches and several others [1]. Further advanced methods by
coupling two or more of the above-mentioned methods can also be found in literature [2]. However, all the methods
discussed above conveniently neglect the presence of uncertainties in the system. In this context, it is to be noted
that all physical systems have inherent uncertainties associated. Ignoring these uncertainties can lead to systems that
exhibit poor performance, or even fail. Therefore, it is of utter importance to consider the effect of uncertainties
while optimizing a system.
In this work, a novel approach for topology optimization under uncertainty has been proposed. Out of two available
options for optimization under uncertainty [3], namely robust design optimization (RDO) and reliability-based
design optimization, the former has been pursued in this study. In order to reduce the computational effort associated
with robust topology optimization, an improved version of Gaussian process, referred to as hybrid polynomial
correlated function expansion (H-PCFE) [4] has been integrated within the topology optimization framework. H-
PCFE is used for computing the moments associated with the objective function in robust topology optimization
(RTO). It is argued that re-training the H-PCFE at every iteration will make the overall algorithm computationally
expensive and time consuming. To alleviate this issue, an adaptive algorithm that automatically decides whether to
retrain the H-PCFE or to use the available H-PCFE model, is proposed. The proposed approach is highly flexible
and can be integrated with already available deterministic topology optimization codes from literature.
Mathematically, RTO can be represented as:
=d ∗ argmin cRTO ( Ε(c(d , X )), var(c(d , X )) )
 V (d )
 = f (1)
V0

subject to  KU = F ,
0 < d
i ,min ≤ d i , i =1,..., N d


where d represents the design variables, di ,min is the minimum relative density for the i th design variable, and N d is
the total number of design variables. X represents the variables with uncertainty. E(•) and var(•) denote the mean
and the variance operator, respectively. F, K and U in equation (1) represent the force vector, stiffness matrix and
the displacement vector of the global system respectively. V ( d ) and V0 represent the material volume and design
domain volume, respectively, f indicates the prescribed volume fraction and c(d , X ) denotes the compliance of the
system.
Performance of the proposed approach has been investigated on the well-known MBB beam. The geometrical
domain and the boundary conditions are shown in Figure 1(a). Due to symmetrical boundary conditions, only half
on the structure, shown in Figure 1(b), is analyzed. Among the material properties, Poisson’s ratio, υ=0.3 and
young’s modulus, E is uncertain with spatial variability and is modelled as a Gaussian random field. The overall
domain is discretized into 60x40 four noded unit square element. Like [1], the mean of the random field is

51 DACOMA-2019
considered as unity. The random field is discretized into 16 random variables by using Karhunen- Loève expansion.
The objective of the problem is to compute the optimum topology of the system. The RTO problem has 2400 design
variables. Figure 1(d) shows the results obtained using the proposed approach. The results obtained using the
deterministic optimization is shown in Figure 1(c). We can see that considering uncertainty results in a more
conservative design (additional limbs). This is expected and is consistent with results in RDO literature. This, in turn,
provides insurance against structural failure due to uncertainties in the system.
(a) Complete design domain for topology optimization (b) Symmetrical boundary conditions
(c) Deterministic topology optimization (d) Robust topology optimization

Figure 1
Keywords: RDO, Topology optimization, Uncertainty, H-PCFE.

Acknowledgement
SG acknowledges the support of the German Academic Exchange Service (DAAD).
References
1. Sigmund, O., & Maute, K. (2013). Topology optimization approaches. Structural and Multidisciplinary
Optimization, 48(6): 1031-1055.
2. Yamada, T., Izui, K., Nishiwaki, S., & Takezawa, A. (2010). A topology optimization method based on the
level set method incorporating a fictitious interface energy. Computer Methods in Applied Mechanics and
Engineering, 199(45-48): 2876-2891.
3. Chen, S., Chen, W., & Lee, S. (2009). Level set based robust shape and topology optimization under random
field uncertainties. Structural and Multidisciplinary Optimization, 41(4): 507-524.
4. Chakraborty, S., & Chowdhury, R. (2015). Polynomial Correlated Function Expansion for Nonlinear
Stochastic Dynamic Analysis. Journal of Engineering Mechanics, 141(3): 04014132.
An intelligent detection model based on fully convolutional neural network

for pavement crack
Duo Ma1,2,3, Hongyuan Fang1,2,3,*, Binghan Xue1,2,3 and Fuming Wang1,2,3
1College of Water Conservancy & Environmental, Zhengzhou University, Zhengzhou 450001, China.
2National Local Joint Engineering Laboratory of Major Infrastructure Testing and Rehabilitation Technology, Zhengzhou
450001, China.
3Collaborative Innovation Center of Water Conservancy and Transportation Infrastructure Safety, Henan Province, Zhengzhou
450001, China.
*Corresponding Author: Hongyuan Fang. Email: 18337192244@163.com.
Abstract: Crack is an often appearing pavement disease, if not timely treatment, will result in the pavement damage
deepening, and affect the normal use of the road. Therefore, it is of great significance to establish an efficient
intelligent identification model for pavement cracks. Neural network is a method of simulating animal nervous
system using gradient descent [1] to predict results by learning weight matrix, which has been widely used in
geotechnical engineering [2], computer vision [3-5], medicine [6] and other fields. However, there are three major
problems in the application of neural network method for crack identification: Too few layers; Extracted crack
features are not complete; Low efficiency to calculate the whole picture. In this paper, a fully convolutional neural
network [7] based on resnet-101[8] is used to establish an intelligent identification model of pavement crack region.
This method using convolutional layer instead of fully connected layer realizes full convolution and speeds up
calculation. The sub-region comes from the feature map at the end of the base network, which avoids multiple
computations of the same picture. In order to improve the recognition accuracy of the model, OHEM and
data-augmentation techniques are adopted. Concrete Crack Images for Classification (CCIC) which is a public data
set collected by using smart phone and the Crack Image Database (CIDB) automatically collected by
vehicle-mounted CCD camera are trained and tested, the identification accuracy reaches 91.4% and 86.4%,
respectively. Comparing with Faster RCNN and different depth models, the proposed model has higher recognition
accuracy and recall rate as shown in fig.1, and can extract more complete and accurate crack features in CIDB.
Furthermore, rotation, translation and scaling images are analyzed. The proposed model has strong robustness and
stability, and can automatically identify image cracks of different forms. The proposed model has broad application
prospects in practical engineering problems.
Keywords: Fully convolutional neural network, Pavement crack, Intelligent detection, Crack Image Database.
Figure 1

53 DACOMA-2019
Acknowledgement
The authors thank the supports from the National Key Research and Development Program of China
(No.2017YFC1501200) and the National Natural Science Foundation of China (No.51678536), the Scientific and
Technological Research Program of Henan Province (No.171100310100), the Outstanding Young Talent Research
Fund of Zhengzhou University (1621323001).
References
1. Souvik Chakraborty, Somdatta Goswami, Timon Rabczuk. (2019). A surrogate assisted adaptive framework
for robust topology optimization. Computer Methods in Applied Mechanics and Engineering, 346.
2. Chen J, Zhu HH, Li X. (2016). Automatic extraction of discontinuity orientation from rock mass surface 3D
point cloud. Computers & Geosciences, 95: 18-31.
3. Kelvin C P Wang, Allen Zhang. (2017). Deep Learning for Asphalt Pavement Cracking Recognition Using
Convolutional Neural Network. Airfield and Highway Pavements
4. Zhang A, Wang K C P, Li B, et al. (2017). Automated Pixel-Level Pavement Crack Detection on 3D Asphalt
Surfaces Using a Deep-Learning Network. Computer aided Civil & Infrastructure Engineering.
5. Somin Park, Seongdeok Bang. Patch-Based Crack Detection in Black Box Images Using Convolutional Neural
Networks.Journal of Computing in Civil Engineering.
6. Hirshberg A, Adar R. (1997). Artificial neural networks in medicine. Delay learning in artificial neural
networks.
7. Dai J, Li Y, He K, et al. (2016). R-FCN: object detection via region-based fully convolutional networks. IEEE
Conference on Computer Vision and Pattern Recognition (CVPR).
8. He K, Zhang X, Ren S, et al. (2015). Deep Residual Learning for Image Recognition. IEEE Conference on
Computer Vision and Pattern Recognition (CVPR).
Bayesian solution for inverse elasticity problems with hybrid uncertainties

using Polynomial Chaos Surrogate dictionaries
Tittu Varghese Mathew1, S. Natarajan1,* and S.P.A. Bordas2
1
Department of Mechanical Engineering, Indian Institute of Technology-Madras, India.
2
Faculté des Sciences, de la Technologie et de la Communication, University of Luxembourg, Luxembourg.
*
Corresponding Author: S. Natarajan. Email: snatarajan@iitm.ac.in.
Abstract: A method is proposed for inferring the presence of an inclusion/void inside an elastic domain under both
material and geometric uncertainties. Marzouk.et.al [2] tried, for the first time, to accelerate the Bayesian inference
by using Polynomial Chaos Expansions (PCE) to represent random variables. Advantage of such a method is that
the evaluation of integrals over the unknown parametric space gets recast into random variable space that underlies
the PCE. Conteras.et.a [1] extended this approach to develop PCE based offline dictionaries for one single inclusion
case of different shapes and sizes within an elastic domain. During the online stage, a model selection approach
based on the evidence provided by the Bayes factor for each of the offline dictionaries were determined and the most
suitable dictionary was chosen that had the highest evidence. Yan.et.al [3] proposed a guided Bayesian inference
approach for detecting multiple flaws in structures using the extended finite element method. Clearly, there exists a
gap in literature where a Bayesian inversion is carried out on elasticity problems were uncertainties in both
geometric and material properties do exist. In this work, a novel method is proposed to address this gap wherein, in
the offline stage, a few PCE surrogates are developed for different combinations of material uncertainties and
inclusion/void geometry. Both the material and geometric uncertainties are taken into account in PCE surrogate
construction using a novel stochastic XFEM method developed in our lab. Thereafter, under the assumption that
there exist only one inclusion/void in the elastic domain, the uncertain material parameters and geometric features
are inferred using the Bayesian evidence estimation with the help of some synthetic data and model averaging. The
proposed novel method is exemplified using a simple 2D elastic domain under static loading case.
Keywords: Bayesian evidence, Model selection, Stochastic XFEM, Material uncertainty, Geometric uncertainty.
Acknowledgement
The first author would like to thank Dr. Hussein Rappel from the University of Luxembourg for the valuable
discussions on Bayesian modelling and model selection.
References
1. Contreras, A., Le Maitre, O.P., Aquino, W., Knio, O.M., (2016). Multi-Modal Polynomial Chaos Surrogate
Dictionary for Bayesian Inference in Elasticity Problem. Probabilistic Engineering Mechanics, 46: 107-119.
2. Marzouk, Y.M., Najm H.N., Rahn, L.A. (2007). Stochastic spectral methods for efficient Bayesian solution of
inverse problems. Journal of Computational Physics, 224(2): 560-586.
3. Yan, G., Sun, H., & Waisman, H. (2015). A guided Bayesian inference approach for detection of multiple
flaws in structures using the extended finite element method. Computers and structures, 152: 27-44.

Adaptive grid refinement in a 3D peridynamic model

Arman Shojaei1,*
1Institute of Materials Research, Materials Mechanics, Helmholtz-Zentrum Geesthacht, Geesthacht, Germany.
*
Corresponding Author: Arman Shojaei. Email: arman.shojaei@hzg.de.
Abstract: The differentiability of displacement field is a rudimentary assumption in the classical continuum
mechanics. In this theory the equations of motion are derived based on partial differential equations. Therefore, this
assumption breaks downs when solution of problems containing discontinuities, such as crack propagation, is of
concern [1]. In fact, the situation becomes more complicated when real life problems, such as 3D propagation of
cracks, come into the picture [2-5].
More recently, peridynamics is receiving a great amount of attention by many researchers in the computational
mechanics community. It is considered to be as an alternative and promising theory appropriately formulated for
discontinuous problems [6-10]. Peridynamic is well designed to describe failure in structures as the theory makes
use of integral equations rather than partial differential equations. In fact, in peridynamics, the equation of motion is
described by a nonlocal integral operator rather than the divergence of the stress tensor involved in the classical
theory [6,7].
The most common way of implementing and discretizing peridynamics is a meshfree approach which uses a
uniform grid of nodes [7, 11]. The peridynamic models most often are much more expensive that methods based on
the classical theory. This lies in the fact that, in peridynamic models, each computational node interacts with many
neighboring nodes over a finite region. To this end, performing realistic problems (in particular 3D problems) with
peridynamic models requires a vast amount of computational resources. In addition, the restriction of the
discretization to uniform grid of nodes leads to inefficiency of the method when a very dense grid for a localized
area is required [11, 12].
Based on the aforementioned issue, in this study a new adaptive refinement technique for 3D peridynamic models is
developed. We shall show that the method can be applied to crack propagation problems in both brittle and ductile
materials; moreover, it permits to increase the resolution of the analysis only in critical zones. The proposed method
is based on a new technique that couples two peridynamic grids (with different grid sizes) using a morphing strategy.
The performance of the method in solution of some 3D real life problems including quasi static propagation of
cracks in composite plates, impact loading conditions on fragile industrial structures is investigated. We compare the
solutions of the present method with those of a conventional peridynamic model, which employs uniform
discretization, and show that the same solution is obtained at a much smaller computational cost.
Keywords: Peridynamic, Meshfree, Nonlocal theory, Adaptive grid refinement, failure, 3D problems.
References
1. Mossaiby, F., Bazrpach, M. and Shojaei, A. (2015). Extending the method of exponential basis functions to
problems with singularities. Engineering Computations, 32(2): 406-423.
2. Mossaiby, F., Shojaei, A., Zaccariotto, M. and Galvanetto, U. (2017). OpenCL implementation of a high
performance 3D Peridynamic model on graphics accelerators. Computers & Mathematics with Applications,
74(8): 1856-1870.
3. Rabczuk, T. and Belytschko, T. (2007). A three-dimensional large deformation meshfree method for arbitrary
evolving cracks. Computer Methods in Applied Mechanics and Engineering, 196(29-30): 2777-2799.
4. Shojaei, A., Galvanetto, U., Rabczuk, T., Jenabi, A. and Zaccariotto, M. (2019). A generalized finite difference
method based on the Peridynamic differential operator for the solution of problems in bounded and unbounded
domains. Computer Methods in Applied Mechanics and Engineering, 34: 100-126.
5. Bazazzadeh, S., Shojaei, A., Zaccariotto, M. and Galvanetto, U. (2018). Application of the peridynamic
differential operator to the solution of sloshing problems in tanks. Engineering Computations, 36: 45-83.
6. Silling, S.A. (2000). Reformulation of elasticity theory for discontinuities and long-range forces. Journal of the

56 DACOMA-2019
Mechanics and Physics of Solids, 48(1): 175-209.

7. Silling, S.A. and Askari, E. (2005). A meshfree method based on the peridynamic model of solid mechanics.
Computers & structures, 83(17-18): 1526-1535.
8. Ren, H., Zhuang, X. and Rabczuk, T. (2017). Dual-horizon peridynamics: A stable solution to varying horizons.
Computer Methods in Applied Mechanics and Engineering, 318: 762-782.
9. Shojaei, A., Mudric, T., Zaccariotto, M. and Galvanetto, U. (2016). A coupled meshless finite
point/Peridynamic method for 2D dynamic fracture analysis. International Journal of Mechanical Sciences,
119: 419-431.
10. Shojaei, A., Zaccariotto, M. and Galvanetto, U. (2017). Coupling of 2D discretized Peridynamics with a
meshless method based on classical elasticity using switching of nodal behaviour. Engineering Computations,
34(5): 1334-1366.
11. Shojaei, A., Mossaiby, F., Zaccariotto, M. and Galvanetto, U. (2018). An adaptive multi-grid peridynamic
method for dynamic fracture analysis. International Journal of Mechanical Sciences, 144: 600-617.
12. Zaccariotto, M., Mudric, T., Tomasi, D., Shojaei, A. and Galvanetto, U. (2018). Coupling of FEM meshes with
Peridynamic grids. Computer Methods in Applied Mechanics and Engineering, 330: 471-497.
Feasibility of machine learning methods for the prediction of roadheader

performance in underground excavation project-A case study
Hadi Fattahi 1,*
1
Department of Mining Engineering, Arak University of Technology, Arak, Iran.
*
Corresponding Author: Hadi Fattahi. Email: h.fattahi@arakut.ac.ir.
Abstract: Estimation of roadheader performance is one of the main subjects in determining the economics of the
underground excavation projects. The poor performance estimation of the roadheaders can lead to costly contractual
claims. In this paper, the application of machine learning methods for data analysis named artificial neural network
(ANN) optimized by hybrid ant colony optimization (ACO) to estimate of roadheader performance is demonstrated.
The estimation abilities offered using ANN-ACO were presented by using field data of achieved from tunnels for
Istanbul’s sewerage system, Turkey. In this model, Schmidt hammer rebound values and rock quality designation
(RQD) were utilized as the input parameters, while net cutting rates was the output parameter. Various statistical
performance indexes were used to compare the performance of those estimation model. The results achieved
indicate that the ANN-ACO model has strong potential to estimation of roadheader performance with high degree of
accuracy and robustness.
Keywords: Machine learning, Roadheader performance, ant colony optimization, Artificial neural network.
References
1. Abdolreza, Y.-C., & Siamak, H. Y. (2013). A new model to predict roadheader performance using rock mass
properties. Journal of Coal Science and Engineering (China), 19(1): 51-56.
2. Ahmadi, M. H., Aghaj, S. S. G., & Nazeri, A. (2013). Prediction of power in solar stirling heat engine by using
neural network based on hybrid genetic algorithm and particle swarm optimization. Neural Computing and
Applications, 22(6): 1141-1150.
3. Bilgin, N., Seyrek, T., Erdinc, E., & Shahriar, K. (1988). Roadheader performance in Istanbul, Golden Horn
clean-up contributes valuable data. Tunnels & Tunneling, 6: 41-44.
4. Bilgin, N., Seyrek, T., Erdinc, E., & Shahriar, K. (1990). Roadheaders clean valuable tips for Istanbul Metro.
Tunnels & Tunneling, 10: 29-32.
5. Bilgin, N., Dincer, T., Copur, H., & Erdogan, M. (2004). Some geological and geotechnical factors affecting
the performance of a roadheader in an inclined tunnel. Tunnelling and Underground Space Technology, 19(6):
629-636.
6. Braik, M., Sheta, A., & Arieqat, A. A comparison between GAs and PSO in training ANN to model the TE
chemical process reactor. Paper presented at the Proceedings of the AISB 2008 symposium on swarm
intelligence algorithms and applications.
7. Chen, S. F. (2011). Redundant Feature Selection Based on Hybrid GA and BPSO. Paper presented at the
Communication Software and Networks (ICCSN), 2011 IEEE 3rd International Conference on.
8. Chiu, S. L. (1994). Fuzzy model identification based on cluster estimation. Journal of intelligent and Fuzzy
systems, 2(3): 267-278.
9. Chopra, S., Mitra, R., & Kumar, V. (2006). Reduction of fuzzy rules and membership functions and its
application to fuzzy PI and PD type controllers. International Journal of Control, Automation and Systems,
4(4): 438.
10. Copur, H., Ozdemir, L., & Rostami, J. (1998). Roadheader applications in mining and tunneling industries.
Mining Engineering, 50: 38-42.
11. Douglas, W. (1985). Roadheaders open new horizons at San Manuel. Engineering & Mining Journal, 186(8):
22-25.
12. Eberhart, R., & Kennedy, J. (1995). A new optimizer using particle swarm theory. Paper presented at the Micro
Machine and Human Science, 1995. MHS'95., Proceedings of the Sixth International Symposium on.
13. Ebrahimabadi, A., Azimipour, M., & Bahreini, A. (2015). Prediction of roadheaders' performance using
artificial neural network approaches (MLP and KOSFM). Journal of Rock Mechanics and Geotechnical
Engineering, 7(5): 573-583.

58 DACOMA-2019
14. Ebrahimabadi, A., Goshtasbi, K., Shahriar, K., & Seifabad, M. C. (2011a). A model to predict the performance
of roadheaders based on the Rock Mass Brittleness Index. South African Institute of Mining and Metallurgy.
Journal, 111(5), 355-364.
15. Ebrahimabadi, A., Goshtasbi, K., Shahriar, K., & Seifabad, M. C. (2011b). Predictive models for roadheaders’
cutting performance in coal measure rocks. Yerbilimleri, 32(2), 89-104.
16. Fattahi, H. (2016a). Application of improved support vector regression model for prediction of deformation
modulus of a rock mass. Engineering with Computers, 32(4), 567-580. doi:10.1007/s00366-016-0433-6
17. Fattahi, H. (2016b). Indirect estimation of deformation modulus of an in situ rock mass: an ANFIS model
based on grid partitioning, fuzzy c-means clustering and subtractive clustering. Geosciences Journal, 20(5),
681-690.
18. Fowel, R. J., & Johnson, S. T. Cuttability assessment applied to drag tool tunnelling machines. Paper presented
at the Proceeding of the 7th International Congress on Rock Mechanics, ISRM, Aachen.
19. García-Pedrajas, N., Hervás-Martínez, C., & Muñoz-Pérez, J. (2003). COVNET: a cooperative coevolutionary
model for evolving artificial neural networks. IEEE Transactions on Neural Networks, 14(3), 575-596.
20. Gehring, K. H. (1989). A cutting comparison. Tunnels and Tunnelling, 21, 27-30.
21. Goktan, R., & Gunes, N. (2005). A comparative study of Schmidt hammer testing procedures with reference to
rock cutting machine performance prediction. International Journal of Rock Mechanics and Mining Sciences,
42(3), 466-472.
22. Goshtasbi, K., Monjezi, M., & Tourgoli, P. (2009). Evaluation of boring machine performance with special
reference to geomechanical characteristics. International Journal of Minerals, Metallurgy and Materials, 16(6),
615-619.
23. Hassan, R., Cohanim, B., De Weck, O., & Venter, G. (2005). A comparison of particle swarm optimization and
the genetic algorithm. Paper presented at the Proceedings of the 1st AIAA multidisciplinary design
optimization specialist conference.
24. Holland John, H. (1975). Adaptation in natural and artificial systems: an introductory analysis with
applications to biology, control, and artificial intelligence: MIT Press Cambridge, MA, USA.
25. Juang, C.-F. (2004). A hybrid of genetic algorithm and particle swarm optimization for recurrent network
design. IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, 34(2), 997-1006.
26. Karimpouli, S., & Fattahi, H. (2018). Estimation of P-and S-wave impedances using Bayesian inversion and
adaptive neuro-fuzzy inference system from a carbonate reservoir in Iran. Neural Computing and Applications,
29(11), 1059–1072.
27. Kaveh, A., & Malakouti Rad, S. (2010). Hybrid genetic algorithm and particle swarm optimization for the
force method-based simultaneous analysis and design. Iranian Journal of Science and Technology, Transaction
B: Engineering, 34(B1), 15-34.
28. Khalaj, G., Yoozbashizadeh, H., Khodabandeh, A., & Nazari, A. (2013). Artificial neural network to predict
the effect of heat treatments on Vickers microhardness of low-carbon Nb microalloyed steels. Neural
Computing and Applications, 22(5), 879-888.
29. Madan, M. (2008). Underground excavation with road headerscase studies. Paper presented at the World
Tunnel Congress.
30. Majdi, A., & Beiki, M. (2010). Evolving neural network using a genetic algorithm for predicting the
deformation modulus of rock masses. International Journal of Rock Mechanics and Mining Sciences, 47(2),
246-253.
Applying adaptive neuro fuzzy inference system to predict surface settlement

caused by mechanized tunneling-A case study
Hadi Fattahi1,*
1Department of Mining Engineering, Arak University of Technology, Arak, Iran.
*
Corresponding Author: Hadi Fattahi. Email: h.fattahi@arakut.ac.ir.
Abstract: Surface settlement (SS) is an important parameter for the design and operation of earth pressure balance
(EPB) shields that should be determined before tunneling. Machine learning methods are accepted as a technology
that offers an alternative way to tackle highly complex problems that can’t be modeled in mathematics. They can
learn from examples and they are able to handle incomplete data and noisy data. The adaptive network–based fuzzy
inference system (ANFIS) is kind of machine learning methods that was used in this study to build a prediction
model for the SS caused by tunneling. The estimation abilities offered using model was presented by using field data
from the Bangkok Subway Project in Thailand. In this model, depth, distance from shaft, ground water level from
tunnel invert, average face pressure, average penetrate rate, pitching angle, tail void grouting pressure and percent
tail void grout filling were utilized as the input parameters, while the SS was the output parameter. The results
achieved indicate that the ANFIS model has strong potential to prediction of SS with high degree of accuracy and
robustness.
Keywords: Surface settlement, Machine learning, Mechanized tunneling, Adaptive neuro fuzzy inference system.
References
1. Ahmadi, M. H., Aghaj, S. S. G., & Nazeri, A. (2013). Prediction of power in solar stirling heat engine by using
neural network based on hybrid genetic algorithm and particle swarm optimization. Neural Computing and
Applications, 22(6), 1141-1150.
2. Avci, E. (2008). Comparison of wavelet families for texture classification by using wavelet packet entropy
adaptive network based fuzzy inference system. Applied Soft Computing, 8(1), 225-231.
3. Avci, E., & Akpolat, Z. H. (2006). Speech recognition using a wavelet packet adaptive network based fuzzy
inference system. Expert Systems with Applications, 31(3), 495-503.
4. Avci, E., Hanbay, D., & Varol, A. (2007). An expert discrete wavelet adaptive network based fuzzy inference
system for digital modulation recognition. Expert Systems with Applications, 33(3), 582-589.
5. Avci, E., Turkoglu, I., & Poyraz, M. (2005). Intelligent target recognition based on wavelet adaptive network
based fuzzy inference system. Paper presented at the Iberian Conference on Pattern Recognition and Image
Analysis.
6. Bezdek, J. C. (1973). Fuzzy mathematics in pattern classification. Ithaca: Cornell university.
7. Boyacioglu, M. A., & Avci, D. (2010). An adaptive network-based fuzzy inference system (ANFIS) for the
prediction of stock market return: the case of the Istanbul stock exchange. Expert Systems with Applications,
37(12), 7908-7912.
8. Chiu, S. L. (1994). Fuzzy model identification based on cluster estimation. Journal of intelligent and Fuzzy
systems, 2(3), 267-278.
9. Clough, G. W., & Leca, E. (1993). EPB shield tunneling in mixed face conditions. Journal of Geotechnical
engineering, 119(10), 1640-1656.
10. Finno, R. J., & Clough, G. W. (1985). Evaluation of soil response to EPB shield tunneling. Journal of
Geotechnical engineering, 111(2), 155-173.
11. García-Pedrajas, N., Hervás-Martínez, C., & Muñoz-Pérez, J. (2003). COVNET: a cooperative coevolutionary
model for evolving artificial neural networks. IEEE Transactions on Neural Networks, 14(3), 575-596.
12. Hajihassani, M., Armaghani, D. J., Marto, A., & Mohamad, E. T. (2015). Ground vibration prediction in quarry
blasting through an artificial neural network optimized by imperialist competitive algorithm. Bulletin of
engineering geology and the environment, 74(3), 873-886.
13. Hornik, K., Stinchcombe, M., & White, H. (1989). Multilayer feedforward networks are universal
approximators. Neural Networks, 2(5), 359-366.

60 DACOMA-2019
14. Hossein, M., Ahmadi, M. A., & Ghare, S. S. (2011). Prediction of Power in Solar Stirling Heat Engine by
Evolving Particle Swarm Optimization and Neural Network. International Journal of Computer Applications,
34(1).
15. Jassar, S., Liao, Z., & Zhao, L. (2009). Adaptive neuro-fuzzy based inferential sensor model for estimating the
average air temperature in space heating systems. Building and environment, 44(8), 1609-1616.
16. Peila, D., Picchio, A., & Chieregato, A. (2013). Earth pressure balance tunnelling in rock masses: Laboratory
feasibility study of the conditioning process. Tunnelling and Underground Space Technology, 35, 55-66.
17. Shao, C., & Lan, D. (2014). Optimal control of an earth pressure balance shield with tunnel face stability.
Automation in Construction, 46, 22-29.
18. Simon, D. (2008). Biogeography-based optimization. IEEE Transactions on Evolutionary Computation, 12(6),
702-713.
19. Suwansawat, S., & Einstein, H. H. (2006). Artificial neural networks for predicting the maximum surface
settlement caused by EPB shield tunneling. Tunnelling and Underground Space Technology, 21(2), 133-150.
20. Togan, V. (2013). Design of pin jointed structures using teaching-learning based optimization. Structural
Engineering and Mechanics, 47(2), 209-225.
21. Toğan, V. (2012). Design of planar steel frames using teaching–learning based optimization. Engineering
Structures, 34, 225-232.
22. Tsou, D., & MacNish, C. (2003). Adaptive particle swarm optimisation for high-dimensional highly convex
search spaces. Paper presented at the Evolutionary Computation, 2003. CEC'03. The 2003 Congress on.
23. Uzlu, E., Kankal, M., Akpınar, A., & Dede, T. (2014). Estimates of energy consumption in Turkey using neural
networks with the teaching–learning-based optimization algorithm. Energy, 75, 295-303.
24. Wei, M., Bai, B., Sung, A. H., Liu, Q., Wang, J., & Cather, M. E. (2007). Predicting injection profiles using
ANFIS. Information Sciences, 177(20), 4445-4461.
25. Zhang, Y., Phillips, P., Wang, S., Ji, G., Yang, J., & Wu, J. (2016). Fruit classification by biogeography-based
optimization and feedforward neural network. Expert Systems, 33(3), 239-253.
Classification of surrounding rock of highway tunnel based on migration

learning
Xuefeng Chen*
1
Tongji University, 1239 Siping Road, 200092 Shanghai, China.
*Corresponding Author: Xuefeng Chen. Email: Chenxf2017@hotmail.com.
Abstract: At present, the main construction method of domestic mountain tunnels still adopts the new Austrian
method. The initial flexible support must be applied immediately after the tunnel is excavated, and the strength of
the support method depends on the real-time classification of the surrounding rock of the tunnel surface. Existing
methods of surrounding rock classification need to use complex equipment to extract the surrounding rock
classification parameters. In fact, relying on the subjective experience of engineers and technicians in the
construction phase is still the main way of surrounding rock classification. The reason is due to cost and efficiency.
However, without rich professional knowledge and experience in surrounding rock classification, there are often
cases of misjudgment. Therefore, it is an inevitable trend to find a new method for classify tunnel surrounding rock
that overcomes the traditional drawbacks.
Based on the image and mechanical parameters of the collected highway tunnel surface, using edge calculation,
deep learning and image recognition technology[see Fig.1], the Python language is used to integrate the above
functions to develop a set of surrounding rock image integrity recognition system for auxiliary surrounding rock
classification. The edge calculation and deep learning technology upgrades the traditional surrounding rock
classification to the level of artificial intelligence. The accumulated experience of machine learning can lead to the
classification result automatically, which reduces the dependence on the subjective experience of engineers. The
main advantage of this method is that with the advantage of artificial intelligence in data storage and analysis, the
machine language can quickly classify and judge the integrity of surrounding rock. It is expected to change the
present situation of surrounding rock classification that relies too much on professional knowledge and experience,
and the accuracy and timeliness of surrounding rock classification can be improved.
Figure 1: YOLO 3 Integral Structure of Depth Neural Network Used in Pre-research
By collecting more than 800 surface rock images, a deep convolutional neural network model was established to
identify the distributed features of surrounding rock, such as joints, fissures fracture degree, integrity, mud and
groundwater, and to identify the surrounding rock classification parameters. The discriminating factors of the
surrounding rock classification are converted into the modified BQ values for classification, and then combined with
the measured rock mechanical parameters for quantitative classification, and finally used to assist in judging the
surrounding rock level of the site.

62 DACOMA-2019
The following results were obtained through research:

(1) The grading result is automatically generated by relying on surface rock images, and the surrounding rock integrity
grading can be quickly and automatically performed after the image of the face is uploaded to the application.
The upload process also updates and integrates the data, and gradually judges and corrects the results. 

(2) Factors affecting the integrity of rock mass in traditional surrounding rock classification- Groundwater impact
correction coefficient (K1 value), main soft rock structural surface correction coefficient (K2 value), initial
stress state correction coefficient (K3 value) is reclassified by machine learning, and the hardness of the rock
(Kv value) is manually intervened in combination with the measured mechanical parameters. Finally, the
corrected BQ value is obtained. This method is suitable for tunnel excavation. Fast grading with high
quantitative grading characteristics.
(3) Successfully developed the highway tunnel surrounding rock image integrity identification system, and applied it
to the actual projects such as the Dongxiang Tunnel of Xingyi Huancheng highway and the No. 1 and No. 2
tunnels of Dadongpo, and compared and verified with the original design surrounding rock grade. It proves
that the system is suitable for the classification of surrounding rock of highway tunnels.
Keywords: Highway tunnel, Edge calculation, Deep learning, Image recognition.
References
1. Chen Chun-ping. Design and Implementation of Tunnel Face 3D_Reconstruction Based on OpenGL.
Southwest Jiao tong University, 2005.
2. Leng Biao. Research and Implementation of Digital Image Processing in Tunnel Face Image. Southwest Jiao
tong University, 2005.
3. ZHOU Chun-lin,ZHU He-hua ， LI Xiao-jun. (2008). Application of infrared photography and image
processing to tunnel construction with new Austrian tunneling method. Chinese Journal of Rock Mechanics
and Engineering, 27(S1): 3166-3172.
4. ZHUANG Xiao-ying,CAI Yong-chang,ZHU He-hua,ZHOU Ding-heng. (2008). Rock slope stability analysis
based on meshless method and shortest path algorithm. Chinese Quarterly of Mechanics, 29(4): 537-543.
5. ZHUANG Xiao-ying，HUANG Run-qiu，ZHU He-hua. (2012). Simulation for 2D compression-shear joint
dynamic propagation process using meshless methods based on level sets coordinates. Chinese Journal of Rock
Mechanics and Engineering, 31(11): 2187-2196.
6. LIU Hou-xiang，LI Wang-shi，ZHA Zhuan-yi，JIANG Wu-jun，XU Teng. (2018). Method for surrounding
rock mass classification of highway tunnels based on deep learning technology，Chinese Journal of Rock
Mechanics and Engineering, 40(10): 1809-1817.
7. Vitthal M.Khatik, Arup Kr.Nandi. (2018). A generic method for rock mass classification. Journal of Rock
Mechanics and Geotechnical Engineering, 10: 102-116.
8. Rabczuk T., Belytschko T. (2004). Cracking particles: a simplied meshfree method for arbitrary evolving
cracks, International Journal for Numerical Methods in Engineering, 61(13): 2316-2343.
9. Rabczuk T., Belytschko T. (2007). A three dimensional large deformation meshfree methodfor arbitrary
evolving cracks, Computer Methods in Applied Mechanics and Engineering, 196(29-30): 2777-2799.
10. Rabczuk T., Zi G., Bordas S., Nguyen-Xuan H. (2010). A simple and robust three-dimensional cracking-particle
method without enrichment, Computer Methods in Ap-plied Mechanics and Engineering, 199(37-40): 2437-2455.
11. Rabczuk T., Gracie R., Song J. H., Belytschko T. (2010). Immersed particle method for fluid-structure
interaction, International Journal for Numerical Methods in Engineering, 81(1): 48-71.
12. Amiri F., Anitescu C., Arroyo M, Bordas S., Rabczuk T. (2014). XLME interpolants,a seamless bridge
between XFEM and enriched meshless methods, Computational Me-chanics, 53(1): 45-57.
13. Ghorashi S., Valizadeh N., Mohammadi S., Rabczuk T. (2015). T-spline based XIGA for Fracture Analysis of
Orthotropic Media, Computers & Structures, 147: 138-146.
14. Anitescu C., Hossain N., Rabczuk T. (2018). Recovery-based error estimation and adaptivity using high-order
splines over hierarchical T-meshes, Computer Methods in Applied Mechanics and Engineering, 328(1): 638-662.
Automatic lithology identification based on machine learning: A case study of

TZ4 well
Siyao Guo1, Zhihui Ye1,*, Han Wang1, Dong Chen2,3 and Dandan Zhu4
1
College of Safety and Ocean Engineering, China University of Petroleum, 18 Fuxue Road, 102249 Beijing, China.
2
College of Petroleum Engineering, China University of Petroleum, 18 Fuxue Road, 102249 Beijing, China.
3
College of Artificial Intelligence, China University of Petroleum, 18 Fuxue Road, 102249 Beijing, China.
4
College of Information Science and Engineering, China University of Petroleum, 18 Fuxue Road, 102249 Beijing, China.
*
Corresponding Author: Zhihui Ye. Email: zhihui.ye@cup.edu.cn.
Abstract: Lithology identification of hydrocarbon reservoir based on well logging curve is significant for drilling
planning and reservoir development. The conventional stratigraphic identification methods usually require large
amount of labour work. The machine learning is able to deal with the data-intensive tasks and enhancing the
lithology/formation recognition, and make it easier to build a lithotype profile. Four typical algorithms, i.e.
Adaboost, decision tree, random forest and linear support vector machine are introduced in present study. By
comparing the prediction accuracy of each algorithm on TZ4 well logging data, an optimal algorithm is chosen for
further study. Then, the sensitivity analysis of data dimensionality reduction and training set adjustment is carried
out, so as to explore the main factors affecting the accuracy of formation identification. The results show that
decision tree algorithm has higher accuracy in the selected four algorithms and when the training ratio reaches 50%
and the number of logging parameters is at least four, the accuracy can reach 90%.
Keywords: Lithology identification, Machine learning, Sensitivity analysis, Prediction accuracy.

Lithology identification is an indispensable part of oil and gas exploration and development. The recognition results
allow better understanding on the stratigraphic profile and oil and gas distribution, providing reference for oil and
gas field development. The efficiency of formation identification technology is required to improve to deal with the
data intensive tasks. Machine learning methods are introduced to meet this requirement and to build the lithotypes
profile efficiently and accurately.
In 1991, Ruiz first proposed the use of artificial neural networks to divide the strata, which laid the foundation for
stratigraphic intelligent identification [1]. Many scholars began to use neural network methods. After nearly two
decades of development, the intelligent recognition technology of the formation has become more and more
developed. Roy used the linear PCA reduction method for classification of seismic facies [2]. T. Zhao used SVM to
classify 3D seismic data by lithology [3]. Tang applied artificial neural networks for the automated lithotyping [4].
John uses the Bayesian approach to estimate the distribution of facies [5]. Al-Anazi and Gates attempted to use the
SVM method for lithology prediction [6]. Zazoun R.S used neural networks to predict fractured reservoirs [7]. The
above research proves that the intelligent identification of formation has important application value and wide
application prospects. However，although the existing stratigraphic intelligent identification method improves the
work efficiency, there are still some problems such as difficulty in data collection and insufficient accuracy.
Therefore, it is necessary to propose a stratigraphic intelligent identification method that can widely identify the well
logging data and has high precision and applicability.
The intelligent recognition method of formation is based on the machine learning classification algorithm, which can
classify and identify the logging data, effectively learn and remember the characteristics of the rock formation in the
reservoir. This method can compare the four algorithms of Adaboost, decision tree, random forest and SVM, and
select the highest accuracy for lithology identification. The TZ4 well is located in the Tarim Basin. The location is
the Devonian Donghetang Formation, which contains three lithologies such as fine sandstone. In order to explore the
impact of different algorithms on accuracy, the TZ4 well was tested with 50% training ratio and eight logging data.
The test results are shown in Fig.1. The left column stands for lithology from real core sample, which is the ground
truth of the lithology/formation recognition. Prediction results of four different methods e.g. Adaboost, Decision tree,
Random forests and SVM are shown in the middle column.
The results from Fig. 1 show that the decision tree algorithm has the highest accuracy of 91.32%, The accuracy of

64 DACOMA-2019
Adaboost, random forest, and SVM algorithms are 82.96%, 70.42%, and 54.98%, respectively. Therefore, the
decision tree algorithm is used to analyze the sensitivity of training data volume and data dimension.
Figure 1: Model comparison and training proportions based on well logging data
The accuracy can be greatly influenced by training data volume in total data set. Five groups of scenarios are
conducted using the decision tree algorithm to investigate the impact of different training proportions on the
prediction result. Typical results for different training set volume are shown in the right column of Fig. 1. It is found
that by training only 30% of the total data set, the accuracy can reach above 80%. And the accuracy is positively
correlated with the training data ratio, the accuracy increases as the training ratio rises.
In order to explore the impact of different logging parameters on accuracy, several groups of scenarios are
conducted, using decision tree algorithm and 50% training proportion. By adding the number of logging parameters,
the test results are presented in Tab. 1.
Table 1: Data dimension based on lithology labels
Case# Parameters Accuracy, %
1 ILD 60.25
2 ILD, SP 79.81
3 ILD, SP, CN 87.78
4 ILD, SP, CN, ILM 88.42
5 ILD, SP, CN, ILM, GR 91.27
6 ILD, SP, CN, ILM, GR, CAL 91.32
7 ILD, SP, CN, ILM, GR, CAL, AC 91.36
8 ILD, SP, CN, ILM, GR, CAL, AC, DEN 91.61
Tab. 1 shows that the accuracy gradually increases from 60% by adding number of parameters. Using the same
training ratio and algorithm, the combinations of the most accurate parameters are shown. The ILD is the most
influential factor and contributes most on the accuracy. It is found that accuracy can reach 90% by using four or
more logging parameters. When the number parameters are greater than four, the accuracy tends to stay steady.
Acknowledgement
The present research is supported by the Strategic Priority Research Program of the Chinese Academy of Sciences，
Grant No. XDA14040402.
References
1. Ruiz R.A. Litho Stratigraphic correlation in the Lawrence Field, llinois, using the CORRELATOR. Computer
system.,1991:462-473
2. Roy, A., Jayaram, V., Marfurt, K., 2013. Active learning algorithms in seismic facies classification. Seg. Tech.
Program Expand. Abstr. 1467-1471.
3. Zhao, V., Jayaram, K., Marfurt, J., 2014. Lithofacies classification in barnett shale usingproximal support
vector machines. Seg. Tech. Program Expand. Abstr. 1491-1495.
4. Tang, H., 2008. Improved carbonate reservoir facies classification using artificial neural network method. In:
Canadian international Petroleum Conference (17-19 June),Calgary, Alberta, Canada.
5. John, A., Lake, L., Torres-Verdin, C., Srinivasan, S., 2008. Seismic-based facies identification and
classification using simple statistics. SPE Reserv. Eval. Eng. 11: 984-990.
6. Al-Anazi, A., Gates, I., 2010. On the capability of support vector machines to classify lithology from well logs.
Nat. Resour. Res. 19: 125-139.
7. Zazoun R.S. Fracture density estimation from core and conventional well logs data using artificial neural
networks: The Cambro-Ordovician reservoir of Mesdar oil field, Algeria. Journal of African Earth Sciences,
2013, 83: 55-73.
Automatic identification of gas hydrate formation using machine learning

algorithms
Yuqiang Ning1, Zhihui Ye1,*, Han Wang1, Dong Chen2,3, Shouding Li4 and Dandan Zhu5
1College of Safety and Ocean Engineering, China University of Petroleum, 18 Fuxue Road, 102249 Beijing, China.
2College of Petroleum Engineering, China University of Petroleum, 18 Fuxue Road, 102249 Beijing, China.
3College of Artificial Intelligence, China University of Petroleum, 18 Fuxue Road, 102249 Beijing, China.
4Institute of Geology and Geophysics, Chinese Academy of Sciences, 19 Beitucheng Road, 100029 Beijing, China.
5College of Information Science and Engineering, China University of Petroleum, 18 Fuxue Road, 102249 Beijing, China.
*Corresponding Author: Zhihui Ye. Email: zhihui.ye@cup.edu.cn.
Abstract: Ensemble learning integrates the results of multiple weak learners through certain rules, so it has better
learning effect than single weak learner. As a method of machine learning, it has been widely used in various fields.
Existing hydrate recognition methods are highly empirical and lack of efficiency. Gas hydrate has the characteristics
of high resistivity, abnormal potential and low gamma parameters, so it can be considered to be identified by using
machine learning method with well logging parameters. Ensemble learning has the characteristics of fast speed and
high accuracy, and has a wide application prospect. In this paper, a complete data analysis process and
corresponding data analysis algorithm are used to analyze and process logging data, and the accuracy of integrated
learning in hydrate recognition is verified by comparing with the real formation conditions. Comparing different
ensemble learning algorithms for the accuracy of final prediction results, the ensemble learning algorithm presents
the highest accuracy, which can be applied for further analysis, e.g. dimension reduction and data training, thus
reducing the workload of data processing.
Keywords: Ensemble learning, Well logging data, Hydrate recognition, Machine learning.
Natural gas hydrate is a non-stoichiometric cage-like crystalline compound formed by natural gas and water at high
pressure and low temperature. The main components are methane, including ethane, propane and other substances.
How to improve the accuracy and timeliness of hydrate recognition has become the primary problem.
Current hydrate identification methods are divided into direct identification method and indirect identification
method. Direct identification is judged by the visual observation of formation sampling. The cost of this method is
high, and hydrate is easy to decompose in this process, resulting in deviation of judgement. Indirect identification
method is realized by physical and chemical properties of hydrate and logging and seismic parameters [1]. Gas
hydrate has the characteristics of high resistivity, increasing P-wave time difference and negative potential anomaly
[2]. At present, Archie formula is commonly used to estimate hydrate saturation by resistivity [3]. Schlumberger
explores formation hydrate by Carbon-Oxygen Ratio Method [4]. The principle of seismic method is to use the
difference of seismic reflection velocity in different strata to detect target strata [5]. The methods of identifying
hydrate by seismic parameters include BSR, blank seismic reflection zone, AVO inversion technology, polarity
inversion and so on, etc. [6].
At present, there are still some problems in hydrate identification methods, such as poor timeliness and accuracy.
For example, BSR does not correspond to hydrate one by one. BSR does not exist in all hydrates [7]. Inks et al.
found that the interpretation results of acoustic logging on the northern slope of Alaska are inconsistent with those of
three-dimensional seismic prediction [8].
In this paper, the integrated learning algorithm is used to program, and four integrated learning algorithms, Adaboost,
SVM, random forest and decision tree, are used to predict whether the formation contains hydrate. Logging data are
from Northwest Eileen State-2 well. Strata are marked separately according to whether hydrates are present or not.
Then the program is written with Python, and the formation data and labels are divided into training set and testing
set respectively. The model is trained with training set and the stratum is predicted with test set. The results of
prediction and the actual stratum situation are compared. The test results show that the AdaBoost algorithm has the
highest prediction result for formation. As shown in Fig.1, the test results illustrate that the method based on

66 DACOMA-2019
ensemble learning has higher accuracy and application value for formation hydrate recognition.
Figure 1: Model comparison based on logging data
Precision rate and recall rate are two indicators to measure the prediction results. Precision rate is the proportion of
the correct prediction in the predicted sample, while recall rate is the proportion of the correct prediction in the
actual sample. And F1 score is an index used to synthesize the two.
The binary average of F1 fraction predicted by AdaBoost algorithm is 0.89, while that predicted by SVM algorithm
is only 0.628. The low value of F1 indicates that there is a big difference between precision and recall rate. The fact
proves that the strata predicted by SVM algorithm are not in good agreement with the actual strata. The following
table lists the F1 scores, precision and recall rates calculated by various methods.
Table 1: Performance metrics comparison between different algorithms

Algorithm F1 score precision recall rate
Adaboost 0.890 0.962 0.862
SVM 0.628 0.474 0.931
Decision tree 0.807 0.821 0.793
Random forest 0.766 0.950 0.655
Acknowledgement
The present research is supported by the Strategic Priority Research Program of the Chinese Academy of Sciences,
Grant No. XDA14040401 and Grant No. XDA14040402.
References
1. Bahk J J, Kim G Y, Chun J H, et al. (2013). Characterization of gas hydrate reservoirs by integration of core
and log data in the Ulleung Basin, East Sea. Marine and Petroleum Geology, 47(Complete), 30-42.
2. Kim H S, Riedel M, Ryu B J, et al. (2013). Improving gas hydrate saturation estimates using P-wave velocity
log data by incorporating XRD-data for detailed matrix-mineralogy definition. Marine and Petroleum Geology,
47(Complete): 155-167.
3. HuaiminDong, JianmengSun, JinjiangZhu, et al. (2019). Developing a new hydrate saturation calculation
model for hydrate-bearing sediments. Fuel, 27-37.
4. Collect T S, Wendlandt R F. (2000). Formation evaluation of gas hydrate-bearing marine sediments on the
Blake Ridge with downhole geochemical log measurements. Proceedings of the Ocean Drilling Program:
Scientific Results, 164(164).
5. Xiao K, Zou C, Lu Z, et al. (2017). Gas hydrate saturations estimated from pore-and fracture-filling gas
hydrate reservoirs in the Qilian Mountain permafrost, China. Scientific Reports, 7(1): 16258.
6. Liu J, Zhang J, Ma F, et al. (2017). Estimation of seismic velocities and gas hydrate concentrations: A case
study from the Shenhu area, northern South China Sea. Marine and Petroleum Geology, S0264817217303239
7. Collett T S, Lee M W, Zyrianova M V, et al. (2012). Gulf of Mexico Gas Hydrate Joint Industry Project Leg II
logging-while-drilling data acquisition and analysis. Marine and Petroleum Geology, 34(1).
8. Inks T L, Collett T, Taylor D J, et al. (2005). Prospecting for Gas Hydrate Accumulations using 2D and 3D
Seismic Data, Milne Point, North Slope Alaska. Search & Discovery. 2005(2005).
Automatic damage analysis of cracked concrete specimens using deep

learning-based crack detection techniques
Pan Sun1, 2,*
*Corresponding Author: Pan Sun. Email: sunpan_tj@tongji.edu.cn.
Abstract: During their service life, civil engineering materials such as concrete have a strong possibility of cracking
which causes a reduction in ultimate load bearing capacity and stiffness. To correctly predict the remaining load
bearing capacity of cracked engineering structures is of great importance for not only the safety concern but also
economic considerations. However, one of the big challenges is the accurate and efficient detection of existing
cracks in civil infrastructures. As a substitution of the traditional human-based inspection methods and image
processing techniques, machine learning, which has been extensively studied worldwide in the field of computer
science in past decades and shown strong ability in image recognition, has been applied for damage detection in
concrete structures, among which, the deep learning method using convolutional neural networks shows very robust
and accurate performance in image analysis. In this work, an indoor experiment of damaged reinforced concrete box
is analyzed. At first a deep learning model using convolutional neural networks is trained using other sources of
cracked and intact concrete images. Then the pictures taken in the indoor experiment once the crack forms during
loading are split into smaller equal-sized square pictures and these pictures are analyzed using the developed deep
leaning model to detect whether the squares are cracked or not. Afterwards a numerical model using mixed cover
meshless interpolation and the continuum damage mechanics are automatically established taking these generated
squares as meshes, in which those detected cracked elements are given a large damage factor while the others not.
The subsequent loading analysis is carried out using the concept of a continuum damage model regularized by
fracture mechanics. The obtained force displacement curve and final crack patterns are then compared to the
experimental results. This method can be served as an application to the safety assessment of real damaged
reinforced concrete structures.
Keywords: Image recognition, Deep learning, Cracks, Reinforced concrete.
Acknowledgement
The author gratefully acknowledges the financial support by China Scholarship Council, the guidance from Prof.
Timon Rabczuk and the experimental data from Dongyang Feng.
References
1. Anitescu, C., Atroshchenko, E., Alajlan, N., & Rabczuk, T. (2017). Deep Learning-Based Crack Damage
Detection Using Convolutional Neural Networks. Computer-Aided Civil and Infrastructure Engineering, 32:
361-378.
2. Protopapadakis, E., Voulodimos, A., Doulamis, A., Doulamis, N & Stathak, T. (2019). Automatic crack
detection for tunnel inspection using deep learning and heuristic image post-processing. Applied
Intelligence, 49: 2793-2806.
3. Bao, Y., Feng, D., Ma, N., Zhu, H & Rabczuk, T. (2018). Experimental and numerical study on structural
performance of reinforced concrete box sewer with localized extreme defect. Underground Space, 3: 166-179.
4. Kurumatani, M., Soma, Y., Terada, K. (2019). Simulations of cohesive fracture behavior of reinforced concrete
by a fracture-mechanics-based damage model. Engineering Fracture Mechanics, 206: 392-407.

A new method of tunnel leakage water identification based on terrestrial laser

scanning
Keshen Zhang1,*, Hehua Zhu1 and Wei Wu1
*Corresponding Author: Keshen Zhang. Email: zhangkeshen@tongji.edu.cn.
Abstract: Terrestrial laser scanning is widely used to obtain point cloud data and describe the deformation and
surface geometric conditions of tunnels. In particular, the reflection intensity values of point cloud can be used to
identify tunnel leakage water. In this paper, a new method based on laser reflection intensity and coordinates of
point clouds is proposed to identify the leakage water location in tunnels and remove the geometric interference of
tunnel facilities such as ventilation ducts, cables and bolt holes. The method is performed in the following steps: (1)
use machine learning to fit the relationship between distance, incidence angle, roughness and laser reflection
intensity values of point clouds, (2) compare the fitting laser reflection intensity value with the real laser reflection
intensity value of each point in the area to be measured and identify the possible leakage points, (3) use piecewise
parabola to fit the cross section curve of tunnel and exclude geometric interference points, and (4) the remaining
possible leakage points represent the real leakage water location. The proposed method have been applied to a
highway tunnel project in Sichuan province, China. The results show that the leakage water location can be well
identified and the geometric interference of tunnel facilities can be removed effectively.
Keywords: Tunnel, Leakage Water, Terrestrial Laser Scanning, Machine Learning.
Acknowledgement
This work was supported by the Special Fund for Basic Research on Scientific Instruments of the National Natural
Science Foundation of China [4182780021].
References
1. Fang, W., Huang, X., Zhang, F., & Li, D. (2014). Intensity Correction of Terrestrial Laser Scanning Data by
Estimating Laser Transmission Function. IEEE Transactions on Geoscience & Remote Sensing, 53(2): 942-951.
2. HöFle, B., & Pfeifer, N. (2007). Correction of laser scanning intensity data: Data and model-driven approaches.
ISPRS Journal of Photogrammetry and Remote Sensing, 62(6): 415-433.
3. Kaasalainen, S., Niittymaki, H., Krooks, A., Koch, K., Kaartinen, H., Vain, A., & Hyyppa, H. (2010). Effect of
Target Moisture on Laser Scanner Intensity. IEEE Transactions on Geoscience and Remote Sensing, 48(4):
2128-2136.
4. Kukko, A., Kaasalainen, S., & Litkey, P. (2008). Effect of incidence angle on laser scanner intensity and
surface data. Applied Optics, 47(7): 986-992.
5. Li, X., Chen, J., & Zhu, H. (2016). A new method for automated discontinuity trace mapping on rock mass 3D
surface model. Computers & Geosciences, 89: 118-131.
6. Nield, J., Wiggs, G., & Squirrell, R. (2015). Aeolian sand strip mobility and protodune development on a
drying beach: examining surface moisture and surface roughness patterns measured by terrestrial laser
scanning. Earth Surface Processes & Landforms, 36(4): 513-522.
7. Sanna, K., Anttoni, J., Mikko, K., & Anssi, K. (2011). Analysis of Incidence Angle and Distance Effects on
Terrestrial Laser Scanner Intensity: Search for Correction Methods. Remote Sensing, 2011, 3(10): 2207-2221.
8. Tan, K., Cheng, X., Ju, Q., & Wu, S. (2017). Correcstion of Mobile TLS Intensity Data for Water Leakage
Spots Detection in Metro Tunnels. IEEE Geoscience & Remote Sensing Letters, 13(11): 1711-1715.
9. Wang, W., Zhao, W., Huang, L., & Wang, Z. (2014). Applications of terrestrial laser scanning for tunnels: a
review. Journal of Traffic and Transportation Engineering (English Edition), 1(5): 325-337.

69 DACOMA-2019
10. Xu, T., Xu, L., Li, X., & Yao, J. (2018). Detection of water leakage in underground tunnels using corrected
intensity data and 3D point cloud of terrestrial laser scanning. IEEE Access: 1-1.
Dynamic Classification of Surrounding Rock Mass in Tunneling Boring

Process Based on Big Operational Data
Mengqi Zhu1,* and Hehua Zhu1
1
College of Civil Engineering, Tongji University, 1239 Siping Road, 200092 Shanghai, China.
*
Corresponding Author: Mengqi Zhu. Email: mqzhu@tongji.edu.cn.
Abstract: Tunnel Boring Machine (TBM) has become an essential part in long infrastructural tunneling excavation
since 19th century, because of its high safety, less effect on surroundings and rapid excavation speed [1, 2].
However, TBMs are more sensitive to geological conditions (i.e. rock bursts, water bursts and other geological
hazards) than other tunneling method [3]. Considering that the actual geological conditions are nearly unknowable
before excavation because observation and measurement cannot be conducted during tunneling nowadays [4-6]. The
high-speed excavation capability of TBMs cannot be fully utilized, especially in rock-soil mixed-face ground or
other unfavorable geological conditions [7]. To depict the real-time profile of the geological conditions of tunnel
face, a series of studies have been conducted based on the operational data of TBMs via state-of-the-art machine
learning methods [3, 4]. But when they conducted the machine learning methods to predict the accordingly
surroundings, they ignored one problem that in many cases, the rock mass types are imbalanced. Contrarily,
traditional machine learning methods are based on the balanced data hypothesis, which will perform bad in
imbalanced data [8]. To solve this problem and obtain a better view of the tunnel face’s conditions, cost-sensitive
AdaCost algorithm was proposed. The database of this study composed by more than 5 ´ 106 data collects from the
Songhua River Water Conveyance project in Jilin province, China. The typical rock mass information including
rock mass types and their corresponding length recorded by in-situ survey before the tunneling was presented in Tab.
1. The proportion of granite in the rock mass types is 45.21%, while that of diorite in the rock mass types is 10.81%,
showing the characteristic of slight imbalanced.
Table 1: Rock mass information

Rock mass types Length (m) Percentage of length (%)
Diorite 2096 10.81
Granite 8766 45.21
Sandstone 3448 17.78
Tuff 5081 26.20
Summary 19391 100.00
There are 199 kinds of available TBM operational data generated each second during excavating process.
Thus, efficient preprocessing is significant for reducing redundant and improving computational
efficiency. In this study, data preprocessing was conducted as shown in Fig. 1. Ten operational variables
related to cutterhead rotational speed and boot pressure were selected by feature engineering as input
variables, and their relative importance are shown in Fig. 2.
②-1 ②-2 ③
4000
3500
3000
2500
Value
2000
2nd C-P
1500 Troque (kN×m)
Thrust/5(kN)
1000 1st C-P Net advance rate×30(mm/min)
②-1 Empty Pushing Segment
500 ②-2 Trail Excavation Segment
③ Stable Excavation Segment
0
0 200 400 600 800 1000
Epoch (s)
Figure 1: The evolution of advancing speed, thrust force and cutterhead torque in a certain tunneling cycle.

71 DACOMA-2019
1.2
1
1.0
0.84
Feature importance
0.8 0.77 0.75
0.67
0.6 0.58
0.48
0.45
0.4
0.33
0.29
0.2
0.0
e e e eed sure itch sure pitch ngle vel
valu l tim ssur s of p pres la tra
eter ona t pre al sp pre mp
of
gle t roo nder
ntiom operati boo rotatioont pumpot anglel pu ot an boooot cyli
pote bo left bo controright bo b
Figure 2: The relative importance of tunneling parameters to the class of surrounding rock
By dividing the database into 90% training sets and 10% test sets, the accuracy of traditional machine
learning methods and cost-sensitive AdaCost methods was calculated and compared in Table 2. Although
the total accuracy of traditional methods is 91%, its performance in the minority class, i.e., Rank IV and
Rank V, is relatively poor. However, the accuracy of classification rises 16% and 50% in Rank IV and Rank
V respectively by conducting cost-sensitive algorithm, which indicated that the proposed AdaCost
algorithm has strong applicability to imbalanced data than traditional machine learning methods.
Table 2: The class of surrounding rock obtained by RF model and AdaCost model respectively
Rock mass rank
Accuracy
Rank II Rank III Rank IV Rank V Total
Traditional methods 0.88 0.97 0.50 0.00 0.91
AdaCost 0.88 0.98 0.66 0.50 0.94
Keywords: Rock mass classification, Tunneling Boring Machine, Imbalanced big data, AdaCost.
Acknowledgement
This study was funded by the National Natural Science Foundation (4182780021) (“Basic research on scientific
instruments of the national natural science foundation of China”).
References
1. Sun W., Shi M., Zhang C., Zhao J. & Song X. (2018). Dynamic load prediction of tunnel boring machine
(TBM) based on heterogeneous in-situ data. Automation in Construction, 92: 23-34.
2. Okubo S., Fukui K. & Chen W. (2003). Expert System for Applicability of Tunnel Boring Machines in Japan.
Rock Mechanics and Rock Engineering, 36(4): 305-322.
3. Zhang Q., Liu Z. & Tan J. (2019). Prediction of geological conditions for a tunnel boring machine using big
operational data. Automation in Construction, 100: 23-34.
4. Yamamoto T., Shirasagi S., Yamamoto S., Mito Y. & Aoki K. (2003). Evaluation of the geological condition
ahead of the tunnel face by geostatistical techniques using TBM driving data. Tunnelling and Underground
Space Technology, 18(2): 213-221.
5. Li S., Liu B., Xu X., Nie L., Liu Z., Song J., Sun H., Chen L. & Fan K. (2016). An overview of ahead
geological prospecting in tunnelling. Tunnelling and Underground Space Technology, 63: 69-94.
6. Alimoradi A., Moradzadeh A., Naderi R., Salehi M. & Etemadi A. (2008). (Prediction of geological hazardous
zones in front of a tunnel face using TSP-203 and artificial neural networks. Tunnelling and Underground
Space Technology, 23: 711-717.
7. Tóth A., Gong Q. & Zhao J. (2013). Case studies of TBM tunneling performance in rock–soil interface mixed
ground. Tunnelling and Underground Space Technology, 38(9): 140-150.
8. Nojima Y., Mihara S. & Ishibuchi H. (2012). Application of parallel distributed genetics-based machine
learning to imbalanced data sets. IEEE International Conference on Fuzzy Systems, 1-6.
Prediction of Strength of Foamed Concrete under Cyclic Loading Using

Artificial Neural Network
Guanbao Ye1, Jiangting Liu1,2, Zhen Zhang1,2,*, Fengrui Rao1,2 and Jiaqi Chen3
1Department of Geotechnical Engineering, Tongji University, Shanghai, China.

2Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education, Tongji University, Shanghai, China.
3School of Mathematical Sciences, Tongji University, Shanghai, China.
*Corresponding Author: Zhen Zhang. Email: zhenzhang@tongji.edu.cn.
Abstract: Foamed concrete fill has been increasingly used as backfill material in geotechnical engineering.
Sometimes, foamed concrete fill is subjected to cyclic loading, such as traffic loading and machine vibration. Thus,
using unconfined compressive stress of foamed concrete could be unsafe as dynamic loading has potential for
degradation of material properties. This study conducted a series of unconfined compressive test of foamed concrete
under static and cyclic loading to explore the influence of loading type on the strength of foamed concrete. The
results show that the cyclic loading yielded lower strength of foamed concrete than static loading, as its internal
structure is easier to fail under cyclic loading. In an affect to predict the unconfined strength of foamed concrete
under cyclic loading based on its strength under static loading, artificial neural network (ANN) was used. Good
agreements between predictions and the real values were obtained after training the model using the test results.
Keywords: Foamed concrete, Unconfined compressive strength, Cyclic loading, Artificial neural network, Data
training.
Acknowledgement
The authors appreciate the financial support provided by the Fundamental Research Funds for the Central
Universities (Grant No. 22120180106) and the Natural Science Foundation of China (NSFC) (Grant No. 41772281
& No. 51508408).
References
1. Park, H. I., & Kim, Y. T. Prediction of strength of reinforced lightweight soil using an artificial neural network.
Engineering Computations, 2011, 28(5): 600-615.
2. Huang, J. J., Su, Q., Zhao, W. H., Li, T., & Zhang, X. X. Experimental study on use of lightweight foam
concrete as subgrade bed filler of ballastless track. Construction and Building Materials, 2017, 149: 911-920.
3. Wu, Y. D., Zeng, C. C., Liu, J., & Diao, H. Measured settlement of highways improved by lightweight
backfilling without road closure. Arabian Journal for Science and Engineering, 2016, 41(10): 3889-3896.
4. Nguyen, T. T., Bui, H. H., Ngo, T. D., & Nguyen, G. D. Experimental and numerical investigation of influence
of air-voids on the compressive behaviour of foamed concrete. Materials & Design, 2017, 130: 103-119.
5. Chen, Z. P., Wang, J. B., Liu, J. F., & Zhao, W. H. Experimental study on dynamic engineering characteristics
of foamed light soil. Highway, 2019, 2: 77-80.
6. Li, B., Gao, Y. F., & Feng, T. G. Analysis of the influence of vibration frequency on the dynamic
characteristics of LCES and its mechanism. Rock and Soil Mechanics, 2008, 29(10): 2731-2740.
7. Liu, Z. X., Gu, H. D., & Chen, D. Q. Experimental study on dynamic characteristics of foamed mixture
lightweight soil using river sludge. Environmental Science & Technology, 2016, 39(7): 80-85.
8. Theodoridis, S. Machine learning: a Bayesian and optimization perspective. Academic Press, 2015.
9. Chen, M. MATLAB neural network principle and examples of fine solution. Tsinghua University Press, 2013.
10. Schmidhuber, J. Deep learning in neural networks: An overview. Neural networks, 2015, 61: 85-117.

73 DACOMA-2019
Key pictures
(a) Low density (b) High density

Figure 1: Failure mode of foamed concrete with different density
(a) With frequency (b) With density

Figure 2: Variation of ratio of dynamic strength to static strength
(a) Training (b) Testing

Figure 3: A comparison between measured value and value predicted by ANN
A coarse-grained model for carbon nanotube/polymer nanocomposites
Behrouz Arash1,* and Timon Rabczuk2
Leibniz Universität Hannover, Appelstraße 9A, Hannover, 30167, Germany.

1
2
*Corresponding Author: Behrouz Arash. Email: b.arash@isd.uni-hannover.de.
Abstract: Although molecular simulations have been widely used in the modeling of nanocomposites, the huge
computational effort required by the simulations severely limits their applicability to small molecular systems over a
limited time scale. One approach to overcome this drawback is the development of coarse-grained (CG) models [1-
4]. The principle of CG models is to map a set of atoms to a CG bead, which enables to extend the accessible time
and length-scales while maintaining the molecular details of an atomistic system (see Fig. 3). The mapping scheme
relates the atomistic coordinates of a structure to the bead positions in the CG model. The CG models provide a
reduced picture and interpretation of complex molecular phenomena by averaging over unessential atomic details. It
facilitates the study of molecular systems since (1) the total number of particles is reduced, (2) the interaction
potentials are simplified and (3) the potential energy surface on which the molecules move is smoothed, leading to
an acceleration of molecular simulations. The main challenge is therefore to develop a CG model that reproduces the
same physical behavior as the atomistic reference system.
In this study, a CG model for carbon nanotube (CNT) reinforced polymer matrix composites is developed. The CG
force fields for nanotubes and polymer chains are calibrated using the strain energy conservation between CG and
all-atom systems. The ability of the CG model in predicting the material properties of CNT/polymer composites are
evaluated through verification processes with atomistic simulations. The simulation results reveal that the CG model
is able to estimate the elastic properties of the nanocomposites with high accuracy and low computational cost. The
effect of the volume fraction of CNT reinforcements on the Young’s modulus of the nanocomposites is investigated.
The application of the method in the modeling of large unit cells with randomly distributed CNT reinforcements is
examined. The proposed CG model will enable the simulation of reinforced polymer matrix composites across a
wide range of length scales from nano to mesoscale. In the future, the CG model can also be utilized and developed
to predict the mechanical properties of polymer composites reinforced by functionalized SWCNs and multi-walled
CNTs subjected to static and dynamic loading.
Keywords: Polymer nanocomposite, Carbon nanotube, Material properties, Coarse-grained model, Molecular
simulations.
References
1. Reith, D., Meyer, H., Müller-Plathe, F. (2001). Mapping atomistic to coarse-grained polymer models using
automatic simplex optimization to fit structural properties. Macromolecules, 34, 2335-2345.
2. Reith, D., Meyer, H., Müller-Plathe, F. (2003). Deriving effective mesoscale potentials from atomistic
simulations. Physical Chemistry Chemical Physics, 24, 1624-1636.
3. Rzepiela, J., Louhivuori, M., Peter, C., Marrink, S. J. (2011). Hybrid simulations: combining atomistic and
coarse-grained force fields using virtual sites. Physical Chemistry Chemical Physics, 13, 10437-10448.
4. Arash, B., Park, H. S., Rabczuk, T. (2015). Mechanical properties of carbon nanotube reinforced polymer
nanocomposites: A coarse-grained model. Composites Part B: Engineering, 80, 92-100.


Patricio Rodriguez1 and Xiaoying Zhuang1,2,*
1TDepartment of Geotechnical Engineering, Tongji University, Shanghai, China.
2
Institute for continuum mechanics, Leibniz University, AppelStraße 11, 30167 Hannover, Germany.
*
Abstract: Phase field modelling of fracture has revitalized the study of failure phenomena in the scientific
community during the last two decades. Multiple applications in multi-field frameworks [1-3] have been attempted
to exploit the advantages of the approach. Complex crack behaviors like nucleation, merging and branching, which
used to rely on intricate numerical techniques, are naturally handled by an additional partial differential equation.
From a theoretical point of view, a phase-field crack evolution is dictated by three energetical principles [4] namely,
energy conservation, irreversibility and global stability. While irreversibility and energy conservation are grounded
on thermodynamic principles and are widely accepted, the third one is still under debate. As pointed out by [5], a
global stability criterion results in unphysical results in nonconvex variational problems since a jumps over energy
barriers would be necessary in order to reach a global minimum. In terms of fracture modelling, numerical results
show cracks propagating instantaneously through the domain when a continuously evolving crack is expected. A
local stability criterion would be more suitable, however as shown by [5], even in this scenario the formulation
allows jumps during crack growth, breaking the causality principle [5]. Modifications to the theoretical framework
have been proposed by [5] and [6] pointing out the necessity of capturing rate dependent effects during fast crack
propagation to obtain physically meaningful solutions. However the implementation of a numerical scheme capable
of representing viscous and inertial effects in a rate independent framework has not been tackled yet. Here, an
algorithm based on a Machine Learning algorithm is proposed in order to study the space of admissible solutions
during the optimization process.
Machine learning has been successfully applied in other contexts to define the optimization domain in problems
with non-convex objective functions in material science [7]. In [8] a classification of learning algorithm was used to
define an optimization domain based on data from experiments and numerical simulations. In [9] an adaptive
algorithm is defined in order to modify the definition of the space of interest during the optimization process.
Following the same line of idea, a multilayered feed-forward neural network is used to explore the effect of inertial
effects on crack propagation during fast transitions. The finite element method is used to discretize the partial
differential equations resulting from the minimization of the total energy functional. Crack propagation in the quasi-
static and dynamic cases is considered. The Alternating Minimization Algorithm (AMA) is adopted here to solve
the quasi-static problem and implicit generalized alpha integration method for the dynamic case. By feeding pairs of
simulations of the quasi-static and dynamic crack propagation examples a multilayer neural network is trained to
calculate a relaxation ratio of a sequence of solutions of the evolution problem. Afterwards, the relaxation ratio is
used along with the quasi-static scheme to obtain improved solutions that better reflect the rate dependent
phenomena that are expected during jumps in the system evolution.
The results show that the total energy of the system is not conserved during fast crack growth, however the energy
imbalance obtained with the calculated relaxation ratio agrees well, within a certain range of crack propagation
velocities, with the energy associated to inertial effects in the dynamic case. For crack tip velocities beyond this
threshold the quasi-static solution becomes inadequate to model crack propagation since phenomena like branching
and crack propagation promoted by wave reflection take on a dominating roll during the evolution.
Keywords: Phase-fields, variational formulation, fracture, rate independent.
Acknowledgement
The author would like to acknowledge the financial support from the Sofja Kovalevskaja Prize of the Alexander von
Humboldt Foundation (Germany).

76 DACOMA-2019
References
1. Wilson, Z. A., Borden, M. J., & Landis, C. M. (2013). A phase-field model for fracture in piezoelectric
ceramics. International Journal of Fracture, 183(2), 135-153.
2. Wilson, Z. A., & Landis, C. M. (2016). Phase-field modeling of hydraulic fracture. Journal of the Mechanics
and Physics of Solids, 96, 264-290.
3. Borden, M. J., Verhoosel, C. V., Scott, M. A., Hughes, T. J., & Landis, C. M. (2012). A phase-field description
of dynamic brittle fracture. Computer Methods in Applied Mechanics and Engineering, 217, 77-95.
4. Mielke, A., & Roubícek, T. (2015). Rate-independent systems. Theory and Application (in preparation).
5. Alessi, R. (2016). Energetic formulation for rate-independent processes: remarks on discontinuous evolutions
with a simple example. Acta Mechanica, 227(10), 2805-2829.
6. Larsen, C. J. (2010). Epsilon‐stable quasi‐static brittle fracture evolution. Communications on Pure and
Applied Mathematics: A Journal Issued by the Courant Institute of Mathematical Sciences, 63(5), 630-654.
7. Liu, Y., Zhao, T., Ju, W., & Shi, S. (2017). Materials discovery and design using machine learning. Journal of
Materiomics, 3(3), 159-177.
8. Balachandran, P. V. (2019). Machine learning guided design of functional materials with targeted
properties. Computational Materials Science, 164, 82-90.
9. Gopakumar, A. M., Balachandran, P. V., Xue, D., Gubernatis, J. E., & Lookman, T. (2018). Multi-objective
optimization for materials discovery via adaptive design. Scientific reports, 8(1), 3738.
A finite element framework for vibration problem of hybrid nanocomposite

beams reinforced with graphene platelet and carbon fiber
Farzad Ebrahimi1,*, Ali Dabbagh2 and Timon Rabczuk3
1Departmentof Mechanical Engineering, Faculty of Engineering, Imam Khomeini International University, Noroozian Blvd.,
34148-96818, Qazvin, Iran.
2School of Mechanical Engineering, College of Engineering, University of Tehran, North Kargar St., 11155-4563, Tehran, Iran.
3Chair of Computational Mechanics, Bauhaus University Weimar, Marienstrasse 15, 99423, Weimar, Germany.
*Corresponding Author: F. Ebrahimi, febrahimy@eng.ikiu.ac.ir.
Abstract: Dispersion of nanosize reinforcements in an initial matrix has attracted the researchers’ attention in the
last two decades. The main reason for the widespread implementation of nanofibers and nanoparticles in a matrix to
generate nanocomposites is enhanced material properties of nanomaterials in comparison with their bulk specimen.
Through this fact, nanomaterials like carbon nanotube (CNT), graphene, and graphene platelet (GPL) were highly
employed by designers to improve the material properties of an initial resin. For instance, [Ke, Yang and
Kitipornchai (2010)] analyzed the nonlinear vibrational characteristics of CNT-reinforced (CNTR) nanocomposite
beams. The thermo-mechanical postbuckling problem of nanocomposite cylindrical panels reinforced with CNTs
was solved by [Shen and Xiang (2014)] once the structure is subjected to an axial compression. In another
investigation, the large deformation behaviors of CNTR nanocomposite skew plates were monitored by [Zhang,
Liew (2015)] based on the finite element method (FEM). The issue of investigating the impacts of nanofibers’
agglomeration on the natural frequency of CNTR nanocomposite doubly-curved shells was probed by [Tornabene,
Fantuzzi, Bacciocchi et al. (2016)]. On the other hand, [Barati and Zenkour (2017)] carried out a postbuckling study
dealing with stability responses of GPL-reinforced (GPLR) nanocomposite beams regarding for the geometrical
imperfection of the structure. The thermal stability problem of graphene-reinforced (GR) nanocomposite plates was
solved by [Shen, Xiang, Lin et al. (2017)]. Both free and forced frequency characteristics of GPLR nanocomposite
plates were surveyed by [Song, Kitipornchai and Yang (2017)] in the framework of the first-order shear deformation
theory (FSDT) of plates. The dependency of material properties of GR nanocomposites are included in a nonlinear
thermal buckling analysis conducted by [Kiani and Mirzaei (2018)] to investigate the stability behaviors of
nanocomposite beams.
Recently, a novel type of nanocomposites has been developed, fabricated from three phases, in particular an initial
matrix strengthened via both macro- and nano-scale reinforcements. Due to the nature of such newly developed
nanocomposites, they are called multi-scale hybrid nanocomposites. [Ebrahimi and Habibi (2018)] enriched a
numerical answer for the low-velocity impact problem of multi-scale hybrid nanocomposites reinforced with carbon
fibers (CFs) and CNTs. The vibrational characteristics of multi-scale hybrid nanocomposite plates reinforced with CFs
and CNTs were investigated by [Ebrahimi and Dabbagh (2019)] in the framework of the classical theory of plates in
association with the micromechanical homogenization methods. The major objective of this manuscript is to present a
powerful FE-based framework founded on the basis of the Rayleigh-Ritz method for the purpose of solving the natural
frequency problem of multi-scale hybrid nanocomposite structures strengthened by dispersion of nanosize GPLs and
macroscale CFs in a polymeric matrix. The under observation geometry is a beam with rectangular cross-section which
is assumed to be modeled implementing a refined shear deformable beam hypothesis. The strain-displacement relations
will be achieved using the definition of the infinitesimal strain tensor in the continuum mechanics. Afterward,
dependent expressions will be developed for both kinetic and strain energies of the nanocomposite beam. The next step
belongs to achievement of the maximum amounts of the abovementioned energies by the means of the Rayleigh-Ritz
method. Finally, the partial differentiations of the Rayleigh parameter with respect to the unknown coefficients will be
calculated to obtain the final eigenvalue of the problem. It is worth mentioning that the presented methodology will be
able to consider for various edge supports at the ends of the structure. The results of this work indicate on the
unbelievable role of the nanoscale reinforcements on the natural frequency of the nanocomposite structure as well as
the volume fraction of the utilized macroscale CFs.

78 DACOMA-2019
Keywords: Free vibration, finite element (FE), multi-scale hybrid nanocomposites, Graphene platelet (GPL),
Refined shear deformable beam model.
References
1. Barati, M. R., Zenkour, A. M. (2017). Post-buckling analysis of refined shear deformable graphene platelet
reinforced beams with porosities and geometrical imperfection. Composite Structures, 181, 194-202.
2. Ebrahimi, F., Dabbagh, A. (2019). Vibration analysis of multi-scale hybrid nanocomposite plates based on a
Halpin-Tsai homogenization model. Composites Part B: Engineering, 173, 106955.
3. Ebrahimi, F., Habibi, S. (2018). Nonlinear eccentric low-velocity impact response of a polymer-carbon
nanotube-fiber multiscale nanocomposite plate resting on elastic foundations in hygrothermal environments.
Mechanics of Advanced Materials and Structures, 25, 425-438.
4. Ke, L. L., Yang, J., Kitipornchai, S. (2010). Nonlinear free vibration of functionally graded carbon nanotube-
reinforced composite beams. Composite Structures, 92, 676-683.
5. Kiani, Y., Mirzaei, M. (2018). Enhancement of non-linear thermal stability of temperature dependent laminated
beams with graphene reinforcements. Composite Structures, 186, 114-122.
6. Shen, H. S., Xiang, Y. (2014). Postbuckling of axially compressed nanotube-reinforced composite cylindrical
panels resting on elastic foundations in thermal environments. Composites Part B: Engineering, 67, 50-61.
7. Shen, H. S., Xiang, Y., Lin, F., Hui, D. (2017). Buckling and postbuckling of functionally graded graphene-
reinforced composite laminated plates in thermal environments. Composites Part B: Engineering, 119, 67-78.
8. Song, M., Kitipornchai, S., Yang, J. (2017). Free and forced vibrations of functionally graded polymer
composite plates reinforced with graphene nanoplatelets. Composite Structures, 159, 579-588.
9. Tornabene, F., Fantuzzi, N., Bacciocchi, M., Viola, E. (2016). Effect of agglomeration on the natural
frequencies of functionally graded carbon nanotube-reinforced laminated composite doubly-curved shells.
Composites Part B: Engineering, 89, 187-218.
10. Zhang, L., Liew, K. (2015). Large deflection analysis of FG-CNT reinforced composite skew plates resting on
Pasternak foundations using an element-free approach. Composite Structures, 132, 974-983.
Modeling effect of thin film’s surface roughness on nanoindentation test using

random forest
Mohammad Javad Rabiei Faradonbeh1,*, Saleh Akbarzadeh1 and Mohammad Silani1
1Isfahan
University of Technology, Isfahan, Iran.
*Corresponding Author: Mohammad Javad Rabiei Faradonbeh. Email: mj.rabiei@me.iut.ac.ir.
Abstract: Nanoindentation test is one of the tests to measure the mechanical properties of materials, especially
macro-composites, which have been used since the mid-1970s. This test was carried out by a device of the same
name that during this test, the tip-shaped of the device penetrates on the surface of the specimen to a size of a few
nanometers, and the force-displacement curve is obtained as a direct output from this test. Experimental results
indicate that this test is dependent on the roughness of the tested part [1]. In the study of mechanical properties using
nanoindentation test, the perfect surface finish on the test specimen is often required [2]. Reaching to a perfect
surface finish is also depends on the limitation of the cutting tool devices [3]. Greenwood and Williamson defined
one of the firsts elastic contact models [4]. However this model can predict the mechanical properties in too many
cases without coatings, this model is not an efficient one in some cases with coatings [5]. In addition to these cases,
access to this device, as well as performing this test in special circumstances, such as high-temperature
nanoindentation test, is one of the instrumental limitations of this test. Also, the production of samples that have
coatings due to the availability of the required devices, is also a timely process to achieve the desired mechanical
properties for the users. In this study, this test defined with a 3D finite element model in Abaqus CAE and the
validity was proved by experimental cases like coating materials. Although this simulation has taken an effective
step in speeding up this process, as well as the ability to carry out this test on a wide range of materials defined by
the user, this study attempts to even allow the user to not to perform simulations and also let the user to achieved the
parameters like stiffness and surface hardness in a fraction of a second. This part of the research is carried out by the
machine learning process. Due to the wide variety of base materials and various coatings that exist, because of the
conventional use of steel in the industry, the specimens like steel base materials with commercial coatings become
the purpose of this research is to continue the process by using machine learning process. After all this, for achieving
to the parameters like stiffness and surface hardness, The user even does not need to simulate this test on steel base
materials with conventional industrial coatings, and only uses the properties that are defined to machine learning
process input, and get stiffness and hardness values in fractions of a second to desired surface.
Keywords: Nanoindentation, Surface roughness, Stiffness, Hardness, Machine learning.
References
1. L. Chen, A. Ahadi, J. Zhou, and J. Ståhl. (2013). Modeling effect of surface roughness on nanoindentation
tests. Procedia CIRP, 8, 334-339.
2. A. W. Warren and Y. B. Guo. (2006). Machined surface properties determined by nanoindentation:
Experimental and FEA studies on the effects of surface integrity and tip geometry. Surf. Coatings Technol..
3. A. C. Fischer-Cripps. (2011). Nanoindentation. The third edition ed. Mechanical Engineering, ed. New
South Wales, Australia: Springer.
4. J. A. Greenwood and J. B. P. Williamson. (1966). Contact of Nominally Flat Surfaces. Proc. R. Soc. A Math.
Phys. Eng. Sci..
5. J. Lee, A. Beheshti, and A. A. Polycarpou. (2017). Rough Surface Normal Nanocontact Stiffness:
Experimental Measurements and Rough Surface Contact Model Predictions. J. Appl. Mech., 84(3), 31006.

Assessing the spalling risk of fire-loaded tunnel linings by multifield model

and machine learning strategy
Zhiran. Gao1 and Yiming. Zhang1,*
1School of Civil and Transportation Engineering, Hebei University of Technology, Tianjin, China
*Corresponding Author: Yiming Zhang. Email:Yming.Zhang@hebut.edu.cn
Abstract: Concrete structures may experience spalling when subjected to fire loading. Spalling is the violent
splitting of concrete pieces from the surface of the structure, greatly jeopardizing the durability of the structure.
Spalling is closely related to various chemical and physical processes such as dehydration of heated concrete, phase
change of pore water, formation of moisture clog, and consequent fast increasing of pore pressure. Different
concrete mixtures, ambient humidity, and fire scenarios provide different spalling risks. The strongly coupled
multi-field model can take into account a single working condition and need considerable computing efforts, which
is unable to consider the uncertainty of concrete materials and environments.
Aiming at solving these problems, this work proposes a fast spalling risk evaluation method, combining the
multi-field model and the machine learning method. Firstly, a large number of material parameters, humidity and
fire scenarios are considered as input parameters for simulations by means of the coupled multi-field model [1-3],
providing the data samples for machine learning. Then the machine learning method is used to study the correlations
between the input parameters and the obtained results such as maximum pore pressure of concrete, see Figure 1.
Finally, by comparing six different algorithms: 1) Logistic Regression, 2) Support Vector Machine, 3) Gaussian
Naive Bayes, 4) Random Forest, 5) Decision Tree, 6) K-Nearest Neighbor, we found that regarding the maximum
pore pressure the prediction accuracies of all algorithms are between 84% and 100% in which the decision tree
algorithm can reach up to 99.18%. The method presented in this study can very quickly predict key factors affecting
spalling risks such as the maximum pore pressure and the temperature at the same depth based on different material
properties and environmental conditions.
Keywords: Tunnel fire, spalling risk, multifield model, machine learning.
Figure 1: Correlation analysis of the spalling factors

References
1. D. Gawin, F. Pesavento, and B. Schrefler. (2006). Towards prediction of the thermal spalling risk through a
multi-phase porous media model of concrete, Computer Methods in Applied Mechanics and Engineering, 195,
5707-5729.
2. Y. Zhang, M. Zeiml, C. Pichler, and R. Lackner, (2014). Model-based risk assessment of concrete spalling in
tunnel linings under fire loading. Engineering Structures, 77, 207-215.
3. Y. Zhang, M. Zeiml, M. Maier, Y. Yuan, and R. Lackner (2017). Fast assessing spalling risk of tunnel linings
under RABT fire: From a coupled thermo-hydro-chemo-mechanical model towards an estimation method.
Engineering Structures, 142, 1-19.
DOI: 10.32604/acm.2019.00107 www. .techscience.com/acm

Computational Modeling for Effect of Crack Healing Pattern in

Self-Healing Concrete
John Hanna

Corresponding Author: John Hanna. Email: john.nabil.mikhail.hanna@uni-weimar.de
Abstract: Concrete is the most widely used construction material in the world, therefore, is considered the second-
most consumed materials on earth, after water. But the weakest point that it subjected to cracking and deterioration
with the time, and the renovation of the buildings made of it costs a lot especially for infrastructure maintenance
which is not easily accessible. Recently, the biological systems become the inspiration for the material scientists.
This fundamental change in material design philosophy has resulted in the creation of ‘smart’ materials, like self-
healing materials. The development of self-healing concrete (SHC) has recently attracted a lot of attention due to its
inherent ability of automatic crack detection and crack repair with the goal of significantly prolong the service life
and reduce the cost of maintenance [1]. The aim of this research is to study the healing efficiency, life expectancy of
the concrete for the given design of micro-capsules and distribution, as well as material degradation over time. In
fact, studies on crack healing pattern are more important when incorporated in research on design of self-healing
structure because different positioning of healing capsules can lead to different crack healing pattern. However, if
the applied load is sustainable or the crack propagation is fast, partial healing at crack mouth will appear because the
capillary force might not be large enough to suck healing agent to fulfill the space above the self-healing capsule. In
addition there are possibilities where fully bonding is not established in the healing zone because chemical
properties of healing agent can be changed during storing or manufacturing process. There are a lot of laboratories
studies and experiments were done to study either the fracture interaction between the capsules and the concrete
matrix or the healing efficiency and healing performance such as [2]. Recently a lot of computational modeling in
SHC is done to study the fracture interaction between the capsules and the concrete matrix with different modeling
techniques; using cohesive elements [3] and using XFEM with cohesive surface which showed high accuracy [4].
But the computational modeling of the healing efficiency or healing performance of SHC is still in its infancy. In
this study, numerical simulations with XFEM, where properties of elements represent for healing agent are
randomly assigned, adopted to simulate the reliability of healing effect.
Keywords: Self-healing concrete, Healing efficiency, Healing pattern, XFEM.
(a) (b) (c)
Figure 1: Schematic sketch of modeling direct tension test: (a) Notched virgin material; (b) Crack propagation; (c)
Healed material
References
1. Gupta, Souradeep & Kua, Harn. (2016). Encapsulation technology and techniques in self-healing concrete.
Journal of Materials in Civil Engineering. 28(12), 04016165. 10.1061.
2. Snoeck, D., Malm, F., Cnudde, V., Grosse, C. U. & Van Tittelboom, K. (2018). Validation of self-healing
properties of construction materials through nondestructive and minimal invasive testing. Advanced Materials
Interfaces 5(17), 1800179.
3. Mauludin M. Luthfi, Zhuang, X. & Rabczuk, T. (2018). Computational modeling of fracture in encapsulation-

82 DACOMA-2019
based self-healing concrete using cohesive elements. Composite Structures, 196, 63-75.
4. F.A. Gilabert, D. Garoz & W. Van Paepegem (2017). Macro- and micro-modeling of crack propagation in
encapsulation-based self-healing materials: Application of XFEM and cohesive surface techniques. Materials
& Design 130, 459-478.
A data-driven solution for the governing equation of flexoelectricity

Bo He1 and Xiaoying Zhuang1,2,*
1Institute of Continuum Mechanics, Leibniz Universität Hannover, Appelstr. 11, 30167 Hannover, Germany
2
College of Civil Engineering, Tongji University, 1239 Siping Road, 200092 Shanghai, China
*
Abstract: Electromechanical coupling effect has been widely utilized in various applications ranging from building
structures [1,2], sensors [4,5] to biomedical devices [6,7]. Flexoelectricity, a type of electromechanical coupling
effect, has shown great potential for the electromechanical device at small length-scale thanks to its size-dependent
effect [7]. However, the involvement of the strain gradient term in the governing equation of flexoelectricity
imposes C¹ continuity for solving the partial differential equation. Meshfree method [8] and Mixed finite element
method (MFEM) [9] overcomes this obstacle by applying higher-order shape function and imposing extra degree of
freedoms, respectively, which largely increase the computational load. This makes the simulation of the
electromechanical response of the complex 3D flexoelectric device by the Meshfree or MFEM computational
expensive and ineffective.
A data-driven technique for solving nonlinear partial differential equation has been successfully implemented in [10].
The physics-informed neural network in [10] is trained to solve supervised learning tasks with respecting to the
given law of physics. In this study, we implement the mixed finite element method model for flexoelectricity in
AceGen [11] to generate the training data for the physics informed neural network in [10]. With TensorFlow, we
implement the data-driven framework to build the physics-informed surrogate model for analyzing the
electromechanical response of certain flexoelectric devices. Our results show that the physics-informed neural
network captures the electromechanical response of the studied flexoelectric devices with marginal errors. It's also
found that the learning speed and accuracy of the physics-informed neural network depends on the given batch size.
Keywords: Flexoelectricity, Data-driven computing, Strain gradient, Neural network, Coupled problem
Figure 1: (a) Electromechanical response of a cantilever beam due to flexoelectric effect by MFEM; (b) The
schematic illustration of the physics-informed neural network applied in this work

84 DACOMA-2019
Acknowledgement
The authors gratefully acknowledge the sponsorship from the ERC Starting Grant COTOFLEXI (No. 802205).
References
1. Chen, J., Qiu, Q., Han, Y., & Lau, D. (2019). Piezoelectric materials for sustainable building structures:
Fundamentals and applications. Renewable and Sustainable Energy Reviews, 101(1), 14-25.
2. Ramesh, T., Prakash, R., & Shukla, K. K. (2010). Life cycle energy analysis of buildings: An overview.
Energy and buildings, 42(10), 1592-1600.
3. Tzou, H. S., & Tseng, C. I. (1990). Distributed piezoelectric sensor/actuator design for dynamic
measurement/control of distributed parameter systems: a piezoelectric finite element approach. Journal of
sound and vibration, 138(1), 17-34.
4. Dosch, J. J., Inman, D. J., & Garcia, E. (1992). A self-sensing piezoelectric actuator for collocated control.
Journal of Intelligent material systems and Structures, 3(1), 166-185.
5. Lin, X., & Yuan, F. G. (2001). Diagnostic Lamb waves in an integrated piezoelectric sensor/actuator plate:
analytical and experimental studies. Smart Materials and Structures, 10(5), 907.
6. Zhou, Q., Lau, S., Wu, D., & Shung, K. K. (2011). Piezoelectric films for high frequency ultrasonic
transducers in biomedical applications. Progress in Materials Science, 56(2), 139-174.
7. Tagantsev, A. K. (1986). Piezoelectricity and flexoelectricity in crystalline dielectrics. Physical Review B,
34(8), 5883.
8. Abdollahi, A., Peco, C., Millan, D., Arroyo, M., & Arias, I. (2014). Computational evaluation of the
flexoelectric effect in dielectric solids. Journal of Applied Physics, 116(9), 093502.
9. Mao, S., Purohit, P. K., & Aravas, N. (2016). Mixed finite-element formulations in piezoelectricity and
flexoelectricity. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences,
472(2190), 20150879.
10. Qin, T., Wu, K., & Xiu, D. (2019). Data driven governing equations approximation using deep neural networks.
Journal of Computational Physics, 395(1), 620-635
11. Korelc, J. (2002). Multi-language and multi-environment generation of nonlinear finite element codes.
Engineering with computers, 18(4), 312-327.
Synthesis and study of copper ferrite-copper oxide nanocomposites

Kambiz Hedayati1,* and Tavus Hosseinabad1
1Department of Science, Arak University of Technology, Arak, Iran.
*Corresponding Author: Kambiz Hedayati. Email: k-hedayati@arakut.ac.ir.
Abstract: In this research copper ferrite as a core magnetic nanostructure were fabricate by a simple precipitation
method using green capping. Some fruits like lemon and orange were applied as biocompatible surface active agents.
Regarding their biocompatibility, low toxicity and adjustable magnetic properties, magnetic copper ferrite
nanoparticles have special physical and chemical properties because of their super-paramagnetism, quantum size
effect and surface-boundary effect [1-4]. Then copper oxide nanoparticles and copper ferrite-copper oxide
nanocomposites were synthesis by a fast chemical procedure. The X-ray diffraction pattern has suitable agreement
with standards and confirms purity of the nanocomposites. Scherrer equation was estimated of crystallite size under
100nm. The SEM image for various CuFe2O4, CuO and CuFe2O4/ CuO indicate mono- disperse particles with
mediocre size under 100 nm. FT-IR spectrum of nanoparticles indicate the absorption peak of CuFe2O4/ CuO. The
hysteresis loop of CuFe2O4and CuFe2O4/ CuO exhibits super paramagnetic behaviour. The photocatalytic activity of
the CuFe2O4/ CuO nanocomposites were degradation Acid black and methyl orange azo dyes above 90% at 60 min.
Due to their unique optical properties it possess various applications in lasers, light-emitting diodes, multi-exciton
generation solar cell detectors, nonlinear optics, single electron devices and biological imaging [5-10].
Keywords: Nanostructures, Photocatalyst, Copper ferrite, Copper oxide.
References
1. Tang, S. C., Hui, S. Y., Chung, H. S. H. (2002). Evaluation of the shielding effects on printed-circuit-board
transformers using ferrite plates and copper sheets. IEEE Transactions on Power Electronics, 17, 1080-1088.
2. Alrashed, A. A. A., Karimipour, A., Bagherzadeh, S. A., Safaei, M. R., Afrand, M. (2018). Electro- and
thermophysical properties of water-based nanofluids containing copper ferrite nanoparticles coated with silica:
Experimental data, modeling through enhanced ANN and curve fitting, International Journal of Heat and
Mass Transfer, 127, 925-935.
3. Chen, S., Yu, X., Huang, J., Yang, J., Lin, S. (2019). Interfacial ferrite band formation to suppress
intergranular liquid copper penetration of solid steel, Journal of Alloys and Compounds, 773, 719-729.
4. Prakash, P., Kumar, R. A., Miserque, F., Geertsen, V., Gravel, E., Doris, E. (2018). Carbon Nanotube
Supported Copper Ferrite Catalyst for Huisgen Cycloaddition, Chem. Commun, 54, 3644-3647.
5. Gu, W., Li, X., Xing, M., Fang, W., Wu, D. (2017). Removal of phosphate from water by amine-functionalized
copper ferrite chelated with La(III), Science of The Total Environment, 619-620, 42-48.
6. Lopez-Ramon, M. V., Alvarez, M. A., Moreno-Castilla, C., Fontecha-Camara, M. A., Yebra-Rodriguez, A.,
Bailon-Garcia, E. (2018). Effect of calcination temperature of a copper ferrite synthesized by a sol-gel method
on its structural characteristics and performance as Fenton catalyst to remove gallic acid from water, Journal
of Colloid and Interface Science, 511, 193-202.
7. Monig, H., Amirjalayer, S., Timmer, A., Hu, Z., Liu, L., Arado, O. D., Cnudde, M., Strassert, C. A., Ji, W.,
Rohlfing, M., Fuchs, H. (2018). Quantitative assessment of intermolecular interactions by atomic force
microscopy imaging using copper oxide tips, Nature Nanotechnology, 13, 371-375.
8. Li, W., Cui, X., Junge, K., Surkus, A. E., Kreyenschulte, C., Bartling, S., Beller, M. (2019). General and
Chemoselective Copper Oxide Catalysts for Hydrogenation Reactions, ACS Catal, 95, 4302-4307.
9. Lan, D., Qin, M., Yang, R., Chen, S., Wu, H., Fan, Y., Fu, Q., Zhang, F. (2019). Facile synthesis of
hierarchical chrysanthemum-like copper cobaltate-copper oxide composites for enhanced microwave
absorption performance, Journal of Colloid and Interface Science, 533, 481-491.

86 DACOMA-2019
10. Zhang, J., Zou, Z., Wang, B., Xu, G., Wu, Q., Zhang, Y., Yuan, Z., Yang, X., Yu, C. (2018). Lysosomal
deposition of copper oxide nanoparticles triggers HUVEC cells death, Biomaterials, 161, 228-239.
Comprehensive Study of Crack Propagation of different Percentage of Doping

of Boron Atoms on the Mechanical Properties of Polycrystalline Graphene
Mohammadreza Izadifar1, Soheil Gohari2, Colin Burvill2 and Timon Rabczuk3,*
1
Institute of Functional Interface, KIT Campus Nord, Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldschafen,
Germany.
2
Department of Mechanical Engineering, The University of Melbourne, Parkville, VIC 3010, Australia.
3
Institute of Research & Development, Duy Tan University, Quang Trung, Da Nang, Vietnam.
*
Corresponding Author: Timon Rabczuk. Email: timon.rabczuk@uni-weimar.de
Abstract: This article investigates the effect of different initial crack sizes on the mechanical response of single-
layer boron doped polycrystalline graphene nanosheets by Molecular Dynamics (MD) simulations. We study 1%,
3%, 6% and 10% of boron doped polycrystalline graphene nanosheets with grain sizes of 10 and 15 nm for eight
different initial crack lengths of 0.02L, 0.04L, 0.08L, 0.12L, 0.16L, 0.2L, 0.24L, and 0.32L, where L is the initial
length of the nanosheet. We found that 1. brittle fracture for boron doped polycrystalline graphene as the failure
occurs without any sign of plastic deformation and low energy absorption and 2. the ultimate tensile strength is
independent of the initial crack size. For identical grain sizes and doping of boron atoms, the ultimate tensile stress
and strain decrease as the crack lengths increases. In contrast, a clear trend was not observed in the ultimate tensile
strength for the same crack length and doping of boron atoms as the grain size increases. The highest difference of
13.4% in the ultimate tensile strength was observed between grain sizes of 10 and 15 nm for 3% of boron doped
nanosheets with an initial crack size of 20 Å.
Keywords: Boron nanosheet, doped polycrystalline graphene, fracture behavior, molecular dynamic simulation,
Initial crack.
References
1. Novoselov, K., S., Geim, A., K., Morozov, S., V., Jiang, D., Zhang, Y., Dubonos, S., V., Grigorieva, I., V., &
Firsov, A. A. (2004). Electric field effect in atomically thin carbon films. Science, 306(5696), 666-669.
2. Izadifar, M., Abadi, R., Jam, A. N., & Rabczuk T. (2017). Investigation into the effect of doping of boron and
nitrogen atoms in the mechanical properties of single-layer polycrystalline graphene. Computational Materials
Science, 138, 435-447.
3. Jam, A., N., Abadi, R., Izadifar, M., & Rabczuk, T. (2018). Molecular dynamics study on the mechanical
properties of carbon doped single-layer polycrystalline boron-nitride nanosheets. Computational Materials
Science, 153, 16-27.
4. Abadi, R., Uma, R., P., Izadifar, M., & Rabczuk, T. (2016). The effect of temperature and topological defects
on fracture strength of grain boundaries in single-layer polycrystalline boron-nitride nanosheet. Computational
Materials Science, 123, 277- 286.
5. Abadi, R., Shirazi, A., H., N., Izadifar, M., Sepahi, M., & Rabczuk, T. (2018). Fabrication of nanopores in
polycrystalline boron-nitride nanosheet by using si,sic and diamond clusters bombardment. Computational
Materials Science, 145, 280-290.
6. Abadi, R., Izadifar, M., Sepahi, M., Alajlan, N., & Rabczuk T. (2018). Computational modeling of graphene
nanopore for using in dna sequencing devices. Physica E: Low-dimensional Systems and Nanostructures,
103, 403-416.
7. Lee, C., Wie, X., Kysar, J. W., & Hone, J. (2008). Measurement of the elastic properties and intrinsic
strength of monolayer graphene. Science, 321(5887), 385-388.
88 DACOMA-2019
8. Golberg, D., Costa, P., M., F., J., Lourie, O., Mitome, M., Bai, X., Kurashima, K., Zhi, C., Tang, C., & Bando,
Y. (2007). Direct force measurements and kinking under elastic deformation of individual multiwalled boron
nitride nanotubes. Nano Letters, 7(7), 2146-2151.
9. Ghosh, S., Bao, W., Nika, D., L., Subrina, S., Pokatilov, E., P., Lau, C., N., Balandin, A., A. (2010).
Dimensional crossover of thermal transport in few-layer graphene. Nature Materials, 9, 555-558.
10. Mortazavi, B., Fan, Z., Pereira, L., F., C., Harju, A., & Rabczuk, T. (2016). Amorphized graphene: A stiff
material with low thermal conductivity. Carbon, 103, 318-326.
11. Silani, M., Talebi, H., Hamouda, A., M., & Rabczuk, T. (2016). Nonlocal damage modelling in clay/epoxy
nanocomposites using a multiscale approach. Journal of Computational Science, 15, 18-23.
12. Silani, M., Ziaei-Rad, S., Talebi, H., & Rabczuk, T. (2014). A semi-concurrent multiscale approach for
modelling damage in nanocomposites. Theoretical and Applied Fracture Mechanics, 74(1), 30-38.
13. Stankovich, S., Dikin, D., A., Dommett, G., H., B., Kohlhaas, K., M., Zimney, E., J., Stach, E., A., Piner, R.,
D., Nguyen, S., T., & Ruo, R., S. (2006). Graphene-based composite materials. Nature, 442(7100), 282-286.
14. Westervelt, R., M. (2008). Graphene nanoelectronics. Science, 320(5874), 324-325.
15. Bunch, J., S., van der Zande, A., M., Verbridge, S., S., Frank, I., W., Tanenbaum, D., M., Parpia, J., M.,
Craighead, H., G., & McEuen, P., L. (2007). Electromechanical resonators from graphene sheets. Science,
315(5811), 490-493.
16. Standley, B., Bao, W., Zhang, H., Bruck, J., Lau, C., N., & Bockrath, M. (2008). Graphene-based atomic-scale
switches. Nano Letters, 8(10), 3345-3349.
17. Peyvandi, A., Soroushian, P., Abdol, N., & Balachandra, A., M. (2013). Surface-modified graphite
nanomaterials for improved reinforcement efficiency in cementitious paste. Carbon, 63, 175-186.
18. Musso, S., Tulliani, J., M., Ferro, G., & Tagliaferro, A. (2009). Influence of carbon nanotubes structure on the
mechanical behaviour of cement composites. Composites Science and Technology, 69(11), 1985-1990.
19. Raee, M., A., Narayanan, T., N., Hashim, D., P., Sakhavand, N., Shahsavari, R., Vajtai, R., & Ajayan, P., M.
(2013). Hexagonal boron nitride and graphite oxide reinforced multifunctional porous cement composites.
Advanced Functional Materials, 23(45), 5624-5630.
20. Wang, M., Wang, R., Yao, H., Farhan, S., Zheng, S., & Du, C. (2016). Study on the three dimensional
mechanism of graphene oxide nanosheets modified cement. Construction and Building Materials, 126, 730-
739.
21. Hou, D., Lu, Z., Li, X., Ma, H., & Li, Z. (2017). Reactive molecular dynamics and experimental study of
graphene-cement composites: Structure, dynamics and reinforcement mechanisms. Carbon, 115, 188-208.
22. Izadifar, M., Kniger, F., Gerdes, A., Woll, C., Thissen, P. (2019). Correlation between composition and
mechanical properties of calcium silicate hydrates identified by infrared spectroscopy and density functional
theory. The Journal of Physical Chemistry C, 123(17), 10868-10873.
23. Giraudo, N., Bergdolt, S., Wohlgemuth, J., Welle, A., Schuhmann, R., Kniger, F., & Thissen, P. (2016).
Calcium silicate phases explained by high-temperature-resistant phosphate probe molecules. Langmuir, 32(51),
13577-13584.
24. Singh, V., Joung, D., Zhai, L., Das, S., Khondaker, S., I., & Seal, S. (2011). Graphene based materials: Past,
present and future. Progress in Materials Science, 56(8), 1178-1271.
25. Abadi, R., Uma, R., P., Izadifar, M., & Rabczuk, T. (2017). Investigation of crack propagation and existing
notch on the mechanical response of polycrystalline hexagonal boron-nitride nanosheets, Computational
Materials Science, 131, 86-99.
26. Mortazavi. B., & Cuniberti, G. (2014). Mechanical properties of polycrystalline boron-nitride nanosheets. RSC
Adv., 4, 19137-19143.
27. Hassan, M., H., Haikal, R., R., Hashem, T., Rinck, J., Koeniger, F., Thissen, P., Heiler, S., Woll, C., & Alkordi,
M., H. (2019). Electrically conductive, monolithic metalorganic frameworkgraphene (mof@g) composite
coatings. ACS Applied Materials & Interfaces, 11(6), 6442-6447.
28. Lee, S., U., Belosludov, R., V., Mizuseki, H., & Kawazoe, Y. (2009). Designing nanogadgetry for
nanoelectronic devices with nitrogen-doped capped carbon nanotubes. Small, 5(15), 1769-1775.
Phase field modelling of microstructural length scale effects on stressed grain

growth in polycrystalline copper thin films
Mostafa Jamshidian1,2,* and Timon Rabczuk2
1
Department of Mechanical Engineering, Isfahan University of Technology, Isfahan 84156-83111, Iran.
2
Institute of Structural Mechanics, Bauhaus-University Weimar, Marienstrasse 15, 99423 Weimar, Germany.
*
Corresponding Author: Mostafa Jamshidian. Email: jamshidian@cc.iut.ac.ir
Abstract: The phase field modeling approach has recently emerged as a powerful simulation tool for microstructure
evolution applications [1-8]. In this approach, the microstructure is represented by a set of non-conserved phase field
variables which are continuous in space and time. Each variable represents the local volume fraction of its
corresponding phase i.e. it takes unity in its corresponding grain and zero elsewhere. Grain boundaries are modeled
as diffuse interfaces across which the phase field variables have a smooth transition. Evolution of the phase field
variables under the effect of thermodynamic driving forces implicitly represents grain boundary migration and
microstructure evolution. In the present research, phase field simulations are employed to investigate the effect of
microstructural length scale on the evolution of microstructure and texture in a polycrystalline copper thin films. It is
generally demonstrated that increasing the microstructural length scale results in a transition from normal to
abnormal grain growth [2]. Such a transition is attributed to the relative magnitude of the two interacting driving
forces i.e. the strain energy minimizing driving force and the curvature driving force. The strain energy minimizing
driving force is kept constant under a fixed applied strain and acts in favor of the texture component with lower
strain energy. On the other hand, the curvature driving force is inversely proportional to the microstructural length
scale and is impartial to grain orientations. As a result, the relative magnitude of the two driving forces is sensitive
to the microstructural length scale and hence the evolution of microstructure and texture is significantly affected by
the microstructural length scale. For the case of small microstructural length scale, computer simulations have
revealed that (i) the dominance of curvature driving force results in the evolution of microstructure by normal grain
growth, (ii) Stressed grain growth is impartial to different texture components i.e. no change in texture occurs, (iii)
The final microstructure contains equiaxed grains with log-normal grain size distribution, and (iv) the overall
stiffness of the polycrystalline aggregate is unchanged during stressed grain growth. On the other hand, for the case
of large microstructural length scale we have observed that (i) the dominance of strain energy minimizing driving
force results in the evolution of microstructure by abnormal grain growth, (ii) stressed grain growth results in the
rapid abnormal growth of the texture component with lower strain energy, (iii) the final microstructure contains
abnormal-shaped grains with bimodal grain size distribution, (iv) the overall stiffness of the polycrystalline
aggregate decreases by stressed grain growth. For intermediate values of the microstructural length scale, a
transitional behavior between the two extreme regimes was observed.
Keywords: Stressed grain growth, Phase field modelling, Polycrystalline microstructure, Microstructural length
scale
References
1. P. Thamburaja, M. Jamshidian. (2014). A multiscale Taylor model-based constitutive theory describing grain
growth in polycrystalline cubic metals, J. Mech. Phys. Solids. 63: 1-28.
2. M. Jamshidian, T. Rabczuk. (2014). Phase field modelling of stressed grain growth: Analytical study and the
effect of microstructural length scale, J. Comput. Phys. 261: 23-35.
3. M. Jamshidian, G. Zi, T. Rabczuk. (2014). Phase field modeling of ideal grain growth in a distorted
microstructure, Comput. Mater. Sci. 95: 663-671.
4. M. Jamshidian, P. Thamburaja, T. Rabczuk. (2016). Modeling the effect of surface energy on stressed grain
growth in cubic polycrystalline bodies, Scr. Mater. 113: 209-213.
5. M. Jamshidian, P. Thamburaja, T. Rabczuk. (2016). A multiscale coupled finite-element and phase-field
90 DACOMA-2019
framework to modeling stressed grain growth in polycrystalline thin films, J. Comput. Phys. 327.
6. M. Jafari, M. Jamshidian, S. Ziaei-Rad, D. Raabe, F. Roters. (2017). Constitutive modeling of strain induced
grain boundary migration via coupling crystal plasticity and phase-field methods, Int. J. Plast. 99.
7. M. Jafari, M. Jamshidian, S. Ziaei-Rad, B. Lee. (2019). Modeling length scale effects on strain induced grain
boundary migration via bridging phase field and crystal plasticity methods, Int. J. Solids Struct.
8. E. Shahnooshi, M. Jamshidian, M. Jafari, S. Ziaei-Rad, T. Rabczuk. (2019). Phase field modeling of stressed
grain growth: Effect of inclination and misorientation dependence of grain boundary energy, J. Cryst. Growth.
518, 18-29.
Parametric Investigation of Large Stretchability of Graphene Nanoribbon

Springs
Brahmanandam Javvaji1, Bohayra Mortazavi1, Xiaoying Zhuang1,2, *
1
Institute of Continuum Mechanics, Leibniz Universität Hannover, Applestr. 11, 30167 Hannover, Germany.
2
College of Civil Engineering, Tongji University, 1239 Siping Road, 200092 Shanghai, China.
Abstract: Graphene has several interesting electronic, thermal and mechanical properties [1,2]. The very high
tensile modulus (1 TPa) and tensile strength (130 Gpa) [3] are useful in various structural applications. However,
graphene is also known to be quite brittle, where tensile fracture strain does not exceed a few percents [4], which
restricts the use of graphene when dealing with large deformation strains. A periodic cutting pattern has been used to
enhance the fracture strain of graphene (graphene kirigami) [5]. The distance between cuts and cut orientations
strongly define the stretchability of graphene kirigami [6]. For straight cuts, the increase in longitudinal spacing
between cuts decreases the out-of-plane deflection and decrease the fracture strain. The shorter distance between
curved cuts decreases the fracture strain [7]. The larger stretchability requires a careful design of cut length and cut
orientations. Hence, in search of a set of optimal parameters which produce higher stretchability require a large
number of simulations or experiments. Machine learning (ML) is an optimal choice for dealing with these large data
sets. A recent study employed ML to accelerate the search of these parameters for graphene kirigami [8]. In a similar
direction of graphene kirigami, a periodic undulating graphene nanoribbon is a promising choice for better
stretchability. There are different types of possibilities of undulations such as sinus, U shape, half-circles, elliptical
and, horseshoe shape [9]. The graphene ribbons with the periodic modifications are similar to a two-dimensional
spring, known as graphene nanoribbon springs (GNRS). A recent experimental report demonstrates the lithography
technique to reliable manufacturing of GNRS [10]. The effect of structural parameters on the mechanical properties
of GNRS is unknown.
Figure 1: (a) GNRS atomic configuraiton. (b) Stress-strain response for GNRS. Markers indicate the
fracture stress and fracture strain. (c) Fracture stress and fracture strain values for different St and Sa.
Color indicates the strain and value denote the stress.
In this work, we consider studying the mechanical properties of GNRS in sinus shape and the effect of structural
parameters. Fig. 1(a) represent the GNRS model. The structural parameters for GNRS are spring pitch (Sp), spring
thickness (St) and spring amplitude (Sa). These three parameters strongly define the mechanical response of GNRS

92 DACOMA-2019
in sinus shape. To study the mechanical properties of GNRS, we employed classical molecular dynamics (MD)
simulations. The atomic interaction between carbon atoms was modeled using the Tersoff potential. We consider Sp
as 6 nm and vary St and Sa. We consider the lower bound of St in a way that it can accommodate at least two
hexagonal carbon rings. Similarly, Sa varies from 0.2 nm to 6 nm. Overall about 200 structures of GNRS are
simulated within in the given range of St and Sa. We evaluate the virial stress and strain at equal intervals of time
under tensile loading with periodic boundary conditions.
Fig. 1(b) shows the stress-strain curves for a selected set of St and Sa. The fracture stress for GNRS with Sa smaller
than St is very high and the fracture strain is lower than pristine graphene (~0.2). When Sa is 0.5 times St, the
fracture strain is nearly equal to pristine graphene with 25% of graphene fracture stress. In this GNRS, we observe
many structural defects, which avoids the brittle failure by coalescence of defects. The ultimate failure is at 0.6
strain. The response of GNRS with Sa 0.7 times St is similar to the previous case. The fracture strain is very much
enhanced (about 1) when Sa is nearly equal to St. The GNRS is going to out-of-plane deflections with tensile
deformation, which helps to increase the strain with very low-stress values. For Sa = 1.6St, the fracture strain is even
more (about 1.6) with nearly 30 GPa fracture stress. Fig. 1(c) reports the fracture stress and fracture strain for
different combinations of St and Sa. From these results, it is clear that a good choice of St and Sa yields a GNRS
system with very large stretchability.
The present results only focused on varying the St and Sa parameters. The effect of Sp and shape are other factors
that also need to be considered. Therefore, a very large data set has to be analyzed and obtain the fracture properties.
As the number of possible configurations between these structural parameters increases, it is impractical to find the
optimal set with good stretchability. ML represents an alternative and promising approach for this type of problems.
Systematic production of training data for ML process using MD simulations to find the fracture stress and fracture
strain is considered as a future aspect in this research.
Keywords: Graphene nanoribbon springs, tensile strength, tensile strain, molecular dynamics.
Acknowledgement
We acknowledge the support of the cluster system team at the Leibniz University of Hannover, Germany in the
production of this work.
References
1. Castro Neto, A. H., Guinea, F., Peres, N.M.R., Novoselov, K.S., & Geim, A.K. (2009). The electronic
properties of graphene. Reviews of Modern Physcis, 81(1), 109-162.
2. Geim, A. K. (2009). Graphene: status and prospects, Science, 324(5934), 1530-1534.
3. Lee, C., Wei, X., Kysar, J. W., & Hone, J. (2008). Measurement of the elastic properties and intrinsic strength
of monolayer graphene. Science, 321(5887), 385-388.
4. Zhang, P., Ma, L., Fan, F., Zeng, Z., Peng, C., Loya, P. E., Liu, Z., Gong, Y., Zhang. J., Zhang. X., & Ajayan,
P. M. (2014). Fracture toughness of graphene. Nature communications, 5, 3782.
5. Blees, M. K., Barnard, A. W., Rose, P. A., Roberts, S. P., McGill, K. L., Huang, P. Y., Ruyack, A. R., Kevek,
J. W., Kobrin, B., Muller, D. A., & McEuen, P. L. (2015). Graphene kirigami. Nature, 524(7564), 204.
6. Qi, Z., Campbell, D. K., & Park, H. S. (2014). Atomistic simulations of tension-induced large deformation and
stretchability in graphene kirigami. Physical Review B, 90(24), 245437.
7. Mortazavi, B., Lherbier, A., Fan, Z., Harju, A., Rabczuk, T., & Charlier, J. C. (2017). Thermal and electronic
transport characteristics of highly stretchable graphene kirigami. Nanoscale, 9(42), 16329-16341.
8. Hanakata, P. Z., Cubuk, E. D., Campbell, D. K., & Park, H. S. (2018). Accelerated search and design of
stretchable graphene kirigami using machine learning. Physical review letters, 121(25), 255304.
9. Gonzalez, M., Axisa, F., Bulcke, M. V., Brosteaux, D., Vandevelde, B., & Vanfleteren, J. (2008). Design of
metal interconnects for stretchable electronic circuits. Microelectronics Reliability, 48(6), 825-832.
10. Liu, C., Yao, B., Dong, T., Ma, H., Zhang, S., Wang, J., Xu, J., Shi, Y., Chen, K., Gao, L., & Yu, L. (2019).
Highly stretchable graphene nanoribbon springs by programmable nanowire lithography. npj 2D Materials and
Applications, 3(1), 23.
Multiscale computation based on FNN and RNN

Bin Li1 and Xiaoying Zhuang1, *

1
Corresponding Author: Xiaoying Zhuang. Email: xiaoyingzhuang@tongji.edu.cn.

*
Abstract: Computational homogenization method offers high accuracy at the expense of high computational cost.
The neural networks can be used to construct fully decoupled approaches in nonlinear multiscale methods by
mapping macroscopic loading and microscopic response. Computational homogenization method for nonlinear
material and implementation of offline multiscale computation are studied to generate data set. This article intends
to model the multiscale constitution using feedforward neural network (FNN) and recurrent neural network(RNN),
and proper loading paths are selected to effectively generalized to unknown paths. Finally, applications to 2-
dimensional multiscale analysis are addressed in detail.
Data-driven methods were thus developed to compute the response of heterogeneous microstructures using an
interpolation technique in a higher dimensional space from a prior set of databases computed from offline nonlinear
calculations on the representative volume element (RVE) [1]. Artificial neural networks (ANNs) have also been
used to approximate various constitutive models because of the capability to learn complex nonlinear relationships.
In the early 90s, Ghaboussi et al. [2] proposed to model the material behavior with neural network.
The use of ANN for direct representation macroscopic mechanical properties has been studied by many researchers
[3-5]. However, choosing the loading case for adequate training of multiscale constitutive model is a challenging
task, which is not precisely known at this time. Most neural networks for multiscale constitutive model are based on
feedforward neural network (FNN) and BP algorithm. Recurrent neural network (RNN) is a type of neural network
where the output from previous step are fed as input to the current step, so that it may be used to model the nonlinear
multiscale constitutive model considering material history dependency. However, only limited papers have adopted
RNN for constitutive modeling. This article intends to model the multiscale constitution using FNN and RNN, and
proper loading paths are selected to effectively generalize to unknown paths.
In the nonlinear computational homogenization scheme, the macroscale incremental strain is applied on RVE, while
the homogenized incremental effective stress and homogenized tangent constitutiove tensor are return back to
macroscale model. For offline multiscale calculation, the macroscopic strain, obtained prior from a given loading
path without calculated from macroscale model, is applied on a single RVE with constrain equations. The
microscale model is simulated incrementally by the restart analysis in ABAQUS until the final step of the loading
path is reached. To calculate the macroscopic stress and tangent tensor, one general three perturbation steps for
plane-stress problems will be carried out on the microscale model. For restart analysis, the analysis results from the
last step are used for the analysis of the current step and are overwritten for the next step, which greatly reduces the
calculation file storage.
Figure 1: The stacked GRU neural networks unrolled through time

As the nonlinear response of macroscale model is depend on the loading history and current load increment, the
macroscopic stress and strain of the previous step and the current incremental strain are used as the input parameters for
FNN. The motivation to use RNN for multiscale nonlinear computation is that the loading history information can be

94 DACOMA-2019
retained. Different from training discrete data points for each path in FNN, a loading path is used as a time-dependent
data chain in RNN, and the previous data will affect the later outputs. Within this paper, a stacked GRU neural network
(SRNN) connected with fully connected layers at each time step is adopted, which is illustrated in Fig. 1.
The question of what constitutes sufficient data sets for adequate training of neural network material model is not
precisely known at this time, and a well-trained neural network model doesn't guarantee a good generalization to
other paths. Within this paper, macroscopic strain tensor increment is set to be cosine function and total macroscopic
strain tensor component is given as
𝑘𝑘
ε𝑡𝑡𝑖𝑖 = ε𝑡𝑡−1
𝑖𝑖 + 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅(𝜃𝜃𝑖𝑖𝑡𝑡 ) 𝑘𝑘 = 1, ⋯ 𝑛𝑛 (1)
𝑛𝑛
ε0𝑖𝑖
where = 0, R and 𝜃𝜃 are the loading radius and angle. The applied macroscopic strain tensor components are
increased with independent constant increments until k=n, then the loading angles will change randomly within the
set {0° , 1° ⋯ , 180° }. Therefore, the training data can reflect the characteristics of different loading direction. The
procedure for applying the strain component is illustrated in Fig.2.
Figure 2: The procedure for applying the strain component

In neural network, hyperparameters refer to prior parameters that need to be tuned to optimize it, including the
network structure, optimized parameters and regularization coefficient. Grid search and manual search are the most
widely used strategies for hyperparameter optimization. Within this paper, grid search with orthogonal experimental
design (OED) is used to obtain optimal result in a short time. Finally, applications to 2-dimensional multiscale
analysis are addressed in detail. It's found that the trained FNN and RNN model both have good generalization
ability on test set, and the stacked RNN model is more capable of mapping a given path, but also more sensitive to
hyperparameters and initial parameters.
Keywords: Multiscale method, constitutive model, FNN, RNN.
References
1. Geers, M. G. D., Yvonnet, J. (2016). Multiscale modeling of microstructure–property relations. MRS Bulletin,
41(8), 610-616.
2. Ghaboussi, J, Garrett Jr, J. H., Wu, X. (1991). Knowledge-based modeling of material behavior with neural
networks. Journal of engineering mechanics, 117(1), 132-153.
3. Unger, J. F., Könke, C. (2008). Coupling of scales in a multiscale simulation using neural networks. Computers
& Structures, 86(21-22), 1994-2003.
4. Lefik, M., Boso, D. P., Schrefler, B. A. (2009). Artificial neural networks in numerical modelling of
composites. Computer Methods in Applied Mechanics and Engineering, 198(21-26), 1785-1804.
5. Le, B. A., Yvonnet, J., He, Q. C. (2015). Computational homogenization of nonlinear elastic materials using
neural networks. International Journal for Numerical Methods in Engineering, 104(12), 1061-1084.
A Manifold Learning Approach for Multiscale Phase Field Evolution for

Fracture
Yangyuanchen Liu1, Kexin Weng1 and Yongxing Shen1,*
1University of Michigan – Shanghai Jiao Tong University Joint Institute, Shanghai Jiao Tong University, 800 Dongchuan Road,
Shanghai, 200240, China.
*Corresponding Author: Yongxing Shen. Email: yongxing.shen@sjtu.edu.cn.
Abstract: The multi-scale fracture simulation of heterogeneous materials is a popular and important subject in solid
mechanics and materials science due to the wide application of composite materials. In this paper we propose a new
computational strategy for solving multiscale fracture problem efficiently through homogenization in a FE2-like [1]
framework. To take into account the phase field induced by fracture at the microstructure level, the Representative
Volume Element (RVE), which consists of an elastic matrix and inclusions (fiber), is endowed. We randomly place
an initial crack in the RVE in order to represent the possible microscopic crack paths. The crack is described by the
regularized variational theory of fracture, or so-called phase field approach, which is able to predict crack nucleation,
propagation and branching without extra criteria [3]. Then the RVE response is homogenized and upscaled to the
macroscale as a phase field constitutive model previously developed by the authors [2], which is then used in a
framework of a phase field modeling of propagating fracture.
This computational scheme for solving multiscale fracture problems can be inefficient and inflexible because the
phase field method requires an unpredictable number of iterations. In this paper we consider an alternative route that
adopts techniques well known in the machine learning community in order to extract the manifold that contains the
inputs to the RVE, namely the initial crack path and the load. Then output data, which is the evolved phase fields,
can be interpolated accurately with minimum online computation. In particular, the locally linear embedding is
chosen in this work for the local phase field evolution for fracture of heterogeneous microstructures. We construct a
manifold which can efficiently and accurately interpolate micro crack evolution laws with given microstructures. In
this paper, we perform a manifold learning approach based on the use of LLE technique [6]. Locally linear
embedding (LLE) [6], a particular example of kernel principal component analysis (kernel PCA) [4,5,7], is an
unsupervised learning algorithm that computes low-dimensional, topology-preserving embeddings of high-
dimensional data points. The offline procedure consists of two stages: (1) dataset construction with the phase field
analysis for the RVE. (2) data manifold construction with the LLE. The online interpolation procedure then readily
delivers the phase field evolution. In the presentation, an accuracy check between FEM results and manifold
interpolation is provided.
Keywords: Multiscale simulation, Manifold learning, Locally linear embedding, Phase field for fracture,
Computational homogenization
References
1. Feyel, F., Cailletaud, G., École nationale supérieure des mines (Paris)., & ENSMP MAT. (1998). Application
du calcul parallèle aux modèles à grand nombre de variables internes.
2. Cheng, C., Shen, Y. (2018). A micromechanics-based phase field approach to fracture. In The Third
International Conference on Damage Mechanics, The Chinese Society of Theoretical and Applied Mechanics.
3. Shen, Y., Mollaali, M., Li, Y., Ma, W., Jiang, J. Implementation Details for the Phase Field Approached to
Fracture. Journal of Shanghai Jiao Tong University (Science), 23(1), 166-174.
4. Scholkopf, B., Smola, A., Muller, KR. (1998). Nonlinear component analysis as a kernel eigenvalue problem.
Neural Computation 10(5), 1299-1319.
5. Scholkopf, B., Smola, A., Muller, KR. (1999). Kernel principal component analysis. In: Advances in kernel
methods: support vector learning. MIT Press, Cambridge (pp. 327-35).

96 DACOMA-2019
6. Roweis, ST., Saul, LK. (2000). Nonlinear dimensionality reduction by locally linear embedding. Science 290,
2323-2326.
7. Zimmer, VA., Lekadir, K., Hooendoorn, C., Frangi, AF., Piella, G. (2015). A framework for optimal kernel-
based manifold embedding of medical image data. Computerized Medical Imaging and Graphics, 41, 93-107.
Numerical simulations on the material point method of impact damage

problems
Jing-xin Ma*
College of Aerospace and Civil Engineering, Harbin Engineering University, 145 Nantong Street, 150001, Harbin, China.
*Corresponding Author: Jingxin ma. Email: majingxin1993@hrbeu.edu.cn.
Abstract: The impact damage problems was studied based on the material point method (MPM). The high-energy
explosive explosion calculation model is established, the underwater and air explosion calculation model is
discussed, the material point method non-reflective boundary treatment method is perfected, the structural plastic
dynamic response and failure model is given, and the numerical simulation of the structural damage effect of
explosives in water and air media is calculated. Using pressure, density, etc. as a typical result analysis, a concrete
solution for calculating the dynamic response of structural plasticity by material point method is given.
Keywords: Material Point Method, Numerical Simulation, Impact problems, Damage.
References
1. Bardenhagen, SG, JU Brackbill, & Deborah Sulsky. (2000). The material-point method for granular materials.
Computer Methods in Applied Mechanics and Engineering, 187: 529-41.
2. Belytschko, Ted, Yury Krongauz, Daniel Organ, Mark Fleming, & Petr Krysl. (1996). Meshless methods: an
overview and recent developments. Computer Methods in Applied Mechanics and Engineering, 139: 3-47.
3. Brackbill, Jeremiah U, & Hans M Ruppel. (1986). FLIP: A method for adaptively zoned, particle-in-cell
calculations of fluid flows in two dimensions. Journal of Computational physics, 65: 314-43.
4. Chen, Weidong, Yaqin Shi, Han Yan, Jingxin Ma, Yuzhuo Yang, & Chunlong Xu. (2019). A stochastic
material point method for probabilistic dynamics and reliability. Computational Mechanics: 1-14.
5. Chen, Z, L Shen, Y-W Mai, and Y-G Shen. (2005). A bifurcation-based decohesion model for simulating the
transition from localization to decohesion with the MPM. Zeitschrift für angewandte Mathematik und Physik
ZAMP, 56: 908-30.
6. Chen, Zhen, & Rebecca Brannon. (2002). An evaluation of the material point method, SAND Report,
SAND2002-0482,(February 2002).
7. Chen, Zhen, W Hu, L Shen, X Xin, & R Brannon. (2002). An evaluation of the MPM for simulating dynamic
failure with damage diffusion, Engineering Fracture Mechanics, 69: 1873-90.
8. Gan, Yong, Zhen Chen, & Stephen Montgomery-Smith. (2011). Improved material point method for
simulating the zona failure response in piezo-assisted intracytoplasmic sperm injection. Computer Modeling in
Engineering and Sciences, 73: 45.
9. Guo, Yajun, & John A Nairn. (2004). Calculation of J-integral and stress intensity factors using the material
point method. Computer Modeling in Engineering and Sciences, 6: 295-308.
10. Harlow, Francis H. 1964. The particle-in-cell computing method for fluid dynamics. Methods Comput. Phys., 3:
319-43.
11. Ma, Shang, Xiong Zhang, Yanping Lian, & Xu Zhou. (2009). Simulation of high explosive explosion using
adaptive material point method', Computer Modeling in Engineering and Sciences (CMES), 39: 101.
12. Rabczuk, Timon, & T Belytschko. (2004). Cracking particles: a simplified meshfree method for arbitrary
evolving cracks, International Journal for Numerical Methods in Engineering, 61: 2316-43.
13. Rabczuk, Timon, Stéphane Bordas, & Goangseup Zi. (2010). On three-dimensional modelling of crack growth
using partition of unity methods. Computers & Structures, 88: 1391-411.
14. Rastkar, Siavash, Maryam Zahedi, Igor Korolev, & Arvind Agarwal. (2017). A meshfree approach for
homogenization of mechanical properties of heterogeneous materials. Engineering Analysis with Boundary
Elements, 75: 79-88.
98 DACOMA-2019
15. Sulsky, Deborah, Zhen Chen, & Howard L Schreyer. (1994). A particle method for history-dependent
materials. Computer Methods in Applied Mechanics and Engineering, 118: 179-96.
16. Zhang, Wei, Xuanming Cai, Nan Ye, & Yubo Gao. (2017). Experimental study on the initiation and energy
release behavior of polymer bonded explosive materials. In AIP Conference Proceedings, 040024. AIP
Publishing.
Numerical Modeling of Crack Propagation in the Presence of Inclusion

Using Extended Finite Element Method
Luthfi M. Mauludin1,2,*
1 Institute
of Structural Mechanics, Bauhaus-Universität Weimar, Marienstrasse 15, D-99423, Weimar, Germany
2Civil Engineering Department, Politeknik Negeri Bandung (POLBAN) Jl. Gegerkalong Hilir Ds. Ciwaruga, Kotak Pos 40012,
Bandung, Indonesia
*
Corresponding Author: Luthfi Muhammad Mauludin. Email: luthfi.muhammad.mauludin@uni-weimar.de.
Abstract: The characteristic of crack propagation in the presence of inclusion is investigated numerically. In this
study, the crack growth is modeled using extended finite element method (XFEM). Two-dimensional rectangular
plate with single inclusion embedded off-centered is modeled. The specimen is subjected to uniaxial tension. The
motion of the crack is modeled by XFEM based on traction-separation cohesive behavior for 2D mixed mode
problem. In addition, enrichment procedure is used to implicitly determine predefined crack in XFEM framework.
Two different inclusions, which are soft and hard inclusions, are considered on crack propagation scheme. The
effects of soft and hard inclusions on crack propagation are studied and observed. It is found that the presence of
inclusion plays a crucial role on the crack trajectory inside the material. The presence of inclusion in the material is
proved to be one of the dominant factors which can alter the crack growth characteristics.
Keywords: Crack propagation, Extended finite element method, Traction-separation, Inclusions.
Acknowledgement
This work is supported and financed by RISTEK-DIKTI (Directorate General of Resources for Science, Technology
and Higher Education. Ministry of Research, Technology and Higher Education of Indonesia) under funding agree-
ment No: 153.39/E4.4/2014. The supports are gratefully acknowledged.
References
1. Chen L, Rabczuk T, Bordas S P A, Liu G, Zeng K, Kerfriden P. (2012). Extended finite element method
with edge-based strain smoothing (ESm-XFEM) for linear elastic crack growth. Computer Methods in
Applied Mechanics and Engineering, 209: 250–265.
2. Nguyen-Vinh H, Bakar I, Msekh M, Song J H, Muthu J, Zi G, Le P, Bordas S, Simpson R, Natarajan S,
Lahmer T, Rabczuk T. (2012). Extended finite element method for dynamic fracture of piezo-electric
materials. Engineering Fracture Mechanics, 92: 19–31.
3. Zhang C, Wang C, Lahmer T, He P, Rabczuk T. (2016). A dynamic XFEM formulation for crack
identification. International Journal of Mechanics and Materials in Design, 12(4): 427–448.
4. Vu-Bac N, Nguyen-Xuan H, Chen L, Bordas S, Kerfriden P, Simpson R, Liu G, Rabczuk T. (2011). A
node-based smoothed extended finite element method (NS-XFEM) for fracture analysis. Computer
Modeling in Engineering & Sciences, 73(4): 331–356.
5. Mauludin, L. M., Zhuang, X., & Rabczuk, T. (2018). Computational modeling of fracture in encapsulation-
based self-healing concrete using cohesive elements. Composite Structures, 196: 63-75.
6. Mauludin, L. M., & Oucif, C. (2019). The effects of interfacial strength on fractured
microcapsule. Frontiers of Structural and Civil Engineering, 13(2): 353-363.

Active Machine Learning to Design Novel Two-dimensional Carbon Nitrides

Bohayra Mortazavi* and Xiaoying Zhuang
Institut für Kontinuumsmechanik, Leibniz Universität Hannover, Appelstraße 11, 30167 Hannover, Germany.
*Corresponding Author: Email: bohayra.mortazavi@gmail.com.
Abstract: Two-dimensional (2D) materials are currently among the most attractive family of materials with highly
promising application prospects for a wide range of advanced technologies. Graphene as the most prominent
member of 2D materials family has been proven to exhibit exceptionally high mechanical and thermal conduction
properties along with unique optical and electronic features. In general, for many applications in nanoelectronics and
nanooptics presenting a narrow and direct band-gap is among the most important requirements, however graphene is
a zero band-gap semimetal. Worthy to remind via chemical doping with nitrogen or boron atoms, mechanical
straining or creation of nanomesh structures, a band-gap can be opened in the graphene. Nevertheless, these
aforementioned methods require additional processing steps after the first synthesize of graphene, which question
their practical application due to the increase in the fabrication time and cost as well. Such that designing novel 2D
materials that are semiconductors in their pristine form has been among the most appealing approached for the
employment of 2D materials in nanoelectronics and nanophotonics.
Among the various classes of 2D materials, carbon-nitride nanomembranes have been among the most successful
nanomaterials with inherent semiconducting electronic characters. These 2D allotropes show a common chemical
formula CxNy, where x and y represent the number of C and N atoms in the unit cell, respectively. Carbon-nitride
nanomaterials offer stiff and stable components owing to the formation of strong covalent bonds between the C-C
and C-N bonds. Depending on the composition of C and N atoms in various atomic lattices, carbon-nitride
allotropes can exhibit diverse electronic, optical, electrochemical, mechanical and thermal conduction properties.
Unlike the graphene, carbon nitride nanosheets show porous atomic lattices and more importantly are inherent
semiconductors, highly promising to design novel nanoelectronics. Carbon nitride 2D structures also show good
mechanical properties and have been proven as promising candidates for the employment; in energy
conversion/storage, catalysis, photocatalysis and oxygen reduction systems. In this work motivated by the
exceptional properties of carbon-nitride nanosheets, we predicted three novel porous lattices; C7N6, C9N4 and C10N3.
Stability, mechanical properties and electronic/optical properties of these novel nanomembranes were explored by
the first-principles density functional theory (DFT) simulations. Acquired theoretical results confirm the stability
and highlight highly attractive properties of C7N6 and C9N4 nanosheets, and may hopefully motivate and guide
future experimental and theoretical studies. In the following illustration the predicted nanosheets are illustrated.
Figure 1: Atomic structure of C7N6, C9N4 and C10N3 monolayers. Contours illustrate the electron localization function within the
unit-cell.
The acquired first-principles results confirm the thermal, dynamical and elastic stability of C7N6 and C9N4
nanosheets and also highlight that these novel 2D systems can endure under severe loading conditions, owing to
their strong covalent bonding nature. Our extensive calculations confirm the stability and highlight highly attractive
101 DACOMA-2019
properties of C7N6 and C9N4 nanosheets, and may hopefully motivate and guide future experimental and theoretical
studies. In the following illustration the predicted nanosheets are illustrated. Nonetheless, these observations cannot
confirm that the predicted lattices are the global minimum structures. To address this important issue, we will
conduct extensive search and prediction of various possible structures, by employing crystal structure prediction
methods. In this approach, novel materials prediction will be accomplished by developing machine-learning
interatomic potential to approximate quantum-mechanical energies and an active learning algorithm for the
automatic selection of an optimal training dataset [1,2]. This approach substantially reduces the amount of highly
demanding DFT calculations, as the DFT will be only used to produce the training data, while structural
optimization is performed using the interatomic potentials. In this method, we first conduct ab-initio molecular
dynamics simulations for around 30 well-known carbon-nitride monolayers, in order to train a moment tensor
potential for approximating the quantum-mechanical energies. Using the proposed method, we will predict novel
carbon-nitride nanosheets with computational costs by around two orders of magnitude faster than the common DFT
based material prediction techniques. Moreover, we will be hopefully able to plot the phase diagram of this
attractive class of 2D semiconductors. These results can serve as a unique guide for the future experimental and
theoretical studies.
Keywords: Carbon-nitride, 2D materials, Machine-learning, Interatomic potential, Quantum-mechanics.
Acknowledgement
Authors appreciate the funding by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)
under Germany’s Excellence Strategy within the Cluster of Excellence PhoenixD (EXC 2122, Project ID
390833453).
References
1. Oganov, A. R.; Glass, C. W. Crystal Structure Prediction Using Ab Initio Evolutionary Techniques:
Principles and Applications. J. Chem. Phys. 2006, 124 (24), 244704. https://doi.org/10.1063/1.2210932.
2. Gubaev, K.; Podryabinkin, E. V.; Hart, G. L. W.; Shapeev, A. V. Accelerating High-Throughput Searches
for New Alloys with Active Learning of Interatomic Potentials. Comput. Mater. Sci. 2019, 156, 148–156.
https://doi.org/10.1016/j.commatsci.2018.09.031.
Prediction of bending angle of laser bent tailor welded blanks using artificial
neural network (ANN)
Mohammad Ali Mousavi1, Mehdi Safari1,* and Mohammad Gohari1
1Arak University of Technology, Arak 3818141167, Iran.
*Corresponding Author: Mehdi Safari. Email: m.safari@arakut.ac.ir.
Abstract: Laser bending is a flexible forming process that needs no hard tooling or external forces [1]. It is used in
some industrial applications such as shipbuilding, automobile, microelectronics and aerospace [2-4]. In this paper,
artificial neural network (ANN) is used for prediction of the effects of laser and sheet parameters on bending angle
of laser bent tailor welded blanks. In addition, the ANN is compared with multivariate regression in prediction of the
effects of process parameters on bending angle. The input parameters are laser power, irradiating speed and laser
diameter. The results show that the correlation coefficient of the neural network is 0.9999 and the average square
error of the results is 0.12666 % which these results indicate the high accuracy of the predicted model by ANN. In
addition, the results of predicted model by multivariate regression are 0.9961 and 0.6738% for correlation
coefficient and average square error respectively. Comparison of obtained results by artificial neural network and
multivariate regression show that the accuracy of the predicted model by neural network is very high and the error
rate is very low in comparison with multivariate regression.
Keywords: Laser bending process, Bending angle, Artificial neural network (ANN), Multivariate regression.
References
1. Safari, M., & Joudaki, J. (2018). Prediction of Bending Angle for Laser Forming of Tailor Machined Blanks by
Neural Network. Iranian Journal of Materials forming, 5 (1), 47-57.
2. Safari, M., Farzin, M., & Mostaan, H. (2017). A Novel Method for Laser Forming of Two-Step Bending of a
Dome Shaped Part. Iranian Journal of Materials forming, 4 (2), 1-14.
3. Safari, M., & Mostaan, H. (2016). Experimental and Numerical Investigation of Laser Forming of Cylindrical
Surfaces with Arbitrary Radius of Curvature. Alexandria Engineering Journal, 55 (3), 1941-1949.
4. Safari, M., & Farzin, M. (2015). Experimental Investigation of Laser Forming of a Saddle Shape with Spiral
Irradiating Scheme. Optics & Laser Technology, 66, 146-150.

Multi-objective optimization of perforated variable-stiffness plates using

NSGA-II
Saeid Nikbakht1,*
*Corresponding Author: Saeid Nikbakht. Email: Saeid.Nikbakht69@gmail.com.
Abstract: Variable stiffness composites are a new type of composite materials in which the fibers follow particular
curves. As the stiffness of these structures could be controlled by changing the curves of the fibers, their strength
under various types of loadings is higher than constant stiffness composites. This is more observable when the
laminates contain holes since the stress concentration in the vicinity of the cutouts in variable stiffness composites is
considerably smaller than constant stiffness composite structures. This paper represents the optimization of variable
stiffness composite plates containing a central cutout. The fiber curves are considered as the main design variable
while increasing the maximum failure load and decreasing the weight of the laminates are taken as the objective
functions. The stress analysis of these plates is conducted using Layer-wise finite element method. Non-dominated
Sorting Genetic Algorithm (NSGA-II) which has been proved to be a robust meta-heuristic algorithm is employed as
the optimizer in this study. Several examples of rectangular variable stiffness composite plates with circular and oval
cutouts are presented to show the influence of the shape of holes on their load bearing capacity.
Keywords: Variable-stiffness composite plates, NSGA-II, Layer-wise theory.
References
1. Akmar, A. I., Kramer, O., & Rabczuk, T. (2018). Probabilistic multi-scale optimization of hybrid laminated
composites. Composite Structures, 184, 1111-1125.
2. Ghasemi, H., Kerfriden, P., Bordas, S. P., Muthu, J., Zi, G., & Rabczuk, T. (2015). Probabilistic
multiconstraints optimization of cooling channels in ceramic matrix composites. Composites Part B:
Engineering, 81, 107-119.
3. Ghasemi, H., Rafiee, R., Zhuang, X., Muthu, J., & Rabczuk, T. (2014). Uncertainties propagation in
metamodel-based probabilistic optimization of CNT/polymer composite structure using stochastic multi-scale
modeling. Computational Materials Science, 85, 295-305.
4. Nikbakt, S., Kamarian, S., & Shakeri, M. (2018). A review on optimization of composite structures Part I:
Laminated composites. Composite Structures, 195, 158-185.
5. Ghasemi, H., Brighenti, R., Zhuang, X., Muthu, J., & Rabczuk, T. (2014). Optimization of fiber distribution in
fiber reinforced composite by using NURBS functions. Computational Materials Science, 83, 463-473.
6. Ghasemi, H., Brighenti, R., Zhuang, X., Muthu, J., & Rabczuk, T. (2015). Optimal fiber content and
distribution in fiber-reinforced solids using a reliability and NURBS based sequential optimization approach.
Structural and Multidisciplinary Optimization, 51(1), 99-112.
7. Lopes, C. S., Gürdal, Z., & Camanho, P. P. (2008). Variable-stiffness composite panels: Buckling and first-ply
failure improvements over straight-fibre laminates. Computers & Structures, 86(9), 897-907.
8. Alhajahmad, A., Abdalla, M. M., & Gürdal, Z. (2010). Optimal design of tow-placed fuselage panels for
maximum strength with buckling considerations. Journal of Aircraft, 47(3), 775-782.
9. Setoodeh, S., Abdalla, M. M., IJsselmuiden, S. T., & Gürdal, Z. (2009). Design of variable-stiffness
composite panels for maximum buckling load. Composite structures, 87(1), 109-117.
10. Van Campen, J. M., Kassapoglou, C., & Gürdal, Z. (2012). Generating realistic laminate fiber angle
distributions for optimal variable stiffness laminates. Composites Part B: Engineering, 43(2), 354-360.
11. Hao, P., Liu, C., Yuan, X., Wang, B., Li, G., Zhu, T., & Niu, F. (2017). Buckling optimization of variable-

104 DACOMA-2019
stiffness composite panels based on flow field function. Composite Structures, 181, 240-255.
12. Soremekun, G., Gürdal, Z., Haftka, R. T., & Watson, L. T. (2001). Composite laminate design optimization by
genetic algorithm with generalized elitist selection. Computers & structures, 79(2), 131-143.
13. Wu, Z., Raju, G., & Weaver, P. M. (2015). Framework for the buckling optimization of variable-angle tow
composite plates. AIAA Journal, 53(12), 3788-3804.
14. Wu, Z., Weaver, P. M., Raju, G., & Kim, B. C. (2012). Buckling analysis and optimisation of variable angle
tow composite plates. Thin-walled structures, 60, 163-172.
15. Wu, Z., Weaver, P. M., & Raju, G. (2013). Postbuckling optimisation of variable angle tow composite plates.
Composite Structures, 103, 34-42.
16. Ijsselmuiden, S. T., Abdalla, M. M., & Gürdal, Z. (2010). Optimization of variable-stiffness panels for
maximum buckling load using lamination parameters. AIAA journal, 48(1), 134-143.
17. IJsselmuiden, S. T., Abdalla, M. M., & Gürdal, Z. (2010). Thermomechanical design optimization of variable
stiffness composite panels for buckling. Journal of Thermal Stresses, 33(10), 977-992.
18. Duran, A. V., Fasanella, N. A., Sundararaghavan, V., & Waas, A. M. (2015). Thermal buckling of composite
plates with spatial varying fiber orientations. Composite structures, 124, 228-235.
19. Legrand, X., Kelly, D., Crosky, A., & Crépin, D. (2006). Optimisation of fibre steering in composite laminates
using a genetic algorithm. Composite structures, 75(1-4), 524-531.
20. Sliseris, J., & Rocens, K. (2013). Optimal design of composite plates with discrete variable stiffness.
Composite Structures, 98, 15-23.
21. Khani, A., Abdalla, M. M., Gürdal, Z., Sinke, J., Buitenhuis, A., & Van Tooren, M. J. L. (2017). Design,
manufacturing and testing of a fibre steered panel with a large cut-out. Composite Structures, 180, 821-830.
22. Dodwell, T. J., Butler, R., & Rhead, A. T. (2016). Optimum fiber steering of composite plates for buckling
and manufacturability. AIAA Journal, 54(3), 1146-1149.
23. Ghasemi, H., Kerfriden, P., Bordas, S. P., Muthu, J., Zi, G., & Rabczuk, T. (2015). Interfacial shear stress
optimization in sandwich beams with polymeric core using non-uniform distribution of reinforcing
ingredients. Composite Structures, 120, 221-230.
24. Huang, G., Wang, H., & Li, G. (2016). An efficient reanalysis assisted optimization for variable-stiffness
composite design by using path functions. Composite Structures, 153, 409-420.
25. Setoodeh, S., Abdalla, M. M., & Gürdal, Z. (2006). Design of variable–stiffness laminates using lamination
parameters. Composites Part B: Engineering, 37(4-5), 301-309.
26. Tabakov, P. Y., & Walker, M. (2010). A technique for stiffness improvement by optimization of fiber steering
in composite plates. Applied Composite Materials, 17(5), 453-461.
Investigation of nonlinear self-healing theory in continuum damage-healing

mechanics
Chahmi Oucif1,* and George Z Voyiadjis2
1 Institute of Structural Mechanics (ISM), Bauhaus-Universität Weimar, Marienstraße 15, D-99423 Weimar, Germany.
2 Department of Civil and Environmental Engineering Louisiana State University, Baton Rouge, LA 70803, USA.
*Corresponding Author: Chahmi Oucif. Email: chahmi.oucif@uni-weimar.de
Abstract: The theoretical investigation of self-healing variables is presented here. Special focus is addressed to the
nonlinear self-healing theory. New healing variables are defined based on elastic modulus, shear modulus, Poisson’s
ratio and bulk modulus. These variables are defined such that they fit the upper and lower limits of continuum
damage-healing mechanics and the generalized nonlinear and quadratic self-healing models. The formulation is
applied for both isotropic and anisotropic materials using scalar and tensor variables, respectively. The hypotheses
of elastic strain equivalence and elastic energy equivalence are used to elucidate the evolution of the healing variable
calculated based on cross-section as function of the healing variable calculated based on the elastic stiffness. In
addition, examples of isotropic elasticity, plane stress and plane strain are applied using the nonlinear healing
models along with the calculation of the components of the fourth-rank healing tensor is each example.
Keywords: Damage, Healing, Nonlinear, Quadratic, Anisotropy.
Acknowledgement
The first author would like to acknowledge the Deutscher Akademischer Austauschdienst (DAAD) for the financial
support of this work.
References
1. Oucif, C., Voyiadjis, G, Z., Rabczuk, T. (2018). Modeling of damage-healing and nonlinear self-healing
concrete behavior: Application to coupled and uncoupled self-healing mechanisms. Theoretical and Applied
Fracture Mechanics, 96: 216-230.
2. Voyiadjis, G, Z., Kattan, P, T. (2009). A Comparative Study of Damage Variables in Continuum Damage
Mechanics. International Journal of Damage Mechanics, 8: 1-26.
3. Barbero, E, J., Greco, F., Lonetti, P. (2005). Continuum damage-healing mechanics with application to self
healing composites. International Journal of Damage Mechanics, 14: 51-81.

Application of an adaptive neuro fuzzy inference system for modelling of

temperature and force in robotic bone drilling process
Vahid Tahmasbi1,*, Amir Hossein Rabiei1 and Mehdi Safari1
1Arak University of Technology, Arak, Iran.
*Corresponding Author: Vahid tahmasbi. Email: Tahmasbi@arakut.ac.ir.
Abstract: The bone drilling is one of the most important, common in the field of biomedical engineering. It is also
very important in dentistry and bone sampling operations. Bone is a very complex material and the process of
drilling is very sensitive. Thus, bone drilling process is indispensable in orthopaedic surgeries and treating bone
breakages. The bone drilling process can be promoted using automatic drilling machines and surgery-assisting
robots. The problematic issue during operation is the high increase in drilling process temperature (higher than 47 0C)
which leads to the so-called ‘thermal necrosis’ or cell death, and local burn in bone tissue. Furthermore, imposing
higher forces to bone might yield to breaking or cracking, and consequently causes serious damages in bone. In this
paper, the tool rotational speed, feed ratio and tool diameter were taken into account as process input parameters,
and process temperature and thrust force were taken as output parameters. An adaptive neuro fuzzy inference
system (ANFIS) is used for modelling of the effects of process input parameters on temperature and force in bone
drilling. The ANFIS network is learned by 75% of data and 25% of data have been employed for testing the
appropriateness of the models into ANFIS network models and results were compared by two statistical criterions
(R , RMSE). The results show that the proposed model by ANFIS is efficient and valid and it can be used for more
general states.
Keywords: Machining; orthopaedic surgery; adaptive neuro fuzzy inference system (ANFIS).
References
1. Gupta, V., Pandey, P. M., Silberschmidt, V. V. (2017). Rotary ultrasonic bone drilling: Improved pullout
strength and reduced damage. Medical Engineering & Physics, 41, 1-8.
2. Lee, J., Chavez, C. L., & Park. S, (2018). Parameters affecting mechanical and thermal responses in bone
drilling: A review. Journal of Biomechanics,71, 4-21
3. Feldmann, A., Wandel, J., Zysset, P. (2016). Reducing temperature elevation of robotic bone drilling.
Medical Engineering & Physics, 38(12), 1495-1504.

Modelling of creep age forming of aluminum 7075 tailor machined blanks

using adaptive neuro fuzzy inference system (ANFIS)
Mehdi Safari1,*, Vahid tahmasbi1 and Amir Hossein Rabiee1
1Arak University of Technology, 3818141167 Arak, Iran.
*Corresponding Author: Mehdi Safari. Email: m.safari@arakut.ac.ir.
Abstract: In this paper, an adaptive neuro fuzzy inference system (ANFIS) is used for modelling of the effects of
creep age forming of aluminum 7075 tailor machined blanks parameters on characteristics of deformed blank. Creep
age forming of tailor machined blanks is more difficult than monolithic plates due to presence of two different
thicknesses in theses blanks [1-4]. Therefore, prediction of forming behavior of tailor machined blanks with creep
age forming process with a powerful tool is important especially in the aerospace industries. In the present work, the
effects of two main creep forming process parameters namely time and temperature are investigated on spring-back
of both thin and thick sections of a tailor machined blank with adaptive neuro fuzzy inference system (ANFIS). The
ANFIS network is instructed by 75% of data. Twenty-five percent of primary data which is considered for testing
the appropriateness of the models is entered into ANFIS network models and results were compared by two
statistical criterions (R, RMSE). Considering the results, it is obvious that the proposed modeling by ANFIS is
efficient and valid and it can be expanded for more general states.
Keywords: Creep age forming process, Al 7075 tailor machined blank, adaptive neuro fuzzy inference system
(ANFIS).
References
1. Safari, M., & Joudaki, J. (2018). Prediction of Bending Angle for Laser Forming of Tailor Machined Blanks by
Neural Network. Iranian Journal of Materials forming, 5 (1), 47-57.
2. Safari, M., Mostaan, H., & Farzin, M. (2016). Laser Bending of Tailor Machined Blanks: Effect of Start Point
of Scan Path and Irradiation Direction Relation to Step of the Blank. Alexandria Engineering Journal, 55 (2),
1587-1594.
3. Safari, M., & Farzin, M. (2014). A Study on Laser Bending of Tailor Machined Blanks with Various
Irradiating Schemes. Journal of Materials Processing Technology, 214 (1), 112-122.
4. Safari, M., Farzin, M., & Ghaei, A. (2013). Investigation into the Effects of Process Parameters on Bending
Angle in the Laser bending of Tailor Machined Blanks based on a Statistical Analysis. Journal of Laser
Applications, 25 (5), 1-10.
5. Safari, M., Farzin, M. (2013). Experimental and Numerical Investigation of Laser Bending of Tailor Machined
Blanks. Optics & Laser Technology, 48, 513-522.

Isogeometric Analysis for Phase-Field Models of High-Order Surface PDEs

with Application to biomembrane modeling
Navid Valizadeh1 *
1 Bauhaus University Weimar, Institute of Structural Mechanics, Marienstraße 15, 99423 Weimar, Germany.
*Corresponding Author: Navid Valizadeh. Email: navid.valizadeh@uni-weimar.de
Abstract: We present a NURBS-based isogeometric analysis for phase-field models of high-order surface partial
differential equations (PDEs) [1]. We first consider mean curvature flow and Willmore flow problems, which are
prime examples of geometric PDEs, and discuss their corresponding phase-field approximations. Based on the
method of manufactured solutions, we study the convergence behavior of isogeometric formulation for the phase-
field models of geometric PDEs. The numerical convergence of these phase-field approximations to the sharp-
interface solutions is also discussed. Next, we solve the Cahn-Hilliard equation on a unit sphere as a model problem
for high-order PDEs on stationary surfaces, where we model the surface using a phase-field approach. Finally, we
consider a phase-field model of multi-component biomembranes which represents a model problem for high-order
coupled surface PDEs. We perform several numerical tests in 2D, 3D and axisymmetric 3D settings and make
comparisons with available numerical or experimental results. Our simulation results show the robustness of IGA
for solving the considered phase-field models.
Keywords: Isogeometric analysis, phase-field modeling, high-order surface PDEs, geometric PDEs, multi-
component vesicles
Acknowledgement
This research was supported by the German Research Foundation (DFG), project number 405890576. This support
is gratefully acknowledged.
References
1. Valizadeh, N., Rabczuk, T. (2019). Isogeometric analysis for phase-field models of geometric PDEs and high-
order PDEs on stationary and evolving surfaces, Computer Methods in Applied Mechanics and Engineering,
351, 599-642.

Influence of Thermostatting on Nonequilibrium Molecular Dynamics

Simulations of Heat Transport in Solids
Zhen Li1, Shiyun Xiong2, ∗, Charles Sievers3, Yue Hu4, Zheyong Fan5, 6,*, Ning Wei1, Hua Bao7,*,
Shunda Chen3, Davide Donadio3, * and Tapio Ala-Nissila6, 8
1
Jiangsu Key Laboratory of Advanced Food Manufacturing Equipment and Technology, Jiangnan University, 214122, Wuxi,
China.
2
Functional Nano and Soft Materials Laboratory (FUNSOM) and Collaborative Innovation Center of Suzhou Nano Science and
Technology, Soochow University, 215123 Suzhou, P.R.China.
3
Department of Chemistry, University of California at Davis, One Shields Avenue, Davis, California 95616, USA.
4
University of Michigan-Shanghai Jiao Tong University Joint Institute, Shanghai Jiao Tong University, Shanghai, China, 200240.
5School of Mathematics and Physics, Bohai University, Jinzhou, China.
6
QTF Centre of Excellence, Department of Applied Physics, Aalto University, FI-00076 Aalto, Finland.
7University of Michigan-Shanghai Jiao Tong University Joint Institute, Shanghai Jiao Tong University, Shanghai, China,200240.
8
Centre for Interdisciplinary Mathematical Modeling and Department of Mathematical Sciences, Loughborough University,
Loughborough, Leicestershire LE11 3TU, UK.
*Corresponding Author: Ning Wei. Email: nwei@nwsuaf.edu.cn. Zheyong Fan. Email: brucenju@gmail.com. Shiyun Xiong.
Email: s.y.xiong216@gmail.com. Hua Bao. Email: hua.bao@sjtu.edu.cn. Davide Donadio. Email: ddonadio@ucdavis.edu.
Abstract: Nonequilibrium molecular dynamics (NEMD) has been extensively used to study thermal transport at
various length scales in many materials. In this method, two local thermostats at different temperatures are used to
generate a nonequilibrium steady state with a constant heat flux. Conventionally, the thermal conductivity of a finite
system is calculated as the ratio between the heat flux and the temperature gradient extracted from the linear part of
the temperature profile away from the local thermostats. Here we show that, with a proper choice of the thermostat,
the nonlinear part of the temperature profile should not be excluded in thermal transport calculations [1,2]. We
compare NEMD results against those from the atomistic Green’s function [3,4] method in the ballistic regime, and
those from the homogeneous nonequilibrium molecular dynamics method in the ballistic-to-diffusive regime. These
comparisons suggest that in all the transport regimes, one should directly calculate the thermal conductance from the
temperature difference between the heat source and sink and, if needed, convert it to the thermal conductivity by
multiplying it with the system length. Furthermore, we find that the Langevin thermostat outperforms the Nosé-
Hoover chain thermostat in NEMD simulations because of its stochastic and local nature. We show that this is
particularly important for studying asymmetric carbon-based nanostructures, for which the Nosé-Hoover thermostat
can produce artifacts leading to unphysical thermal rectification. Our findings are important to obtain correct results
from MD simulations of nanoscale heat transport as the accuracy of the interatomic potentials is rapidly improving.
Keywords: Nonequilibrium molecular dynamics, Thermal transport, Local thermostats, Temperature gradient.
Acknowledgement
ZL and NW were supported by the National Natural Science Foundation of China (Grant No 11502217). SX
acknowledges the support from the National Natural Science Foundation of China (Grant No. 11804242) and the
Jiangsu Provincial Natural Science Foundation (Grant No. BK20160308). YH and HB acknowledge the support
from the National Natural Science Foundation of China (Grant No. 51676121). ZF and TA acknowledge the support
from the Academy of Finland Centre of Excellence program QTF (Project 312298) and the computational resources
provided by Aalto Science-IT project and Finland’s IT Center for Science (CSC).

110 DACOMA-2019
References
1. Cepellotti, A., & Marzari, N. (2017). Boltzmann transport in nanostructures as a friction effect. Nano Letters,
acs.nanolett. 7b01202.
2. Allen, & Philip, B. (2014). Size effects in thermal conduction by phonons. Physical Review B, 90(5), 054301.
3. Mingo, N., & Yang, L. (2003). Phonon transport in nanowires coated with an amorphous material: an atomistic
green’s function approach. Physical Review B, 68(24), 245406.
4. Wang, J. S., Wang, J., & J. T. Lü. (2008). Quantum thermal transport in nanostructures. The European
Physical Journal B - Condensed Matter, 62(4), 381-404.
Crack Propagation with Different Radius Local Random Damage based on

peridynamic theory
Jinhai Zhao1,2,* and Meng Jia2,3
1
Shanghai Baoye Group Corp., Ltd, Shanghai, China.
2
Civil Engineering, Tongji University, Shanghai China.
3
Tongji Architectural Design (Group) Co., Ltd, Shanghai ,China.
*
Corresponding Author: Jinhai Zhao. Email: zhaojinhai87@126.com.
Abstract: Inheriting the advantages of traditional theories, the theory of peridynamics can well clarify the
mechanism of crack propagation by the integral solution. Moreover, it is not necessary to set the initial crack and
crack path in advance. In this paper, specimens with bilateral cracks are tested using the theory of peridynamics in
different conditions, namely the specimen without local damage, that with small radius local damage, and that with
large radius local damage. The crack propagation path under the uniaxial tension and the displacement in the Y-
direction are compared and analysed, in order to figure out the influence of local stochastic damage with different
radiuses on the crack propagation path and Y-direction displacement.
The calculating example 2, the crack propagation path in picture 9 as below
0.03 0.03 0.03
0.02 0.02 0.02
0.01 0.01 0.0 0.01

length(m)
0.0 0.0
length(m)
length(m)
0.094 0.12 0.13

0.19 0.24 0.25
0.28 0.35 0.38
0.00 0.38 0.00 0.47 0.00 0.50
0.47 0.59 0.63
0.56 0.71 0.75
0.66 -0.01 0.82 -0.01 0.88
-0.01 0.75 0.94 1.0
-0.02 -0.02 -0.02
-0.03 -0.03 -0.03
-0.02 -0.01 0.00 0.01 0.02 -0.02 -0.01 0.00 0.01 0.02 -0.02 -0.01 0.00 0.01 0.02
width(m) width(m) width(m)
(a) (b) (c)
Figure 9: Crack propagation path of specimen 10-00 with r=2 mm
Keywords: Fracture mechanics, peridynamics, stochastic damage, bilateral crack.
References
1 Suresh, S. (1999).Material fatigue. Beijing: National Defense Industry Press1.
2 Griffith, A.A. (1921).The phenomena of rupture and flow in solids. Philosophical transactions of the
royal society of london. Series A, containing papers of a mathematical or physical character,
221(2),163-198.
3 Paris P, Erdogan F.(1963). A critical analysis of crack propagation laws. Journal of basic engineering,
85(4), 528-533.
4 Rice, J.R. (1968).A path independent integral and the approximate analysis of strain concentration by
notches and cracks. Journal of applied mechanics, 35(2),379-386.
5 Huanding Wang, Wei Wang.(2003). Finite Element Method. Harbin: Harbin Institute of Technology
Press.
6 Gonzalez-Herrera A, Zapatero J.(2008).Tri-dimensional numerical modelling of plasticity induced
fatigue crack closure. Engineering Fracture Mechanics, 75(15),4513-4528.
7 McClung R C, Sehitoglu H. (1989).On the finite element analysis of fatigue crack closure-1. Basic
modeling issues. Engineering Fracture Mechanics, 33(2), 237-252.
8 Belytschko T, Black T.(1999). Elastic crack growth in finite elements with minimal remeshing.
International journal for numerical methods in engineering, 45(5), 601-620.

112 DACOMA-2019
9 Daux C, Moës N, Dolbow J, et al.(2000).Arbitrary branched and intersecting cracks with the extended
finite element method. International journal for numerical methods in engineering, 48(12),1741-1760.
10 Fang X, Jin F, Wang J. (2007).Cohesive crack model based on extended finite element method.
Journal of Tsinghua University (Science and Technology), 47(3),344.
11 Xiujun F, Feng J I N, Jinting W.(2008).Seismic fracture simulation of the Koyna gravity dam using an
extended finite element method. Journal of Tsinghua University (Science and Technology), 12, 2065-
2069.
12 Zhuang Z, Cheng B B. (2011).Equilibrium state of mode-I sub-interfacial crack growth in bi-materials.
International journal of fracture, 170,27-36.
13 Silling S A, Askari E. (2005).A meshfree method based on the peridynamic model of solid mechanics.
Computers & structures, 83(17-18),1526-1535.
14 Silling S A, Bobaru F. (2005). Peridynamic modeling of membranes and fibers. International Journal
of Non-Linear Mechanics, 40(2-3),395-409.
15 Ayatollahi M R, Aliha M R M.(2009). Analysis of a new specimen for mixed mode fracture tests on
brittle materials. Engineering Fracture Mechanics, 76(11),1563-1573.
16 Foster J T , Silling S A , Chen W.(2015). An energy based failure criterion for use with peridynamic
states. International Journal for Multiscale Computational Engineering, 9, 675-688.
17 Zeng S. (2014).Bilateral crack growth behavior of Q345 steel under uniaxial tensile load. Guangxi
University.
18 Feng S, Zhang Q, Huang D.(2012).Peridynamics modeling of failure process of concrete structure
subjected to impact loading. Engineering Mechanics, 29, 12-15.
19 Zhang Qing, Gu Xin, Yu Yangtian.(2016). Peridynamics Simulation For Dynamic Response Of
Granular Materials Under Impact Loading. Chinese Journal of Theoretical & Applied Mechanics,
48(1), 56-63.
20 Gu X, Zhou X. (2017).Numerical simulation of propagation and coalescence of cracks using
peridynamic theory. Rock and Soil Mechanics, 38,610-616.
21 Ren H, Zhuang X, Cai Y, et al.(2016). Dual-horizon peridynamics. International Journal for Numerical
Methods in Engineering, 108,1451-1476.
22 Silani M, Talebi H, Hamouda A M, et al.( 2016).Nonlocal damage modelling in clay/epoxy
nanocomposites using a multiscale approach. Journal of Computational Science, 15,18-23.
23 Talebi H, Silani M, Rabczuk T.(2015). Concurrent multiscale modeling of three dimensional crack and
dislocation propagation. Advances in Engineering Software, 80, 82-92.
24 Amiri F, Millán D, Arroyo M, et al.(2016). Fourth order phase-field model for local max-ent
approximants applied to crack propagation. Computer Methods in Applied Mechanics and
Engineering, 312,254-275.
25 Amiri F, Millán D, Shen Y, et al.(2014).Phase-field modeling of fracture in linear thin shells.
Theoretical and Applied Fracture Mechanics, 69, 102-109.
26 Areias P, Rabczuk T, de Sá J C.( 2016). A novel two-stage discrete crack method based on the
screened Poisson equation and local mesh refinement. Computational Mechanics, 58(6),1003-1018.
27 Thai T Q, Rabczuk T, Bazilevs Y, et al.(2016). A higher-order stress-based gradient-enhanced damage
model based on isogeometric analysis. Computer Methods in Applied Mechanics and Engineering,
304,584-604.
28 Budarapu P R, Gracie R, Bordas S P A, et al.( 2014). An adaptive multiscale method for quasi-static
crack growth. Computational Mechanics, 53(6),1129-1148.
29 Budarapu P R, Gracie R, Yang S W, et al.(2014). Efficient coarse graining in multiscale modeling of
fracture. Theoretical and Applied Fracture Mechanics, 69,126-143.
Topology optimization of piezoelectric materials

Chuong Nguyen1 and Xiaoying Zhuang1,*
1
Institute of Continuum Mechanics, Leibniz Universität Hannover, Hannover, Germany.
*
Corresponding Author: Xiaoying Zhuang. Email: zhuang@ikm.uni-hannover.de
Abstract: Piezoelectric materials are characterized by the electromechanical coupling phenomenon, i.e., able to
produce an electric field when being stressed, and deform when subjected to an electric field. Such materials are
found in various engineering application such as sensors, actuators and transducers. Auxetic materials, i.e., materials
have negative Poissons ratio, have show potential applications from biomedical to defense problems. Kinematic
deformation are affected by the Poissons ratio can enhance the performance of piezoelectric transducer. Under a
compressive load on surface of the piezocomposites device, the auxetic polymer matrix contracts laterally and allow
the ceramic rods expand, resulting in enhanced electromechanical coupling device [1].
In this work, we focus on design of piezocomposites by mean of topology optimization based on the parameterized
level set method [2, 3]. Level set funciton implicitly represent the design boundaries as the zero level set of the one-
higher-dimension level set function, directly evolves the design boundaries during the topology optimization and
ensures the optimized designs have clear boundaries and avoid numerical problems such as checkerboard patterns,
gray-scale elements. The charge coefficients or electromechanical coupling factor are optimized for the hydrophone
application. The matrix material are optimized in order to have auxetic behavior. Figure 1 shows the three
dimensional unit cell which are optimized with the level set method. The next steps is to embed the material
structure in to classes of 1-3 and 2-2 piezocomposites, we expect the performance piezoelectric devices are
improved.
Keywords: Piezoelectric materials, Auxetic materials, Topology optimization
References
1. Smith W. A. Optimizing electromechanical coupling in piezocomposites using polymers with negative
Poisson’s ratio. In IEEE 1991 Ultrasonics Symposium, pages 661-666 vol.1, dec 1991.
2. Ghasemi, Hamid, Harold S. Park, and Timon Rabczuk. "A multi-material level set-based topology optimization
of flexoelectric composites."Computer Methods in Applied Mechanics and Engineering 332 (2018): 47-62.
3. Nanthakumar, S. S., et al. "Topology optimization of flexoelectric structures."Journal of the Mechanics and
Physics of Solids 105 (2017): 217-234.

A prediction model of debris flow susceptibility in Southwest China based on

a quantitative data processing method
Feng Ji1 and Zili Dai2, *
1
State Key Laboratory of Geohazard Prevention and Geoenvironment Protection, Chengdu University of Technology, Chengdu
610059, China.
2
Department of Earth Science, Shimane University, 1060 Nishikawatsu-cho, Matsue, Shimane 690-8504, Japan.
*
Corresponding Author: Zili Dai. Email: zili.dai@riko.shimane-u.ac.jp
Abstract: Southwest China is charactered by steep mountains and deep valleys, and is strongly affected by the uplift
activity of the Qinghai-Tibet Plateau. Furthermore, there is an abundance of loose material and rainfall, making it a
severe disaster zone in terms of debris flow. Evaluation on the debris flow susceptibility requires the quantitative
assessment of the factors leading to the initiation, propagation and deposition of debris flows. In the existing
literatures [1-3], however, the debris flow susceptibility is mainly determined using a qualitative method with
relevant specifications, and little reliable quantitative prediction models were presented. To evaluate the
susceptibility to the debris flow in this area, in total of 70 typical debris flows distributed in five water catchments
(the Brahmaputra River, Nujiang River, Yalong River, Dadu River, and Ming River) were investigated in the past
three years.
In the five water catchments, a detailed field work was carried out from the valley outlet to the valley watershed,
including the investigation of watershed terrain, geological structure, outbreak scale, loose material distribution,
processes of occurrence and movement, frequency of debris flows. A bulk density test (specimen size 50 cm × 50
cm × 50 cm) and screening test of the loose material were conducted in the deposit zone of the debris flow.
According to the superelevation and flow depth of the curved gully zone, the speed of the debris flow was calculated
to provide the basic data for the dynamic parameter calculation. For the active debris flow gullies, the geologic
condition in the source areas was invested in drilling and geophysical prospecting, so that the source material
volume, soil composition, structure, and content of the fine-grained soil could be determined precisely. In addition,
the precipitation data before the outbreak of debris flows were collected from local meteorological bureaus. The
impulse force, the sediment discharge, and other dynamic parameters were calculated.
Considering debris flow features and index-acquisition conditions, 9 indexes were selected to evaluate the
susceptibility of debris flow in this work. These were the catchment area, longitudinal grade, average gradient of
slope on both sides of the gully, catchment morphology, valley slope orientation, loose material reserves, loose
material position, antecedent precipitation, and H1p rainfall intensity. Each factor was classified into certain
categories according to their values. The values of the 9 indexes of the 70 debris flows in Southwest China were
obtained through the field investigation.
In Japan, the Hayashi methods of quantification are well known and widely used in various fields such as coal mine
flooding risk assessment, earthquake-triggered damage assessment of tunnels, and the gas leakage prediction model
of large explosions in hard rocks. According to a reasonable principle, the quantitative and qualitative variables can
be mutually transformed using this method. Therefore, with respect to the processing of the quantitative and
qualitative influencing factors of earthquake-induced hazards, this method has a considerable degree of flexibility
and applicability.
In the quantification theory, qualitative variables are termed items and all possibilities for each item are termed
categories. To express the response of an item and the category for each sample, the dummy variable δi (j, k) is
introduced into the model; here, sample i represents the factor in category k of item j. The matrix composed of all
categories δi (j, k) is referred to as the response matrix X:
æ d1 (1,1)!d1 (1, r1 ) d1 (2,1)!d1 (2, r2 )!d1 (m,1)!d1 (m, rm ) ö
ç ÷ (1)
ç d (1,1)!d 2 (1, r1 ) d 2 (2,1)!d 2 (2, r2 )!d 2 (m,1)!d 2 (m, rm ) ÷
X =ç 2 ÷
" " " " " "
ç ÷
ç d (1,1)!d (1, r ) d (2,1)!d (2, r )!d (m,1)!d (m, r ) ÷
è n n 1 n n 2 n n m ø

115 DACOMA-2019
To establish a quantitative prediction model, we use qualitative and quantitative experimental observations to fit the
linear relationship between the concerned independent variable and the dependent variable affecting its variations.
Quantification theory type I supposes that the random variable changes with the m variables:
m rj
yi = åå d i ( j , k ) b jk + e i (2)
j =1 k =1
where yi denotes the susceptibility of the ith debris flow gully and bjk is a constant coefficient that relies on category
k in item j. Ɛi represents a random error in the ith spot-check and consists of all δi (j, k).
Using Eq. 2 and matrix derivation regression calculation, the contribution values of each item can be
obtained, and the susceptibility prediction model of debris flow is established and is represented as
follows:
Y = 0.573x11 + 0.821x12 + 0.910x13 + 0.875x21 + 0.955x22 + 0.320x23 - 0.107x32 - 0.163x41 + 0.135x42 +
0.213x43 - 0.136x51 - 0.174x52 + 0.246x62 + 0.454x63 - 0.220x71 - 0.161x72 + 0.034x82 + 0.071x83 - 0.038x91 +
0.043x92 (3)
In Eq. 3, Y is the susceptibility for the debris flow. When the predicted value (Y) is less than 1.5, the susceptibility is
low. When Y is greater than or equal to 1.5 but less than 2.5, the susceptibility for the debris flow is medium. When
Y is greater than or equal to 2.5, the susceptibility for the debris flow is high.
The values of each index are obtained from the field investigation, and the susceptibility of the 70 debris
flows can be calculated using the proposed model. Fig. 1 compares the predicted susceptibilities with the
actual ones. It is shown that the coincidence rate is 78.53% for low-susceptibility debris flow valleys,
92.38% for medium-susceptibility debris flow valleys, 82.01% for high-susceptibility debris flow valleys,
and 86.38% for all the samples, which indicates that the regression model can predict the susceptibility
well.
Figure 1: Comparison of measured and predicted values

Keywords: Debris flow, Susceptibility, Prediction model, Factor index system, Quantification theory type I.
Acknowledgement
This work was supported by the National Natural Science Foundation of China (51308082), Independent
Exploration Fund of State Key Laboratory of Geohazard Prevention and Geoenvironment Protection
(SKLGP2011Z009), the Key Fund Project of Sichuan Provincial Department of Education (15ZA0075), the Open
Fund of State Key laboratory of Geohazard Prevention and Geoenvironmental Protection (Chengdu University of
Technology; grant no. SKLGP2017K005).
References
1. Qin, S., Lv, J., Cao, C., Ma, Z., Hu, X., Liu, F., ... & Dou, Q. (2019). Mapping debris flow susceptibility based
on watershed unit and grid cell unit: a comparison study. Geomatics, Natural Hazards and Risk, 10(1): 1648-
1666.
2. Kang, S., & Lee, S. R. (2018). Debris flow susceptibility assessment based on an empirical approach in the
central region of South Korea. Geomorphology, 308: 1-12.
3. Blais-Stevens, A., & Behnia, P. (2016). Debris flow susceptibility mapping using a qualitative heuristic method
and Flow-R along the Yukon Alaska Highway Corridor, Canada. Natural Hazards and Earth System Sciences,
16(2): 449-462.
Prediction of shield moving performance during tunnelling based on

optimized LSTM neural network
Song-Shun Lin1, Shui-Long Shen2, *, Annan Zhou3
1 SJTU-NUS PhD student, Department of Civil Engineering, School of Naval Architecture, Ocean, and Civil
Engineering, Shanghai Jiao Tong University, 800 Dong Chuan Road, Minhang District,
Shanghai 200240, China
2 Department of Civil and Environmental Engineering, College of Engineering, Shantou University,
Shantou, Guangdong 515063, China.

3 Discipline of Civil and Infrastructure, School of Engineering, Royal Melbourne Institute of
Technology (RMIT), Victoria 3001, Australia

*Corresponding Author: Shui-Long Shen. Email: shensl@stu.edu.cn
Abstract: Shield moving performance prediction is of a significant importance to the safety of tunnelling
construction process. The analysis of shield moving performance, such as advance rates and thrust values, can
improve shield tunnelling efficiency. With a rapid development of monitoring techniques, a huge amount of data
including geological and operational data can be collected during the tunnelling construction process. The
analysis and mining of these complex coupled and nonlinear parameters related to shield moving performance are
challenging for project managers and tunnelling engineers. To improve the accuracy of prediction and achieve the
dynamic prediction of advance rates and thrust values of the shield machine, this paper proposes the improved
particle swarm optimization (IPSO) and a IPSO-LSTM hybrid algorithm which integrates the particle swarm
optimization (PSO) with the long-short term memory (LSTM) neural network. The IPSO is utilized to optimize
parameters (geological and operational variables) since it is complex and difficult to determine the parameters of
LSTM, such as the number of neurons, iterations and learning rate. Besides, a nonlinear inertia weight is
provided to improve the convergence speed and global research of PSO. The case study of Shenzhen Metro
project is adopted to validate the proposed model (IPSO-LSTM). Results demonstrated that the developed model
achieves well prediction performance of shield machine during tunnelling when the LSTM has two hidden layers.
This work provides a feasible and accurate tool for shield performance prediction in shield tunnelling projects.
Keywords: Shield tunnelling, Advance rates and thrust values prediction, Geological and operational data, IPSO-
LSTM, Nonlinear inertia weight.
Acknowledgement
The research work described herein was funded by the National Natural Science Foundation of China (NSFC)
(Grant No. 41672259) and the Innovative Research Funding of the Science and Technology Commission of
Shanghai Municipality (Grant No. 18DZ1201102).
References
1. Khanh T.P. N., & Kamal M. (2019). A new dynamic predictive maintenance framework using deep learning
for failure prognostics. Reliability Engineering and System Safety. 188: 251–262.
2. Sundaravelpandian, S., Johan, S., & Philipp, G. (2018). Deep-learning neural-network architectures and
methods: using component-based models in building-design energy prediction. Advanced Engineering
Informatics, 38: 81-90.
3. Tao, J., Sun, G., Wang, X.Y., & Guo, L.Q. (2019). Robust optimization for a wing at drag divergence Mach
number based on an improved PSO algorithm. Aerospace Science and Technology, 92: 653–667.
4. Chen, Y., Zhang, S., Zhang, W.Y., Peng, J.J., & Cai, Y.S. (2019). Multifactor spatio-temporal correlation
model based on a combination of convolutional neural network and long short-term memory neural network
for wind speed forecasting. Energy Conversion and Management, 185: 783-799.
117 DACOMA-2019
5. Zhang, Z.D., Ye, L., Qin, H., & Liu, Y.Q. (2019). Wind speed prediction method using Shared Weight Long
Short-Term Memory Network and Gaussian Process Regression. Applied Energy, 247: 270-284.
Data-driven approach to Inverse Problems using Deep Learning Networks

Ngoc-Hien Nguyen1, Vinh Phu Nguyen2,*, Thi-Hong-Hieu Le3, and Yan Ding1
1Department of Mathematical Sciences, RMIT University, Melbourne, Victoria 3000, Australia.
2Department of Civil Engineering, Monash University, Clayton, Victoria 3800, Australia.
3Department of Aerospace Engineering, Ho Chi Minh City University of Technology, Vietnam.
*Corresponding Author: Vinh Phu Nguyen. Email: phu.nguyen@monash.edu.
Abstract: Inverse source problems such as the determination of the contaminant source’s locations and properties
from measurement data in the field of air/water quality management are challenging and ill-posed due to sparsity
information. Optimization based approaches (e.g., Levenberg–Marquardt algorithm [1]) and statistical based
approaches (e.g., Bayesian inferences [2]) are normally used for such problems. However, they require the
knowledge of mathematical models which govern the physical phenomenon under investigation (aka., a forward
model). Furthermore, solving repeatedly the forward model in an optimization procedure is computationally
expensive. This paper presents a complete framework to perform inverse source problems using deep learning
networks. The framework consists of two Long Short-Term Memory (LSTM) [3] networks to perform two different
tasks. The first network is an LSTM-Classifier to classify the number of contaminant source components, while the
second one is an LSTM-Regressor to predict the sources’ properties, namely location, width, and strength of the
sources. The LSTM networks are chosen in the context of sensor measurement, where recorded data is time-variant
responses in the sense of continuous or discontinuous sequences. An experimental study based on a two-dimensional
water surface model is conducted and an extensive study of hyper-parameters (e.g. number of hidden layers,
learning rate, etc.) is also performed. The results show that the proposed framework is very efficient in terms of
computational time and prediction accuracy and expected to function as a tool of studying inverse source problems
for water/air quality management as well as similar applications.
Keywords: Inverse source problems, Deep learning networks, LSTMs, Classification, Regression.
References
[1] Hoang K.C., Khoo B.C., Liu G.R., Nguyen N.C. & Patera A.T. (2013). Rapid identification of material
properties of the interface tissue in dental implant systems using reduced basis method, Inverse Problems in Science
and Engineering, 21(8): 1310-1334.
[2] Nguyen N.H., Khoo B.C., & Willcox K. (2014). Model order reduction for Bayesian approach to inverse
problems, Asia Pacific Journal on Computational Engineering, 1(2): 1-17.
[3] Hochreiter S., & Schmidhuber J. (1997). Long Short-Term Memory. Neural Comput. 9(8): 1735-1780.

Data-driven methods for quantifying the effects of geological variables on

performance degradation of metro tunnel in operation
Xin Wang1, * and Hehua Zhu1
1
*
Corresponding Author: Xin Wang. Email: wxjames@tongji.edu.cn
Abstract: The construction speed of the urban rail transit is increasing in recent years [1, 2], leading to the problem
that the long-term serviceability of shield tunnel is often ignored. In order to avoid the irreversible destruction
caused by structure deterioration, it is necessary to have a study on serviceability degradation model of shield tunnel
based on its performance assessment, thus providing guidance for daily maintenance. Therefore, this paper
conducted clustering analysis to identify different degradation models of shield tunnel. On this basis, the influence
of geological variables on degradation model was quantitatively evaluated using statistical learning methods. These
data-driven methods were applied to 5 tunnel intervals of Shanghai ML 1 and 2.
Specifically, an existing assessment indicator, Tunnel Serviceability Index (TSI), was employed to evaluate the
condition of shield tunnels in soft soils in this study [3]. Currently, TSI is the official index to evaluate the tunnel
serviceability according to Shanghai Engineering Technical Standard. The value range of TSI is 0 to 5, where 0
indicates the best condition and 5 is the worst condition. The final formula of TSI was shown in Eq. (1).
TSI = 0.16 save + 0.01sdiff ,ave + 0.09cave + 0.08dl + 0.05d c + 0.50d s - 0.23 (1)
Since the cumulative settlement of ML 1 and 2 is observed twice a year while other measurable variables are not
monitored as frequently as cumulative settlement, TSIs which is merely related to settlement factors instead of TSI is
calculated and adopted for further analysis. 5 metro intervals including 196 tunnel sections are investigated. Each
section includes 50 segment rings and is 60 m long. The length of TSIs time series for ML 1 and 2 is 32 and 26,
separately. Fig. 1 shows the observations of TSIs time series. It can be seen that with the increase of service age,
TSIs increases and the condition of tunnel sections becomes worse.
1.4
TSIs time series
1.2 Mean vaule of ML1
Mean value of ML2
1.0
0.8
TSIs
0.6
0.4
0.2
0.0
0 2 4 6 8 10 12 14 16
Age (year)
Fig. 1. The observations of TSIs time series

Generally, TSIs changes dynamically over time. Thus, the identification of performance degradation model should
consider the change of TSIs in the whole historical period, which can be achieved by time series clustering analysis.
In view of the low dimensionality of TSIs time series, 196 sets of TSIs time series data are processed and analyzed
by the k-means algorithm based on DTW distance. Compared to Euclidean distance, DTW can find out the distance
which has the minimal warping cost under different conditions and allows time series data being compressed or
stretched [4], which is suitable for modeling clustering problems when the length of time series data is different. The
clustering quantity parameter K is determined and optimized by ten-fold cross method by comparing Mean Squared
Distortion (MSD) (Fig. 2). Finding the “elbow” where the MSD stops decreasing significantly is one of the criteria
employed to choose the optimum number of components [5]. The results show that the clustering analysis can
identify 4 various clusters objectively, each of which shows similar trends and characteristics (Fig. 3).
Next, five independent variables, namely, average depth of burial DEPTH, thickness of soil passing through the
third layer SOIL3, thickness of soil passing through the forth layer SOIL4, the river length RIVER and the station
length STATION are selected as geological variables according to geological prospecting data. After Z-score

120 DACOMA-2019
standardization method, the influence of these five variables on the performance degradation model of shield tunnels
is studied quantitatively by using the Ordinal Logistic Regression Model. The results show that the statistical
influences of these five independent variables are significant and the regression parameters are estimated to be -
0.866, -0.655, -0.799, -0.594 and -0.332, respectively (Tab. 1). These parameters indicate that with the increase of
DEPTH, SOIL3, SOIL4, RIVER and STATION, the probability of tunnel sections which stay in worse performance
degradation model will increase. The practical significance of the five independent parameters is explained in
combination with mechanical concepts and existing empirical knowledge.
0.27 0.8
Test 1 Cluster 1
0.24 Test 2 Cluster 2
Test 3 Cluster 3
0.21 Test 4 0.6 Cluster 4
Test 5
Test 6
0.18
Test 7
MSD
Test 8
TSIs
0.15 Test 9 0.4
Test 10
0.12
0.09 0.2
0.06
0.03 0.0
2 3 4 5 6 7 8 9 0 2 4 6 8 10 12 14 16
K Age (year)
Fig. 2. Mean squared distortion of ten sets for k-means Fig. 3. Mean TSIs time series of the finial 4 clusters
Tab. 1. Parameter estimation results of Ordinal Logistic Regression method
Parameter 95% Confidence Interval
Variables Standard Error OR P-value
Estimation Lower Upper
DEPTH -0.866 0.248 0.421 0.000 -1.352 -0.379
SOIL3 -0.655 0.233 0.519 0.005 -1.112 -0.198
SOIL4 -0.799 0.253 0.450 0.002 -1.296 -0.303
RIVER -0.594 0.193 0.552 0.002 -0.973 -0.216
STATION -0.332 0.144 0.717 0.021 -0.613 -0.050
Besides, Tab. 2 shows the results of parallel line test. P-value equals 0.060 which is bigger than 0.05, indicating the
regression equations are parallel to each other, thus validating the availability of Ordinal Logistic process.
Tab. 2. Results of parallel line test
Model -2 Log Likelihood Chi-square df P-value
Zero Hypothesis 482.164 — — —

Generalized 464.430 17.734 10 0.060
Keywords: Tunnel Serviceability Index (TSI), Degradation model, Clustering analysis, Ordinal Logistic Regression,
Geological variables.
Acknowledgement
This study was funded by National Natural Science Foundation (51478341).
References
1. Wang, S., Jiang, X., & Bai, Y. (2019). The influence of hand hole on the ultimate strength and crack pattern of
shield tunnel segment joints by scaled model test. Frontiers of Structural and Civil Engineering, 1-14.
2. Huang, Q., Huang, H., Ye, B., Zhang, D., & Zhang, F. (2018). Evaluation of train-induced settlement for metro
tunnel in saturated clay based on an elastoplastic constitutive model. Underground Space, 3(2): 109-124.
3. Li, X., Lin, X., Zhu, H., Wang, X., & Liu, Z. (2017). Condition assessment of shield tunnel using a new
indicator: The tunnel serviceability index. Tunnelling and Underground Space Technology, 67: 98-106.
4. Ding, H., Trajcevski, G., Scheuermann, P., Wang, X., & Keogh, E. (2008). Querying and mining of time series
data: experimental comparison of representations and distance measures. Proceedings of the VLDB Endowment,
1(2): 1542-1552.
5. Yajima, A. , Wang, H. , Liang, R. Y. , & Castaneda, H. . (2015). A clustering based method to evaluate soil
corrosivity for pipeline external integrity management. International Journal of Pressure Vessels and Piping,
126-127: 37-47.
Study on K-means Evaluation Method for Service Performance of Highway

Tunnels
Chong-bang Xu1, *, Fa-you Deng2, Xiao Xu1 and Gang Liu1
1 Bridge And Tunnel Research Center, Research Institute of Highway Ministry of Transport, Beijing 100088, China.
2 College of Civil Engineering, Shaoxing University, Zhejiang Shaoxing 312000, China.
*Corresponding Author: Chongbang Xu. Email: xubang-2003@163.com.
Abstract: The service performance of highway tunnel is the comprehensive reflection and evolution of many
engineering data and parameters of tunnel engineering[1,2]. In order to reflect the influence of these data on the
service performance of tunnels, according to the idea of big data analysis method[3-6], a big data evaluation index
system for the service performance of existing highway tunnels is constructed, and a K-means evaluation method for
the service performance of existing highway tunnels is proposed by introducing K-means clustering algorithm[7-10].
The main results are as follows: (1) According to the state of tunnel engineering data in the service period of tunnel,
the tunnel engineering data can be divided into static data and dynamic data. The static data include engineering
prospection data, engineering design data and engineering construction data, which are basically unchanged when
the tunnel project enters service period; The dynamic data are mainly the maintenance and inspection data of the
tunnel in service, which are constantly changing during the service period. (2) There are more than 100 initial types
of static and dynamic data in tunnel engineering. After preliminary screening by expert experience method and
secondary screening by machine learning (static data adopts Spielman correlation coefficient algorithm and dynamic
data adopts Apriority association rule algorithm), 19 evaluation indexes are finally screened out. Among them, there
are 10 static indexes: stratum lithology, geological structure, unfavorable geology, special rock and soil,
groundwater, climate characteristics, tunnel length, secondary lining concrete, pavement structure type, construction
disaster accident; 9 dynamic indexes: tunnel portal level, leakage level, pavement level, maintenance road facility
level, drainage facility level, ceiling level, interior decoration level and marking line level.(3) On the basis of 19
evaluation indexes, a tunnel service performance evaluation model based on K-means clustering algorithm is
constructed. The best clustering effect is determined by Xie-Beni value. The clustering centers of each evaluation
index are determined by 186 engineering cases with different tunnel service performance. (4) By 36 engineering
case tests, it shows that the accuracy rate of the K-means tunnel service performance evaluation model is 75%,
which indicates that the K-means tunnel service performance evaluation can generally reflect the tunnel service
performance.
Keywords: Highway Tunnels, Big Data, Service Performance, Evaluation Model, Evaluation Index
Acknowledgement
The research is supported by National Key R&D Program of China (Grant No.2017YFC0840202).
References
1. He, B., & Li, Y. (2014). Big data reduction and optimization in sensor monitoring network. Journal of Applied
Mathematics, 2014(2), 155-175.
2. Shi, M., Zhang. L., Sun, W., & Song, X. (2019). A fuzzy c-means algorithm guided by attribute correlations
and its application in the big data analysis of tunnel boring machine. Knowledge-Based Systems, In Press.
3. Kumar, B. (2015). An encyclopedic overview of 'big data' analytics. International Journal of Applied
Engineering Research, 10(3): 5681-5705.
4. Chen, C., & Zhang, C. (2014). Data-intensive applications, challenges, techniques and technologies: A survey
on Big Data. Information Sciences, 275: 314-347.
5. Sun, Z., & Wang, P. (2017). Big data, analytics and intelligence: an editorial perspective. Journal of New
Mathematics and Natural Computation, 13(2): 75-81.
122 DACOMA-2019
6. Kantardzic, M. (2011). Data mining: concepts, models, methods, and algorithms (2th edition). Wiley-IEEE
Press.
7. Zimmermann, H. (2001). Fuzzy set theory and its applications (4th edition). Kluwer Academic Publishers.
8. Jain, A. (2010). Data clustering: 50 years beyond K-means. 19th International Conference on Pattern
Recognition (ICPR 2008), 31(8): 651-666.
9. Xu, R., & Wunsch, D. (2005). Survey of clustering algorithms. IEEE Transactions On Neural Networks, 16(3),
645-678.
10. Kanungo, T., Mount, D., & Netanyahu, N. (2002). An efficient k-means clustering algorithm: Analysis and
implementation. IEEE Transactions On Pattern Analysis And Machine Intelligence, 24(7): 881-892.
A LSTM framework in predicting changes in digital parameters of granite

compositions under uniaxial compression load
Yalei Wang1, Jinming Xu1,*, and Xuejie Tao1
1
Department of Civil Engineering, Shanghai University, Shanghai 200444, China.
*
Corresponding Author: Jinming Xu. Email: xjming@shu.edu.cn.
Abstract: Using the video images photographed during the laboratory uniaxial compression tests, the regions of
various compositions in granite specimen was determined and labeled using the threshold segmentation. Digital
parameters of each composition with time were then obtained. After constructing and training a LSTM (Long-Short
Term Memory) framework, we estimated the changes in digital parameters of the compositions using RMSE (Root
Mean Square Error) to measure the reliability. It shows that there is a good agreement between the observed and
predicted digital parameters.
Existing studies. Ben-Yosef et al [1] constructed a full interpretation model to identify the semantic features of the
object images. Klyuchnikov et al. [2] presented an identification technique in determining rock type at the drilling
bit using machine learning classification methods. Imamverdiyev and Sukhostat [3] developed a model for
geological facies classification in wells using the deep learning. Sagheer and Kotb [4] proposed a deep long-short
term memory (DLSTM) architecture using the production data of two actual oilfields. Gao et al [5] used the
traditional recurrent neural (RNN) networks, long-short term memory (LSTM) networks and gated recurrent unit
(GRU) networks to conduct the real-time prediction of operating parameters in tunnel boring machines (TBM).
Zhou et al [6] presents a predictive framework, containing a wavelet transform noise filter, convolutional neural
network feature extractor, and long short-term memory predictor, for determining the attitude and position of the
shield machine. Yang et al [7] used the time series theory and long and short term memory neural network (LSTM)
to predict the transient landslide displacement. Zuo et al [8] reviews the state-of-the-art application of deep learning
algorithms for processing geochemical exploration data and mining the geochemical patterns.
We first extract digital parameters of compositions in granite at each constant. The size of the rock specimen
was around 50 mm × 50 mm × 100 mm. The video images were photographed during the laboratory uniaxial
compression tests, by leveling the camera normal to the specimen surface with a distance of about 0.5m. For each
frame in the test video image, the threshold segmentation was used to extract the regions including the compositions,
or biotite, feldspar, quartz, and crack. The labeling was then performed for each composition. Digital features of a
composition were characterized by area A, eccentricity E, and inscribed radius RI. Figure 1 shows the area A with
time of biotite in granite specimen under axial load. In the figure, the horizontal label is the second transferred from
the total 732 frames with the fps of 25.
Figure 1. Inscribed radius RI (in pixel) with time of compositions in granite specimen under axial load.
We then constructed a LSTM framework. All digital parameters of the compositions were considered as the data
set, in which 90% and 10% were respectively partitioned into the training and test ones. For later convenience, these
parameters were standardized with the zero mean and unit variance. A simple LSTM (Long-Short Term Memory)
framework (see Figure 2) was established to predict the changes in digital parameters of compositions. In the LSTM
framework, we set the parameters of input size (in Sequence Input Layer), number of hidden units (in LSTM Layer),
and number of responses (in Fully Connected Layer), as 1, 1, and 200, respectively.

124 DACOMA-2019
Digital Parameters of compositions

at each constant from video imge
Sequence Input LSTM Fully Connected Regression

Layer Layer Layer Layer
LSTM Framework
Figure 1. LSTM framework for predicting changes in digital parameters of granite compositions
We trained the LSTM framework and estimated the changes in digital parameters of compositions. To train
the LSTM framework, the ADAM technique was used as the optimization algorithm, the maximum epochs, gradient
threshold, initial learn rate, learn rate drop period, and learn rate drop factor were set as 250, 1, 0.005,125, 0.2,
respectively. RMSE (Root Mean Square Error) was used to measure the difference between the observed and
predicted values. Figure 3 shows the observed and predicted eccentricities of feldspar using the LSTM framework. It
can be seen that there is a good agreement between the observed and predicted eccentricities.
Eccentricity of feldspar
Time /s (in prediction)
Figure 3. Observed and predicted eccentricities of feldspar using constructed LSTM framework.
The estimation of other digital parameters of other compositions, influences of LSTM parameters and composition
locations, and relations between these parameter changes with mechanical features of the rock should be
investigated in more detail.
Keywords: LSTM Framework, Digital Parameter, Granite Composition, Uniaxial Compression Load.
Acknowledgement
Financial support for the study was provided by the Natural Sciences Foundation Committee of China under Grant
No. 41472254. This support is gratefully acknowledged.
References
1. Ben-Yosef G., Assif L., Ullman S. (2018). Full interpretation of minimal images. Cognition, 171: 65–84.
2. Klyuchnikov N., Zaytsev A., Gruzdev A. et al. (2019). Data-driven model for the identification of the rock type
at a drilling bit. Journal of Petroleum Science and Engineering, 178: 506–516.
3. Imamverdiyev Y., Sukhostat L. (2019). Lithological facies classification using deep convolutional neural
network. Journal of Petroleum Science and Engineering, 174: 216–228.
4. Sagheer A., Kotb M. (2019). Time series forecasting of petroleum production using deep LSTM recurrent
networks. Neurocomputing, 323: 203–213.
5. Gao X., Shi M., Song X., et al. (2019). Recurrent neural networks for real-time prediction of TBM operating
parameters. Automation in Construction, 98: 225–235.
6. Zhou C., Xu H., Ding L., et al. (2019). Dynamic prediction for attitude and position in shield tunneling: A deep
learning method. Automation in Construction, 105: 102840.
7. Yang B., Yin K., Du J. (2018). A model for predicting landslide displacement based on time series and long
and short term memory neural network. Chinese Journal of Rock Mechanics and Engineering, 37(10): 2334–
2343.
8. Zuo R., Xiong Y., Wang J., et al. (2019). Deep learning and its application in geochemical mapping. Earth-
Science Reviews, 192: 1–14.
A newly-developed alloy gas atomization process and the powder product used
in selective laser melting process
Pengfei Yan1 and Biao Yan2, *
1School of Physics and Engineering, Tongji University, Shanghai 200092, China
2School of Materials and Engineering, Tongji University, Shanghai 201804, China
*Corresponding Author: Biao Yan. Email: 84016@tongji.edu.cn
Abstract: Selective laser melting process(SLM), as a common additive manufactory (AM or 3D printing) process,
has been widely used in high-end manufacturing, such as dental and bone implants for medical treatment,
lightweight structure and functional components of aerospace, prototype manufacturing of automotive parts, etc.
However, the high cost of powder raw materials is one of the reasons that hinder the application of this process in
batch-production industries. Our group combined with the existing work foundation of atomization powder process,
powder metallurgy and SLM technology. Firstly we carried out the innovation of atomization process to obtain low-
cost new SLM printing powder. Subsequently, we developed matching printing process based on the performance of
the printing powder. And we secondly optimized topology of lightweight lattice structures from the developed
printing parameters. In this paper, the design principles and results of the new atomization process are described[1,
2]. And the case of our typical aluminum alloy powder product was used to demonstrate the characteristics of the
powder product, the printing results and the secondary topology optimization example of lightweight lattice
structure.
Keywords: Additive manufactory, Gas atomization, Aluminum alloy powders, Lattice structure.
References
1. Ruan, C., Xing, F., Huang, Y., Yu, X., Zhang, J., & Yao, Y. (2017). The influence of acoustic field induced by
HRT on oscillation behavior of a single droplet. Energies, 10(1): 48.
2. Zu, HB, Zhou, ZW, Wang, ZL (2012). The Influence of Acoustic Field Induced by HRT on Oscillation
Behavior of a Single Droplet. Applied Mathematics and Mechanics, 33(12): 1481–1492.

Prediction of Undrained Shear Strength of Soft Clays Using LightGBM

Method Based on Bayesian Optimization
Wengang Zhang123* Chongzhi Wu3 Runhong Zhang3 Yongqin Li3

1Key Laboratory of New Technology for Construction of Cities in Mountain Area, Chongqing University, Chongqing
400045, China
2National Joint Engineering Research Center of Geohazards Prevention in the Reservoir Areas, Chongqing University,
Chongqing 400045, China

3School of Civil Engineering, Chongqing University, Chongqing 400045, China
*Corresponding Author: Wengang Zhang. Email: zhangwg@cqu.edu.cn
Abstract: This study applies a novel data-driven lightGBM machine learning method [1] for capturing the
relationships between the undrained shear strength and various basic soil parameters. Based on the method,
a general approach is developed to predict the undrained shear strength (𝑠𝑢 ) of soft clays by employing soil
data sets from sites including Finland, Sweden and Norway [2], available in the TC304 database. The 𝑠𝑢 is
predicted in consideration of six feature variables including the preconsolidation stress, vertical effective
stress, liquid limit, plastic limit, natural water content, and sensitivity [3][4].In order to reduce the
dependence on the rule of thumb and bruteforce search, the Gaussian Bayesian optimization method is used
to find the model hyperparameters [5]. The developed model is used as a contrastive analysis with other
approaches (including traditional empirical equations and other machine learning methods) under 10-fold
cross-validation for better persuasive results. It is shown that lightGBM model outperforms these
approaches in terms of both accuracy and computational efficiency, which makes it a satisfactory tool in
prediction of geotechnical parameters. For geotechnical engineering applications, a new lightGBM -based
model can be readily obtained from the proposed approach with reasonable accuracy, provided that the new
site-specific test data are available.
Keywords: Undrained Shear Strength, Computational Efficiency, LightGBM, Machine Learning Method,
Bayesian Optimization.
Acknowledgement
The authors would like to thank the members of the TC304 Committee on Engineering Practice of Risk
Assessment & Management of the International Society of Soil Mechanics and Geotechnical Engineering
for developing the database 304dB used in this study and making it available for scientific inquiry. We also
wish to thank M. D’Ignazio, T.T. Länsivaara, J. Ching and K.K. Phoon for contributing this database to the
TC304 compendium of databases. The study is partially supported by the Natural Science Foundation of
Chongqing, China (cstc2018jcyjAX0632) and Venture & Innovation Support Program for Chongqing
Overseas Returnees (cx2017123).
References
1. Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., ... & Liu, T. Y. (2017). Lightgbm: A
highly efficient gradient boosting decision tree. In Advances in Neural Information Processing
Systems (pp. 3146-3154).
127 DACOMA-2019
2. D’Ignazio, M., Phoon, K. K., Tan, S. A., & Länsivaara, T. T. (2016). Correlations for undrained shear
strength of Finnish soft clays. Canadian Geotechnical Journal, 53(10): 1628-1645.
3. Ching, J. and Phoon, K.K. (2014a). Transformations and correlations among some clay parameters —
the global database. Canadian Geotechnical Journal, 51(6): 663–685.
4. Ching, J. and Phoon, K.K. (2014b). Correlations among some clay parameters — the multivariate
distribution. Canadian Geotechnical Journal, 51(6): 686–704.
5. Snoek, J., Larochelle, H., & Adams, R. P. (2012). Practical bayesian optimization of machine learning
algorithms. In Advances in neural information processing systems (pp. 2951-2959).
A Review in Recent Development of 5 Axis CNC Milling Machine Tool

Operations
Mohsen Soori* and Mohammed Asmael
Department of Mechanical Engineering, Eastern Mediterranean University, Famagusta, North Cyprus, Via Mersin 10, Turkey.
*Corresponding Author: Mohsen Soori. Email: mohsen.soori@gmail.com, mohsen.soori@emu.edu.tr.
Abstract: To produce sophisticated parts with free form surfaces such as turbine blades, airfoils, impellers and
aircraft components, 5-axis CNC machine tools are used. To increase the accuracy as well as efficiency in process of
part production using 5-axis CNC machine tools, effective parameters and errors such as geometric and gouging
errors are considered in the research works. Also, machining operations of sculptured surfaces by using 5-axis CNC
machine tools are analyzed and modified to increase quality of part production. Feed rate scheduling and
optimization systems in 5-axis CNC machine tool operations are developed in order to increase surface quality in
machined parts. Cutting forces prediction and analysis are presented to provide ability of cutting force and cutting
tool monitoring system along machining paths. Cutting Toolpath and cutter orientation in 5-axis CNC machine tool
operations are analyzed and optimized to increase efficiency in process of part production. To create and analyze
actual machined parts in virtual environments, virtual machining systems are developed. Rotary axis of 5-axis CNC
machine tools are analyzed and modified in order to increase accuracy of the axis in machining operations of free
form surfaces. In order to increase efficiency in process of part production, optimized cutting tool paths with regard
to minimum energy consumption are generated. A review of recent development in 5-axis CNC machine tool
operations is presented in this paper to develop the process of part production using machining operations.
Keywords: 5-axis CNC machine tools, Geometrical errors, Cutting forces, Process optimization.
References
1. Farouki, R. T., Han, C. Y., & Li, S. (2014). Inverse kinematics for optimal tool orientation control in 5-axis
CNC machining. Computer Aided Geometric Design, 31(1): 13-26.
2. Hsu, Y. Y., & Wang, S. S. (2007). A new compensation method for geometry errors of five-axis machine tools.
International Journal of Machine Tools and Manufacture, 47(2): 352-360.
3. Bohez, E. L. (2002). Compensating for systematic errors in 5-axis NC machining. Computer-Aided Design,
34(5), 391-403.
4. Bohez, E. L., Ariyajunya, B., Sinlapeecheewa, C., Shein, T. M. M., & Belforte, G. (2007). Systematic
geometric rigid body error identification of 5-axis milling machines. Computer-Aided Design, 39(4): 229-244.
5. Langeron, J. M., Duc, E., Lartigue, C., & Bourdet, P. (2004). A new format for 5-axis tool path computation,
using Bspline curves. Computer-Aided Design, 36(12): 1219-1229.
6. Lei, W. T., & Hsu, Y. Y. (2003). Accuracy enhancement of five-axis CNC machines through real-time error
compensation. International journal of machine tools and manufacture, 43(9): 871-877.
7. Lei, W. T., & Sung, M. P. (2008). NURBS-based fast geometric error compensation for CNC machine tools.
International Journal of Machine Tools and Manufacture, 48(3-4): 307-319.
8. Lasemi, A., Xue, D., & Gu, P. (2016). Accurate identification and compensation of geometric errors of 5-axis
CNC machine tools using double ball bar. Measurement Science and Technology, 27(5): 055004.
9. Ibaraki, S., Nagai, Y., Otsubo, H., Sakai, Y., Morimoto, S., & Miyazaki, Y. (2015). R-Test Analysis software
for error calibration of Five-Axis machine tools–Application to a Five-Axis machine tool with two rotary axes
on the tool side–. International Journal of Automation Technology, 9(4): 387-395.
10. Zhong, L., Bi, Q., Huang, N., & Wang, Y. (2018). Dynamic accuracy evaluation for five-axis machine tools
using S trajectory deviation based on R-test measurement. International Journal of Machine Tools and
Manufacture, 125: 20-33.

129 DACOMA-2019
11. Čep, R., Malotová, Š., Kratochvíl, J., Stančeková, D., Czán, A., & Jakab, T. (2018). Diagnosis of machine tool
with using Renishaw ball-bar system. In MATEC Web of Conferences (Vol. 157, p. 01006). EDP Sciences.
12. Usop, Z., Sarhan, A. A., Mardi, N. A., & Wahab, M. N. A. (2015). Measuring of positioning, circularity and static
errors of a CNC Vertical Machining Centre for validating the machining accuracy. Measurement, 61: 39-50.
13. Fussell, B. K., Jerard, R. B., & Hemmett, J. G. (2003). Modeling of cutting geometry and forces for 5-axis
sculptured surface machining. Computer-Aided Design, 35(4): 333-346.
14. Baptista, R., & Simoes, J. A. (2000). Three and five axes milling of sculptured surfaces. Journal of materials
processing technology, 103(3): 398-403.
15. Yoon, J. H., Pottmann, H., & Lee, Y. S. (2003). Locally optimal cutting positions for 5-axis sculptured surface
machining. Computer-Aided Design, 35(1): 69-81.
16. Ma, W., He, G., Zhu, L., & Guo, L. (2016). Tool deflection error compensation in five-axis ball-end milling of
sculptured surface. The International Journal of Advanced Manufacturing Technology, 84(5-8): 1421-1430.
17. Sencer, B., Altintas, Y., & Croft, E. (2008). Feed optimization for five-axis CNC machine tools with drive
constraints. International Journal of Machine Tools and Manufacture, 48(7-8): 733-745.
18. Beudaert, X., Lavernhe, S., & Tournier, C. (2012). Feedrate interpolation with axis jerk constraints on 5-axis
NURBS and G1 tool path. International Journal of Machine Tools and Manufacture, 57: 73-82.
19. Sun, Y., Bao, Y., Kang, K., & Guo, D. (2013). An adaptive feedrate scheduling method of dual NURBS curve
interpolator for precision five-axis CNC machining. The International Journal of Advanced Manufacturing
Technology, 68(9-12): 1977-1987.
20. Chen, M., & Sun, Y. (2019). Contour error–bounded parametric interpolator with minimum feedrate
fluctuation for five-axis CNC machine tools. The International Journal of Advanced Manufacturing
Technology, 1-18.
21. Luo, S., Dong, Z., & Jun, M. B. (2017). Optimal tool orientation generation and chip volume/cutting force
predictions for 5-axis CNC machining of curved surfaces using flat-end mills. Computer-Aided Design and
Applications, 14(3): 331-342.
22. Luo, S., Jun, M. B., & Dong, Z. (2016). Numerical simulation of chip ploughing volume and forces in 5-axis
CNC micro-milling using flat-end mills. Procedia Manufacturing, 5: 348-361.
23. Aslan, D., & Altintas, Y. (2018). Prediction of cutting forces in five-axis milling using feed drive current
measurements. IEEE/ASME Transactions on Mechatronics, 23(2): 833-844.
24. Luo, S., Dong, Z., & Jun, M. B. (2017). Optimal tool orientation generation and chip volume/cutting force
predictions for 5-axis CNC machining of curved surfaces using flat-end mills. Computer-Aided Design and
Applications, 14(3): 331-342.
25. Jun, C. S., Cha, K., & Lee, Y. S. (2003). Optimizing tool orientations for 5-axis machining by configuration-
space search method. Computer-Aided Design, 35(6): 549-566.
26. Farouki, R. T., & Li, S. (2013). Optimal tool orientation control for 5-axis CNC milling with ball-end cutters.
Computer Aided Geometric Design, 30(2): 226-239.
27. Lavernhe, S., Tournier, C., & Lartigue, C. (2008). Optimization of 5-axis high-speed machining using a surface
based approach. Computer-Aided Design, 40(10-11): 1015-1023.
28. Sun, C., & Altintas, Y. (2016). Chatter free tool orientations in 5-axis ball-end milling. International Journal of
Machine Tools and Manufacture, 106: 89-97.
29. Bo, P., Bartoň, M., Plakhotnik, D., & Pottmann, H. (2016). Towards efficient 5-axis flank CNC machining of
free-form surfaces via fitting envelopes of surfaces of revolution. Computer-Aided Design, 79: 1-11.
30. Altintas, Y., Brecher, C., Weck, M., & Witt, S. (2005). Virtual machine tool. CIRP annals, 54(2): 115-138.
31. Merdol, S. D., & Altintas, Y. (2008). Virtual simulation and optimization of milling applications-part II:
optimization and feedrate scheduling. Journal of Manufacturing Science and Engineering, 130(5): 051005.
32. Soori, M., Arezoo, B., & Habibi, M. (2013). Dimensional and geometrical errors of three-axis CNC milling
machines in a virtual machining system. Computer-Aided Design, 45(11): 1306-1313.
Mineral Segmentation from Polarized Light Microscopy Images of Rock

Slices via Deep Neural Networks
Qimin Wang1 , Yanlu Ding2 and Xiaoying Zhuang1,*
1 Department of Geotechnical Engineering, Tongji University, Shanghai 200092, China.
2 Department of Civil Engineering, The University of Hong Kong, Hong Kong, China.
Abstract: There are about 3000 known minerals on earth. A mineral is an element or chemical compound that is
normally crystalline and that has been formed as a result of geological processes [1]. All rocks are made up of 2 or
more of these minerals. Different mineral bends light in different ways, which enables identifying the types of
mineral by colors of the transmitted light and the reflected light, as shown in Fig. 1. Polarization microscopy is
routinely applied in material sciences and geology to identify minerals on the basis of characteristic refraction
properties and colors. This phenomenon is the so-called birefringence, which is the optical property of a material
having a refractive index that depends on the polarization and propagation direction of light, as shown in Fig.2. This
effect was first described by the Danish scientist Rasmus Bartholin in 1669, who observed it [2] in calcite, a crystal
having one of the strongest birefringence. Crystals with non-cubic crystal structures are often birefringent, as are
plastics under mechanical stress. Whereas polarizing microscopes are commonly used in imaging of cellulose in cell
walls of plants and starch grains in biology, polarizing microscopes are also used for identification or imaging of
birefringent structures like crystals in rock slices. We operated the Leica DM750 microscope in Fig. 3 for polarized
light imaging. Utilizing the image processing software Leica QWin, all the images of rock slices were generated in a
similar photography condition regarding the light source, the exposure time, the focus and the gaining.
(a) 0° (b) 45° (c) 90° (d) 135°

Figure 1: Images of the same rock slice region at different relative polarized angles (magnified by 10 times)
Figure 2: Explanation of birefringence Figure 3: Polarized light microscope configuration
Using the microscopy system, we have generated a rock slice images database of about 200, where the samples are
from the different location of the same site. However indistinguishable on the macro scale, the rock slices are

131 DACOMA-2019
unrevealed remarkably heterogeneous on the micro scale by the polarized light microscopy. By shifting the relative
polarized angle of the two polarized gratings, we observed dramatic colors changing of different minerals (Fig. 4),
which could be good basis for minerals classification. However, this approach is often hindered by the efficiency to
rotate the gratings, and by the accuracy of polarized angle. During certain range of polarized angle , colors of
minerals changes enormously. We proposed a data driven method to render the underlying features, or the semantic
content of a rock slice image in different polarized angles, given only the image at 0°. Arguably, a major limiting
factor for previous approaches has been the lack of image representations that explicitly represent semantic
information and, thus, allow to separate image content from style [3]. Here we use rock image features derived from
Convolutional Neural Networks (CNN) optimised for object recognition, which make high level rock information
explicit, as shown in Fig. 4. The method allows us to produce new images of different polarized angles that requires
limited prior images.
Figure 4: Rock slice image style transfer in a Convolutional Neural Network (CNN)
Keywords: Image segmentation, Polarized lighting microscopy, Rock slices, Deep neural networks.
Acknowledgement
The Author Qimin Wang thankfully acknowledge the scholarship from the HKU Civil Centennial Future Scholars
(HKU-CCFS 2019).
References
1. Bartholin, E. (1983). Experimenta crystalli Islandia disdiaclastici quibus mira et insolita refractio detegitur.
Daniel Paullus.
2. Nickel, E. H. (1995). Definition of a mineral: definició de mineral. Boletín de la Sociedad Española de
Mineralogía, 18: 207-212.
3. Gatys, L. A., Ecker, A. S., & Bethge, M. (2016). Image style transfer using convolutional neural networks.
In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2414-2423.
A Novel Method for Health Assessment of Subway Shield Tunnel Structure

Based on Knowledge Graph
Yadong Xue1,*, Hongfei Zhang2 and Peizhe Shi1
1TongjiUniversity, 1239 Siping Road, 200092 Shanghai, China.
2ShaoxingUniversity,508 Ring road,312000 Shaoxing,China.
*Corresponding Author: Yadong Xue. Email:yadongxue@tongji.edu.cn.
Abstract: With the rapid increasing traffic pressure in large cities, Fast urban rail transit has shown more and more
importance to relieve traffic, and it has gradually become the preferred travel mode. Most of the urban rail transit
adopts the form of underground shield tunnel. Due to the large-scale construction of subway shield tunnels, the
construction environment around the tunnel is becoming more and more complicated. As an underground structure,
the subway structure is inevitably degraded with the influence of external loads and service time. In order to
improve the subway service, its daily operation time is getting longer and longer, which puts forward higher
requirements for the health inspection, evaluation and maintenance of tunnel structure. In order to evaluate the
structural performance or health of tunnels in a timely and scientific manner, the researchers have carried out a lot of
research work in testing methods, equipment and evaluation methods. At present, the commonly used health
evaluation methods are indicator-based methods, including single factor, multiple factors, comprehensive evaluation
and so on. However there are exist many and complicated influence factors of tunnel structure, such as construction
ways, service time, structure scale, structure form, structural defects (leakage water, crack, staggered segment,
segment spalling, etc.), material aging, state of bolt tightening, tunnel section shape, tunnel structure repair and
reinforcement, surrounding strata, surrounding construction and so on [1]. These factors are always difficult to
consider, and many factors are difficult to describe quantitatively, so the most commonly used evaluation method is
only based on the tunnel structure itself, using multiple factors to carry out weighted synthetic evaluation [2].
Limited to the complexity of underground tunnel engineering, the weight of each factor is still mainly based on the
experience of experts, resulting in the factor that can`t be considered completely in structural evaluation, and the
degree of influence of factors is greatly affected by human experience and the assessment takes long time which is
also difficult to consider the dynamic evolution characteristics of tunnel health. The concept of Knowledge Graph
（KG）was proposed by Google in 2012[3]. It converts factual text content into interconnected graph structures and
establishes connections between massive chaotic information. It can provide users with a concise, easy to understand
and visualization data support [4]. Based on KG, the detailed information of large-scale rail transit tunnel structures
can be fully constructed, which provides the basic data for the structure health assessment. This paper focuses on the
shield tunnel constructed in soft ground in Shanghai. Firstly, based on the concept of subway tunnel to construct the
KG [5], which mainly includes the information of shield tunnel construction (including construction contractor,
construction method, construction date, construction process and acceptance, etc.), operating environment (including
the surrounding environment, geological and hydrology, the adjacent construction, internal environment, train traffic
load, station platform type and subways transfer, etc.), the structure characteristics (including the structure section
size, tunnel section shape, the segment assembling modal, etc.), segment ring unit (per five rings or connected aisle
as a unit, including internal force and deformation of the structure, defect characteristics, reinforcement and repair,
etc.). Then, taking conceptual KG as the template, taking the concrete subway tunnel project as the object, using the
Neo4j [6] diagram database to construct the subway KG which covers all the 16 operating subway lines in Shanghai
(total 415 subway sections, 705*2 kilometers of shield tunnel). The Multi-index weighted comprehensive evaluation
method is adopted to evaluate the structural health degree of each segment ring unit. To improve the reliability of
the evaluation results, a Web-based Delphi method will be used. The health degree of segment ring unit then be
added to the KG as new nodes. Thirdly, a Graph Neural Network (GNN) model was built with Pytorch [7,8]. Taking
the KG of segment ring unit as input and the its health degree index as output, the training and testing samples were
prepared. And then, the model is trained and validated. Finally, when the knowledge information of segment ring
unit changed, using the trained model to evaluate its structural health degree, then update its KG. Compared with the
traditional methods, the novel method of shield tunnel structure health assessment proposed in this paper, can obtain
more comprehensive information that affects tunnel structure health, on the other hand, it can use the trained model
to evaluate the health degree fast and automatically. With the advantages of Knowledge Graphs, it is convenient to

133 DACOMA-2019
manage and query information related to subway tunnels, and provide support for artificial intelligence services
based on big data. Based on the subway KG, it is expected to provide intelligent support for the tunnel operation and
maintenance decision-making.
Keywords: Shield tunnel, Structural health, Knowledge Graph, Graph neural network.
Acknowledgement
The work presented in this article was supported by the Science and Technology Commission of Shanghai
Municipality (Grant No. 18DZ1205902) and the Key innovation team program of innovation talents promotion plan
by MOST of China (No. 2016RA4059).
References
1. Huang, H., Xue, Y., Shao, H., & Du, J. (2019). Rapid disease detection theory and engineering practice of
urban subway shield tunnel. Shanghai Scientific & Technical Publishers.
2. Su, H. (2018). Research on structural safety assessment for shield tunnel in Shanghai. Chinese Journal of
underground Space and Engineering, 14(S2): 940-946.
3. Zhao, J., Liu, K., & He, S. (2018). Knowledge Graph. Higher Education Press.
4. Lin, N., Li,P., & Xie, X. (2015). Research on Evolution Mechanism of shield Tunnel Disease Based on
Segment Performance Analysis. Chinese Journal of underground Space and Engineering,11(S2): 802-809.
5. Wira-Alam, A., & Mathiak, B. (2012). Mining Wikipedia's snippets Graph-First step to build a new knowledge
base. KNOW@ LOD, 43-48.
6. Zhang, Z., Pang, G., & Hu, J. (2017). Neo4j the definitive guide. Tsinghua University Press.
7. Annervaz, K. M., Chowdhury, S. B. R., & Dukkipati, A. (2018). Learning beyond datasets: Knowledge graph
augmented neural networks for natural language processing. arXiv preprint arXiv:1802.05930.
8. Minervini, P., Fanizzi, N., d'Amato, C., & Esposito, F. (2015). Scalable learning of entity and predicate
embeddings for knowledge graph completion. 2015 IEEE 14th International Conference on Machine Learning
and Applications: 162-167.
An Adaptive Collocated Method for PHT Splines with Optimal Selection of

Collocation Points
Yue Jia1, Cosmin Anitescu2, *, Yongjie Jessica Zhang3 and Timon Rabczuk2
1
School of Mechanics, Civil Engineering and Architecture, Northwestern Polytechnical University, Xi’an, 710129, China.
2
Institute for Structure Mechanics, Bauhaus University Weimar, Geschwister-Scholl-Straße 8, 99423 Weimar, Germany.
3
Department of Mechanical Engineering, Carnegie Mellon University, Pittsburgh, Pennsylvania, 15213 USA.
*Corresponding Author: Cosmin Anitescu. Email: cosmin.anitescu@uni-weimar.de.
Abstract: We propose an adaptive isogeometric analysis (IGA) collocation method with a recovery based error
estimator. The proposed collocation method is based on a new basis for polynomial hierarchical splines over T-
meshes (PHT-splines) [1]. The improved method collocates at Gauss points [2] in the interior of the domain, which
are the optimal for the new PHT elements with C1 continuity. Near the boundary, the locations of the collocation
points are determined by an optimization method based on machine learning algorithms.
The new PHT basis is based on local knot vectors and has improved conditioning compared to the original PHT
basis [3]. Furthermore, based on the new PHT basis, the improved IGA collocation can be extended to arbitrary
degree polynomials. The local refinement strategy is driven by a recovery-based error estimator by computing a
higher order recovered solution. The improved IGA collocation method is also compared with the Galerkin method
as well as the Greville-abscissae collocation. The improved IGA collocation has been tested on several 2D and 3D
problems and results in optimal convergent rates and high accuracy and robustness of the approximation.
Keywords: isogeometric collocation, PHT splines, Gauss points, recovery-based error estimator.
References
1. Deng J. S., Chen F., Li X. et al. (2008). Polynomial splines over hierarchical T-meshes. Journal Graphical
Models, 70(4): 76-86.
2. De Boor C., Swartz B. (1973). Collocation at Gaussian Points. SIAM Journal on Numerical Analysis, 10(4):
582-606.
3. Kang H, Xu J, Chen F, Deng J. (2015). A new basis for PHT-splines. Craphical Models, 80: 149-159.

Optimal Placement of Active Members in Adaptive Cable-Strut tensile

Structure Using an Optimal Combination Algorithm
Sha Li1,* and Nan Xiao2
1Beijing Institute of Technology, Zhuhai, 6 Jinfeng Road, 519000 Zhuhai, China.
2College of Civil Engineering and Architecture, Zhejiang University, Hangzhou, 310058, China
*Corresponding Author: Sha Li. Email: sha_li@bitzh.edu.cn.
Abstract: An adaptive cable-strut tensile structure is defined as a controllable structure which can actively alter its
configuration or self-stress in response to variation of the outer environment by changing the length of active
members with actuators. The realization of such alteration is involved with optimization and control processes. The
shape control or the internal force control of such structure had already been studied in some researches [1-3].
However, the number and placement of active members in their control were usually determined subjectively or on
random search. Since the optimization of the placement active members concerns about discrete variables,
intelligent optimization algorithms based on random search are often introduced to solve such problems, such as
genetic algorithm, simulated annealing algorithm, particle swarm optimization, ant colony algorithm and so on [4-6].
By analyzing the coefficient matrix of the incremental calculation of the control equations, this paper reveals a
mathematical relationship between the number and placement of active members and the solution of control
equations. The criteria of the number and placement of active elements are proposed. A mathematic formulation for
optimal placement with the objective of minimizing control efforts is established, and a solution strategy based on
genetic algorithm is proposed. A simple-supported tensegrity beam has been used for the validation of the proposed
strategy (Fig. 1).
6 4 10 8 12
2
1 11
5 3 Δf 9 7
Figure 1: The target shape of a tensegrity beam
Furthermore, taking minimizing shape control errors, the length changes and the number of active members as
objective functions, a hybrid variable multi-objective optimization model with active member distribution vector and
the length change as variables has been established, and an optimal combination algorithm with non-dominant sorting
genetic algorithm(NSGA-Ⅱ)and sequential quadratic programming(SQP) is introduced. The controlling validity is
investigated via case study. The results present the contradiction relationship between the optimal solutions of each
objective function, and indicate that the higher accuracy demanded the more active members needed.
Keywords: Optimal placement, Active member, NSGA-Ⅱ, SQP, Adaptivestructure.
Acknowledgement
The authors acknowledge the financial support of the Beijing Institute of Technology, Zhuhai (Project XK-2019-06).
References
1. Li, S., Xiao, N., Dong S.L.(2014). Research on simultaneous control of force and shape for adaptive cable-strut
tensile structure. J. Huazhong Univ. of Sci. & Tech.(National Edition), 42(8): 119-127.

136 DACOMA-2019
2. Li, S., Xiao, N., Dong S.L. (2014).Sensitivity Analysis on responses of adaptive cable-strut tensile structure
with length changeable elements. J. Huazhong Univ. of Sci. & Tech.(National Edition), 42(10): 120-123.
3. Kmet S., Platko P., Mojdis M. (2012). Analysis of Adaptive Light-Weight Structures. Procedia Engineering,
40: 199-204.
4. Animesh N., Debabrata C., Mahesh S. (2019). Optimal Sensors/Actuators Placement in Smart Structure Using
Island Model Parallel Genetic Algorithm. International Journal of Computational Methods, 16(6): 1840018.
5. Dhuri K., Seshu P.(2009). Multi-Objective Optimization of Piezo Actuator Placement and Sizing Using
Genetic Algorithm. Journal of Sound and Vibration, 323(3-5): 495-514.
6. Zhang L., Xia R., Huang H.(2010). Optimal Placement of Actuator Truss Adaptive Structures. Chinese Journal
of Mechanical Engineering, 23(2): 233-241.
Two Applications of Isogeometric Boundary Element Method for

Axisymmetric Singular Helmholtz Acoustic Problems
Ahmed Mostafa Shaaban1,*
1Institute of Structural Mechanics, Bauhaus University Weimar, Weimar, Germany.
*Corresponding Author: Ahmed.mostafa.shaaban. Email: mohamed@uni-weimar.de.
Abstract: In this article, two applications are developed for axisymmetric singular time harmonic wave propagation
(Helmholtz equation) in acoustic problems [1]. The boundary element methods (BEM) [2] is coupled with
isogeometric analysis (IGA) with parameterizing the computational domain and approximating the unknown
solution using non-uniform rational B-splines basis functions (NURBS) [3].
In the first application, the horn problem with cylindrical symmetry is solved. Due to the axisymmetric horn body,
the three-dimensional problem is simplified in BEM from a surface integral of Helmholtz equation to a combination
of line integral - similar to a two-dimensional BEM problem – with another integral over the angle of revolution.
The discretization is applied only to the former integration. This leads to significant computational savings [4].
Similar to the first axisymmetric application, the muffler problem is studied. In addition, the muffler body includes
internal components (the extended inlet and outlet tubes in mufflers). To model these internal components in BEM,
another equation is needed in addition to the main direct integral equation (DIE). The additional equation is
formulated by considering the normal derivative of DIE on the surface of the internal component [5]. Since both
applications have singular kernels, special singularity treatments are utilized for the integration of these kernels [6].
Several numerical examples are considered and the results are compared against previously published numerical
methods to validate the present formulation and verify the efficiency of the proposed approaches.
Keywords: BEM, Isogeometric, Acoustics, Axisymmetric, Singularity.
Acknowledgement
This project is supported by the Bauhaus-Universität Weimar with a scholarship within the framework of the
“Thüuringer Graduiertenförderung”.
References
1. Keller J. B. & Givoli D. (1989). Exact Non-reflecting Boundary Conditions. Journal of Computational Physics,
82(1): 172-192.
2. Wrobel L.C. (2002). The Boundary Element Method: Applications in Thermo-fluids and Acoustics. John
Wiley & Sons, 1.
3. Simpson R.N., Bordas S.P.A., Trevelyan J. & Rabczuk T. (2012). A Two-dimensional Isogeometric Boundary
Element Method for Elastostatic Analysis. Computer Methods in Applied Mechanics and Engineering, 209-
212:87-100.
4. Udawalpola R., Wadbro E. & Berggren M. (2011). Optimization of a Variable Mouth Acoustic Horn.
International Journal for Numerical Methods in Engineering, 85: 591-606.
5. Wu T. W. & Cheng. C. Y. R. (2003). Boundary Element Analysis of Reactive Mufflers and Packed Silencers
with Catalyst Converters. Electronic Journal of Boundary Elements, Vol. 1, No.2, 218-235.
6. Guiggiani M., Krishnasamy G., Rudolphi T.J. & Rizzo F.J. (1992). A General Algorithm for the Numerical
Solution of Hypersingular Boundary Integral Equations. Journal of Applied Mechanics, 59(3).

Parameter Optimization in Penalty-Based Implicit Discontinuous Approach

for Efficient Rock Failure Analysis
Fei Zheng1,* and Xiaoying Zhuang1,*
1Instituteof Continuum Mechanics, Leibniz University Hannover, 30167 Hannover, Germany.
*Corresponding Author: Fei Zheng (fzheng@ikm.uni-hannover.de); Xiaoying Zhuang (zhuang@ikm.uni-hannover.de).
Abstract: The penalty function method is a commonly used approach in discontinuous/contact computation methods
for its simplicity in implementation. However, the choice of the penalty parameters is an important task in its
application in geotechnical problems, in terms of computational accuracy and efficiency. In discontinuum-based
approaches, such as discrete element method (DEM) [1,2] and discontinuous deformation analysis (DDA) [3,4], the
contact models are applied in mainly two scenarios, i.e., (a) the contact of real joints/cracks and (b) the bond between
particles/blocks. For contact of real joints, the contact parameter values calibrated by field measurements or lab-scale
tests results are the preferred choice. In simulation cases without the realistic data, the empirical data associated with
the geotechnical materials can also be applied. In addition to the explicit representation of jointed rock block systems,
the bonded block/particle models are also used to represent the meso-scale rock structures (e.g., grain structure of
crystalline rocks) [5]. In these applications, the contact penalty parameters (normal and tangential stiffness) and the
contact strength parameters influence the computation in the following aspects: (1) the stiffness and deformation of the
global model; (2) the potential failure pattern of a model; (3) the computational efficiency in solution of the global
stiffness matrix in implicit DDA approach; (4) the critical time step size in explicit DEM approach.
Figure 1: An illustration of the bond and contact model in discontinuum-based approach
In this work, the accuracy and efficiency of bonded block models using DDA approach are investigated. With the
implicit time integration approach, DDA methods have the advantages of unconditionally stable time step and
efficient algorithmic damping for quasi-static analysis of contact problems. An illustration model for the
investigated bonded block model is shown in Fig. 1. In this triangular block-based representation, the edge-edge
bonds, edge-edge contact and vertex-vertex contact controls the deformation and failure mechanism of the model.
The key parameters, including the normal and tangential stiffness in bonds and contacts, are investigated with
various combinations. The associated computational efficiency, the relationship of meso-scale contact stiffness and
strength parameters with the macro-scale deformation properties and failure patterns are evaluated. A multi-level
sensitivity analysis approach is used to investigated both the impact of changing input parameters and the sensitivity

139 DACOMA-2019
of input parameters on the computational efficiency and the macro-scale system bulk response. The strategies to
select the optimized input penalty parameters are proposed in this paper.
Keywords: Parameter Optimization, Penalty function, Discontinuous approach, Rock failure analysis.
Acknowledgement
This study was financially supported by Sofja Kovalevskaja Program from Alexander von Humboldt Foundation.
References
1. O'Sullivan, C., & Bray, J. D. (2004). Selecting a suitable time step for discrete element simulations that use the
central difference time integration scheme. Engineering Computations, 21(2/3/4): 278-303.
2. Yan, Z., Wilkinson, S. K., Stitt, E. H., & Marigo, M. (2015). Discrete element modelling (DEM) input
parameters: understanding their impact on model predictions using statistical analysis. Computational Particle
Mechanics, 2(3): 283-299.
3. Jiang, Q., Chen, Y., Zhou, C., & Yeung, M. C. R. (2013). Kinetic energy dissipation and convergence criterion
of discontinuous deformations analysis (DDA) for geotechnical engineering. Rock Mechanics and Rock
Engineering, 46(6): 1443-1460.
4. Zhu, H., Wu, W., Zhuang, X., Cai, Y., & Rabczuk, T. (2016). Method for estimating normal contact
parameters in collision modeling using discontinuous deformation analysis. International Journal of
Geomechanics, 17(5): E4016011.
5. Zhang, Y., & Wong, L. N. Y. (2018). A review of numerical techniques approaching microstructures of
crystalline rocks. Computers & Geosciences, 115: 167-187.
Application of Machine Learning to the Prediction of Displacement Capacities

of Non-ductile RC Beam-Columns
Reza Allahvirdizadeh1, *, Mohammad Reza Azadi Kakavand2, De-Cheng Feng3 and Ertugrul
Taciroglu4
1ISISE, Institute of Science and Innovation for Bio-Sustainability (IB-S), Department of Civil Engineering, University of Minho,
Guimarães, Portugal.
2 Unit of Strength of Materials and Structural Analysis, Institute of Basic Sciences in Engineering Sciences, University of
Innsbruck, Innsbruck, Austria.
3 School of Civil Engineering, Southeast University, Nanjing 210096, China.
4 Department of Civil and Environmental Engineering, University of California, Los Angeles, CA, USA.
*Corresponding Author: Reza Allahvirdizadeh. Email: allahvirdizadeh@gmail.com.
Abstract: Provisions in modern design standards attempt to preclude less ductile or brittle failure modes in lateral
load resisting systems, as their occurrence may lead to catastrophic failures under extreme loading events, such as
earthquakes. Such modes depend on the type and characteristics of the structural element/connection. For reinforced
concrete (RC) beam-column components, the most desirable failure mode is the flexural (and not the shear, or axial)
mode. Observations from previous earthquakes reveal that a significant portion of existing inventories of structural
frames do not necessarily satisfy modern expectations, as shear and axial failure modes were frequently recorded in
post-event surveys. Given the existing vulnerabilities, there is a need to efficiently and accurately predict the seismic
behavior of large numbers of buildings. Prior efforts in this realm focused on experimental characterization of brittle
failure modes, data from which fed combinations of analytical and statistical studies that yielded numerous
predictive formulae/models [1-3]. These prior studies have identified the key parameters that control the hierarchy
of failure models, which include longitudinal/transversal reinforcement ratios, axial load level, plastic-to-nominal
shear force ratio, sectional dimensions, and transverse reinforcement spacing [4-8]. The numbers of these parameters
hints at the complex nature of the problem, and despite a fair number of attempts, considerable errors are still
encountered in validation studies on presently available predictive models [3]. In the present study, we make use of
an extensive database that contained data from 326 rectangular RC beam-columns and carry out an exhaustive
training of an Artificial Neural Network (ANN) to predict the lateral displacement (drift) capacities. A variety of
ANN architectures with different hidden layers and neurons are considered and the data set is randomly divided into
training and testing subsets producing a large number of ANN realizations for evaluation. The accuracy of the most
precise architecture is significantly higher than any of the prior predictive models considered in this study.
Keywords: ANN, RC beam-column elements, Displacement capacity prediction, Flexure-shear-axial failure,

Statistical analyses.
References
1. Elingwood, B.R. (2001). Earthquake risk assessment of building structures. Reliability Engineering & System
Safety, 74(3): 251-262.
2. Chao, S.H., Goel, S.C. & Lee, S.S. (2007). A seismic design lateral force distribution based on inelastic state of
structures. Earthquake Spectra, 23(3): 547-569.
3. Azadi Kakavand, M.R. & Allahvirdizadeh, R. (2019). Enhanced empirical models for predicting the drift capacity
of less ductile RC columns with flexural, shear or axial failure modes. Frontiers of Structural and Civil Engineering
(Accepted).
4. Kato, D. & Ohnishi, K. (2002). Axial load carrying capacity of reinforced concrete columns under lateral load
reversals. In: Proceedings of the 3rd US-Japan Workshop on Performance-Based Earthquake Engineering
Methodology for Reinforced Concrete Building Structures, 247-255.

141 DACOMA-2019
5. Pjuol, S., Ramirez, J.A. & Sozen, M.A. (1999). Drift capacity of reinforced concrete columns subjected to cyclic
shear reversals. Seismic Response of Concrete Bridges, Special Publication, American Concrete Institute, 187: 255-
274.
6. Pjuol, S., Sozen, M.A. & Ramirez, J.A. (2000). Transverse reinforcement for columns of frames to resist
earthquakes. Journal of Structural Engineering, 126(4): 461-466.
7. Elwood, K.J. (2002). Shake table tests and analytical studies on the gravity load collapse of reinforced concrete
frames. Dissertation for the Doctoral Degree, University of California, USA.
8. Zhu, L., Elwood, K.J., Haukaas, T. (2007). Classification and seismic safety evaluation of existing reinforced
concrete columns. Journal of Structural Engineering, 133(9): 1316-1330.
Damage Evolution and Energy Absorption Modeling of Laminated GFRP

Plate with Weave Pattern Variation Under High Velocity Impact
Ilyani Akmar Abu Bakar1*, Muhammad Fakhrullah Ramli2, Zarina Itam3 and Mohd Khairul
Kamarudin1
1Universiti Teknologi MARA Shah Alam, 40450 Shah Alam, Selangor, Malaysia.
2Quantum Springs Sdn. Bhd., 36000 Teluk Intan, Perak, Malaysia.
3College of Engineering, Universiti Tenaga Nasional, Jalan Ikram-Uniten, 43000 Kajang, Selangor, Malaysia.
*Corresponding Author: Ilyani Akmar Abu Bakar. Email: ilyan104@uitm.edu.my
Abstract: The effectiveness design of composite structures should be understand the impact induced damage
mechanism in laminates especially in military and structural applications. It starts with the development of
numerous experimental and numerical analysis to study the dynamic response of the composite structures in the
dynamic loading phenomena. It is considered that the numerical analysis would significantly saving the
computational cost in predicting the ballistic impact as experimental work of damage evolution can be prohibitively
expensive. This study aims to investigate the damage evaluation and energy absorption performance characteristics
of laminated GFRP plate with weave structure as the variable under high velocity impact in numerical analysis view
point. TexGen software is used to generate the weave patterns whilst ABAQUS is utilized to simulate the explicit
three-dimensional finite element analysis. A series of ballistic impact simulations has been performed on proposed
woven GFRP plates with a conical nose shaped projectile. Selecting a proper material models and contact definition
is the major criteria to ensure the accuracy of the modeling. The damage evolution model based on Hashin’s criteria
has been assigned to predict the failure of the laminates. The projectile has a conical nose shaped with 6.24 mm of
diameter, 10 g mass and 13.83 mm in length. The impact velocity is ranged from 200 to 450 m/s. A good correlation
between the simulated composite plate with selected weave patterns has been drawn in terms of damage evolution
and energy absorption criteria during impact.
Keywords: Damage evolution, Weave patterns, Energy absorption, Ballistic impact.
Acknowledgement
The authors would like to give and endless gratitude to Faculty of Civil Engineering UiTM, Shah Alam for the
support given in completing this study. This study was carried out in collaboration between all authors. Ilyani
Akmar Abu Bakar and Mohd Khairul Kamarudin were involved in conceptualizing the problem, and worked on the
whole framework of this study. Muhammad Fakhrullah Ramli conducted the numerical analysis and Zarina Itam
contributed to the reviewing of the manuscript. Last but not least, to the anonymous reviewers of this study for their
comments and suggestions.
References
1. Cantwell, W. J., & Morton, J. (1989). Comparison of the low and high velocity impact response of
CFRP. Composites, 20: 545-551.
2. Prewo, K. M. (1980). The importance of fibres in achieving impact tolerant composites. Philosophical
Transactions of the Royal Society of London. Series A, Mathematical and Physical Sciences, 294: 551-558.
3. Aslan, Z., Karakuzu, R., & Okutan, B. (2003). The response of laminated composite plates under low-velocity
impact loading. Composite structures, 59: 119-127.
4. Liu, D., & Malvern, L. E. (1987). Matrix cracking in impacted glass/epoxy plates. Journal of Composite
Materials, 21: 594-609.
5. Liou, W. J., & Sun, C. T. (1987). Dynamic response of laminated composite plates using a three-dimensional
hybrid-stress finite-element formulation.
143 DACOMA-2019
6. Gama, B. A., Rahman, M., & Gillespie Jr, J. W. (2004). Energy absorbing damage mechanisms in quasi-static
punch shear of thick-section composites. CD Proceedings, SAMPE, 909-21.
7. Onyechi, P. C., Edelugo, S. O., Chukwumuanya, E. O., & Obuka, S. P. N. (2014). Ballistic penetration response
of glass fibre reinforced polyester (GFRP) composites: Body amour. International Journal of Scientific &
Technology Research, 3: 226-237.
8. Wen, H. M. (2001). Penetration and perforation of thick FRP laminates. Composites Science and Technology,
61, 1163-1172.
9. Siva Kumar, K., & Balakrishna Bhat, T. (1998). Response of composite laminates on impact of high velocity
projectiles. In Key Engineering Materials (Vol. 141, pp. 337-348). Trans Tech Publications.
10. Velmurugan, R., & Sikarwar, R. S. (2014). Ballistic impact on glass/epoxy composite laminates. Defence
Science Journal, 64: 393-399.
11. Chapra, S. C., & Canale, R. P. (1998). Numerical methods for engineers (Vol. 2). New York: Mcgraw-hill.
12. Barre, S., Chotard, T., & Benzeggagh, M. L. (1996). Comparative study of strain rate effects on mechanical
properties of glass fibre-reinforced thermoset matrix composite. Composites Part A: Applied Science and
Manufacturing, 27: 1169-1181.
13. Zheng, D. (2007). Low velocity impact analysis of composite laminated plates (Doctoral dissertation, University
of Akron).
14. Choi, H. Y., & Chang, F. K. (1991). Impact damage resistance of graphite/epoxy laminated
composites. Polymer Engineering & Science, 31: 1294-1300.
15. Akmar, A. I., Kramer, O., & Rabczuk, T. (2018). Probabilistic multi-scale optimization of hybrid laminated
composites. Composite Structures, 184: 1111-1125.
16. Majeed, W. I., & Tayeh, F. H. (2015). Stability and Dynamic Analysis of Laminated Composite Plates. Journal
of Engineering, 21: 139-159.
17. Pius, C., Nnaemeka, O., Cornelius, O., & Chinenye, A. (2014). Energy Absorption Capacity of A GRFP
Composite under Impact of High Velocity Projectiles. International Journal of Scientific & Technology
Research, 3: 1-12.
An experimental study on natural frequencies measurement in steel

plates/beams using PAK-Mobile MKII system
Soheil Gohari1,*, Mohammadreza Izadifar2, Timon Rabczuk3, Colin Burvill1, Saeed Mouloodi1 and
Naeim Akbari Shahkhosravi1
1 College of Engineering and Science, Victoria University, Melbourne, VIC 8001, Australia
2 Institute of functional interfaces, KIT Campus Nord, Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldschafen,
Germany
3 Division of Computational Mechanics, Ton Duc Thang University, Ho Chi Minh City, Viet Nam
*Corresponding Author: Soheil Gohari, Email: soheil.gohari@unimelb.edu.au
Abstract: PAK Mobil MKII is a compact mobile multi-channel measuring system used for vibration and acoustic
analysis of engineering structures. It can be universally configured for any common measurement pickups and is
easy to adapt to the task. In this paper, an experimental investigation into obtaining the natural frequencies of thin
plates made of steel is conducted. The experimental results are then compared with the theoretical results. Good
agreement between the results is observed. The results presented in the paper can be used as a useful guideline for
engineers and industries.
Keywords: PAK Mobil MKII system, Experimental study, Natural frequencies, Steel Plates/Beams.
References
1. Kistler Instrument Corporation, 75 John Glenn Drive, Amherst NY 14228, Phone 716-691-5100, Fax 716-691-
5226, e-mail: kicsales@kistler.com, www.kistler.com.
2. www. imechanica.org/node/5208.
3. www.vibrationdata.com/StructuralDC.htm.
4. www.iop.org/EJ/article/0964-1726/15/4/020/sms6_4_020.pdf.

Assessment of pipe-pile integrity using transverse velocity signal -- Numerical

simulation
Zhitang Lu1,*, Dongjia Liu1, Zili Dai2, Yang Guo3 and Zhaibang Ke3
1
Department of Geotechnical Engineering, Hefei University of Technology, 230009 Hefei, China.
2
Department of Geoscience, Shimane University, 6908504 Matsue, Japan.
3
Anhui institute of Building Research and Design, 230000 Hefei, China.
*
Corresponding Author: Zhitang Lu. Email: zhitang_lu@163.com.
Abstract: In order to study the feasibility of transverse velocity response for the evaluation of pipe-pile integrity,
3D (a three-dimensional) finite integration model of pile-soil system is established, and the numerical analysis of
low-strain integrity testing of the pipe-pile containing cracks is carried out. The vertical and transverse velocity
responses of the pile-top are obtained using the EFIT (elastodynamic finite integration technique), and the accuracy
of the algorithm is verified via comparing with the measured data. The characteristics of the reflected waves due to
the cracks are analyzed. It is found that the transverse velocity response can be used to determine the integrity of
pipe-piles and has advantages in identifying vertical cracks. The main component of reflected waves in transverse
velocity signal is from the shear waves. The circumferential velocity at 180° from excitation point is not affected by
3D interference. The wider the excitation pulse is, the worse the identification effect of transverse crack becomes;
however, the location of transverse crack derived from the traditional 1D theory is shallow. The proposed correction
method can help to improve the performance in locating cracks. This research enriches the theory of low strain
testing and provides references for low strain testing of the pipe-pile containing cracks.
Major cracks can affect the quality and reliability of pipe-piles. Cracks might be a results of poor construction, or
can be a results of damages during transportation, and installation. Pile integrity can provide valuable information
about the presence of such cracks. However, the inspection and evaluation of integrity problems are often
challenging, mainly because these cracks are surrounded by soil and not easily accessible for visual inspection. Low
strain integrity testing can shed light on this hidden part of concrete piles, which has been widely used for assessing
the quality of piles in civil engineering. Generally, this method provides information about continuity, defects such
as necking, soil incursions, changes in cross section.
The objective of this paper is to study the feasibility of transverse velocity response for the evaluation of pipe-pile
integrity. Firstly, a pile-soil interaction model in cylindrical coordinate system is established under the low strain
integrity testing conditions. By means of the appropriate treatment of boundary conditions, the numerical solution of
an initial-boundary value problem for pipe-pile testing is obtained using the EFIT. Secondly, the wave propagation
process in the pipe-pile containing cracks is shown, and the transverse velocity histories at pile-top are illustrated.
Besides, the influences of sensor location, contact time of the hammer, sampling position, and the stiffness of
surrounding soil on testing results are analyzed in order to identify cracks more accurately.
The location of the stress and the velocity components produced by the EFIT are shown in Fig. 1, in which these
components are located at different positions. This “staggered-grid” arrangement outperforms conventional grids
concerning the implementation of boundary conditions and numerical stability.

146 DACOMA-2019
Δr
r v
Δz φ
/ v
2r(i) Δφ
z r(c) v
r(o)
Fig. 1 Location of velocity and stress components on the EFIT cell
A pipe-pile containing an axial crack is modeled as shown in Fig. 2. The velocity histories received from both the
intact pipe-pile and the defect one are plotted in Fig. 3. From Fig. 3a, it can been seen that there is little difference
between the two curves, which indicates that the vertical-velocity signals cannot be used to identify axial cracks.
However, it is shown that in Fig. 3b, reflected wave from crack can be clearly recognized in the circumferential-
velocity curve, which shows that the circumferential-velocity has a strong advantage in crack identification.
Therefore, 3D sensors are urgently needed for pipe-pile integrity testing.
Excitation position 7 000

2 000 800
300 500
Pile-top 4 mm wide × 100 mm Pile-toe

deep
Fig. 2 Dimensions of a pipe-pile containing crack (All dimensiors are in mm)
2.0
1.5
Dimensionless circumferential-velocity
Defect pipe-pile
Dimensionless vertical-velocity
1.5
1.0
Defect pipe-pile
Intact pipe-pile Intact pipe-pile
Reflected wave
1.0 from pile-toe
0.5
0.5 0.0
0.0 -0.5
Reflected wave Reflected wave
from crack from pile-toe
-0.5 -1.0
0 2 4 0 2 4 6 8 10
t/ms t/ms
(a) Vertical-velocity historyies (b) Circumferential-velocity histories

Fig. 3 Velocity histories received from pipe-pile top
Keywords: Transverse velocity response, Elastodynamic finite integration technique, Pipe-pile integrity, 3D
interference, Low-strain integrity testing.
Acknowledgement
This work was supported by the National Natural Science Foundation of China (41807266).
References
1. Davis AG. The nondestructive impulse response test in North America: 1985-2001. NDT & E International
2003; 36(4),185–193.
2. Rausche F. Non-destructive evaluation of deep foundations. In: Proceedings of the Fifth International
Conference on Case Histories in Geotechnical Engineering, New York, NY, OSP-5, 2004, 1–9.
3. Chai HY, Phoon KK, Zhang DJ. Effects of the source on wave propagation in pile integrity testing. Journal of
Geotechnical and Geoenvironmental Engineering 2010; 136(9): 200–1208.
4. Nirbhay M , Dixit A , Misra R K. (2017). Finite Element Modelling of Lamb Waves Propagation in 3D Plates
and Brass Tubes for Damage Detection. Russian Journal of Nondestructive Testing. 53(4): 308-329.
Experimental and numerical analyses on mixing uniformity of water and

saline in pipe flow
Bin Sun1, Yuanbo Lu1, *, Hongyuan Fang1, Yisheng Zhang1 and Yihua Chang1
1Zhengzhou University, No.100 Science Avenue, Zhengzhou 450001, China.
*Corresponding Author: Yuanbo Lu. Email: 18137852539@163.com.
Abstract: Two-phase flow is commonly observed in many industrial applications [1, 2]. In order to study the mixing
process of soluble two-phase in turbulent pipeline, saline was chosen as a tracer and was injected into the pipeline [3].
In the present study, the computational fluid dynamics (CFD) software ‘FLUENT’ was employed to simulate flow
fields in water–saline pipelines. Three variates (volumetric flow rate ratio, pipe diameter, and cross-flow flux) were
considered to investigate the effects of multivariate on mixing uniformity [4, 5]. The coefficient of variation (COV)
was selected as the evaluation index of mixing uniformity, effective mixing length (EML, the distance from the
fertilizer inlet to the fully mixed position) was chosen to quantitatively analyze the fully mixed position of water and
saline in pipelines[6, 7]. The results of this numerical model agree well with experimental measurements and it shows
that this model can well predict the concentration field of water and saline in pipeline. Based on the experimental
and simulated results, it was found that for the fixed flow rate ratio and cross-flow flux, the values of EML increased
significantly with increasing pipe diameters. Furthermore, dimensional analysis (D-A) and regression analysis (R-A)
were adopted to examine the influences of the three variates on EML, and their correlation coefficients of curve-
fitting equation were calculated to be 0.997 and 0.967.
Keywords: Concentration distribution, Dimensional analysis, Effective mixing length, Numerical simulation, Pipe
flow.
Acknowledgement
This research was supported by The National Natural Science Foundation of China (No. 51509224) and The
Scientific and Technological Research Program of Henan Province (No. 162102310522), for which the authors are
grateful.
References
1. Cheng, L., Ribatski, G., & Thome, JR. (2008). Two-phase flow patterns and flow-pattern maps: Fundamentals
and applications. Applied Mechanics Reviews, 61(5): 050802-050802-28.
2. Hartmann, J., Martmann, Aa., Wuijts, S., Husman, A & Hoek, J. (2018). Risk governance of potential emergin
g risks to drinking water quality: Analysing current practices. Environmental Science & Policy, 84: 97-104.
3. Samadi, F., Kruzic, A., & Prabakar. S. (2019). An Investigation of Monochloramine Recoveries under Differen
t Mixing Condition. World Environmental and Water Resources Congress, 160-172.
4. Zughbi, H.D., Khokhar, Z.H., & Sharma, R.N. (2003). Mixing in pipelines with side and opposed tees. Industri
al & Engineering Chemistry Research, 42(21): 5333-5344.
5. Zughbi, H.D. (2006). Effects of Jet Protrusion on Mixing in Pipelines with Side-tees. Chemical Engineering Re
search and Design, 84(11): 993-1000.
6. Tang, P., Hong, L., Zakaria, I., & Chao, C. (2018). Effect of manifold layout and fertilizer solution concentrati
on on fertilization and flushing times and uniformity of drip irrigation systems. Agricultural Water Manageme
nt, 200: 71-79.
7. Stanev, V. G., Iliev, F. L., Alexandrov, B. S., Hansen, S., & Vesselinov, V. V. (2018). Identification of release
sources in advection–diffusion system by machine learning combined with Green’s function inverse method. A
pplied Mathematical Modelling, 60: 64-76.

On the Generation of Design Response Spectrum for Muscat Region in Oman

Mohamed Alaa Tabnaj1, Himanshu Gaur 1, 2,* and Ram Kishore Manchiryal1
1Department of Civil Engineering, Middle East College, Muscat, Oman.
2Institute of Structural Mechanics, Bauhaus–Universität Weimar, Marienstrasse 15, D-99423 Weimar, Germany.
*Corresponding Author: Himanshu Gaur. Email: himanshugaur82@gmail.com.
Abstract: When a structure is subjected to an earthquake loading, it faces time varying loads. As a result, time-
varying stresses and displacements are developed. From the design point of view, only maximum values of these
quantities are of interest as the structure should withstand the maximum possible load subjected to it. To generate
these results, response spectrum analysis is abundantly performed, also categorized as dynamic analysis of structures.
Response spectra for a particular region is of great importance for design engineers as it is used to simulate the
behaviour of the structure when subjected to earthquake loading in that region. This analysis is also performed in
estimating the performance of existing structures. In order to generate the design response spectrum for Oman
region, a single degree of freedom model is prepared using SAP 2000 software. The model is subjected to
earthquake time histories that is generated by software program ProShake. Available earthquake records of nearby
locations of Oman (GCC countries) are also utilized in order to generate ‘Design Response Spectrum’ for Oman
region. An effort is made to generate a smooth design spectrum that can be used for the analysis. Generated ‘Design
Response Spectrum’ is capable of providing adequate earthquake resistance design of structures for Oman region.
Keywords: Design Response Spectrum, Dynamic Analysis, Earthquake Engineering.
Acknowledgement
Authors gratefully acknowledge the financial assistance provided by Middle East College, Muscat, Sultanate of
Oman for carrying out this work.
References
1. Al-Shijbi, Y., El-Hussain, I., Al-Kalbani, A., Mohamed, M. E.(2018) ‘Probabilistic seismic Hazard assessment
for the Arabian Peninsula’ Pure and Applied Geophysics.
2. Chopra. K.A (2012), Dynamics of structures: Theory and applications to earthquake engineering, Prentice Hall,
Boston.
3. Coburn.A and Spence. R (2002), Earthquake Protection, 2nd Edition, John Wiley & Sons.
4. EduPro Civil Systems (2019) ProShake 2.0 – Educational Version .
5. El-Hussain I. A. Deif, K. Al-Jabri, N. Toksoz, S. El-Hady, S. Al-Hashmi, K. Al-Toubi, Y. AlShijbi, M. Al-
Saifi, and S. Kuleli, (2012) ‘Probabilistic seismic hazard maps for the Sultanate of Oman’ MIT, Nat Hazards
64(1): 73–210.
6. El-Hussain, I., Deif , A. Al-Jabri ,K., Mohamed A. M. E. , Al-Rawas G., Tokso¨z. M.. N., Sundararajan N., El-
Hady S. Al-Hashmi S, Al-Toubi K. Al-Saifi M., and Al- Habsi Z. (2013)’ Seismic microzonation for Muscat
region, Sultanate of Oman’ Net Hazards.
7. Elnashai, A. S., and Sarno, L. (2008) Fundamentals of Earthquake Engineering. United Kingdom, WIELY.
8. Freeman, S. (2007)’ Response spectra as a useful design and analysis tool for practising structural engineers’
ISET Journal of Earthquake Technology [online] 77(1): 25-37.

149 DACOMA-2019
9. Gioncu .V and Mazzolani .F.M (2011), Earthquake engineering for structural design, Spon Press, London.
10. Gupta. K.A (1999), Response Spectrum Method in Seismic Analysis and design of structures.
11. Hancilar, K., El-Hussain, I., Sesetyan, K., Deif, A., Cakti, E. Al-Rawas, G., Safak, E., and Al-Jabri, K.
(2018)’Earthquake risk assessment for the building inventory of Muscat, Sultanate of Oman’ Nat Hazards.
12. Islam.S, Ahmed. S, and Jameel.M (2011), ‘Generation of Response Spectra Along with Time History for
Earthquake in Dhaka for Dynamic Analysis of Structure’ SUST studies 14(2): 56-68.
13. Mohamedzein. Y, Abdalla. J, and Abdelwahab.A (2004), ‘Development of Design Response Spectral for
Central Khartoum Sudan’ World Conference on Earthquake Engineering.
14. Pascucci .V, Free.M.W ,and Lubkowski.Z.A (2008) ‘Seismic Hazard and seismic design requirements for the
Arabian peninsula region’ The 14th World Conference on Earthquake Engineering.
Effects of Radial Charge Parameters of Liner on the Velocity of
Shaped Charge Jet

Mengwen Xu *
Nanjing University of Science and Technology, 210094 Nanjing, China.

Corresponding Auther: XU Mengwen. Emial: 15105142159@163.com.
Abstract: In this paper, the explosive impulse in the detonation process of shaped charge is analyzed theoretically.
Taking the shaped charge with a diameter of 56mm as an example, the influences of the ratio of the outer diameter
of the charge to the liner on the velocity of shaped charge jet is investigated with the cone angles from 30 to 70
degrees. The effects of radial charge parameters and cone angles of liner on jet velocity was further studied by
numerical simulation. The results of theoretical analysis and numerical simulation show that when the ratio of the
outer diameter of the charge to the liner is greater than 4, the velocity of the jet formed by the liner tends to be stable,
which accounts for about 17.9% of the total height of liner. The maximum velocity of jet is approximately linear
with the cone angle of liner. The theory is applicable to the calculation of jet velocity under different cone angles,
which provides a theoretical basis for the design of shaped charge.
Keywords: Liner, Shaped Charge jet, Velocity, Cone angle.
Acknowledgement
The authors gratefully acknowledge the supports by the Project Establishment of Jiangsu Postgraduate Research and
Practice Innovation Program in 2018 (No: KYCX18_0470).
References
1. Chou P.C, Carleone J, Flis W.J, et al. Improved formulas for velocity, acceleration and projection angle of
explosively driven liners [J]. Propellants Explosive Pyrotechnics, 1983:175-183.
2. Eichelberger R.J. Re-examination of the nonsteady theory of jet formation by lined cavity charges [J]. J Appl
Phy, 1955, 26(4):398-402.
3. Gurney R W. The initial velocities of fragments from bombs, shells, and grenades[R]. U. S. Army Ballistic
Research Lab, BRL Report 405, 1943. Report ARBRL-CR-00461, 1981, AD A104682.
4. Hennequin E.M. Analytical model of the shaped charge liner collapse [C]. Proc. 7th Int. Symp. On Ballistics,
The Hague, Netherlands, 19-21 April 1983.
5. Randers-Pehrson G. An improved equation for calculating fragment projection angle[C]. Proc. 2nd Int. Symp
on Ballistics, Daytona Beach, FL, 1976:9-11.
6. HUANG Zhengxiang. Theroy and practice of shaped charge [M]. Beijing institute of technology press,
2014:20-28. (in Chinese)
7. OU Yangchuping, GAO Senlie, WANG Ruce, XU Xuehua. Some problems of geometric similarity law in the
application of energy-accumulating effect engineering [J]. Explosion and Shock Waves, 1986.4(6):351-357.
DOI :10.32604/acm.2019.00203 www.techscience.com/acm

Application of Multi Objective Bees Algorithm to Access the Optimal

Generalization of Nonlinear Full Vehicle Model
Reza Khademi Zahedi1, *, Abbas Moradi 2 and Afshin Ghanbarzadeh 2

1
Bauhaus University Weimar, Marien-Strasse 15, 99423 Weimar City, Germany.
2 Shahid Chamran University of Ahvaz, Department of Mechanical engineering, Blv. Golestan, Ahwaz, Iran.
*Corresponding Author: R.KHZ. Email: reza.khademi.zahedi@gmail.com.
Abstract: A generalized nonlinear model was formulated and proposed for the dynamic analysis of nine degree of
freedom (9 DOF) model road vehicle with passive and semi-active suspension systems. The model incorporates a
nonlinear suspension spring and nonlinear damping force for passive and semi-active dampers, while the
contribution due to biodynamic of the human operator was considered to be negligible. The semi-active damper is
characterized by force generators in accordance with the control laws based on suspension mass velocity.
There is a set of optimal solutions, well-known as Pareto optimal solutions [1-3], which distinguishes significantly
the inherent natures between single- objective and multi-objective optimization problems. In this paper, we used the
Bees algorithm [4-5] for multi-objective optimization. Griffin et. al. [6], Rakheja [7], and Barak [8] have shown that
the interior vibration of a vehicle has a significant effect in comfort and road holding capability. Bouazara and
Richard [9] studied three types of suspension system (active, semi-active and passive) for an eight-degree of freedom
vibration model. Gündoğdu [10] presented an optimization of a four-degree of freedom quarter car seat and
suspension system using genetic algorithms. Alkhatib et. al. [11] applied genetic algorithm (GA) to the optimization
problem of a linear one-degree of freedom (1-DOF) vibration isolator. The general optimization problem was stated
mathematically in [12].
A systematic methodology was applied in an effort to select optimum values for the suspension damping and
stiffness parameters of 9 DOF full-car nonlinear models, subjected to road excitation. First, models involving
passive suspension dampers with constant or dual rate characteristics were considered. In addition, models with
semi-active suspensions were also examined. Moreover, special emphasis was put in modeling possible temporary
separations of the wheel from the ground. For all these models, appropriate methodologies are employed for
capturing the motions of the vehicle resulting from passing with a constant horizontal speed over roads involving an
isolated or a distributed geometric irregularity. The optimization process was based on three suitable performance
criteria, related to ride comfort, suspension travel and road holding of the vehicle and yielding the most important
suspension stiffness and damping parameters. As these criteria are conflicting, a suitable multi-objective
optimization methodology is set up using the Bees Algorithm (BA) [13]. As a result, a series of diagrams with
typical numerical results are presented and compared in both the corresponding objective spaces (in the form of
classical Pareto fronts) and parameter spaces.
Keywords: Multi Objective Bees Algorithm, Ride Comfort, Road Holding, Linear and Nonlinear Model,
Suspension System.
Reference
1. Fonseca, C. M., & Fleming, P. J. (1993, June). Genetic Algorithms for Multiobjective Optimization:
FormulationDiscussion and Generalization. In Icga (Vol. 93, No. July, pp. 416-423).
2. Coello, C. A., & Christiansen, A. D. (2000). Multiobjective optimization of trusses using genetic algorithms.
Computers & Structures, 75(6): 647-660.
3. Coello, C. A. C., Lamont, G. B., & Van Veldhuizen, D. A. (2007). Evolutionary algorithms for solving multi-
objective problems (Vol. 5, pp. 79-104). New York: Springer.
4. Pham, D. T., Ghanbarzadeh, A., Koc, E., Otri, S., Rahim, S., & Zaidi, M. (2005). The bees algorithm.
Technical Note, Manufacturing Engineering Centre, Cardiff University, UK.
152 DACOMA-2019
5. Pham, D. T., Ghanbarzadeh, A., Koç, E., Otri, S., Rahim, S., & Zaidi, M. (2006). The bees algorithm—a novel
tool for complex optimisation problems. In Intelligent Production Machines and Systems (pp. 454-459).
Elsevier Science Ltd.
6. Griffin, M. J., Parsons, K. C., & Whitham, E. M. (1982). Vibration and comfort IV. Application of
experimental results. Ergonomics, 25(8): 721-739.
7. Rakheja, S. (1985). COMPUTER AIDED DYNAMIC ANALYSIS AND OPTIMAL DESIGN OF
SUSPENSION SYSTEMS FOR OFF-ROAD TRACTORS.
8. Barak, P. (1991). Magic numbers in design of suspensions for passenger cars. SAE transactions, 1698-1733.
9. Bouazara, M., & Richard, M. J. (2001). An optimization method designed to improve 3-D vehicle comfort and
road holding capability through the use of active and semi-active suspensions. European journal of mechanics-
A/Solids, 20(3): 509-520.
10. Gündoğdu, Ö. (2007). Optimal seat and suspension design for a quarter car with driver model using genetic
algorithms. International Journal of Industrial Ergonomics, 37(4): 327-332.
11. Alkhatib, R., Jazar, G. N., & Golnaraghi, M. F. (2004). Optimal design of passive linear suspension using
genetic algorithm. Journal of Sound and vibration, 275(3-5): 665-691.
12. Osyczka, A. (1985). Multicriteria optimization for engineering design. In Design optimization (pp. 193-227).
Academic Press.
13. Pham, D. T., Ghanbarzadeh, A., Koc, E., & Otri, S. (2006, July). Application of the bees algorithm to the
training of radial basis function networks for control chart pattern recognition. In Proceedings of 5th CIRP
international seminar on intelligent computation in manufacturing engineering (CIRP ICME’06), Ischia, Italy
(pp. 711-716).
Machine learning in the engineering crack identification problems

Chao Zhang1 and Cuixia Wang1, *
1Zhengzhou University, No.100 Science Avenue, 450001 Zhengzhou, China.
*Corresponding Author: Cuixia Wang. Email: cuixia.wang@outlook.com.
Abstract: Machine learning (ML) method is used for the numerical solution of crack identification problems in two-
dimensional piezoelectric structures under dynamic loads. The machine learning method is a flexible numerical
approach to make predictions based on some pre-obtained training data without the need of any explicit functions.
Fracture is modeled by the extended finite element method (XFEM), which allows the use of a single regular mesh
for a large number of iterations with different crack geometries. The effects of various dynamical test loads on the
crack identification are investigated. In this work, the XFEM-ML methodology is applied to two-dimensional
electromechanical problems where different number of cracks is considered. The results show that this methodology
can be effectively employed for detecting multiple cracks in piezoelectric materials.
Keywords: Machine learning, Crack identification, XFEM, Dynamic load.
References
1. Zhang, C., Wang, C., Lahmer, T., He, P., & Rabczuk, T. (2016). A dynamic XFEM formulation for crack
identification. International Journal of Mechanics and Materials in Design, 12(4): 427-448.
2. Zhang, C., Nanthakumar, S. S., Lahmer, T., & Rabczuk, T. (2017). Multiple cracks identification for
piezoelectric structures. International Journal of Fracture, 206(2): 151-169.
3. Zhang, C., Zhao, J., & Rabczuk, T. (2018). The interface strength and delamination of fiber-reinforced
composites using a continuum modeling approach. Composites Part B: Engineering, 137: 225-234.
4. Wan, J., Tan, Y. W., Jiang, J. W., Chang, T., & Guo, X. (2018). Machine learning valence force field model.
arXiv preprint arXiv:1808.01714.

Road surface damage detection using neural network

Sara Changizi1, Jafar Amani-Dashlejeh2, *
1Faculty of Architecture, Payame Noor University of Hormozgan, Bandar-Abbas branch, Iran.
2Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, the Netherlands.
*Corresponding Author: Jafar Amani-Dashlejeh. Email: J.AmaniDashlejeh@tudelft.nl.
Abstract: In the road safety inspection, a road surface damage detection is a time consuming and an expensive
process. In this contribution, a deep learning model based on Convolutional Neural network (CNN) algorithm is
employed for prediction of presence or absence of damage in a road surface. Also for detection of damage type, the
model is trained and tested for different damage types: longitudinal crack, lateral crack and alligator damage. The
CNN model is implemented using the Keras library at front-end and TensorFlow at back-end and also openCV
library is used for image processing. The comparison of the damage type predicted using the proposed model with
the experimental data proves the fidelity of the proposed model for prediction of damage type in a road surface and
its application for road safety inspection.
Keywords: Damage type detection, Image processing, Deep learning, Neural network.
References
1. Maeda, H., Sekimoto, Y., Seto, T., Kashiyama, T., & Omata, H. (2018). Road Damage Detection and
Classification Using Deep Neural Networks with Smartphone Images. Computer-Aided Civil & Infrastructure
2. Amani, J., & Moeini, R. (2012). Prediction of shear strength of reinforced concrete beams using adaptive
neuro-fuzzy inference system and artificial neural network. Scientia Iranica A, 19: 242-248.

Modeling the stability of a soil-rock-mixture slope based on the digital image

technology and numerical manifold method
Tao Chen1, 3, Yongtao Yang1, * and Hong Zheng2
1 State
Key Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics, Chinese Academy
of Sciences, Wuhan 430071, China
2 Key Laboratory of Urban Security and Disaster Engineering, Ministry of Education, Beijing University of Technology, Beijing
100124, China
3 University of Chinese Academy of Science, Beijing, 100049, P. R. China
*Corresponding Author: Yongtao Yang. Email: ytyang@whrsm.ac.cn
Abstract: As a widely distributed geotechnical material, the soil-rock mixtures (SRM), which are composed of
high-strength block stones, fine-grained soil and pores, had formed since the Quaternary period. In regarding to the
physical and mechanical properties, there are significant differences between the soil and rock materials. Hence the
SRMs show extremely heterogeneous character. With the development of large and even giant hydropower projects
in the southwest areas of China, many SRM slopes with different scales are encountered in the reservoir areas or/and
near the dam areas [1-3]. The failure of these SRM slopes may result in great casualties and economic losses.
Therefore, evaluating the stability and proposing reinforcing measures for the weak areas of the SRM slopes are of
great importance to the safety of the project during the process of construction and operation. With the rapid
development of computer hardware technology and image theory, digital image processing has provided an
indispensable technical means for the development of modern science and technology. It has been applied to various
subject areas, such as in the medicine, biology, aerospace and geotechnical engineering, and has achieved satisfying
results. In the present paper, the digital image processing technology is employed to establish the real structure
model of a soil-rock-mixture slope. The numerical manifold method (NMM) can solve the problems of continuity
and discontinuity in geotechnical engineering in a unified way [4]. For the analysis of discontinuous problems, the
unique and superior features of the NMM include the following [5] : (1) large displacement, sliding along joint faces
and even discrete blocky movement are allowed; (2) by simply cutting mathematical patches with joints or material
interfaces, discontinuity can be captured in a natural way without employing jump functions; and (3) the employed
discontinuity representation algorithm is suitable for any number and any geometrical complexities such as
intersections, junctions and the branching of discontinuities in one element. On this basis, the numerical manifold
method is used to simulate the stability of a soil-rock mixture slope. An improved shear strength reduction technique
is implemented into the numerical manifold method to evaluate the stability of the slope, as well as to obtain the
factor of safety (FOS). In this technique [6], with the application of the relationship between the internal friction
angle φ and the Poisson's ratio υ, υ is also adjusted for the purpose of eliminating the spurious plastic deformation
that may happen in the SRM slope. The simulation results show that: (1) the improved numerical manifold method
is able to accurately predict the FOSs of slopes; (2) it is very difficult to form a slipping surface passing from the toe
of the SRM slope to the top of the slope; (3) the rock blocks in the SRM slope make the internal stress field of the
slope became very heterogeneous; (4) the sliding zone of the soil-rock mixed slope obtained by shear strength
reduction method is very zigzag.
Keywords: Soil-Rock Mixture Slope, Digital Image Processing, Shear Strength Reduction Method, Numerical
Manifold Method, Slope Stability Analysis.
Acknowledgement
This study is supported by the National Natural Science Foundation of China, under the grant 51609240, 11572009,
and 51538001 and the National Basic Research Program of China (973 Program), under the grant 2014CB047100.
References

156 DACOMA-2019
1. Xu, WJ., Yue, ZQ., & Hu, RL. (2008). Study on the mesostructure and mesomechanical characteristics of the
soil-rock mixture using digital image processing based finite element method. International Journal of Rock
Mechanics and Mining Sciences, 45(5): 749-762.
2. Xu, WJ., Wang, YJ., Chen, ZY., &Hu, RL. (2008). Stability analysis of soil-rock mixed slope based on digital
image technology. Rock and Soil Mechanics, 28: 341-346.
3. CHEN, S., YUE, ZQ., & THAM, LG. (2004). Digital Image-based numerical modeling method for prediction
of inhomogeneous rock failure. Rock Mechanics and Mining Sciences, 41: 939－957.
4. Shi, GH. (1991). Manifold method of material analysis. In: Proceedings of the transactions of the Ninth Army
Conference on Applied Mathematics and Computing, 57-76.
5. Wu, Z., & Wong, LNY. (2013). Modeling cracking behavior of rock mass containing inclusions using the
enriched numerical manifold method. Engineering Geology, 162(4): 1-13.
6. Zheng, H., Li, CG., Li, ZF., & Ge, XR. (2002). Finite element method for solving the factor of safety. Chinese
Journal of Geotechnical Engineering, (05): 626-628.
DEM study of modeling method and thermal cracking of sandstone

Zhang Wengang1, 2, 3, *, Wang Zhenyu2, Wu Chongzhi2, Gao Xuecheng2
1Key Laboratory of New Technology for Construction of Cities in Mountain Area, Chongqing University, Chongqing 400045,
China
2School of Civil Engineering, Chongqing University, Chongqing 400045, China
3National Joint Engineering Research Center of Geohazards Prevention in the Reservoir Areas, Chongqing University,
Chongqing 400045, China

*Corresponding Author: Zhang Wengang. Email: zhangwg@ntu.edu.sg / cheungwg@126.com
Abstract: Sandstone is a common sedimentary rock composed mainly of sand-sized mineral particles with some
cement (clay or quartz etc.). Generally, the sandstone structure, as well as the shape of the mineral particles have a
significant influence on the properties of sandstone. This paper presents a novel method to simulate sandstone with
any shaped mineral particles and investigates the effects of particle shape and sandstone structure on the thermal
cracking responses. A random angular bend (RAB) algorithm and overlap detection algorithm is proposed to build
the numerical model of sandstone. Three representative parameters are used to quantitatively control the shape
feature of mineral particles in term of three major aspects, including form, roundness and surface texture,
respectively. The overlap detection algorithm is developed to address the difficulties associated with spatial
allocation of irregularly shaped particles. Finally, the effect of particle shape and sandstone structure on thermal
cracking of sandstone are investigated based on discrete element method. Some useful conclusions for numerical
simulation on Sandstone are arrived at.
Keywords: Particle shape, Discrete element method, Sandstone structure, Overlap detection, Thermal cracking
Acknowledgement
The study is partially supported by the Natural Science Foundation of Chongqing, China (cstc2018jcyjAX0632) and
Venture & Innovation Support Program for Chongqing Overseas Returnees (cx2017123).
References
1. Zahn, C.T. & Roskies, R.Z. (1972). Fourier descriptors for plane closed curves. IEEE Transactions on
computers, 100: 269-281.
2. He, X., Li, Q., Chen, H.K., et al. (2017). Experimental analysis on physical and mechanical properties of
thermal shock damage of granite. Frattura Ed Integrita Strutturale, 181-190, DOI: 10.3221/Igf-Esis.39.18.
3. Bakun-Mazor, D., Hatzor, Y.H., Glaser, S.D., et al. (2013). Thermally vs. seismically induced block
displacements in Masada rock slopes. International Journal of Rock Mechanics and Mining Sciences, 61: 196-
211, DOI: 10.1016/j.ijrmms.2013.03.005.

Hip Fracture Risk Assessment based on Different Failure Criteria using QCT-
based Finite Element Modeling
Hossein Bisheh1,* and Yunhua Luo1,2
1 Department of Mechanical Engineering, University of Manitoba, Winnipeg, Manitoba, R3T 5V6, Canada.
2 Department of Biomedical Engineering, University of Manitoba, Winnipeg, Manitoba, R3T 5V6, Canada.
*Corresponding Author: Hossein Bisheh. Email: Hossein.Bisheh@umanitoba.ca.
Abstract: The most common injury of the elderly during the sideways fall is hip fracture. It was reported that hip
fracture may lead to a long term disability and death of individuals [1]. The total number of hip fracture is increasing
over the world [2]. Therefore, a special attention must be dedicated on this important issue in order to provide
appropriate plans for prevention and treatment of hip fracture. Precise evaluation of hip fracture risk leads to reduce
hip fracture occurrence in individuals and assist to check the effect of a treatment. Accurate assessment of hip
fracture risk in the elderly also helps us to consider proper preventing schemes such as effective design of hip
protectors and or providing proper treatment plans to protect the elderly against future hip fracture.
By integrating an imaging technology such as Dual-Energy X-ray Absorptiometry (DXA) or Quantitative Computed
Tomography (QCT) and a numerical method such as the Finite Element (FE) Method, a category of more reliable
tools for assessing hip fracture risk has been developed which does not have the limitations of statistical models and
methods which are based on measuring bone mineral density (BMD). However, in numerical and computational
models such as QCT-based finite element models, choosing a proper failure criterion based on bone microstructure
is very important for accurate assessment of hip fracture risk. The human femur consists of inhomogeneous (porous)
cancellous bone and nearly homogenous cortical bone, so, their failure mechanism is totally different due to their
different microstructures. Failure mechanism of the cancellous bone is often in the form of buckling, and the failure
of denser cancellous bone and the cortical bone is mostly characterized by local cracking [3,4]. Although stress- and
strain-based failure criteria are accurate for ductile materials such as metal, they may not be accurate for bones
because bone is categorized as a brittle material [5]. The tensile strength of bones is smaller than their compressive
strength, indicating bone should be classified as a brittle material [5]. Due to this property of bones, strain energy
criterion which is a combination of both stress and strain effects may lead to more accurate assessment of hip
fracture risk. In the literature, hip fracture risk was usually estimated using the von Mises stress and strain criteria
[6-9], the maximum principle stress and strain criteria [10-13], the maximum shear stress criterion [14], the
maximum distortion energy criterion [14], and the strain energy criterion [15,16]. To the best of our knowledge,
there is no comparative study in hip fracture risk assessment by using different failure criteria. The aim of this study
is to define hip fracture risk index using the strain energy, the von-Mises stress, and the von-Mises strain criteria and
compare the calculated fracture risk indices using these criteria at the critical regions of femur.
A subject-specific QCT-based finite element model is introduced to evaluate hip fracture risk using the strain energy,
the von-Mises stress, and the von-Mises strain criteria during the single-leg stance and the sideways fall
configurations. We construct FE model of the femur from the QCT image of clinical cases and then simulate the
single-leg stance and sideways fall configurations by finite element analyses, and finally fracture risk indices are
assessed in the critical regions of femur, regions with high stress and strain [17, 18], using the strain energy, the von-
Mises stress, and the von-Mises strain criteria, and then we evaluate and discuss about their rate of conservation and
accuracy based on the bone failure mechanism.
It was found that based on these criteria, the hip fracture risk at the femoral neck and the intertrochanteric region is
higher than other parts of the femur, probably due to the larger amount of cancellous bone in these regions. The
results of this study also showed that the strain energy criterion gives more reasonable assessment of hip fracture
risk based on the bone failure mechanism and the von-Mises strain criterion is more conservative than two other
criteria and leads to higher hip fracture risk indices.

159 DACOMA-2019
Keywords: Hip Fracture Risk, Finite Element Model, Strain Energy, von-Mises Stress, von-Mises Strain.
References
1. Resnick, N.M. & Greenspan, S.L. (1989). ‘Senile’ osteoporosis reconsidered. Journal of the American Medical
Association (JAMA), 261(7): 1025-1029.
2. Gullberg, B., Johnell, O., & Kanis, J.A. (1997). World-wide Projections for Hip Fracture. Osteoporosis
International, 7(5): 407-413.
3. Mirzaei, M., Keshavarzian, M., & Naeini, V. (2014). Analysis of strength and failure pattern of human proximal
femur using quantitative computed tomography (QCT)-based finite element method. Bone, 64: 108-114.
4. Stölken, J.S., & Kinney, J.H. (2003). On the importance of geometric nonlinearity in finite-element simulations
of trabecular bone failure. Bone, 33(4): 494-504.
5. Cordey, J., & Gautier, E. (1999). Strain gauges used in the mechanical testing of bones Part I: Theoretical and
technical aspects. Injury, 30 (Supplement 1), SA7-SA13.
6. Lotz, J.C., Cheal, E.J., & Hayes, W.C. (1991). Fracture Prediction for the Proximal Femur Using Finite
Element Models: Part I-Linear Analysis. Journal of Biomechanical Engineering, 113(4): 353-360.
7. Lotz, J.C., Cheal, E.J. & Hayes, W.C. (1991). Fracture Prediction for the Proximal Femur Using Finite Element
Models: Part II-Nonlinear Analysis.” Journal of Biomechanical Engineering, 113(4): 361-365.
8. Keyak, J.H., Rossi, S.A., Jones, K.A., & Skinner, H.B. (1997). Prediction of femoral fracture load using
automated finite element modeling. Journal of Biomechanics. 31(2): 125-133.
9. Luo, Y., Ferdous, Z. & Leslie, W.D. (2013). Precision study of DXA-based patient-specific finite element
modeling for assessing hip fracture risk. International Journal Numerical Methods in Biomedical Engineering,
29(5): 615-629.
10. Ota, T., Yamamoto, I., & Morita, R. (1999). Fracture simulation of the femoral bone using the finite-element
method: How a fracture initiates and proceeds.” Journal of Bone Mineral Metabolism, 17(2): 108-112.
11. Testi, D., Viceconti, M., Baruffaldi, F., & Cappello, A. (1999). Risk of fracture in elderly patients: a new
predictive index based on bone mineral density and finite element analysis. Computer Methods and Programs
in Biomedicine, 60(1): 23-33.
12. Schileo, E., Taddei, F., Cristofolini, L., & Viceconti, M. (2008). Subject-specific finite element models
implementing a maximum principal strain criterion are able to estimate failure risk and fracture location on
human femurs tested in vitro. Journal of Biomechanics, 41(2): 356-367.
13. Gong, H., Zhang, M. Fan, Y., Kwok, W.L., & Leung, P.C. (2012). Relationships between Femoral Strength
Evaluated by Nonlinear Finite Element Analysis and BMD, Material Distribution and Geometric Morphology.
Annals of Biomedical Engineering, 40(7): 1575-1585.
14. Keyak, J.H. & Rossi, S.A. (2000). Prediction of femoral fracture load using finite element models: an
examination of stress- and strain-based failure theories. Journal of Biomechanics, 33(2): 209-214.
15. Kheirollahi, H. & Luo, Y. (2015). Assessment of Hip Fracture Risk Using Cross-Section Strain Energy
Determined by QCT-Based Finite Element Modeling. BioMed Research International, 2015, Article ID
413839, 15 pages, https://doi.org/10.1155/2015/413839.
16. Kheirollahi Nataj Bisheh, H. (2015). Assessment of Hip Fracture Risk Using Cross-Section Strain Energy
Determined from QCT-Based Finite Element Model. M.Sc. Thesis, University of Manitoba.
17. Kheirollahi, H. & Luo, Y. (2015). Identification of High Stress and Strain Regions in Proximal Femur during
Single-Leg Stance and Sideways Fall Using QCT-Based Finite Element Model. International Journal of
Biomedical and Biological Engineering, 9(8): 633-640.
18. Kheirollahi, H. & Luo, Y. (2017). Understanding Hip Fracture by QCT-Based Finite Element Modeling.
Journal of Medical and Biological Engineering, 37(5): 686-694.
Parametric Study of Hip Fracture Risk using QCT-based Finite Element

Model
Hossein Bisheh1,* and Yunhua Luo1,2
1Department of Mechanical Engineering, University of Manitoba, Winnipeg, Manitoba, R3T 5V6, Canada.
2Department of Biomedical Engineering, University of Manitoba, Winnipeg, Manitoba, R3T 5V6, Canada.
*Corresponding Author: Hossein Bisheh. Email: Hossein.Bisheh@umanitoba.ca.
Abstract: Different parameters such as age, height, weight, and body mass index (BMI) may influence on the hip
fracture risk in the elderlies which is the most common injury for them especially during the sideways fall. Hence, a
study considering various parameters can be useful to prevent probable hip fracture in the elderlies and propose an
appropriate treatment for the case of hip fracture occurrence. Parametric studies of hip fracture risk based on the
gender, age, height, weight, and BMI of subjects lead to a precise estimation of hip fracture relation to the
specifications and lifestyle of a community.
In this study, the correlations between the hip fracture risk and different parameters such as age, body height, body
weight, and BMI are investigated and attained for 30 females (totally 60 right and left femurs) and 30 males (totally
60 right and left femurs). To this purpose, we construct subject-specific QCT-based finite element model of the
femur using Mimics and ANSYS software based on the QCT data of subjects and then hip fracture risk index is
calculated based on the strain energy criterion [1, 2] using MATLAB codes at the critical regions of the femur,
locations usually receiving high stress and strain within sideways fall [3, 4]. The hip fracture risk index versus age,
body height, body weight, and BMI (representative of body shape) at the three critical cross-sections of femur
during the single-leg stance and the sideways fall for 30 females and 30 males are shown in the scattered plots. The
correlation coefficients (𝑟𝑟) and the corresponding statistical significance (𝑝𝑝-value) are calculated using in-house
MATLAB codes for the parameters considered in this study.
According to the obtained results, there is an increasing trend of hip fracture risk with age, body height, body weight,
and BMI at the three critical regions of the femur in both women and men during the single-leg stance and the
sideways fall. The correlations of hip fracture risk with age and height are not much significant (𝑝𝑝 > 0.05) for the
clinical cases investigated in this study. It means that at least for the clinical cases investigated in this study, we
cannot conclude that the hip fracture risk strongly increases with the age and it may depend on the individual’s
health status and lifestyle. The results of this study also show that there is a strong correlation between the hip
fracture risk with the body weight and BMI (𝑝𝑝 < 0.05). It indicates that subjects with higher body weight and BMI,
overweight and obese subjects, are more prone to the hip fracture risk.
Keywords: Hip Fracture Risk, Parametric Study, Finite Element Model, Strain Energy.
References
1. Kheirollahi, H. & Luo, Y. (2015). Assessment of Hip Fracture Risk Using Cross-Section Strain Energy
Determined by QCT-Based Finite Element Modeling. BioMed Research International, 2015, Article ID
413839, 15 pages, https://doi.org/10.1155/2015/413839.
2. Kheirollahi Nataj Bisheh, H. (2015). Assessment of Hip Fracture Risk Using Cross-Section Strain Energy
Determined from QCT-Based Finite Element Model. M.Sc. Thesis, University of Manitoba.
3. Kheirollahi, H. & Luo, Y. (2015). Identification of High Stress and Strain Regions in Proximal Femur during
Single-Leg Stance and Sideways Fall Using QCT-Based Finite Element Model. International Journal of
Biomedical and Biological Engineering, 9(8), 633-640.
4. Kheirollahi, H. & Luo, Y. (2017). Understanding Hip Fracture by QCT-Based Finite Element Modeling.
Journal of Medical and Biological Engineering, 37(5), 686-694.

Phase field fracture in elasto-plastic solids: a length-scale insensitive model for

constitutive behaviour of quasi-brittle materials
Jianguang Fang1,2, *
1School of Civil and Environmental Engineering, University of Technology Sydney, Sydney, NSW 2007, Australia.
2Institute of Structural Mechanics, Bauhaus-University Weimar, 99423 Weimar, Germany.
*Corresponding Author: Jianguang Fang. Email: Jianguang.Fang@uts.edu.au.
Abstract: The phase-field model for fracture has been coupled with continuum plasticity to capture complex
constitutive behaviour. However, the length scale parameter in the existing coupled models must be interpreted as a
material property to match the material strength in experiments. This may lead to overly diffuse damage and the Γ-
convergence of the regularized variational model cannot be guaranteed. This study presents a coupled phase-field
fracture and plasticity model for quasi-brittle materials with emphasis on the effect on the length scale. The
proposed model is formulated in the variational framework using energetic principles [1]. The numerical
implementation of the proposed model is introduced by using the finite element method [2]. The effective yield
stress in the proposed model is calibrated to vary with the length scale parameter such that the tensile strength of the
quasi-static brittle materials remains the same. The length scale parameter is not determined by the tensile strength
of the material and can take any proper values. Five representative benchmark tests are considered to validate the
capability of the developed model. The numerical force versus displacement curves and crack paths are both in good
agreement with the corresponding experimental results. More importantly, the global response of the structures is
insensitive to the length scale though it may influence the size of the failure zone.
Keywords: Phase-field modelling, Elasto-plastic mechanics, Variational formulation, Fracture, Quasi-brittle

material.
Acknowledgement
This work was supported by The University of Technology Sydney (UTS), Australia through a Chancellor's
Postdoctoral Research Fellowship and by Australian Research Council (ARC) through Discovery Projects
(DP160104661 and DP190103752).
References
1. Fang, J., Wu, C., Li, J., Liu, Q., Wu, C., Sun, G., & Qing, L., (2019). Phase field fracture in elasto-plastic
solids: variational formulation for multi-surface plasticity and effects of plastic yield surfaces and hardening.
International Journal of Mechanical Sciences 156: 382-396.
2. Fang, J., Wu, C., Rabczuk, T., Wu, C., Ma, C., Sun, G., & Li, Q., (2019). Phase field fracture in elasto-plastic
solids: Abaqus implementation and case studies. Theoretical and Applied Fracture Mechanics 103, 102252.

Research on the Design Construction Building Information Exchange

(DCBIE) based on BIM
Juan Ju*
Shanghai JIANKE Engineering Consulting Co., Ltd, Shanghai, 200032, China.
*Corresponding Author: JU Juan. E-mail: jujuan@jkec.com.cn.
Abstract: The collaborative project management method based on BIM benchmark model, will improve our
efficiency of project management, because it represents the project as a multi objective dynamic optimization
process from original fragmented management of investment, quality and schedule to a benchmark BIM
model. The data exchange standards is fundamental and important when we talk about the exchange based on
the benchmark model from one single stage to multi stages of construction projects. In this paper, one
standard of design construction building information exchange based on BIM is put forward to realize the
collaborative project management, which using a standard data exchange file, to solve many problems such as
model reuse, information transfer and multi objective dynamic modification after design changes.
Keywords: Benchmark BIM, Design construction building information exchange, Collaborative project
management.
Renferences
1. HUANG Weizheng. (2014). Research on BIM Information Exchange in Construction Engineering,
Construction Economy, 2014(7): P106-108.
2. LI Jiancheng. (2015). WANG Guangbin. Introduction to BIM Application [M]. Shanghai: Tongji
University Press.
3. National BIM Standard-United States Version 3, National Institute of Building Sciences
buildingSMART alliance, 2015.
4. GB/T25507-2010 Industry foundation classes platform, 2010.
5. ZHOU Cheng. RESEARCH ON IDM-BASED BUILDING INFORMATION DELIVERY
STANDARD [D]. Shanghai: Shanghai Jiao Tong University, 2013.
6. MA Zhiliang, ZHANG Dongdong, MAJiankun. (2014). BIM-based Collaborative Work Model and
Information Utilization Framework for IPD Projects, Journal of Tongji University (NATURAL
SCIENCE), 2014(9): P1325-P1333.
7. CHEN Li-juan, LUO Han-bin，XIN Hong-yan. (2015). Development and Application of Large-scale
Expo Project Whole Life-cycle Management Platform Based on BIM, Journal of Civil Engineering
and Management, 2015(9): P54-P61.
8. GB50854-2013 Standard method of measurement for building construction and fitting-out works,
2013.
9. GB50300-2013 Unified standard for constructional quality acceptance of building engineering,2013.
DOI : 10.32604/acm.2019.00221 ww.techscience.com/acm

Optimization and Studying the Effect of Machining Parameters in the

Cortical Bone Turning by Sensitivity Analysis
Vahid Tahmasbi1, Mehdi Safari1, Jalal Joudaki1,*
1Arak University of Technology, Daneshgah Street, 38181-41167 Arak, Iran.
*Corresponding Author: V. Tahmasbi. Email: tahmasbi@arakut.ac.ir.
Abstract: Machining and cutting of cortical bones are very common and important in the field of orthopedic
surgeries. Considerable advances in bone machining are obtained by using CNC machines and automatic surgery
robots but still, researches are needed to investigate the effects of machining parameters in bone machining. Surface
quality is a decisive parameter and choosing the proper tool and machining parameter is vital [1]. In this paper, for
the first time, the effect of machining parameters on surface quality obtained in the cortical bone turning is studied.
The machining parameters included in the investigation are cutting speed, feed rate and depth of cut and the
response surface methodology is used to analyze the obtained surface quality according to a second order regression
model. The sensitivity of surface quality to the input parameters was measured by applying Sobol sensitivity
analysis and the results are optimized by Dringer algorithm. Finally, the optimum machining condition is
determined as 30 m/s cutting speed, 0.1 mm/rev feed rate, and 0.5 mm depth of cut. Furthermore, the effectiveness
of the input parameters on the surface quality is determined as 19% for cutting speed, 74% for feed rate and 7%
depth of cut.
Keywords: Bone machining, Cortical Bone, Machining parameters, Sensitivity Analysis, Optimization, Surface
quality.
References
1. Safari, M., Joudaki, J., & Emadi, M. (2018). Surface quality in dry machining of 55Cr3 steel bars.
International Journal of ISSI, 15(1): 33-39.

A machine learning-based continuum damage model for predicting complex

crack propagation
Tran Quoc Thai1 and Xiaoying Zhuang1,*
1
Institute of Continuum Mechanics, Leibniz University of Hannover, Appelstr. 11, 30167 Hannover, Germany.
*
Corresponding Author: X. Zhuang. Email: zhuang@ikm.uni-hannover.de.
Abstract: Micro damage is a complicated mechanical behavior related to fracture diffusion process. To describe the
phenomena, some analytic model has been applied such as continuum damage model or phase-field fracture model.
However, the most difficulty in modeling the structural failure and crack propagation is the constitutive relation
between the damage variable and the stress/strain state is difficult to determine. Some fundamental modes of
fracture are assumpted to simplify the analytical models however this is inadequate to capture the realistic behavior
fracture where the structures are usually subjected to mixed modes fracture or multiple cracks. Constructing a
reliable and high accurate model to describe the damage phenomena still a challenge due to the lack of information
of the constitutive relations especially at the microscale where the initiated microdamage can finally lead to the
propagation of macro cracks.
The development of machine learning has been motivated by high-speed computers which allows handling a big
amount of data. The artificial neural network, a branch of machine learning, employs complicated multilayers of
artificial neuron in order to construct a mapping function based on the training data as the input of the system.
Artificial neural networks recently have been adopted in a wide range of engineerings including mechanics to solve
to tackle the difficulties in examining the constitutive relations in solid mechanics [1, 2, 3, 4, 5, 6]. The accuracy of
input data is the core that determines the success of a machine learning model. Because of that, in order to utilize
machine learning as a useful tool for studying solid materials, the prepared data play an important role. Thank to the
development of advanced imaging techniques such as computed tomography (CT), ultrasound imaging (UI), image
scanning microscopy (ISM), optical coherence tomography (OCT), scanning electron microscope (SEM) and
magnetic resonance imaging (MRI), the information of microstructure is revealed which allows a great opportunity
to acquire a large amount of data of microstructure. Because of that reason, image-based machine learning has been
applied to study a wide range of mechanic problems such as inverse elasticity problem [7], heterogeneous
composite materials [8, 9], porous plasticity media [10], designing materials [11].
In this proposed work, the data of deflected microstructure will be obtained via an imaging technique and
constructed as input parameters of the neuron network. The network will be trained in order to be able to predict the
constitutive relation of damage variable for any given stress/strain state. The trained damage model, in turn, is
applied into the conventional framework of an analysis tool such as FEM, IGA to simulate the fracture process and
crack propagation in structures and test for practical applications where the machine learning system has to expose
with some cracking situations that it has not encountered before.
Keywords: Machine learning, Damage model, Image based constitutive relation.
References
1. Ghaboussi, J., Pecknold, D. A., Zhang, M., Haj‐Ali, R. M. (1998). Autoprogressive training of neural network
constitutive models. International Journal for Numerical Methods in Engineering, 42(1): 105-126.
2. Jung, S., Ghaboussi, J. (2006). Neural network constitutive model for rate-dependent materials. Computers &
Structures, 84(15-16): 955-963.
3. Ji, G., Li, F., Li, Q., Li, H., Li, Z. (2011). A comparative study on Arrhenius-type constitutive model and
artificial neural network model to predict high-temperature deformation behaviour in Aermet100
steel. Materials Science and Engineering: A, 528(13-14): 4774-4782.

165 DACOMA-2019
4. Furukawa, T., Yagawa, G. (1998). Implicit constitutive modelling for viscoplasticity using neural
networks. International Journal for Numerical Methods in Engineering, 43(2): 195-219.
5. Hashash, Y. M. A., Jung, S., Ghaboussi, J. (2004). Numerical implementation of a neural network based
material model in finite element analysis. International Journal for numerical methods in engineering, 59(7):
989-1005.
6. Sun, Y., Zeng, W. D., Zhao, Y. Q., Qi, Y. L., Ma, X., & Han, Y. F. (2010). Development of constitutive
relationship model of Ti600 alloy using artificial neural network. Computational Materials Science, 48(3): 686-
691.
7. Patel, D., Tibrewala, R., Vega, A., Dong, L., Hugenberg, N., Oberai, A. A. (2019). Circumventing the solution
of inverse problems in mechanics through deep learning: Application to elasticity imaging. Computer Methods
in Applied Mechanics and Engineering, 353: 448-466.
8. Liu, R., Yabansu, Y. C., Agrawal, A., Kalidindi, S. R., Choudhary, A. N. (2015). Machine learning approaches
for elastic localization linkages in high-contrast composite materials. Integrating Materials and Manufacturing
Innovation, 4(1), 13.
9. Li, X., Liu, Z., Cui, S., Luo, C., Li, C., Zhuang, Z. (2019). Predicting the effective mechanical property of
heterogeneous materials by image based modeling and deep learning. Computer Methods in Applied Mechanics
and Engineering, 347: 735-753.
10. Wang, K., Sun, W. (2018). A multiscale multi-permeability poroplasticity model linked by recursive
homogenizations and deep learning. Computer Methods in Applied Mechanics and Engineering, 334: 337-380.
11. Bessa, M. A., Bostanabad, R., Liu, Z., Hu, A., Apley, D. W., Brinson, Chen W., Liu, W. K. (2017). A
framework for data-driven analysis of materials under uncertainty: Countering the curse of
dimensionality. Computer Methods in Applied Mechanics and Engineering, 320: 633-667.
Methods Based on Artificial Neural Networks for the Solution of Partial

Differential Equations on Domains with Complex Geometries
Cosmin Anitescu1,*, Elena Atroshchenko2 and Somdatta Goswami1
2
University of New South Wales, High St, Kensington, NSW 2052, Australia.
*
Corresponding Author: Author's Name. Email: cosmin.anitescu@uni-weimar.de.
Abstract: Machine learning and methods based on artificial neural networks have become increasingly applied to a
variety of topics in areas such as image processing, voice recognition, and object detection. In this work, we present
a set of algorithms for solving partial differential equations using the approximation properties of deep neural
networks (DNN). The proposed procedure is designed to solve partial differential equations on complex geometries,
such as the ones obtained from spline description and can alleviate the problems encountered due to gaps in
geometry description or other boundary parameterization issues.
We will consider and compare algorithms based on collocation (as in [1]), as well as those based on energy
minimization [2] which require evaluating an interior integral from scattered points in the domain. The relation
between the standard spline-based approximation spaces and DNNs with different activation functions will be
examined in in detail. The method proposed has been tested on benchmark problems with known solutions, which
allows for a better understanding of the approximation and convergence properties of DNNs.
Keywords: Artificial neural networks, Partial differential equations, Collocation, Energy minimization.
Acknowledgement
C. Anitescu acknowledges the support of the German Research Foundation (DFG project number 392023639).
References
1. Raissi, M., Perdikaris, P., Karniadakis, G.E., Physics-informed neural networks: A deep learning framework
for solving forward and inverse problems involving nonlinear partial differential equations, Journal of
Computational Physics, Volume 378, 2019, 686-707.
2. Weinan E, Yu B., The Deep Ritz Method: A Deep Learning-Based Numerical Algorithm for Solving
Variational Problems, Communications in Mathematics and Statistics, Volume 6, 2018, 1-12.

Research on Application of Machine Learning and Optimization Algorithm in

Transparent Soil: A Review
Bo Wang1 and Zhengwei Zhu2,*
1School of civil engineering, Chongqing University, No.174 Shazhengjie, Shapingba, 400044, Chongqing, China.
2School of civil engineering, Chongqing University, No.174 Shazhengjie, Shapingba, 400044, Chongqing, China.
*Corresponding Author: Zhengwei Zhu. Email: zhuzw@cqu.edu.cn.
Abstract: This paper reviews the application and development of machine learning in geotechnical engineering. It is
found that artificial intelligence methods such as machine learning has important application value and development
prospects in geotechnical engineering problems, such as flood, landslide, soil pile friction, slope displacement and
shear strength. For transparent soil, a special synthetic soil, based on the characteristics of transparent soil:
transparency, shear strength, oil content, liquid limit, plastic limit, a combination of machine learning and
optimization algorithms is proposed to analyze and predict the Correlation coefficient of transparent soil. Finally,
the application process, selected parameters, prediction methods and precautions of machine learning in the model
test of transparent soil slope is discussed.
Keywords: Geotechnical engineering, Artificial intelligence, Slope, Transparent soil, Review.
References
1. Sadek S, Iskander M G, Liu J. (2002). Geotechnical properties of transparent silica. Canadian Geotechnical
Journal, 39(1): 111-124.
2. Iskander M, Liu J. (2010). Spatial deformation measurement using transparent soil. Geotechnical Testing
Journal, 33(4): 314-321.
3. Ezzein FM, Bathurst RJ. (2011). A transparent sand for geotechnical laboratory modeling.Geotechnical Testing
Journal, 34(6): 590-601.
4. Sun J, Liu J. (2011). Visualization of tunnelling-induced ground movement in transparent sand. Tunnelling and
Underground Space Technology, 40: 236-240.
5. Bui D T, Pradhan B, Nampak H, et al. (2016). Hybrid artificial intelligence approach based on neural fuzzy
inference model and metaheuristic optimization for flood susceptibilitgy modeling in a high-frequency tropical
cyclone area using GIS. Journal of Hydrology, 540:317-330.
6. Bui D T, Bui Q T, Nguyen Q P, et al. (2017). A hybrid artificial intelligence approach using GIS-based neural-
fuzzy inference system and particle swarm optimization for forest fire susceptibility modeling at a tropical area.
Agricultural and forest meteorology, 233: 32-44.
Order Boundary Value Problems. Computers, Materials & Continua, 59(2): 433-456.
10. Chen W, Panahi M, Pourghasemi H R.(2017). Performance evaluation of GIS-based new ensemble data
mining techniques of adaptive neuro-fuzzy inference system (ANFIS) with genetic algorithm (GA), differential
evolution (DE), and particle swarm optimization (PSO) for landslide spatial modelling. Catena,, 157: 310-324.
11. Pham B T, Hoang T A, Nguyen D M, et al. (2018). Prediction of shear strength of soft soil using machine
learning methods. Catena, 166: 181-191.
12. Snoek J, Larochelle H, Adams R P.(2012).Practical bayesian optimization of machine learning algorithms.

168 DACOMA-2019
Advances in neural information processing systems, 2951-2959.

13. Kiran S, Lal B, Tripathy S S.(2016).Shear strength prediction of soil based on probabilistic neural network.
Indian Journal of Science and Technology, 9(41): 2221-2232.
Thermal conductivity identification of nonlinear functionally graded

materials via a machine learning strategy
Zhuojia Fu1,2,*, Wenzhi Xu1 and Qiang xi1
1Center for Numerical Simulation Software in Engineering and Sciences, College of Mechanics and Materials, Hohai
University, Nanjing 211100, China.
2Institute of Continuum Mechanics, Leibniz University Hannover, Hannover 30167, Germany.
*corresponding author: Zhuojia Fu Email: zhuojiafu@gmail.com.
Abstract: A machine learning strategy based on the genetic algorithm (GA) [1] and the semi-analytical collocation
solver (SACS) [2] is presented for the thermal conductivity identification of nonlinear functionally graded materials
(FGMs) [3]. In this study, only the temperature or heat flow on the FGM surface can be non-invasively measured via
the thermal sensors. Instead of extensive experiments, the semi-analytical collocation solver is used to construct the
database of the relationship between the thermal conductivity and the temperature or heat flow on the FGM surface.
Based on the aforementioned constructed database, the GA-based machine learning strategy is implemented to
identify the thermal conductivity of nonlinear FGMs [4]. In addition, the Tikhonov regularization is introduced to
eliminate the effect of measuring noise in the measurement data. Several benchmark examples are presented to
verify the feasibility, robustness and applicability of the proposed machine learning strategy.
Keywords: Machine learning, Genetic algorithm, Semi-analytical, Parameter identification, Functionally graded
materials.
References
1. Sasaki, M., Wang, Y., Hirano, T. & Hirai, T. (1989). Design of sic/c functionally gradient material and its
preparation by chemical vapor deposition. Journal of the Ceramic Society of Japan, 97(1125): 539-543.
2. Fu, Z. J., Xi, Q., Chen, W., & Alexander, C. (2018). A boundary-type meshless solver for transient heat
conduction analysis of slender functionally graded materials with exponential variations. Computers &
Mathematics with Applications, 76(4): 760-773.
3. Divo, E., Kassab, A., & Frankli, R. (2000). Characterization of space dependent thermal conductivity with a
bem-based genetic algorithm. Numerical Heat Transfer, Part A: Applications, 37(8): 845-875.
4. Chen, B., Chen, W., Wei, X. (2015). Characterization of space-dependent thermal conductivity for nonlinear
functionally graded materials. International Journal of Heat and Mass Transfer, 84: 691-699.

An Alternative Optimization Algorithm to Analysis of Structure by Hybrid

Artificial Neural Network-Isogeometric Finite Element Program
D. Nguyen Kien1, Xiaoying Zhuang1,2 and Timon Rabczuk3,*
1 Department of Geotechnical Engineering, Tongji University, 1239 Siping Road, 200092 Shanghai, China.
2 Institute
of Continuum Mechanics, Leibniz University Hannover, Appelstrasse 11, 30167 Hannover, Germany.
3 Chair of Computational Mechanics, Bauhaus University Weimar, Marienstrasse 15, 99423 Weimar, Germany.
*Corresponding Author: Timon Rabczuk. Email: timon.rabczuk@uni-weimar.de.
Abstract: The artificial neural network (ANN), which is a system of computing, has its ability to learn the
performance task without being programmed and makes attempts to model abstraction in high level in data [1]. And
Isogeometric Analysis (IGA) [2] is a computational approach, which offers the possibility of integrating methods for
numerical analysis into NURBS-based Computer Aided Design tool, studies the case in which the geometry is more
complicated. The hybrid method which is the combination of ANN and IGA (called hybrid IGA/ANN) takes its
advantages of these methods to study numerical problems. In the present paper, beside introducing hybrid IGA/ANN
(Fig. 1), an optimization algorithm to train data in Neural Network, called adadeltamax, is proposed (Algorithm 1).
I-shape beam problem with changing in cross-section geometry was applied. The results, gotten from adadeltamax
in hybrid method, were compared to those from other optimization algorithms-sgd, nesterov, adagrad, adadelta,
rmsprop, rmsprop-nesterov, adam, and nadam, prove to be an acceptable advantage to use (Fig. 2, Tab. 1). In order
to visualize more specifically the result, after training data with more time of iterations, the values of relative error
larger than 10 percent are quite small while those of the relative error less than 2.5 percent mostly make up for the
huge proportion (Figs. 3,4). The optimization algorithm proves to be a choice to train the data in Neural Network.
Keywords: Neural network, Isogeometric analysis, I-shape beam, Optimization algorithm.
Algorithm 1: Adadeltamax Optimization Algorithm
Figure 1: Flow chart of Artificial Neural Network-Isogeometric Analysis Hybrid Method

171 DACOMA-2019
Figure 2: Mean Square Error (MSE) for (a) training data set and (b) testing data set in comparison among number
of optimizers. (Y axis in log scale.)
Table 1: Train and Test MSE (×10-3) in comparison among different optimizers after 8000 epochs
Optimizers adadelta adadelta sgd adagrad rmsprop rmsprop- adam nadam nesterov
max nesterov
Train loss 0.1804 0.1491 0.1403 0.2073 0.1297 0.0807 0.1422 0.1794 0.1311
Test loss 0.1459 0.1388 0.1099 0.2292 0.0911 0.0709 0.1312 0.1423 0.0911
Figure 3: Correlation between the values gotten from real Figure 4: Histograms of relative error
values and predicted values regarding training set and testing regarding trainging set and testing set after
set after 21000 epochs in adadeltamax optimizer. 21000 epochs in adadeltamax optimizer.
References
1. Goodfellow, I. , Bengio, Y., & Courville, A. (2016). Deep Learning. The MIT Press, Cambridge.
2. Hughes, T.J.R., Cottrell, J.A., & Bazilevs, Y. (2005). Isogeometric Analysis: CAD, Finite Element, NURBS,
Exact Geometry and Mesh Refinement. Computer Method in Applied Mechanics and Engineering, 194(39-41):
4135-4195.
Deep learning algorithm for solving PDEs for fluid flow field
Juan Lin1,* and Shuwei Zhou2
1Shanghai Maritime University, 1550 Haigang Avenue, 201306 Shanghai, China.
2
Leibniz University Hannover, Appelstrasse 11, 30167 Hannover, Germany.
*
Corresponding Author: Juan Lin. Email: juanlin@shmtu.edu.cn.
Abstract: How to exactly model fluid flow field in underground engineering has been a longstanding computational
challenge and the deep-learning approach provides new perspectives for solving this. We propose to solve the
governing PDEs of fluid flow by approximating the solution with a deep neural network. We train the network to
satisfactorily obey the differential operator, initial condition, and boundary conditions. The implementation is
meshfree compared with the traditional FEM simulations for fluid flow. Without using a mesh, the neural network is
trained on batches of randomly sampled time and space points and in this way the deep learning algorithm provides
the general solution to the continuity equation of fluid flow. The used method has a similar spirit to Galerkin
methods, with the solution approximated by a neural network instead of a linear combination of basis functions. We
also compare the results obtained from the deep learning approach and the analytical solution, which shows a
satisfactory agreement.
Keywords: Deep learning, PDEs, Fluid flow, Neural network.
References
1. Sirignano, J., & Spiliopoulos, K. (2018). DGM: A deep learning algorithm for solving partial differential
equations. Journal of Computational Physics, 375: 1339-1364.
2. Zhou, S., Zhuang, X., & Rabczuk, T. (2019). Phase-field modeling of fluid-driven dynamic cracking in porous
media. Computer Methods in Applied Mechanics and Engineering, 350: 169-198.
3. Zhou, S., Zhuang, X., & Rabczuk, T. (2018). A phase-field modeling approach of fracture propagation in
poroelastic media. Engineering Geology, 240: 189-203.
Solution of Second Order Boundary Value Problems. Computers, Materials & Continua, 59: 345-359.
Order Boundary Value Problems. Computers, Materials & Continua, 59: 433-456.

DML based artificial neural network approach to solve transient wave

equation using the Collocation method
Abhishek Mishra1, Cosmin Anitescu2 and Pattabhi R. Budarapu1,*
1Indian Institute of Technology Bhubaneswar, Odisha, 752050, India.
*Corresponding Author: Pattabhi Budarapu. Email: pattabhi@iitbbs.ac.in.
Abstract: The working mechanics of a human brain (made-up of neurones) can be mimicked through artificial
neural networks (ANN). ANN operates by reducing a given task to a mathematical function, and hence the ability to
easily apply several mathematical operations using high-level open source libraries like Tensorflow, Keras, Pytorch,
to name a few. Therefore, this technique has been extended to solve several ordinary and partial differential
equations (PDEs) [1]. ANN is also applied to solve non-linear PDEs of higher order, thereby addressing the
problems associated with finite element and finite difference methods. Furthermore, deep machine learning (DML)
consisting of some extra layers (hidden layers), significantly enhances the capacity to resolve complex issues.
Therefore, ANN in conjunction with classical methods can result in efficient solution technique for several physical
problems involving PDEs. Two popular approaches to solve the governing PDEs are: (i) collocation method and (ii)
energy minimisation method [2]. In collocation method, the idea is to choose a finite-dimensional domain with a
specified number of points in the selected domain, known as collocation points, and train the network such that the
governing equations and the boundary and initial conditions are satisfied with minimum error at the selected points.
The energy minimisation approach involves the formulation of governing equations in terms of energy and solve for
the response by minimising the energy [3, 4]. Neural networks combined with collocation approach was employed
to solve problems satisfying the laws of physics, known as physics informed neural networks (PINNs), for instance:
Burger’s equation, Schrödinger equation, Allen-Cahn equation, Navier-Stokes equation [5], to name a few.
In this paper, the collocation method is used to solve one and two dimensional problems governed by transient wave
equation. The collocation points are chosen by dividing the domain uniformly in both space and time. The
methodology is implemented using Tensorflow library in Python language. Xavier Initialisation method to used
initialise the neural network. The mean square error at the collocation points is calculated and minimised while
training the ANN. The relative normalised error is observed to be less than 5% in both of the cases.
Keywords: Artificial neural networks, Deep machine learning, Collocation method, Energy minimisation, Transient
wave equation
Acknowledgement
Abhishek would like to thank Prof. Rabczuk for giving an opportunity to work on the evolving DML based
approaches to solve PDEs and providing the resources. He also would like to acknowledge DAAD for funding the
program.
References
1. Yadav, N., Yadav, A., & Kumar, M. (2015). An Introduction to Neural Network Methods for Differential
Equations. Yadav, N., Yadav, A., & Kumar, M. (2015). An introduction to neural network methods for
differential equations (p. 21). Netherlands: Springer. https://doi.org/10.1007/978-94-017-9816-7.
3. Weinan, E., & Yu, B. (2018). The Deep Ritz Method: A Deep Learning-Based Numerical Algorithm for
Solving Variational Problems. Communications in Mathematics and Statistics, 6(1): 1-12.
4. Golts, A., Freedman, D., & Elad, M. (2018). Deep Energy: Using Energy Functions for Unsupervised Training
of DNNs. arXiv preprint arXiv:1805.12355.

174 DACOMA-2019
5. Raissi, M., Perdikaris, P., & Karniadakis, G.E. (2019). Physics-Informed Neural Networks: A Deep Learning
Framework for Solving Forward and Inverse Problems Involving Nonlinear Partial Differential Equations.
Journal of Computational Physics, 378: 686-707.
6. Evans, J.A., Hiemstra, R.R., Hughes, T.J., & Reali, A. (2018). ScienceDirect Explicit higher-order accurate
isogeometric collocation methods for structural dynamics. Computer Methods in Applied Mechanics and
Phononic Crystal Band-Gap Detection by Machine Learning Technique

Binh Nguyen1 and Xiaoying Zhuang1,2,*
1Institute for Continuum Mechanics, Leibniz University Hannover, Appelstrasse 11, 30167 Hannover, Germany.
2
*
Abstract: Phononic crystal metamaterial has recently received tremendous development to deliver peculiar features
that cannot be found in conventional material, such as cloaking, super-lenses and signal-filtering [2]. Such properties
can be revealed from the determination of dispersion relation of the phononic crystal, which associates the eigen-
frequencies and wave number of a unit cell due to the Bragg's scattering phenomenon or local resonant. In numerical
studies, the dispersion relation, apparently depends on the material properties of the constituents as well as the
topology of the unit cell, is normally computed from solving the eigenvalue problem by means of Finite element
method (FEM) as [4]
where K and M are the finite-element-assembled stiffness and mass matrices, respectively, is the angular
frequency and k is the wave vector defined in the reciprocal space.
In this work, we propose an alternative approach to determine the energy gap in the dispersion relation of 2D
phononic crystal using Deep Learning Neuron Network. The proposed approach consists of two procedures
described as following.
First of all, in 2D phononic crystal, the unit cell can belong to one of the 5 Bravais lattices grouped into four crystal
families [3], that are monoclinic, orthorhombic, hexagonal and tetragonal. Therefore, in the first stage of the method,
the type of unit cell will be determined by a classification-trained Neuron Network. By doing this, we can obtain the
lattice unit vectors as well as the reciprocal lattice unit vectors so that the Brioullin zone or the Irreducible Brioullin
zone can be defined and used for the computation of the dispersion relation. It should be noted that, not only
geometry of the unit cell is identified but also the material distribution. In the limitation of linear isotropic material,
each constituent is assumed to be defined by two Lame’s parameters and mass density.
Once the unit cell setting up is prepared, the training process can start. As described, the material parameters of the
constituents defined discretely are taken to be the inputs. Moreover, as we were able to identify the type of unit cell in
the first step, specific wave number for different type of unit cell can also be taken as the input parameter. We expect
the eigenvalue , which will be collected into a vector corresponding to different eigenmode, to be the output [1].
Figure: Schematic of Neural network to predict phononic energy band gap
Keywords: Machine learning, Phononic crystal, Band-gap, Dispersion relation, Energy band diagram.

176 DACOMA-2019
References
1. Finol, D., Lu, Y., Mahadevan, V., Srivastava, A. (2019). Deep convolutional neural networks for eigenvalue
problems in mechanics. International Journal for Numerical Methods in Engineering, 118(5): 258-275.
2. Laude, V. (2015). Phononic crystals: artificial crystals for sonic, acoustic, and elastic waves, vol. 26. Walter
de Gruyter GmbH & Co KG.
3. Kittel, C. (1976). Introduction to solid state physics, vol. 8. New York: Wiley.
4. Nguyen, B. H., Zhuang, X., Park, H. S., Rabczuk, T. (2019). Tunable topological bandgaps and frequencies in
a pre-stressed soft phononic crystal. Journal of Applied Physics, 125(9): 095106.
Extended isogeometric-meshfree collocation methods without branch

enrichment for cracks analysis under contact loading
Nhon Nguyen-Thanh1,*, Weidong Li2 and Kun Zhou2

1School of Electrical and Electronic Engineering, Nanyang Technological University, 50 Nanyang Ave, Singapore.
2
School of Mechanical and Aerospace Engineering, Nanyang Technological University, 50 Nanyang Ave, Singapore.
*
Corresponding Author: N.Nguyen-Thanh. Email: thanhnhon@ntu.edu.sg.
Abstract: Collocation methods have been developed as a powerful alternative to Galerkin's method in the context of
isogeometric analysis (IGA) characterized by a significantly reduced computational cost, but still guaranteeing
higher order convergence rates. In this work, we develop a novel adaptive isogeometric-meshfree collocation
method (IGAM-C) for simulating cracks in an elastic half space under contact loading. The concept of IGAM-C
method is based upon the correspondence between the isogeometric basis functions and the moving least-squares
meshfree shape functions, which provides the flexibility of meshfree adaptive refinement for isogeometric analysis.
The proposed method does not require special treatment for essential boundary condition and the time-consuming
integration of a weak form. The integrals are evaluated directly in the parametric space following the framework of
IGA. Moreover, the elastic crack problems are successfully solved without branch enrichment. The near-tip field is
accurately captured by special nodal arrangements around the tip. Furthermore, the resolution of the nonlinear
equations governing the contact problem is derived from a strong form to avoid the disadvantages of numerical
integration. The implementation of the proposed method is simplified, and the computational effort is reduced. The
accuracy and robustness of the presented method are demonstrated through its application to several numerical
examples.
Keywords: IGA, Meshfree Method, Collocation, Contact Mechanics, Fracture Mechanics.
Introduction
Computational fracture mechanics has been dominated by weak form-based methods for decades. Among these
methods, the classical finite element method (FEM) has been applied for crack modeling. The extended FEM
(XFEM) [1] was further developed to avoid time-consuming re-meshing in crack problems, whereas it suffers from
low computational efficiencies due to the implementation of enrichment functions. On the other hand, the strong-
form methods do not require numerical integration and thus perform more efficiently than weak-form methods.
Meshfree methods [2] are viewed as next-generation computational techniques for problems involving large
deformation, fracture and fragmentation and contact problems. Since the meshfree methods do not rely on a fixed
topological connectivity between nodes, they are expected to be more adaptive and robust in resolving the problems
where the classic grid-based methods are not suitable. Compared to the finite element method (FEM), meshfree
methods are able to save computational time by avoiding re-meshing after each crack extension. Nevertheless, the
meshfree method usually has less computational efficiency than the FEM, as higher computational cost is required
for meshfree interpolation and numerical integration. The isogeometric analysis (IGA) was first proposed by Hughes
et al. [3] as a powerful and reliable tool for computation and simulation of engineering problems. By adopting the
non-uniform rational B-splines (NURBS) which are widely used in computer-aided designs as basis functions, IGA
is not only applicable to engineering analyses but also capable of delivering accuracies superior to the standard FEM
in many applications including contact mechanics. However, IGA based on NURBS is difficult to locally refine due
to its tensor product nature, which leads to an excessive overhead of control points with increasing refinement. With
the goal of overcoming the limitations of NURBS for IGA, adaptive IGA has been studied with T-splines, PHT-
splines, hierarchical B-splines, LR-splines, etc. However, constraint equations are required on these methods,
increasing the complexity and effort in their implementation. Recently, within the isogeometric collocation (IGA-C)
context, several strategies have been studied to reduce the computational cost for Galerkin methods by using higher-
order isogeometric basis functions [4].
This paper aims to investigate crack problems via an adaptive isogeometric analysis meshfree collocation (IGAM-C)
approach which is a strong form-based approach. In comparison to weak-form approaches, the present approach
exhibits advantages inheriting from the IGAM method and the collocation method, such as the flexibility of adaptive
178 DACOMA-2019
refinement, elimination of numerical integration, and convenient treatment of essential boundary conditions. The
IGAM-C approach utilizes the linear reproducing points which are calculated by a knot vector as the collocation
points and achieves the adaptive refinement in a straightforward manner. The adaptive refinement procedure uses a
gradient-based error estimator to efficiently determine meshes with large errors. In modeling cracks via the IGAM-C
approach, the strong discontinuity along crack surfaces is introduced by the visibility criterion. To achieve a smooth
approximation around crack tips, the diffraction method is further employed. Besides, the singularity around crack
tips is implemented by the adaptive refinement without requiring the enrichment functions [5]. It is shown that the
IGAM-C approach is more efficient and accurate through the adaptive refinement, and successfully captures the
near-tip fields for the accurate calculation of stress intensity factors.
Numerical example
Consider a central-cracked plate frictionally contacting with a cylinder as shown in figure 1. This model is loaded
with a vertical displacement u = 0.002 and a horizontal displacement v = 0.00075. The geometrical configuration
with the following parameters: radius R = 1, width w = 2, high h1 = 0.3, h2=0.5, crack length a = 1/4w, b = 1/4w.
Young’s modulus E = 1, ν = 0.3 and the coefficient of friction μ = 0.2. The penalty parameters are εn = 1000 and εt =
100.
Figure 1: Schematic of the cylinder on plane Hertzian with crack under contact loading: (a) Geometrical modeling,
(b) adaptive mesh after 4 refinement steps, (c) contour plots of von Mises stress.
Acknowledgement
This research work was conducted in the SMRT-NTU Smart Urban Rail Corporate Laboratory with funding support
from the National Research Foundation (NRF), SMRT and Nanyang Technological University (Grant No.
M4061892).
References
1. T. Belytschko and T. Black. (1999). Elastic crack growth in finite elements with minimal remeshing. Int. J.
Numer. Meth. Eng., 45(5): 601-620.
2. T. Belytschko, Y.Krongauz, D.Organ, M.Fleming, and P.Krysl. (1996). Meshless methods: An overview and
recent developments. Comp. Meth. Appl. Mech. Eng. 39: 3-47.
3. T.J.R.Hughes, J.A.Cottrell, and Y.Bazilevs. (2005). Isogeometric analysis: CAD, finite elements, NURBS,
exact geometry and mesh refinement. Comp. Meth. Appl. Mech. Eng. 149: 4135-4195.
4. D. Schillinger, M.J.Borden, and H.K.Stolarski. (2015). Isogeometric collocation for phase-field fracture
models. Comp. Meth. Appl. Mech. Eng. 248: 583-610.
5. G. Zi, T.Rabczuk, and W.Wall. (2007). Extended meshfree methods without branch enrichment for cohesive
cracks. Computational Mechanics 40: 367-382.
Numerical simulation of Von-Karman plate problem by nonlocal operator

method
Huilong Ren1,*
*Corresponding Author: Huilong Ren. Email: huilong.ren@uni-weimar.de.
Abstract: Nonlocal operator method as a recently proposed numerical approach shows great potential in solving
various physical problems. Some features of this method include nodal integration or numerical integration,
variational formulation or weighted residual formulation, free of shape function in conventional method, C^n
continuity of the operators, meshless, non-uniform discretization. The method is ready for explicit/implicit analysis
of solid mechanics involved geometrical nonlinearity. Owning to its high continuity, the method can solve higher
order problems directly without resorting to the mixed formulation, which introduces the intermediate fields and
requires a large number of degree of freedoms. In this paper, we apply the nonlocal operator method to solve the
Von-Karman plate problem. The numerical results in terms of in-plane and out-of-plane displacement are verified
by that of Abaqus with S4R element formulation.
Keywords: Nonlocal operator method, plate theory, Von-Karman plate.
References
1. Ren, H., Zhuang, X., & Rabczuk, T. (2019). Higher order nonlocal operator method. arXiv preprint
arXiv:1905.02809.
2. Rabczuk, T., Ren, H., & Zhuang, X. (2019). A Nonlocal Operator Method for Partial Differential Equations
with Application to Electromagnetic Waveguide Problem. Computers, Materials & Continua 59 (2019), Nr. 1.

Data Driven solution for 1D Cahn-Hilliard Equation

Jorge López-Zermeño1,*, Somdatta Goswami1 and Navid Valizadeh1
1Institute of Structural Mechanics, Bauhaus University Weimar, Marienstraße 15, 99423 Weimar, Germany.
*Corresponding Author: Jorge López-Zermeño. Email: jorge.alberto.lopez.zermeno@uni-weimar.de.
Abstract: The artificial neural network (ANN) is a class of machine learning tools, which is inspired by the
structure and behavior of biological neural systems. ANN recognizes the patterns in a series of input and output
values and using the acquired ‘knowledge’ it then predicts the unknown output values for a given set of input values.
However, despite its excellent performance in the domain of image processing and computer science, it cannot
ensure that the essential physical laws associated with an engineering problem are satisfied. Recent development in
the field of machine learning, propose the use of physics informed/constrained neural networks (PINN) for solving
partial differential equations (PDE) [1]. In PINN methods, the neural networks are trained to solve supervised
learning problems while respecting any given law of physics described by general non-linear partial differential
equations.
The Cahn-Hilliard equation is a time dependent fourth-order PDE that describes the phase separation process. The
high order operators that appear in the Cahn-Hilliard equation are the main obstacle to obtain an accurate solution.
Numerical methods like finite element analysis are often not capable enough to solve the PDEs directly; instead, the
fourth-order equation is decomposed into two second-order equations using various assumptions. Moreover, the
curse of dimensionality, present in the traditional numerical methods is a hindrance for solving higher order PDE’s
in higher dimensions [2]. In this context, we propose a PINN based approach to solve the Cahn-Hilliard equation. In
this work, we consider the one-dimensional Cahn-Hilliard equation with periodic boundary conditions. The
governing equation with periodic boundary conditions could be stated as:
∂u ∂ 4u ∂ ( F' ( u ) )
2
+ − = 0 ,x ∈ [ 0 , 2π ] ,t ∈ [ 0 ,8500] ,
∂t ∂x 4 ∂x 2
1
F = ( u 2 − 1) ,
2
4
1 cos( x+1/10) (1)
u ( 0 ,t ) = cos ( 2x ) + e ,
100
u ( t, 0 ) = u ( t, 2π ) ,
∂u ( t, 0 ) ∂u ( t, 2π )
= ,
∂x ∂x
where, u ∈ [-1, 1] denotes the phase, F is the free energy function and x represents the spatial co-ordinates. In this
problem, we consider ϵ = 0.18. The training data set is generated by solving the Cahn-Hilliard equation with a
spectral Fourier discretization, using the toolbox chebfun [3], with N = 512 points to discretize the domain and a
time step, Δt = 0.1. Fig. 1 shows the training points at time t = 8300s. The neural network is set up with four hidden
layers of 500 neurons each. For optimization, we use the Adam (adaptive momentum) optimizer. For the time
integration, we use the fourth-order implicit Runge-Kutta method with q stages, where q = 100. Using this setup, the
mean squared error of the governing equation at n collocation points is minimized, where n = 512. The loss function
is defined as:

181 DACOMA-2019
L = LSEint + LSEbound ,
q+1 n
(
LSEint = ∑∑ uin ( x nj ,i − u n ,i ) , )
2
i=1 j=1 (2)
( )
q
(2π ) + ( u n+1 (0) − u n+1 (2π ) )
2 2
LSEbound = ∑ u
n+ci n+ci
(0) − u
i=1
2 2
q 
∂u i ( 0 ) ∂u i ( 2π )   ∂u n+1 ( 0 ) ∂u n+1 ( 2π ) 
n+c n+c
+∑  −  + −  ,
 ∂x ∂x   ∂x ∂x 
i=1
 
where LSEint denote the loss in the interior points and LSEint denote the loss in the boundary points. For training the
data set, we use Δt = 0.2. The convergence of the minimization of the loss function is shown in Fig. 1, along with the
comparison of the predicted and expected solution. The time step and number of stages were determined such that
the theoretical error is below the machine precision (Δt2q≤10-16). To measure the accuracy of the proposed approach,
the relative L2 norm is computed using:
(3)
For the proposed approach, L2 = 0.006%. This proves the ability of the method to efficiently predicted the solution
for one-dimensional Cahn-Hilliard equation.
Figure 1: From left to right: Training points over the exact solution of the Cahn-Hilliard equation. Convergence of
the training Loss function. The training is done at time t = 8300 s, and comparison between prediction and expected
solution of the Cahn-Hilliard equation at t = 8300.2 s
Keywords: Machine learning, Differential equations, Cahn-Hilliard equation.
Acknowledgement
JLZ acknowledges the support received from the Country-related cooperation programme with Mexico:
CONACYT-DAAD funding programme number 57177537 and SG acknowledges the support of the German
Academic Exchange Service (DAAD).
References
1. Raissi, M., Perdikaris, P., & Karniadakis, G. E. (2019). Physics-informed neural networks: A deep learning
framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal
of Computational Physics, 378: 686-707.
2. Han, J., Jentzen, A., & Weinan, E. (2018). Solving high-dimensional partial differential equations using deep
learning. Proceedings of the National Academy of Sciences, 115: 8505-8510.
3. Driscoll, T. A., Hale, N., & Trefethen, L. N. (2014). Chebfun guide. Pafnuty Publications, Oxford.
Probabilistic characterization of rock mass deformation modulus from P-

wave velocity based on information criteria
Xianbin Huang1,2, Hehua Zhu3, Keshen Zhang3, Wei Wu3, Lianyang Zhang4*
1
Shanghai Municipal Engineering Design Institute (Group) Co.,Ltd.
2
Shanghai Water Design&Engineering Co.,Ltd.
3
Department of Geotechnical Engineering, Tongji University, Shanghai, China
4
Department of Civil Engineering and Engineering Mechanics, University of Arizona, Tucson, Arizona, USA
*Corresponding Author: Lianyang Zhang. E-mail address: lyzhang@email.arizona.edu.
Abstract: Determination of rock mass deformation modulus (Erm) is a very important but very challenging task in
rock mechanics and rock engineering. Since direct measurements of Erm by in situ tests are expensive, time-
consuming and sometimes even impossible, many empirical methods for estimating Erm have been developed by
using geotechnical classification indices. However, in many cases, accurate determination of classification indices is
not an easy task. Since the nondestructive and easy geophysical method is commonly used to measure the P-wave
velocity (Vp) in the field, this paper has developed an empirical correlation between Erm and Vp based on 834
collected datasets by considering both inherent and transformation uncertainties. To do that, first information criteria
are employed to select the best model for deriving a general empirical correlation; maximum likelihood estimation
(MLE) methods are used to specify the incorporated parameters of each candidate model; and computational
intelligence techniques like particle swarm optimization (PSO), artificial bee colony (ABC) and differential
evolution (DE) algorithms are adopted to derive the optimal solutions for MLE. Then, based on the derived general
empirical equation, a Bayesian equivalent sample approach using Markov Chain Monte Carlo (MCMC) simulation
is applied to conduct probabilistic characterization of Erm by combining prior knowledge and project-specific
observed Vp data. Comparative studies of the results from the Bayesian equivalent sample approach and in situ
measured Erm values suggest that the proposed method is effective and validate the applicability of the derived
general empirical correlation.
Keywords: Rock mass deformation modulus, P-wave velocity, Empirical correlation, Maximum likelihood
estimation, Computational intelligence techniques, Markov Chain Monte Carlo simulation.
80
Zhao 2013 Erm= 0.2523exp(0.8443*Vp)
Zou 2013 3.161
Erm= 0.1053Vp
Zhou et al. 2015
Li and Feng 2014 Erm= 2.227Vp
Song et al. 2011 Erm= 0.5813exp(0.6944*Vp)
Li and Yan 2015
60 Erm= 0.0878Vp
3.353
Wang et al. 2006

Li et al. 2010 Erm= 3.529Vp
Jiang and Jiao 2013
Yang and Zhang 2009
Wang et al. 2013
Quan 2004
Erm(GPa)
40 Zhou et al. 2003

Li et al. 2008
Ding 2009
Li et al. 2014
Huang et al. 2014
Qi et al. 2011
20
0
0 1 2 3 4 5 6 7
Vp(km/s)
Fig. 1 Database employed and derived empirical correlations between rock mass deformation modulus Erm and P-
wave velocity Vp

183 DACOMA-2019
Fig. 2 (a) 25 sets of Erm and Vp data at Longkaikou hydropower station (Zhou et al. 2015); (b) Scatter plot of the
30,000 equivalent samples of Erm from MCMC simulation; (c) Histogram and relative frequency (PDF) of the
equivalent samples for Erm, and (d) Validation of the probability distribution for Erm estimated from the equivalent
samples
Acknowledgment
Acknowledgement is made to the Program of Natural Science Foundation of China (Grant No. 41672268) and
Shanghai Thousand Talents Program for their supports of this research.
References
1. Zhou HF, Nie DX, Wang CS (2015) Correlation between wave velocity and deformation modulus of basalt
masses as dam foundation in hydropower projects. J China Univ Geosci, 11:1904–1912 (in Chinese)
2. Akaike H (1973) Information theory and an extension of the maximum likelihood principle. In: Pertaran BN,
Csaaki F (ed) The Second International Symposium on Information Theory, Budapest, Hugary. Acadeemiai
Kiadi, pp 267–281
3. Beck JL, Au SK (2002) Bayesian updating of structural models and reliability using Markov chain Monte
Carlo simulation. J Eng Mech 128(4): 380–391
4. Feng X, Jimenez R (2015) Estimation of deformation modulus of rock masses based on Bayesian model
selection and Bayesian updating approach. Eng Geol 199: 19–27
5. Gokceoglu C, Sonmez H, Kayabasi A (2003) Predicting the deformation moduli of rock masses. Int J Rock
Mech Min 40(5): 701–710
6. Hoek E, Diederichs MS (2006) Empirical estimation of rock mass modulus. Int J Rock Mech Min 43(2): 203–
215
7. Zhang L (2016) Engineering Properties of Rocks. 2nd edition. Butterworth-Heinemann, Elsevier. p378.
8. Zhang L (2017) Evaluation of rock mass deformability using empirical methods – A review. Undergr Sp 2(1),
1-15.
Stochastic multiscale modeling of thermal properties of Polymeric clay

nanocomposites and global sensitive analysis
Bokai Liu 1, *
1 Bauhaus University Weimar, Marien-Straße 15, 99423 Weimar, Germany
*Corresponding Author: Bokai Liu. Email: bokai.liu@uni-weimar.de.
Abstract: We propose a stochastic multiscale modeling based on sensitivity analysis(SA) methods to quantify the
most significant input parameters influencing the heat conductivity of polymeric clay nanocomposites(PCNs). The
effect of the conductivity of cylinder, the conductivity of epoxy, interficial resistance, volume fraction and aspect
ratio on the conductivity of composites is systematically studied by using Finite Element Method(FEM). We use
global sensitive analysis to rank how much influence the different input parameters can achieve. We also use
surrogate model to describe the true model in this paper. According to our results, all surrogate models in stochastic
methods predict the same conclusion that the biggest sensitive index from key input parameters is aspect ratio.
However,interficial Resistance has no significant effect on thermal properties of the PCNs since sensitive index is
nearly 0. We also compare with different surrogate models, all of them are suit for sensitivity analysis, and they all
have high precision to the thermal model.
Keywords: Uncertainty quantification, Polymeric nanocomposites(PCNs), thermal properties, Stochastic

modeling , Multiscale modeling.
References
1. B. Mortazavi, J. Bardon, S. Ahzi. (2013). Interphase effect on the elastic and thermal conductivity response of
polymer nanocomposite materials: 3d finite element study, Computational Materials Science 69:100-106
2. T. Mori, K. Tanaka. (1973). Average stress in matrix and average elastic energy of materials with misfitting
inclusions, Acta metallurgica 21 : 571-574
3. A. K. Sen, S. Torquato, E_ective conductivity of anisotropic two-phase composite media, Physical Review B
39 (1989) 4504.
4. G. Liu, A step-by-step method of rule-of-mixture of fiber-and particle-reinforced composite materials,
Composite structures 40 (1997) 313-322.
5. A. Pourasghar, Z. Chen. (2019). Hyperbolic heat conduction and thermoelastic solution of functionally graded
cnt reinforced cylindrical panel subjected to heat pulse, International Journal of Solids and Structures 163: 117-
129
6. B. Mortazavi, M. Baniassadi, J. Bardon, S. Ahzi. (2013). Modeling of two-phase random composite materials
by finite element, mori-tanaka and strong contrast methods, Composites Part B: Engineering 45: 1117-1125.
7. N. Vu-Bac, T. Lahmer, Y. Zhang, X. Zhuang. (2014). T. Rabczuk, Stochastic predictions of interfacial
characteristic of polymeric nanocomposites (pncs), Composites Part B: Engineering 59: 80-95.
8. A. Saltelli, M. Ratto, T. Andres, F. Campolongo, J. Cariboni, D. Gatelli, M. Saisana, S. Tarantola, Global
sensitivity analysis: the primer, John Wiley & Sons, 2008.
9. I. M. Sobol'. (2005). Global sensitivity indices for the investigation of nonlinear mathematical models,
Matematicheskoe Modelirovanie 17: 43-52.
10. N. Vu-Bac, T. Lahmer, H. Keitel, J. Zhao, X. Zhuang, T. Rabczuk. (2014). Stochastic predictions of bulk
properties of amorphous polyethylene based on molecular dynamics simulations, Mechanics of Materials 68 :
70 - 84.
11. T. Most, C. Bucher. (2005). A moving least squares weighting function for the element-free galerkin method
which almost fulfills essential boundary conditions, Structural Engineering and Mechanics 21: 315-332.

Uncertainty Analysis of Engineering Structures by Time Domain Spectral

Element Based Modeling
Shuvajit Mukherjee *, Srinivasan Gopalakrishnan and Ranjan Ganguli
Indian Institute of Science, C. V. Raman Road, 560012 Bengaluru, India.

*Corresponding Author: Shuvajit Mukherjee. Email: shuvajit.mukherjee@gmail.com.
Abstract: This work investigates the use of time domain spectral element method (TSEM) for the uncertainty
analysis and modeling of beam structures. We have considered both Timoshenko beam (TB) and sandwich beam to
perform the analysis. Beam is one of the key building blocks of different engineering structures and depending upon
the choice of applications the material properties differ. With the advent of new materials, over the years the
material systems for engineering structures changed from isotropic to composite depending upon the field of
applications. Aerospace industry prefers lightweight structures with desired stiffness and strength. In this age,
composites and Sandwich structures have replaced isotropic materials in many aerospace applications. The analysis
and design of composite structures are complex and the prediction of responses become more difficult in the
presence of uncertainty. Uncertainty is unavoidable in engineering structures. So, to obtain a reliable design, it
should be considered into modeling. The uncertainty can be in material as well as in geometric properties and it
occurs due to the manufacturing process. This kind of uncertainty is termed as aleatory uncertainty which is inherent
and cannot be reduced. The other kind of uncertainty is epistemic uncertainty which is basically due to the
assumptions in the mathematical model we are using. So, to predict the response we have to consider probabilistic
modeling to obtain reliable design. In this work, we explain the effect of aleatory uncertainty on the response
behaviour of beam structures through Monte Carlo simulation [1]. Also, a stochastic time domain spectral element
method (STSEM) is proposed for the uncertainty analysis and modeling of beam structures [2].
The modeling of sandwich beam considers higher-order sandwich panel theory which can address the core
flexibility. The material parameters are modeled as 1-D non-Gaussian random field [3]. Monte Carlo simulation
considering computationally efficient time-domain spectral element method is explored to perform the numerical
experiments for different boundary conditions as well as for different materials in the sandwich face sheet and core.
The effects of material and geometric uncertainty in the response behaviour of a sandwich beam are quantified. Both
individual as well as combined effect of uncertain material and geometric parameters are studied for the static, free
vibration and dynamic response [1]. For the discretization of non-Gaussian random field expansion optimal linear
estimation (EOLE) [3] is used.
In the next phase, the stochastic time domain spectral element method (STSEM) based modeling is proposed.
Discretization of the random field is obtained by optimal linear estimation (OLE) [3]. Both lognormal and Weibull
type distribution is considered for input random field. The OLE-based discretization of a random field simulates the
random fields digitally and the realizations of stiffness, mass matrix, and dynamic stiffness matrix are also obtained.
Computationally efficient time domain spectral element method (TSEM) [4] is used to develop the STSEM
formulation. For TSEM the degrees of freedom (DOF) are very less and it provides a diagonal mass matrix which
requires less computational cost followed by reduction in the CPU time. In this work, STSEM [2] is proposed for
beam structures and the effect of correlation length on the response statistics is also studied. The computational
efficiency of the proposed method and the deflection statistics of the Timoshenko beam and sandwich beam are
obtained. This STSEM formulation can further used for complex structural and computationally demanding wave
propagation problems.
Keywords: Time domain spectral element method, Monte Carlo simulation, Random field, Uncertainty, Sandwich
beam.
References

186 DACOMA-2019
1. Mukherjee, S., Sekhar, B. R., Gopalakrishnan, S., & Ganguli, R. (2018). Static and Dynamic Analysis of
Sandwich Panel with Spatially Varying Non-Gaussian Properties. Journal of Sandwich Structures & Materials,
DOI: 10.1177/1099636218793979.
2. Mukherjee, S., Gopalakrishnan, S., & Ganguli, R. (2019). Stochastic time domain spectral element analysis of
beam structures. Acta Mechanica, 230(5): 1487-1512.
3. LI, C.-C., Kiureghian. A. (1993). Optimal Discretization of Random Fields. Journal of Engineering Mechanics.
119(6): 1136-1154.
4. Patera, A. T. (2008). A Spectral Element Method for Fluid Dynamics: Laminar Flow in A Channel Expansion.
Journal of Computational Physics, 54(3): 468–488.
Probabilistic stability analysis of earth dam slope using extreme gradient

boosting method
Lin Wang1 and Wengang Zhang1, *
1 Chongqing University, 83 Shabei Road, 400045 Chongqing, China.
*Corresponding Author: Zhang Wengang. Email: zhangwg@cqu.edu.cn.
Abstract: Probabilistic risk assessment of geotechnical structures has gained increasing popularity in geotechnical
engineering, which provides valuable information for the prevention and control of geological hazards[1-3]. It is
well recognized that earth dam is the most common type of dam in the world, however, few studies have performed
to investigate the effects of uncertainties underlying soil-water characteristic curve (SWCC) on the earth dam slope
failure risk. This paper develops a probabilistic risk assessment approach for earth dam slope under the Bayesian
framework. The proposed approach shields light on the propagation of SWCC model selection uncertainty and
parameter uncertainty on the earth dam slope failure assessment, and allows providing a model-independent estimate
of failure risk with the aid of Bayesian model averaging. For illustration, the proposed approach is applied to the risk
assessment of Ashigong earth dam located in Qinghai province of China. Results show that the SWCC parameter
uncertainty and model selection uncertainty can be rationally quantified under the Bayesian framework making use
of the limited test data and prior knowledge. Although the SWCC parameter uncertainty plays a dominant role in the
risk assessment compared with the SWCC model selection uncertainty, the earth dam slope failure risk is
significantly affected by the SWCC model selection. In geotechnical risk assessment, it is prudently to adopt the
model independent estimate obtained from the proposed approach for rational decision making, instead of betting on
a single model.
Keywords: Uncertainty quantification, Risk assessment, Soil-water characteristic curve, Earth dam, Bayesian
approach.
Acknowledgement
This work was supported by the Chongqing Engineering Research Center of Disaster Prevention & Control for
Banks and Structures in Three Gorges Reservoir Area (Nos. SXAPGC18ZD01 and SXAPGC18YB03), and Natural
Science Foundation of Chongqing, China (cstc2018jcyjAX0632), and Special Funding for Postdoctoral research in
Chongqing (No. Xm2017007). The financial support is gratefully acknowledged.
References
1. Huang, J., Lyamin, A. V., Griffiths, D. V., Krabbenhoft, K., Sloan, S. W. (2013). Quantitative risk assessment
of landslide by limit analysis and random fields. Computers and Geotechnics, 53: 60-67.
2. Li, D.,Q., Yang, Z.Y., Cao, Z.J., Zhang, L.M. (2019). Area failure probability method for slope system failure
risk assessment. Computers and Geotechnics, 107: 36-44.
3. Zhang, J., Huang, H. W. (2016). Risk assessment of slope failure considering multiple slip surfaces. Computers
and Geotechnics, 74: 188-195.

Probabilistic Assessment of Serviceability Limit State of Diaphragm Walls for

Braced Excavation in Clays using Point Estimate Method
Wengang Zhang1,2*, Runhong Zhang1, Chongzhi Wu1 and Anthony T.C. Goh3
1 School of Civil Engineering, Chongqing University, #83, Shabei Street, 400045, Chongqing, China.
2Key Laboratory of New Technology for Construction of Cities in Mountain Area, Chongqing University, #83, Shabei
Street, Chongqing 400045, China
3 School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, 639798,
Singapore
*Corresponding Author: Wengang Zhang. Email: cheungwg@126.com
Abstract:. Bracing systems for deep excavations are commonly required to ensure stability and prevent against
excessive deformations in the construction of basements for shopping malls and underground transportation
facilities. For excavations in deposits of soft clays, stiff retaining wall systems such as diaphragm walls help to
restrain ground movements and wall deflections in order to prevent damage to nearby buildings and utilities. It is
quite common for designers to limit the maximum wall deflection to 0.5% times the maximum excavation depth.
However, a review of measured diaphragm wall displacements from various published case histories of successful
deep excavations show that wall deflections can be up to more than 2% times the excavation depth. Since the
allowable threshold wall displacement depends on various influencing parameters, wall deflections should not be
limited to an arbitrary value. For the previous reliability studies that on reliability of braced excavation systems,
usually a probabilistic framework combining a simplified wall deflection estimation model with reliability methods
to determine the probability of serviceability limit state failure. This paper adopts the point estimate method (PEM),
which is free from development of a response surface or surrogate models. It then compares the PEM results with
the combination of polynomial regression and Monte Carlo simulation, which proves that PEM is more efficient and
reliable. This study presents an reasonbale methodology that allows engineers to determine the required threshold
limiting normalized wall deflection to meet the different target serviceability reliability indices.
Keywords:. probabilistic assessment, wall deflection; braced excavation; point estimate method
Acknowledgement
The study is supported by the Natural Science Foundation of Chongqing, China (cstc2018jcyjAX0632) and Venture
& Innovation Support Program for Chongqing Overseas Returnees (cx2017123).
References
1. Mishra, S. , Stauffer, F. , Kinzelbach, W. , Kovar, K. , & Hoehn, E. . (2000). Uncertainty propagation using the
point estimate method. Calibration & Reliability in Groundwater Modelling: Coping with Uncertainty
Modelcare Conference Held in Zurich.
2. Connor Langford, J. , & Diederichs, M. S. . (2013). Reliability based approach to tunnel lining design using a
modified point estimate method. International Journal of Rock Mechanics and Mining Sciences, 60: 263-276.
3. Zhang, W., Goh, A. T. C., & Zhang, Y. (2015). Probabilistic Assessment of Serviceability Limit State of
Diaphragm Walls for Braced Excavation in Clays. Asce-Asme Journal of Risk and Uncertainty in Engineering
Systems Part a-Civil Engineering, 1(3), 06015001. doi: Artn 0601500110.1061/Ajrua6.0000827

Tech Science Press
871 Coronado Center Drive, Suite 200
Henderson, Nevada, 89052, USA
www.techscience.com

Data Driven Computing and Machine Learning in Engineering

Uploaded by

Copyright:

Available Formats

Data Driven Computing and Machine Learning in Engineering

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Data Driven Computing and Machine Learning in Engineering

Uploaded by

Copyright:

Available Formats

DACOMA-19

DATA DRIVEN COMPUTING AND MACHINE LEARNING

Machine Learning Based Solution Of Partial Differential Equations In Computational Mechanics:

Data-Driven Simulation of History-Dependent Materials Using Structured Data Sets and

Effect of size of microcapsules on the self-healing concrete

Automatic Identification of Gas Hydrate Formation using Machine Learning Algorithms

Influence of Thermostatting on Nonequilibrium Molecular Dynamics Simulations of Heat

Thermal conductivity identification of nonlinear functionally graded materials via a machine

Machine learning based solution of Partial Differential Equations in

3 Institute of Continuum Mechanics, Leibniz University Hanover, 30167 Hannover, Germany

DOI: 10.32604/acm.2019.00001 www.techscience.com/acm

Machine Learning for Seawall Structures Under Flooding and Erosion

Figure 1. Framework of CNN for predicting damage of a hollow seawall block.

DOI : 10.32604/acm.2019.00003 ww.techscience.com/acm

Data-driven Construction and Maintenance of Urban Tunnels

Logarithm Xi =Φ-1 [Fi (Yi )]

random field discretion

DOI : 10.32604/acm.2019.00005 www.techscience.com/acm

σv=γH Monitoring Data Inspection Data

Tunnel defect diagnosis

Distribution of max Inspecting crack

Diagnosis of Defects Existing crack

(Discrete BN) Distribution of

Solving direct and inverse computational engineering problem through a deep

Applied Mechanics Lab., Dept. of Engineering Mechanics, School of Aerospace Engineering,

DOI : 10.32604/acm.2019.00009 www.techscience.com/acm

An investigation of deep learning in predicting shield tunnel excavation

Keywords: Shield tunnel, construction parameters, deep learning, LSTM.

DOI : 10.32604/acm.2019.00011 www.techscience.com/acm

The application of a deep energy method in the bending and eigenvalue

*Corresponding Author: Xiaoying Zhuang. Email: zhuang@ikm.uni-hannover.de.

DOI : 10.32604/acm.2019.00013 www.techscience.com/acm

Isogeometric analysis of a phase field model for simulating electromechanical

DOI : 10.32604/acm.2019.00015 www.techscience.com/acm

Exploration signals from higher dimensions base on machine learning

Keywords: Machine learning, High-throughput technologies, Gene, Feature selection.

DOI : 10.32604/acm.2019.00017 www.techscience.com/acm

Volumetric parametrization using convolution neural networks

Keywords: Convolution neural networks, Volumetric parametrization, B-Spline filters.

DOI : 10.32604/acm.2019.00019 www.techscience.com/acm

Boundary Shape Identification of the Heat Conduction Problem Based on

School of Aerospace Engineering, Tsinghua University, 100084, Beijing, China.

DOI : 10.32604/acm.2019.00021 www.techscience.com/acm

Free vibration analysis of Timoshenko nanobeam using differential transform

DOI : 10.32604/acm.2019.00023 www.techscience.com/acm

Data-Driven Simulation of History-Dependent Materials Using Structured

Keywords: Data-driven, History-dependence, Structured data sets.

DOI : 10.32604/acm.2019.00027 www.techscience.com/acm

Artificial neural networks for solving coupled PDEs in flexoelectricity

Figure 1: The architecture of the neural network for flexoelectric model

DOI : 10.32604/acm.2019.00029 www.techscience.com/acm

Nonlinear mechanical behavior and failure mechanism of three-dimensional

Keywords: Nonlinear, Coarse-grained potentials, REBO potential, Carbon nanotube networks.

DOI : 10.32604/acm.2019.00031 www.techscience.com/acm

Designing phononic crystal with anticipated band gap through a deep

DOI : 10.32604/acm.2019.00033 www.techscience.com/acm

A comparative study of machine learning methods applied to the singular

DOI : 10.32604/acm.2019.00035 www.techscience.com/acm

Data driven thermo-mechanical calculation during additive manufacturing

DOI : 10.32604/acm.2019.00037 www.techscience.com/acm