Application of Artificial Intelligence in Additive Manufacturing-A Review
Application of Artificial Intelligence in Additive Manufacturing-A Review
Application of Artificial Intelligence in Additive Manufacturing-A Review
Abstract
Additive manufacturing is technique of producing a 3D object layer by layer. This is different from a conventional
machining operation. In a conventional machining, a component is made by removing chips from the raw material.
Additive manufacturing is also called 3D printing. 3D printing has the advantage in that it can produce any complicated
3D object. 3D printing has been deployed in food industries, chemical industries, aeronautical industries, healthcare
industry, etc. Many researchers have been working in the recent past on additive manufacturing. In this research work
an attempt has been made to explore the usefulness of artificial intelligence (AI) in additive manufacturing. AI enabled
Additive manufacturing results in significant cost reduction. Though many researchers have been working in the area
of AI not many researchers have focused in detail on the applicability of AI in the area of additive manufacturing. In
this context, the findings of current research work assume special significance. The research findings from the current
research work provide future directions which are useful not only for academicians but also for practitioners interested
in pursuing further research.
Keywords
Industry 4.0, AM, Artificial Intelligence, Smart manufacturing, Additive manufacturing
1. Introduction
Artificial intelligence, a technique by which computer connected devices mimic human intelligence. Applications of
Additive manufacturing (AM) applications is found currently in food, chemical, aerospace, automotive and healthcare
industries. The biggest benefit of 3D printing is that even complex objects can be made as per customer requirements.
It is better suited for low volume production as of today. The stages in additive manufacturing are 3D model
preparation, component prototyping and component production. The objective of prefabrication stage is to figure out
whether it is technically possible and feasible to print a given 3D model. Artificial intelligence enabled 3D printing
/or AM is also called smart manufacturing. The smart manufacturing would result in improved productivity.
2. Literature Review
An attempt has been made in the following paragraphs to bring out the usefulness of the AI in additive manufacturing.
2.1 Printability
Printability shows the ease with which a 3D object can be made by 3D printing (Telea and Jalba 2011). In theory, any
3-dimensional object can be made by additive manufacturing. The scope of 3D printing is limited by the component
geometry. Applicability of 3D-printing is also restricted by the type of material. Selection of 3D printing is decided
by the amount of time available for product manufacture. Lu (2016) has proposed an algorithm for measuring the
printability of any 3D object. That is by using this method, it is possible to find out whether any given object can be
produced by 3D printing or not. The method consists (Figure 1) of a module for feature extraction, using machine
learning and another module for managing 3D-printer. This method calculates the printability by looking into cost,
size and time. So, printability measure is very useful in production environments. This would help in decision making.
Adaptive slicing algorithm proposed by Gregori et al. (2014) was computationally very efficient. John C et al. (2016)
developed a implied slicing algorithm. The main objective of using AI in slicing algorithm is to make it
computationally more efficient. Though many researchers have been working on improving the efficiency of slicing
algorithms, still much more work needs to be done with regard to reducing the computational time and possibility of
using parallel computation. With the advent of Industry 4.0, especially Big-data, parallel computing has become a
necessity as the implementation of AI require huge computational infrastructure. Literature shows that researchers
have proposed methods consisting of two processors for speeding up of pre-fabrication in 3D printing.
In correct selection of process parameters during AM would result in LOF defects. Yadroitsev et.al. (2007) have
observed that optimizing the hatch distance porosity in AM products may be minimized. They employed a laser spot
size (SZ) of 70 µm and (HW) hatch width of 120 µm for minimum porosity. Detailed study on the effect of hatch
distance on porosity was made by Mireles et.al. (2015).
During additive manufacturing when there is a gas entrapment, would result in gas pores. Ng et.al. (2009) have done
extensive analysis of gas pores and have concluded that it is impossible to eliminate gas pores in AM products.
However, the AM products may have gas pores to the extent of 0.7%. Sercombe et.al. (2008) have studied gas-pores
formation and have concluded that these pores would act like crack propagation centers.
The process of computing optimum weights by the input layer is called training of the ANN. When a comparison is
made between the output of ANN with that of expected output, it is called supervised learning. For the first iteration,
weights are set to random values. So that less deviation is observed between actual and expected results. The process
is repeated and many iterations are made.
AT the time of training the network, input as well as output data are made available to the given network. The amount
of time taken for training of a network varies from problem to problem. As and when a network approaches pre-
defined performance target, training is said to be complete. No more training is required for the network. At this stage,
the network can be employed for testing with data obtained outside trained data samples. If ANN gives satisfactory
results during this stage this means that the network has learnt the general patterns of the application.
Eckart Uhlmanna et.al. (2018) and Jardine et al. (2006) have designed the necessary steps, for identifying clusters,
from sensors data. The steps are - SLM equipment data obtained through the sensor, Pre-processing of data, and
Cluster-based analysis and evaluation. The results are used for subsequent preventive maintenance of SLM machine.
Many researchers have tried by using cheaper material for providing support. The scheme followed here is that once
the manufacturing of the component gets over support material is dissolved.
Hopkins et.al. (2009), have used acrylic co-polymers as support material. By this way, the support can be taken out
easily. Acrylonitrile butadiene styrene material was used for 3D printing by using support painted with polylactic acid
(Domonoky et.al. 2017). The component when immersed in isopropyl alcohol and potassium hydroxide the support
will be removed. Ni, F et.al. (2017) has explored by using polyvinyl alcohol as a support material (because it dissolves
in water) in 3D printing process.
In SLS process, laser light is used as a heat source. During the process, laser light is moved over metal powder which
is evenly spread on metal platform. During the process, the laser, is moved as per the CAD model in incremental
fashion. This will cause sintering of the metal powder. This forms a layer of the 3D object. After the first layer is
formed, roller is used for spreading metal powder evenly and then the entire process is repeated. The SLS process, is
carried out by many researchers (Mansour and Hague 2003) by using various types of polymers such as ceramic
materials polyester, metals, etc.
In SLS process energy is consumed not only for the processing but also for performing non-value adding activities.
Energy required for processing depends upon how much material is to be fused while building the 3D object. In
addition to processing, energy is also consumed while moving piston, re-coater arm and heating. Researchers have
demonstrated that processing consumes about 56% of the energy. This clearly shows that significant amount of energy
is consumed in performing various non-value adding activities. Thus, in future much attention is required from
researchers in minimizing the energy consumed in performing non-value adding activities. This would the become a
significant value addition.
The following factors contribute to the process energy requirement for sintering the material powder Lu (2016)
observed that absorptivity of parent material, the average intensity of the laser, scanning speed of the laser and the
spot diameter of the laser add up to the processing energy consumption. Main limitation of the additive manufacturing
process is that it consumes more energy than the traditional machining process. Thus, much research is required in
this direction in making the AM process eco-friendly.
(2011)
(2011)
Milling 6.8(Mild steel) Kara and Lee
(2011)
Injection
molding
1.47
Duflou et.al .
(2012)
Industries, nowadays, make use of machine learning enabled 3D printing for spare parts manufacturing. For the supply
of spare parts to automobile and defense sectors. Machine learning based technologies would help in predicting the
life of an equipment. Thus, it is possible to help in knowing when the equipment is going to fail and how many sparts
are required. 3D printing would manufacture that many spare parts and deliver at the right time. Thus, spare parts can
be made available at right time and this would make the customer happy. This would also result in unexpected
equipment failure and the resulting downtime. This would enhance the brand image of the manufacturer. Spare parts
manufacturer will have reduced cost in logistics, reduced inventory cost, if he is close to the customer.
communication network, without any human intervention. The intent of such a system is to control a heterogeneous
swarm of cyber-physical systems by artificial intelligence. Norman (2012) developed a solution for a distributes access
control system by using an expert system. Nowadays, swarm intelligence algorithms are made an integral part of a
control system (Perez et al.,2018).
Desnitsky et al. (2016) have proposed a scheme for detecting intruders in the case of cyber-physical systems. The
scheme includes designing the protected embedded system. A good example given here is the indoor security system.
The solution proposed here selects the best combination of security components depending upon the optimization
problem. In this method, after identifying functional and non-functional requirements, optimum number of
components are determined and then figuring out harm full effects on the device based on testing. The author also
claims that the method can be tweaked for managing group of devices for shielding of the area under consideration.
Though many researchers have been working on enhancing the security of cyber-physical systems (CPS) in smart
manufacturing companies, providing security to CPS is a dynamic problem and requires continuous research for
protecting a company’s CPS. What was observed from the literature is that the cyber- attacks happening each time is
greatly different. This will add up to the complexity dimension of the problem. Also, the vulnerability of a cyber
physical system may vary from company to company. This makes the problem even more difficult.
Intruder detection systems (IDS) consists of different types of sensors (e.g., microwave sensors, electric field sensors,
infra-red sensors, etc.) for detecting intruders as well as isolating the system from the outside. Earlier intruder
detection systems are designed in such a way that it required human intervention. These systems are limited by the
capability of human operators. With the advent of vision systems, many researchers have started solving intruder
detection systems and have successfully demonstrated intruder detection, classification, and tracking. Many
researchers have used both optical and thermal cameras for acquiring images for subsequent processing by intelligent
intruder detection systems. This is because optical cameras have the difficulty in detecting intruders during night times
and also when the intruder is camouflaged.
Camera system
Literature has reported using thermal camera having a field of view of 35 mm and having 640X512 resolution. It was
also reported of using an optical camera -SAMSUNG SCO-2120R optical camera, in addition to thermal camera
(Figure 5). Literature has also reported using of a network video server fitted with IP filtering function for joining to
a network.
The camera is controlled by a pan-tilt device equipped with 3500 and a tilt range of rotation (300-800). In order to
overcome tangential distortion due to manufacturing process, Zhang (2000) proposed a calibration algorithm. In his
research work, 26 chess board images are taken from different perspectives for Camera calibration. In his research,
camera calibration algorithm computes distortion coefficients using the chess board images. Then these coefficients
are used for correcting camera distortion using OpenCV function (Bradski et al. 2008).
Virtual Fence(VF)
It is a virtual line (Figure 6) that differentiates system or protected area from surroundings. A virtual fence is setup by
using a spline curve. The control points of spline curve is setup by using the coordinates given by the operator. The
operator normally uses a GUI for feeding the coordinates.
The cubic spline function connecting two points (xi, yi) and (xi+1, yi+1) is given below.
P(x) = a 𝑖𝑖 (x − 𝑥𝑥𝑖𝑖 )3 + 𝑏𝑏𝑖𝑖 (x − 𝑥𝑥𝑖𝑖 )2 + 𝑐𝑐𝑖𝑖 (x − 𝑥𝑥𝑖𝑖 ) + 𝑑𝑑𝑖𝑖 (1)
Intrusion detection is very much significant in surveillance system. It is capable of identifying moving objects entering
into pre-warning area defined by VF. The system is also capable of detecting moving objects in real time.
During intruder detection, the moving object is classified as humans or animals. Researchers have used deep learning
for classifying moving objects. The Convolution Neural Network (CNN) is also a type of deep learning algorithm.
CNN does learn from the coefficients of the convolution filter uses relevant features for classifying images. Seung
One researcher has proposed a CNN based intruder detection algorithm for classifying moving objects. In their
research work, they used six convolution layers. The detection algorithm extracts feature that are unique for a given
image. They used a training image database of 10 class of human intruders as well as wild animals. The training
database consisted of 5000 images in each class. Thus, there were 50000 images. Most of the intruder detection is
done using OpenCV 3.0 library.
One researcher has implemented a virtual fence utilizing a graphical user interface. A virtual fence will differentiate
areas under surveillance from external areas. One research work (Jia et al., 2014) has been reported to have identified
the moving objects in the background by using a convolution neural network (CNN). The researcher also tried to
classify moving objects as either animal or intruder. The IDS are normally designed in such a way that whenever an
intruder is detected while crossing the virtual fence, an alarm is prompted. It was reported in one research work that
Virtual fence was implemented as a spline curve with set of control points defined using a graphic user interface by
an operator. In another research, deep learning was used for classifying moving objects, estimation of pose recognition
of different actions, recognition of different scenes and speech recognition. Deep learning was implemented using a
computer-based model consisting of many layers. Particle filters are used for intruder tracking. Particle filter is a
simulation-based technique. It uses Bayesian type of probability distribution. It gives good results not only for linear
but also for non-linear environments.
One of the objectives of Intruder detection is to increase the accuracy of intruder identification. In order to address the
limitations of IDS, a researcher has proposed intelligent intruder identification system that improves the reliability and
efficiency of intrusion identification in nuclear power plants. This scheme uses two-cameras and deep learning
technologies for tracking intruder behavior detection.
3. Conclusion
Additive manufacturing process as of today is limited by the geometrical feature, type of material, and the time
requirement. Many researchers have been working in evaluating the suitability of a product to be made by 3D printing.
Researchers have proposed a printability algorithm which suggests whether a component is to be produced by 3D
printing or traditional methods by looking at several factors such as feature, material, time, etc. There is a scope for
extending the applicability of 3D printing for other materials as well.
Tool path generated by the slicer algorithm is very much similar to that of CNC machine program. Slicing algorithm
specifies the path along which the printer head of a 3D machine should move, while manufacturing a product. The
efficiency of the 3D printing process depends to a large extent on the slicing algorithm. Many researchers have been
working on improving the efficiency of the slicing algorithm. There is a wide scope for further research in this topic.
Training of the network (ANN) required large number of samples or training data. The efficiency of training ANN,
depends upon the number of parts in the data set. Larger the samples higher will be the prediction rate. Quality of
samples are also very important. Training is said to be complete when the network is capable of making predictions
to the user defined accuracy levels. The problem with this type of training ANN is that it will consume lot of time.
Thus, there is a wide scope for reducing the time required for training ANN without compromising the quality and
accuracy of predictions.
Many researchers have started working on condition-based maintenance of machine tools. These methods make use
of artificial intelligence and machine learning techniques. These techniques would help in preventing the un
anticipated equipment failure and the subsequent downtime. Online machine diagnostic systems have to be more
robust. This would help in enhancing the machine availability, the product quality and subsequently enhancing the
customer satisfaction. Online machine diagnostics requires further research for making it more robust.
One of the limitations of additive manufacturing is that components with overhang requires support and this would
result in wastes and would increase the cost of manufacturing. Many researchers have explored the effect of orientation
of component on support volume requirements. They had used AI based techniques for minimizing the waste of
material. This aspect of AM is still in infancy and further research is required in the time to come.
Additive manufacturing process consumes more energy than conventional machining process. Thus, additive
manufacturing processes are not energy efficient. Thus, further research is required in making the additive
manufacturing more energy efficient. This would also reduce the cost of manufacturing. This would extend the scope
of additive manufacturing process. Additive manufacturing process also generates billions of small particles during
manufacturing a product, if these particles when inhaled would cause harmful effects.
Machine learning based methods have been proposed by many researchers for performing the predictive maintenance
of machine tools. These methods would help in accurately predicting the remaining useful life of machine tool. This
would help in reducing the machine down time. This would also help in estimating for spare parts requirements. So
that, the spare parts would be ready before the machine fails and thus reduce the machine downtime.
Application of AI in additive manufacturing is still in infancy and further research is required in (i) reducing the energy
consumption by additive manufacturing (ii) increasing the productivity during additive manufacturing (iii) increased
use of AI in additive manufacturing for reducing manufacturing cost (iv) increased accuracy in predicting the
remaining life of machine tool (v) minimizing defects during additive manufactured products.
Though many researchers have been working on enhancing the security of cyber-physical systems (CPS) in smart
manufacturing companies, providing security to CPS is a dynamic problem and requires continuous research for
protecting a company’s CPS. What was observed from the literature is that the cyber-attacks happening each time is
greatly different. This will add up to the complexity dimension of the problem. Also, the vulnerability of a cyber
physical system may vary from company to company. This makes the problem even more difficult.
Acknowledgments
The author would like to express his sincere thanks to the management of Pandit Deendayal Petroleum University, for
providing the necessary infrastructure and timely support.
References
Bradski, G., and Kaehler, A., Learning OpenCV: Comput. Vision with the Open CV Library. O’Reilly Media,
Sebastopol, CA, USA., 2008
Buczak, A.L., and Guven, E., A Survey of Data Mining and Machine Learning Methods for Cyber Security Intrusion
Detection, IEEE Communications Surveys and Tutorials, vol.18(2), pp.1746-1757, 2017.
Da Silveira, G., Borenstein, D., and Fogliatto, H. S., Mass customization: Literature review and research directions,
International Journal of Production Economics, vol. 72, no. 49, pp. 1–13, 2001.
Das, P., Mhapsekar, K., Chowdhury, S., Samant, R., and Anand, S., Selection of Build Orientation for Optimal
Support Structures and Minimum Part Errors in Additive Manufacturing, Computer-Aided Design, and
Applications, pp. 1–13, 2017.
Desnitsky, V., Chechulin, A., Kotenko, I., Levshun, D. and Kolomeec, M., Combined Design Technique for Secure
Embedded Devices Exemplified by a Perimeter Protection System, SPIIRAS Proceedings, vol.48, pp.5-31, 2016.
DOI: 10.15622/sp.48.1
Domonoky, and Bonsai Brain, Support - Full Disclosure [WWW Document], 2016.
Duflou, J. R., Kellens, K., Guo, Y., and Dewulf, W., Critical Comparison of Methods to Determine the Energy Input
for Discrete Manufacturing Processes, CIRP Annals- Manufacturing Technology, vol. 61, no. 1, pp. 63-66, 2012.
Eckart Uhlmanna, Rodrigo Pastl Pontes, Claudio Geisert, and Eckhard Hohwieler, Cluster identification of sensor
data for predictive maintenance in a Selective Laser Melting machine tool, Procedia Manufacturing, vol. 24, pp. 60-
65, 2018.
Edward MEHR, Tim Ellis, Jordan NOONE, Real-time adaptive control of additive manufacturing processes using
machine learning, US20180341248A1, 2018.
Frank, D., and G. Fadel., Expert System-Based Selection of the Preferred Direction of Build for Rapid Prototyping
Processes, Journal of Intelligent Manufacturing, vol. 6, pp. 339–345, 1995.
Gregori, R. M. M. H., Volpato, N., Minetto, R. and. Silva, M. V. G. D., Slicing Triangle Meshes: An Asymptotically
Optimal Algorithm,14th International Conference on Computational Science and Its Applications, Guimaraes, pp.
252-255, 2014.
Hassan, Q. F., Aspects of SOA: An Entry Point for Starters., Annals of Computer Science Series, vol. 7, no. 2, pp 125,
2009.
Hassan, Q. F., Demystifying Cloud Security, CrossTalk, pp. 16– 21, 2011.
Hopkins, P. E., Priedeman Jr. W. R. and Bye, J. F., Support Material for Digital Manufacturing Systems.
US8246888 B2, 2009.
Jardine, A. K. S., Lin, D., and Banjevic, D., A review on machinery diagnostics and prognostics implementing
condition-based maintenance, Mech. Syst. Signal Process., vol. 20, pp. 1483–1510, 2006.
Jia, Y., Steelhammer, E., and Donahue, J., et al., Caffe: An open-source convolutional architecture for fast feature
embedding. In: Proceedings of the 22nd ACM International Conference on Multimedia, pp. 675-678, 2014.
John C. Steuben, Athanasios, Iliopoulos, P., John Michopoulos, G., Implicit slicing for functionally tailored additive
manufacturing, Computer-Aided Design, vol. 77, pp. 107-119, 2016.
Kara, S. and Li, W., Unit Process Energy Consumption Models for Material Removal Processes, CIRP Annals-
Manufacturing Technology, vol. 60, no. 1, pp. 37-40, 2011.
Kaur, P., Kumar, M., and Bhandari, A.,2017, A Review of Detection Approaches for Distributed Denial of Service
Attacks. Systems Science and Control Engineering, vol.5(1), pp.301-320, 2017.
Kulkarni, P., Marsan, A., and Dutta. D., A review of process planning techniques in layered manufacturing. Rapid
Prototyping Journal, vol.6, no.1, pp.18–35, 2000.
Leary, M., Merli, L., Torti, F., Mazur, M., and M. Brandt., Optimal Topology for Additive Manufacture: A Method
for Enabling Additive Manufacture of Support-Free Optimal Structures, Materials and Design, vol. 63, pp. 678–
690, 2014.
Lu, T., Towards a fully automated 3D printability checker, Proc. IEEE Int. Conf. Ind. Technol., vol. 2016–May, pp.
922–927, 2016.
Luo, Y., Ji, Z., Leu, M. C., and Caudill, R., Environmental Performance Analysis of Solid Freedom Fabrication
Processes, Proc. of the IEEE International Symposium on Electronics and the Environment, Paper No. 96, pp. 1-6,
1999.
Mansour, S., and Hague, R., Impact of Rapid Manufacturing on Design for Manufacture for Injection Molding,
Proc. of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, vol. 217, no. 4, pp.
453-461, 2003.
Mell, P., and Grance, T., The NIST definition of cloud computing. National Institute of Standards Technology, vol.53,
pp.50 -62, 2009.
Mireles, J., Ridwan, S., and Morton, P.A., Hinojos A, Wicker RB Analysis, and correction of defects within parts
fabricated using powder bed fusion technology. Surface Topography: Metrology and Properties vol.3, no.3, 034002,
2015.
Morgan, H. D., J. A., Cherry, S., Jonnalagadda, D., Ewing, and J. Sienz, Part Orientation Optimisation for the
Additive Layer Manufacture of Metal Components, The International Journal of Advanced Manufacturing
Technology, vol. 86, pp. 1679–1687, 2016.
Ng, G.K.L., Jarfors, A.E.W., Bi, G., and Zheng, H.Y., Porosity formation and gas bubble retention in laser metal
deposition. Applied Physics A, vol.97(3), pp.641-649, 2009.
Ni, F., Wang, G., and Zhao, H., Fabrication of Water-Soluble Poly (Vinyl Alcohol)-Based Composites with
Improved Thermal Behavior for Potential Three-Dimensional Printing Application, Journal of Applied Polymer
Science, vol. 134, 2017.
Norman, J., 2012, OpenSource Physical Security, URL: http://www.layerone.org/wp-content/uploads/2012/07/
LayerOne2012-John_Norman-DIY_Access_Control.
Pham, D. T., Dimov, S. S., and Gault, R. S., Part Orientation in Stereolithography, International Journal of
Advanced Manufacturing Technology, vol. 15, pp. 674–682, 1999.
Principe, J.C., Euliano, N.R. and lefebvre, W.C., Neural, and adaptive systems: fundamentals through simulation.
New York, John Wiley, 2000.
Sercombe, T., Jones, N., Day, R., and Kop, A., Heat treatment of Ti‐6Al‐7Nb components produced by selective laser
melting, Rapid Prototyping Journal, vol.14(5), pp.300-304, 2008.
Strano, G., L., Hao, R. M., Everson, and Evans, K. E., A New Approach to the Design and Optimisation of Support
Structures in Additive Manufacturing, International Journal of Advanced Manufacturing Technology, vol.66, pp.
1247–1254, 2013.
Tao F., Zhang, L., Venkatesh, V. C., Luo, Y., and Cheng, Y., Cloud manufacturing: a computing and service-oriented
manufacturing model, Proceedings of Institute of Mechanical Engineering Part B Journal Engineering Manufacture,
vol. 225, no. 4, pp. 1969–1976, 2011.
Telea, A., and Jalba, A., Voxel-Based Assessment of Printability of 3D Shapes, Mathematical Morphology and Its
Applications to Image and Signal Processing: 10th International Symposium, ISMM 2011, Verbania-Intra, Italy, July
6-8, 2011.
Yadroitsev, I., Thivillon, L., Bertrand, P., and Smurov, I., Strategy of manufacturing components with designed
internal structure by selective laser melting of metallic powder. Applied Surface Science, vol.254(4), pp.980-983.,
2007.
Yam, R. C. M., Tse, P. W., Li, L., and Tu, P., Intelligent Predictive Decision Support System for Condition-Based
Maintenance, International Journal of Advanced Manufacturing Technology, vol.17, pp.383–391, 2001.
Zhang, Z., 2000. A flexible new technique for camera calibration. IEEE Transactions on Pattern Analysis and
Machine Intelligence, vol. 22, no.11, pp.1330–1334, 2000.
Biography
Dr. M.B. Kiran is an Associate Professor in the Department of Mechanical Engineering, School of Technology,
Pandit Deendayal Petroleum University, Gandhinagar, Gujarat, INDIA. He earned his graduation (B.E.) from the
University of Mysore in 1987. He did his post-graduation (M.E.) in Production Engineering from P.S.G. College of
Technology (1991) and Doctoral degree (Ph.D.), in Surface Metrology from Indian Institute of Technology (I.I.T.),
Madras in 1997. He has Industry/Research/Teaching experience of 25 years. He has published technical papers in
many reputed national/international journals and conferences. He is a certified project manager (P.M.P.) from P.M.I.
He has completed many mission-critical projects. He has conducted many training programs for working executives.