6th European Workshop on

Structural Health Monitoring - We.3.E.2

More info about this article: http://www.ndt.net/?id=14171

Lamb Wave Propagation Modeling Using

Cellular Automata
P. KLUSKA1, W. J. STASZEWSKI1, M. J. LEAMY2 and T.UHL1

ABSTRACT

Theory of Lamb Wave propagation in plate-like structures have found many

practical applications in Structural Health Monitoring. However for better
understanding of complex physical phenomena associated with wave propagation and
wave interaction with damage numerical simulations are as important as laboratory
experiments. The paper shows the application of Cellular Automata technique for
modeling of elastic wave propagation. After a brief introduction to Lamb waves, 2-D
triangular Cellular Automata approach for wave propagation is presented. Numerical
simulations are performed for undamaged and damaged aluminium plates. The results
are compared with the Local Interaction Simulation Approach (LISA).

INTRODUCTION

Monitoring for structural damage is important in maintenance of many

engineering structures. Various damage detection techniques have been developed for
Structural Health Monitoring (SHM) applications. Methods based on guided waves
are particularly attractive in plate-like structures. It appears that Lamb waves are the
most widely used guided ultrasonic waves for damage detection in metals and
composites, as reviewed in [1-3]. Despite numerous laboratory implementation SHM
application of Lamb waves in real engineering systems remain limited. The
complexity of physical mechanism associated with these waves is an important factor
when implementation of Lamb waves techniques is considered for damage detection.
_____________
1
P. Kluska, W.J. Staszewski and T. Uhl - Department of Robotics and Mechatronics, AGH
University of Science and Technology, Al. Mickiewicza 30, 30-059 Krakow, Poland;
2
M.J. Leamy - George W. Woodruff School of Mechanical Engineering, Georgia Institute of
Technology, Atlanta, GA 30332-04505, USA

1
Licence: http://creativecommons.org/licenses/by-nd/3.0
Also, signal features produced by defects tend to be embedded in the background
noise associated with material, structural and environmental variability and thus are
difficult to detect reliably. It is widely acknowledged that numerical simulations can
help with the entire damage detection implementation process and interpretation
analysis, leading to accurate SHM diagnosis and prognosis.
Various methods have been developed for modelling and numerical simulations of
elastic wave propagation, including semi-analytic techniques (e.g. methods based on
the theory of diffraction, boundary element methods) and numerical algorithms such
as methods based finite differences, finite elements, spectral elements, elastodynamic
finite integration techniques (EFIT) or local interaction simulation approach (LISA),
as reviewed in [4] . The LISA is particularly attractive for very fast parallel
computation of large models of wave propagation in complex media with sharp
interfaces [5]. More recently the Cellular Automata (CA) approach based on
rectangular [6] and triangular [7] . The latter offers arbitrary meshing geometries and
is an attractive alternative finite difference based methods.
The major objective of the paper is to explore the CA approach for Lamb wave
propagation in metallic structures. A 2-D case study of wave propagation in
undamaged and damaged aluminium plates is investigated. Numerical results are
compared with simulations based on the LISA technique.

LAMB WAVES

Lamb waves are elastic perturbations propagating in solid plates with free
boundaries. These waves arise from coupling of shear and longitudinal waves
reflected at the top and bottom surfaces of the plate, leading to an infinite number of
dispersive modes. The wave propagation problem can be analysed using the classical
elastodynamic wave equation

(1)

where where and are Lamé constants, ρ is the material density and W is the
vector of particle displacements. This equation can be solved using the displacement
potential approach or the partial wave technique, as demonstrated in [8]. The former
decoupled the elastodynamic wave equations leading to the well known Rayleigh-
Lamb frequency relations

q 2  k 2 2

tan(qh) 4k 2 pq
(2)
tan( ph)

for symmetric Sn (n=0,1,2,...) modes

tan(qh)


q2  k 2 
2

(3)
tan( ph) 4k 2 pq

for antisymmetric An (n=0,1,2, ...) modes, where k is the wave number and variables p
and q can be defined in terms of circular frequency ω = 2πf and the shear cT and
longitudinal cL wave velocities as

2
 (4)

The fundamental S0 and A0 modes will only propagate in the plate for small values of
the frequency-and-plate-thickness" product. The elastodynamic and Rayleigh-Lamb
equations can be solved numerically to obtain physical displacements and phase/group
velocities as a function of the frequency-and-plate-thickness" product , respectively.

CELLULAR AUTOMATA

Cellular Automata (CA) were introduced in the early 1940’s and developed
further in the following years by John von Neumann to model self-replicating
systems [9]. At the same time Stanislaw Ulam started his work for Los Alamos
Scientific Laboratory and began research related to biology [10]. Both scientists
shared their experiences and built the base of the CA theory. Following these
development the potential of the method for modelling complex physical phenomena
has been recognised. Stephen Wolfram made a detailed study of CA and proposed
their classification [11].
CA are mathematical idealisations of physical systems and introduce discretisation
of time and space. CA consist of cells array in D-dimension. Every cell has set of
states. These states are usually the same for all cells. The state of i-cell at a time step
t+1 is determined by a rules function R. Set of rules utilize i-cell neighbours’ state and
its own state, i.e.

si (t  1)  R({s j (t )}) , j  N (i) (5)

where R is a function containing set of rules, si is the i cell’s state in the next step,
t is the time step, sji the neighbours’ state in the previous step and N(i) is the
neighbourhood of the i cell.

Figure 1. Neighbourhood structures in two-dimensional CA with comparison between classical

rectangular automata (left) and triangular automata for arbitrary geometries (right) [12].

Two different types of cells are used in practice. Figure 1 shows neighbourhood
types examples for rectangular and triangular CA. In these examples the dark cell in

3
the middle of such structures is the i-cell from Equation (5) and can be described
using the Cartesian coordinates pair (x, y) according to considered 2-D space.
CA are an efficient tool for dynamic physical systems simulations because of
possibility of parallel computation. Numerical implementation of CA for mechanical
phenomena simulation are different in comparison to methods based on finite
elements that involve partial differential equation. In CA applications the global
behaviour of physical system is determined by a local interaction between cells based
on a specified rule set. Figure 2 shows how strains from neighbours acting on a
particular cell can be calculated for the 2-D problem. These strains are used for
specific rule functions introduced in Equation (5).

Figure 2. Strains needed to establish a rule set for CA [12].

LAMB WAVE PROPAGATION MODEL

CA were used to model Lamb wave propagation. A rectangular (400 x 150 mm;
2 mm thickness) aluminium plate (Young's modulus E = 71 GPa, Poisson ratio ν =
0.338 and density ρ = 2711 kg/m3) was used in this model. Both damaged and
undamaged cases were investigated. A circular hole (diameter equal to 4 mm),
located in the middle of the plate, was considered as a simulated damage. Figure 3
illustrates the geometry of the considered model. The excitation and response
positions were selected at the top (x=200 mm and y = 5 mm) and bottom (x = 200 mm
and y = 145 mm) of the plate respectively.

4
 Figure 3. Aluminium specimen used for CA Lamb wave propagation modelling.

Parallel implementation of triangular CA - proposed and described in [12] - was

used in these investigations. This implementation is based an object oriented code
written in Java. The Java-based numerical simulation platform used allows for elastic
wave propagation simulations. The geometry of the plate was modelled using a set of
triangular cells obtained as a finite element mesh from the commercial COMSOL
multiphysics software package. Figure 4 illustrates the mesh fragment for the plate
with a circular defect. The meshes for the undamaged and damaged plates contained
approximately 30 000 cells. The discretisation used was equal to 1e7 steps per second.

Figure 4. Mesh fragment for the plate with a circular defect.

A five-cycle - shifted by 900 sine burst signal with a Hann window envelope, as
shown in Figure 5 - was used for excitation. Although in theory a minimum two
Lamb wave modes propagate in plates, it is always possible to select excitation
frequency for which the amplitude of one fundamental modes is reduced almost to
zero. This so-called single mode excitation was used in the current 2-D investigations.
The excitation frequency - selected experimentally - was equal to 100 kHz. This led to
A0 Lamb wave mode propagation.

5
 Figure 5. Excitation signal used for Lamb wave propagation modelling.

RESULTS AND DISCUSSION

Once the model of the undamaged and damaged aluminium plate implemented
CA were used to simulated Lamb wave propagation. The simulation results for the
(x,y) plane are presented in Figures was developed Figure 6.

(a)

(b)

(c)

Figure 6. CA Lamb wave propagation simulated for the undamaged (left column) and damaged (right
column) plate for various time snapshots.

6
 Here, the normalised displacement amplitude is given for various time moments.
The results - given for the undamaged plate in the left column and for the damaged
plate in the right column - display expected wave propagation features. The incident
wave is clearly broken after it passes through the hole (Figure 6a - right column and
attenuated when it reaches the bottom of the plate (Figure 6b - right column.
Numerous reflections from plate edges can be observed in Figure 6c for both cases
investigated.
The same results but obtained for the LISA method [123] - shown in Figure 7 -
display very similar wave propagation phenomena and wave interaction with damage
features. Direct comparison of Lamb wave responses can be analysed in Figure 8. The
CA and LISA results for the incident wave are almost identical for both cases
investigated. However, wave reflection components simulated by the CA approach
appear to be delayed if compared with the LISA approach. This is probably due to
different meshing geometry used in the relevant models. The amplitude of the incident
wave for the damaged plate is reduced in Figure 7b, as expected.

(a)

(b)

(c)

Figure 7. LISA Lamb wave propagation simulated for the undamaged (left column) and damaged (right
column) plate for various time snapshots.

7
 (a)

(b)

Figure 8. Lamb wave responses obtained for CA and LISA simulations: ( a) undamaged plate; (b)
damaged plate.

CONCLUSIONS

Cellular Automata were explored to simulate Lamb wave propagation in

undamaged and damaged aluminium plates. A simple single-mode propagation was
considered. The results were compared with the LISA method.
Wave propagation phenomena and damage features in Lamb wave responses were
identical for both methods when compared visually. The amplitude and arrival time
for the incident wave component were almost identical. However, some increased
arrival times for reflected components could be observed for CA simulations if
compared with the LISA results. These discrepancies - probably due to different
meshing - need further investigations.
Nevertheless, these preliminary results are very promising. The ability of
triangular CA to model Lamb wave propagation in arbitrary geometries has been
demonstrated.
ACKNOWLEDGEMENTS

8
 The work presented in this paper was supported by funding from the Foundation
for Polish Science under the research WELCOME project no. Project no. 2010-3/2.

REFERENCES

1. W.J. Staszewski, 2004, Structural Health Monitoring Using Guided Ultrasonic Waves, In:
Advances in Smart Technologies in Structural Engineering, J. Holnicki-szulc and C.A. Mota
Soares, eds., Berlin: Springer, pp. 117-162.
2. A. Raghavan and C.E. Cesnik , 2007, Review of Guided-wave Structural Health Monitoring, The
Shock and Vibration Digest, 39, pp 91-114.
3. A.J. Croxford, P.D. Wilcox, B.W. Drinkwater and G. Konstantinidis, 2007, Strategies for Guided-
Wave Structural Health Monitoring, Proc. Roy. Soc. A, 463, pp. 2961-2981.
4. A. Balikin, 2007, Numerical Simulation of Guided Waves for SHM - a Literature Survey, UPSSP
Platform Grant Report, Department of Mechanical Engineering, Sheffield University.
5. P. Paćko, T. Bielak, A.B. Spencer, W.J. Staszewski, T. Uhl and K. Worden, 2012, Lamb Wave
Propagation Modelling and Simulation Using Parallel Processing Architecture and Graphical
Cards, Smart Structures and Materials, accepted for publication.
6. M. J. Leamy, 2008, Application of Cellular Automata Modelling to Seismic Elastodynamics, In:
International Journal of Solids and Structures 45 (2008), pp. 4835-4849
7. R. K. Hopman, M. J. Leamy, 2011, Triangular Cellular Automata for Computing Two-
Dimensional Elastodynamic Response on Arbitrary Domains, Journal of Applied Mechanics,
78(2): 021020, pp. 1-10.
8. J. L. Rose, 1999, Ultrasonic Waves in Solid Media, Cambridge University Press.
9. J. von Neumann, 1966 , Theory of Self-Reproducing Automata, University Illinois Press.
10. W. A. Beyer, P. H, Sellers, M. S. Waterman, 1985, Stanislaw M. Ulam’s Contributions to
Theoretical Theory, Letters in Mathematical Physics 10, pp. 231-242.
11. S. Wolfram, 2002, A New Kind of Science, Wolfram Media, Inc., Champaign, IL, USA.
12. M.J. Leamy, 2011, Parallel Implementation of Traingular Cellular Automata for Computing Two-
Dimensional Elastodynamic Response on Arbitrary Domains, In proceedings of the 10th
International Conference on Vibration Problems, Prague, Czech Republic, September 5-8.