Electro-magneto-mechanical elastodynamic model for Lamb wave

damage quantification in composites

Luke Borkowski and Aditi Chattopadhyay
School for Engineering of Matter, Transport, and Energy
Arizona State University, Tempe, AZ, 85287

ABSTRACT

Physics-based wave propagation computational models play a key role in structural health monitoring (SHM) and the
development of improved damage quantification methodologies. Guided waves (GWs), such as Lamb waves, provide the
capability to monitor large plate-like aerospace structures with limited actuators and sensors and are sensitive to small
scale damage; however due to the complex nature of GWs, accurate and efficient computation tools are necessary to
investigate the mechanisms responsible for dispersion, coupling, and interaction with damage. In this paper, the local
interaction simulation approach (LISA) coupled with the sharp interface model (SIM) solution methodology is used to
solve the fully coupled electro-magneto-mechanical elastodynamic equations for the piezoelectric and piezomagnetic
actuation and sensing of GWs in fiber reinforced composite material systems. The final framework provides the full
three-dimensional displacement as well as electrical and magnetic potential fields for arbitrary plate and transducer
geometries and excitation waveform and frequency. The model is validated experimentally and proven computationally
efficient for a laminated composite plate. Studies are performed with surface bonded piezoelectric and embedded
piezomagnetic sensors to gain insight into the physics of experimental techniques used for SHM. The symmetric
collocation of piezoelectric actuators is modeled to demonstrate mode suppression in laminated composites for the
purpose of damage detection. The effect of delamination and damage (i.e., matrix cracking) on the GW propagation is
demonstrated and quantified. The developed model provides a valuable tool for the improvement of SHM techniques due
to its proven accuracy and computational efficiency.
Keywords: structural health monitoring, Lamb wave, numerical wave propagation modeling, electro-magneto-
mechanical coupling, guided wave, piezoelectric, piezomagnetic

1. INTRODUCTION
Guided wave (GW) techniques for the purpose of damage detection in plate-like geometries have been proven to be
one of the most effective, economical, and accurate means of performing structural health monitoring (SHM) on
aerospace structures, including metallic and composite.1-3 One specific type of GW, Lamb waves, have demonstrated
their effectiveness in in-situ characterization of damage due to their ability to travel great distances in plate-like
structures and sensitivity to small scale damage.2,4-6 Data-driven approaches utilizing Lamb waves,7-9 although proven
accurate, are limited in their effectiveness because of the time-consuming and expensive nature of experiments prohibits
conducting a large number of studies over a short period time. Therefore physics-based models, when combined with
experiments and data driven models to achieve a hybrid sensing approach, can provide information, capabilities, and
flexibility that would otherwise not be available.10 This hybrid framework is referred to as virtual sensing.
The development of numerical models for GW propagation simulation is necessary because of the limitations
associated with analytical models, especially when structures involving complex geometries, material architectures and
properties, and damage are of interest. Numerical techniques used for wave propagation modeling include finite element
method,11-13 finite strip elements,14-16 boundary element method,17-18 spectral element methods,19-21 and local interaction
simulation approach (LISA) / sharp interface model (SIM).22-24 Due to the capability of the LISA/SIM numerical
modeling scheme to accurately model wave propagation across sharp material boundaries, such as damage,22-24 and its
proven computational efficiency,24-25 this technique is well suited for use in a virtual sensing framework for damage
detection and quantification. However, the application and accuracy of this technique has been limited because the
actuation and sensing of wave signals using piezoelectric and piezomagnetic transducers has not been considered until
recently.
In Ref. 25 the authors incorporated piezoelectric coupling within the LISA/SIM framework to allow piezoelectric
actuation and sensing to be explicitly modeled. To further extend the theory, piezomagnetic coupling will be
incorporated to provide the capability to model noncontact magnetic actuation and sensing. Because of the ability of

Health Monitoring of Structural and Biological Systems 2014, edited by Tribikram Kundu, Proc. of SPIE Vol. 9064, 90641B ·
© 2014 SPIE · CCC code: 0277-786X/14/$18 · doi: 10.1117/12.2045232

Proc. of SPIE Vol. 9064 90641B-1

piezomagnetic actuation and sensing to be performed without contact with the structure, these approaches can be used
where piezoelectric actuation and sensing is difficult, such as in cases where sensors or actuators must be embedded in
the structure or where contact is not feasible.26-30 In addition to simulating piezomagnetic actuation and sensing, the new
formulation can be used to model the dynamic behavior of magneto-electro-elastic composites where the combination of
piezoelectric and piezomagnetic material phases results in an overall magnetoelectric coefficient greater than that of any
single phase material.31-36 These composites are used extensively in electronic packaging, transducers (e.g., magnetic
field sensors, ultrasonic/acoustic sensors and actuators), and for energy conversion.37
In the current paper, a fully coupled electro-magneto-mechanical elastodynamic model for wave propagation in a
heterogeneous, anisotropic material system is developed. The objective of developing this novel modeling scheme is to
advance the formulation presented in Ref. 25 for electromechanical coupling to also include piezomagnetic coupling.
This improved actuation and sensing capability will, in turn, ease the monitoring strategy used for damage detection with
GWs. The final set of equations provides the full 3D displacement as well as electrical and magnetic potential fields for
arbitrary plate and transducer geometries and excitation waveform and frequency. The model framework is based on the
LISA/SIM solution methodology framework developed in Ref. 24 for an orthotropic material, but is extended to include
piezoelectric and piezomagnetic coupling and explicit consideration of the actuators and sensors for an anisotropic
material system. The model is validated experimentally by comparing the simulated group velocity to that achieved
experimentally over a wide range of frequency-thickness products. Various studies are investigated to demonstrate the
GW propagation features in composites and the capabilities of the develop model to aid in the detection, localization,
and quantification of damage.

2. THREE-DIMENSIONAL ELECTRO-MAGNETO-MECHANICAL COUPLED

ELASTODYNAMIC MODEL FRAMEWORK
Accounting explicitly for the physics of piezoelectric and piezomagnetic actuation and sensing within the LISA/SIM
solution methodology requires the derivation of a set of incremental equations for the solution of a 3D fully generalized,
fully coupled electro-magneto-mechanical elastodynamic wave propagation model for a heterogeneous, anisotropic
material system. The final set of equations will provide the evolution of the time-varying displacement, electric potential,
and magnetic potential fields for an arbitrary geometry and actuation waveform. This formulation solves the
elastodynamic equations as an initial value problem and Maxwell’s equations (i.e., Gauss’s laws for electric and
magnetic fields) as boundary value problems at each time step. The derivation for the fully coupled electromechanical
LISA/SIM formulation25 is extended in this paper to include magnetic coupling. The incorporation of piezomagnetic
coupling into the theory further expands the capabilities of the LISA/SIM framework to simulate more complex
actuation and sensing in GW damage detection models as well as the dynamic behavior of electromagnetic coupled
composites.

2.1 Governing Equations and Discretization

For this model, the spatial domain is discretized in the x, y, z directions into a cuboidal grid with dimensions x, y,
and z, respectively, as shown in Figure 1. The material properties of each cell are defined at the lower left front corner
of the cell, meaning an element with its center at location   x / 2,   y / 2,   z / 2 will have its properties
defined at  ,  ,   . Additional points are defined in the grid, denoted by a star and a cross in Figure 1, at infinitesimal
distances  and  from the nodal points and interface to enforce continuity of displacement, electrical and magnetic
fields, and traction at the cell interfaces. The distances  and  are exaggerated in Figure 1 for clarity.

Proc. of SPIE Vol. 9064 90641B-2

 Ammo
MIME

LEWMOM

Figure 1: Definition of grid and supplemental points

The governing equations for a linear piezoelectric continuum can be found in Ref. 38 while the governing equations
for a linear piezomagnetic continuum are available in Ref. 39. For an electro-magneto-elastic medium, the constitutive
equation that governs the interaction of the elastic, electric, and magnetic fields can be written as
 ij  Cijkl  kl  ekij Ek  qkij H k , (1)
where ij, Cijkl, kl, ekij, qkij, Ek, Hk are components of the second order stress tensor, fourth order stiffness tensor, second
order strain tensor, third order piezoelectric tensor, third order piezomagnetic tensor, first order electric field tensor, and
first order magnetic field tensor, respectively. In addition, the electric displacement vector can be expressed in terms of
the strain and electric and magnetic field in the form
Di  eijk  jk  ij E j  aij H j , (2)
where Di is a component in the first order electric displacement tensor, ij is a component of the second order dielectric
tensor, and aij is a component of the second order magnetoelectric coefficient tensor. The magnetic flux density vector
can be expressed as
Bi  qijk  jk  aij E j  ij H j (3)
where ij is a component of the magnetic permeability tensor.
The components of the small strain tensor kl are expressed in terms of the displacement components uk using the
strain-displacement relation,

1
 kl  uk ,l  ul ,k ,
2
 (4)

and the components of the electric field Ei and magnetic field Hi are obtained from the electric potential i via
Ei  ,i (5)
and the magnetic potential i via
Hi  ,i (6)
Using the strain-displacement relation (Eq. (4)), definition of electric field (Eq. (5)), definition of the magnetic field
(Eq. (6)), and the symmetry of the stiffness tensor, Eqs. (1), (2), and (3) can be expressed in terms of displacement,
electric potential, and magnetic potential as
 ij  Cijkl uk ,l  ekij, k  qkij, k , (7)

Di  eijk u j ,k  ij, j  aij, j , (8)

and
Bi  qijk u j ,k  aij, j  ij, j . (9)
In an elastic medium, force equilibrium is enforced through the elastodynamic wave equation in the form
Cijkl uk ,lj  ekij, kj  qkij, kj  ui . (10)

Proc. of SPIE Vol. 9064 90641B-3

The effect of viscoelasticity is not included in the current derivation since this has been investigated by previous
researchers, such as in Ref. 40.
In the absence of volume charges, Maxwell’s equation,
∇ D  0 , (11)
must be satisfied, which requires
eijk u j ,ki  ij, ji  aij, ji  0 . (12)
In the absence of magnetic monopoles and assuming static electricity and magneticity, the following equation must
also be satisfied
∇ B  0 , (13)
which requires
qijk u j ,ki  aij, ji  ij, ji  0 . (14)
A central difference scheme is used to approximate the second order derivatives of the displacement and electrical
potential at points defined at   a ,   b ,   c  in the cuboidal grid in terms of their first order derivatives. Here, a,
b, and c represent neighboring nodes and all have the value of ±1 and  represents a small distance away from the node.
The expressions for the second order derivatives of displacement and electric potential are provided in Ref. 25 while
some of the expressions for the second order derivative of magnetic potential are supplied here for clarity.
  a ,  b , c
,1  a /2,  ,  ,1  a , b , c
,11  (15)
ax / 2
  a,  b /2,  ,  b /2,
  a ,  b , c
,2  ,2
,12  (16)
ax
Similarly, the first order derivative of displacement and electric and magnetic potential at points   a / 2,  ,   ,
 ,   b / 2,   , and  ,  ,   c / 2 are also expressed using finite difference. For brevity, these expressions are not
included in the present paper.
Next, continuity of displacement and electric and magnetic potential will be enforced at additional points defined at
a small distance from the grid points. A very small distance, denoted by  will be defined as
   x , x  1 . (17)
Since the procedure for enforcing continuity of the first derivative of displacement is similar to that presented in Ref. 24
and continuity of the first order derivative of electric potential is similar to that presented in Ref. 25, they are not
repeated in this paper. However, the continuity of the first derivative of magnetic potential results in
  a ,  b , c   a, ,   , ,   a ,  b , c 
  a,  ,
  , ,
  a ,  b , c
,1   a ,  b , c
 ,1    
,1
ax ,1
ax
(18) (19) (20)
 ,  b,  ,  ,
  a ,  b , c     a ,  b , c  , b,   , ,
,2    a ,  b , c   a ,  b , c ,2 
by ,2  ,2 by
(21) (22) (23)
 ,  , c
  a ,  b , c    ,  ,   a ,  b , c 
 ,  , c
  ,  ,
,3  ,3    a ,  b , c   a , b , c
,3  ,3
cz cz .
(24) (25) (26)

The expressions for the first order derivatives in Eqs. (18), (22), and (26), in addition to their displacement and
electric counterparts, remain unknown. To solve for equilibrium and Maxwell’s equations, continuity of tractions,
electric displacement, and magnetic flux density across the element interfaces are enforced. This will allow for the
unknown first order derivatives to be eliminated.
2.2 Enforcement of Elastodynamic Equilibrium and Continuity of Traction
Evaluating the elastodynamic equilibrium at the points   a ,   b ,   c  can be expressed as

Proc. of SPIE Vol. 9064 90641B-4

   a ,   b ,  c   a ,   b ,  c
Cijkl uk ,lj  elij  a ,   b ,  c ,lj  a ,  b ,  c  qlij  a ,  b ,  c ,lj  a ,  b ,  c
(27)
    a ,  b ,  c ui  a ,  b ,  c
for a,b,c= ±1.
The stress tensor at points near the nodes can be expressed as
 ij  a ,  b ,  c
(28)
 Cijkl  a ,   b ,  c uk,l a ,   b ,  c  elij  a ,   b ,  c ,l  a ,   b ,  c  qlij  a ,   b ,  c ,l  a ,   b ,  c
for a,b,c= ±1.
Next, traction continuity is imposed across the cell interfaces at points near the nodes while recalling that the
material properties (e.g., stiffness tensor, density, piezoelectric tensor, piezomagnetic tensor, dielectric tensor, and
magnetic permeability tensor) are constant in each cell, for example,
  ,  ,    ,  ,    ,  ,    ,  ,   , ,
Cijkl  Cijkl  Cijkl  Cijkl  Cijkl . (29)
Since the cell faces are orthogonal and aligned, the tractions can be expressed directly as the stress tensor. The vector
equations can be expressed in compacted form as
 i1  , b , c   i1  , b , c (30)
  a ,  , c   a ,  , c
i2  i2 (31)
 i3a , b ,    i3 , b ,  (32)
for a,b,c= ±1.

2.3 Final Expressions for Nodal Mechanical Displacement

After substituting the expressions for stress into Eqs. (30), (31), and (32), replacing the first and second order spatial
derivatives with their respective finite difference expressions in Eqs. (27) and (28), and summing over a, b, and c, the
unevaluated first order derivatives can be eliminated through a linear combination of the traction continuity and
equilibrium equations. The time derivatives of the displacement are then expanded using finite difference, and the final
expression for the nodal displacement at time t+t is achieved, as presented in Eqs. (33), (35), (36), and (37). The
solution of displacement at any point at time t+t, solved using forward integration, is a function of the material
properties of the surrounding elements and the displacement and electric potential of the surrounding nodes at time t and
t-t.
2
 ,  ,  , t  1  2u ,  ,  , t  u ,  ,  , t  1   t
ui i i 8
  f  g  h (33)
a,b,c 1
where
1
  s
8 a,b,c 1 ,
(34)

and
 f fy f 
f  2 x   z
 x 2 y 2 z 2 
 
   
f x  Cis1k1 uk  a,  ,  uk ,  ,  e1si1    a,  ,    ,  ,  q1si1    a,  ,   ,  ,   (35)
f y  Cis2k 2 u  ,   b,
k

 u     e    
k
, , s
2i 2
, b,
  
 ,  ,
 q2si 2  ,  b,
  ,  ,

f z  Cis3k 3 u  
k
, , c
 u     e    
k
, , s
3i 3
, , c
       q    
, , s
3i 3
, , c
   ,  ,  ,
and

Proc. of SPIE Vol. 9064 90641B-5

  g xy g xz g yx g yz g zx g zy 
g  2       
 abxy acxz abxy bcy z acxz bcy z 
g xy  Cis1k 2 u
k 
,  b,
  
 uk ,  ,  e2si1   ,  b,    ,  ,  q2si1  ,  b,   ,  ,  
g xz  Cis1k 3 u    u     e
k
, , c
k
, , s
3i1            q
, , c , , s
3i1       
, , c , ,

g yx  Cis2k1 u    u     e
k
 a, ,
k
, , s
1i 2            q
 a, , , , s
1i 2       
 a, , , , (36)

g yz  Cis2k 3 u    u     e
k
, , c
k
, , s
3i 2            q
, , c , , s
3i 2       
, , c , ,

g zx  Cis3k1 u    u     e
k
 a, ,
k
, , s
1i 3            q
 a, , , , s
1i 3       
 a, , , ,

g zy  Cis3k 2 u    u     e
k
,  b,
k
, , s
2i 3            q
,  b, , , s
2i 3        ,
,  b, , ,

and
hxy hxz hyz
h  
abxy acxz bcy z

 
hxy  Cis1k 2  Cis2k1 uk  a,  b,  uk  a,  ,  uk ,  b,  uk ,  , 
  e1si 2  e2si1     a,  b,     a,  ,    ,  b,    , , 

  q1si 2  q2si1     a,  b,     a,  ,    ,  b,    ,  , 

hxz   Cis1k 3  Cis3k1   uk  a,  ,  c  uk  a,  ,  uk ,  ,  c  uk ,  , 

(37)
  e1si 3  e3si1     a,  ,  c     a,  ,    ,  ,  c    ,  , 

  q1si3  q3si1     a,  ,  c     a,  ,    ,  ,  c    ,  , 

hyz   Cis2k 3  Cis3k 2   uk ,  b,  c  uk ,  b,  uk ,  ,  c  uk ,  , 

  e2si 3  e3si 2    ,  b,  c    ,  b,    ,  ,  c    ,  , 

,

  q2si3  q3si 2    ,  b,  c    ,  b,    ,  ,  c    ,  , 

where superscript “s” denotes the point   a ,   b ,   c  .

2.4 Enforcement of Maxwell’s Equation (Gauss’s Electric Field Law) and Continuity of Electric Displacement
A similar approach is followed to achieve an expression for the electric potential at time t. First, Maxwell’s equation
is enforced at every point   a ,   b ,   c  as
  a ,  b ,  c   a ,  b ,  c
eijk u j ,ki  ij  a ,  b , c ,ji a , b , c  aij  a , b , c ,ji a , b , c  0
(38)
for a,b,c= ±1.
Next the continuity of the electric displacements are enforced at infinitesimal distances from the interface, which
will result in the following equations
D1  , b , c  D1  , b , c (39)
  a ,   , c   a ,  , c
D2  D2 (40)
D3 a , b ,   D3 a , b ,  (41)
for a,b,c= ±1.

Proc. of SPIE Vol. 9064 90641B-6

2.5 Final Expressions for Nodal Electrical Potential
After substituting the expressions for electric displacement (Eq. (8)) into Eqs. (39), (40), and (41)), replacing the
first order spatial derivatives with their respective finite difference expressions in Eqs. (8) and (12), and summing over a,
b, and c, the unevaluated first order derivatives can be eliminated through a linear combination of the electric
displacement continuity and Maxwell’s equation. After simplification, the final expression for the electric potential at
time t is achieved, as seen in Eqs. (42), (43), (44), and (45). The solution of electric potential at any point at time t is a
function of the material properties of the surrounding elements and the displacement and electric and magnetic potential
of the surrounding nodes at time t. Since the coupled equation for electric displacement at the point  ,  ,   at the
current time step is dependent on the electric potential of the nodes surrounding the point  ,  ,   at the current time
step, the solution of the boundary value problem requires a linear algebra technique for the solution of a set of dependent
equations. For the current paper, LU decomposition was utilized to solve for the electric potential for the reasons of
computational accuracy and efficiency.

 q  r  s  0 (42)
a,b,c 1
where
 q qy q 
q  2 x   z
 x 2 y 2 z 2 
 

qx  e1s j1 uj  a,  ,  uj ,  ,  11
s
 
   a,  ,    ,  ,  a11
s
 
  a,  ,   ,  ,  (43)
q y  e2s j 2 u  ,   b,
j

 u        
j
, , s
22
, b,
  
 ,  ,
 a22
s  ,  b,
  ,  ,

qz  e3s j 3 u  
j
, , c
 u         
j
, , s
33
, , c
       a    
, , s
33
, , c
   ,  , ,
and
 rxy rxz ryx ryz rzx rzy 
r  2       
 abxy acxz abxy bcy z acxz bcy z 
 
rxy  e1s j 2 uj ,  b,  uj ,  ,  12
s

  ,  b,    ,  ,  a12
s
 
  ,  b,   ,  , 
rxz  e1s j 3  u     u     
j
, , c
j
, , s
13            a
, , c , , s
13       
, , c , ,

ryx  e2s j1  u     u     
j
 a, ,
j
, , s
21            a
 a, , , , s
21       
 a, , , , (44)

ryz  e2s j 3 u    u     

j
, , c
j
, , s
23            a
, , c , , s
23       
, , c , ,

rzx  e3s j1  u     u     
j
 a, ,
j
, , s
31            a
 a, , , , s
31       
 a, , , ,

rzy  e3s j 2  u     u     
j
,  b,
j
, , s
32            a
,  b, , , s
32        ,
,  b, , ,

and

Proc. of SPIE Vol. 9064 90641B-7

 s xy s xz s yz
s  
abxy acxz bcy z

 
s xy  e1s j 2  e2s j1 uj  a,  b,  uj  a,  ,  uj ,  b,  uj ,  , 

 12
s
  21
s

   a,  b,     a,  ,    ,  b,    , , 
  a12
s
 a21 s
  a, b,  a, ,  , b,   , , 
s xz   e1s j 3  e3s j1   uj  a,  ,  c  uj  a,  ,  uj ,  ,  c  uj ,  , 
(45)
 13 s
  31s
  a, , c   a, ,   , , c   , , 
  a13
s
 a31 s
  a, , c  a, ,  , , c   , , 
s yz   e2s j 3  e3s j 2   uj ,  b,  c  uj ,  b,  uj ,  ,  c  uj ,  , 

  23s
  32s
  , b, c   , b,   , , c   , ,  ,
  a23
s
 a32s
  , b, c  , b,  , , c   , , 
where superscript “s” denotes the point   a ,   b ,   c  .

2.6 Enforcement of Maxwell’s Equation (Gauss’s Magnetic Field Law) and Continuity of Magnetic Flux Density
A similar approach is followed to achieve an expression for the magnetic potential at time t. First, Gauss’s law for
magnetism is enforced at every point   a ,   b ,   c  as
  a ,   b ,  c   a ,   b ,  c
qijk u j ,ki  aij  a ,  b , c ,ji a ,  b , c  ij  a , b , c ,ji a , b , c  0 (46)
for a,b,c= ±1.
Next the continuity of the magnetic flux density is enforced at infinitesimal distances from the interface, which will
result in the following equations
B1  , b , c  B1  , b , c (47)
B2 a ,  , c  B2 a ,  , c (48)
  a , b ,    a , b , 
B3  B3 (49)
for a,b,c= ±1.

2.7 Final Expressions for Nodal Magnetic Potential

After substituting the expressions for magnetic flux density (Eq. (9)) into Eqs. (47), (48), and (49), replacing the
first order spatial derivatives with their respective finite difference expressions in Eqs. (9) and (14), and summing over a,
b, and c, the unevaluated first order derivatives can be eliminated through a linear combination of the magnetic flux
density continuity and equation for Gauss’s magnetism law. After simplification, the final expression for the magnetic
potential at time t is achieved, as seen in Eqs. (50), (51), (52), and (53). The solution of magnetic potential at any point at
time t is a function of the material properties of the surrounding elements and the displacement and electric and magnetic
potential of the surrounding nodes at time t. Similar to the solution of the nodal values of electric potential, LU
decomposition was utilized to solve for the magnetic potential.

 l  m  n   0 (50)
a,b,c 1
where

Proc. of SPIE Vol. 9064 90641B-8

  l ly l 
l  2 x   z 
 x 2 y 2 z 2 
 

l x  q1s j1 uj  a,  ,  uj ,  ,  a11
s
 
   a,  ,    ,  ,  11
s

   a,  ,   ,  ,   (51)
l y  q2s j 2 u  ,   b,
j

 u     a    
j
, , s
22
, b,
  
 ,  ,
 22
s  ,  b,
  ,  ,

l z  q3s j 3  u  
j
, , c
 u     a    
j
, , s
33
, , c
           
, , s
33
, , c
   ,  , ,
and
 mxy mxz m yx m yz mzx mzy 
m  2       
 abxy acxz abxy bcy z acxz bcy z 
 
mxy  q1s j 2 uj ,  b,  uj ,  ,  a12
s

  ,  b,    ,  ,  12
s

 ,  b,   ,  ,  
mxz  q1s j 3 u    u     a
j
, , c
j
, , s
13            
, , c , , s
13       
, , c , ,

m yx  q2s j1 u    u     a
j
 a, ,
j
, , s
21            
 a, , , , s
21       
 a, , , , (52)

m yz  q2s j 3 u    u     a
j
, , c
j
, , s
23            
, , c , , s
23       
, , c , ,

mzx  q3s j1 u    u     a

j
 a, ,
j
, , s
31            
 a, , , , s
31       
 a, , , ,

mzy  q3s j 2 u    u     a

j
,  b,
j
, , s
32            
,  b, , , s
32        ,
,  b, , ,

and
nxy nxz n yz
n  
abxy acxz bcy z

 
nxy  q1s j 2  q2s j1 uj  a,  b,  uj  a,  ,  uj ,  b,  uj ,  , 
  a12
s

 a21
s
   a,  b,     a,  ,    ,  b,   , , 
  12s
 21 s
  a, b,  a, ,  , b,   , , 
nxz   q1s j 3  q3s j1   uj  a,  ,  c  uj  a,  ,  uj ,  ,  c  uj ,  , 
(53)
  a13
s
 a31 s
  a, , c   a, ,   , , c   , , 
  13s
 31 s
  a, , c  a, ,  , , c   , , 
n yz   q2s j 3  q3s j 2   uj ,  b,  c  uj ,  b,  uj ,  ,  c  uj ,  , 

  a23
s
 a32 s
  , b, c   , b,   , , c   , ,  ,
  23s
 32 s
  , b, c  , b,  , , c   , , 
where superscript “s” denotes the point   a ,   b ,   c  .

Proc. of SPIE Vol. 9064 90641B-9

 3. SIMULATION AND EXPERIMENTAL RESULTS AND DISCUSSION
3.1 Experimental Setup and Physical Model Development
A 250 mm x 250 mm x 2 mm laminated composite plate with a [90/0/90/0] s layup was manufactured for
experimental validation of the developed GW model. Symmetrically collocated lead zirconate titanate (PZT) actuators
with a diameter of 6.35 mm were bonded to the plate 150 mm from a single PZT sensor along the plate center line.
Switching between the poling directions of the symmetrically collocated PZTs allowed for Lamb wave mode
suppression, facilitating interpretation of the sensor signal and time of flight (ToF) calculations. The elastic and
piezoelectric properties of the PZT material (APC 850) are presented in Table 1. The composite plate was comprised of
unidirectional IM-7 carbon fibers embedded in a matrix of EPON 863 epoxy. The homogenized elastic properties of the
unidirectional laminae were computed using the MS-GMC micromechanics approach41 and are presented in Table 2. A 5
cycle cosine tone burst signal was used to excite the PZT actuators with a maximum electric potential of 10 V. The
frequency-thickness products between fb/2=175 kHz-mm and 875 kHz-mm were investigated due to the common
utilization of this range for damage detection.

Table 1. PZT (APC 850) properties

Elastic Properties
Elastic Moduli Shear Moduli
Poisson’s Ratio
(GPa) (GPa)
E1 63.0 n12 0.301 G12 23.5
E2 63.0 n13 0.532 G13 23.0
E3 54.0 n23 0.532 G23 23.0
Density (kg m-3) 7500
Piezoelectric Properties (C m-2)
e1 11 0 e2 11 0 e3 11 2.18
e1 22 0 e2 22 0 e3 22 2.18
e1 33 0 e2 33 0 e3 33 23.59
e1 12 0 e2 12 0 e3 12 0
e1 13 27.14 e2 13 0 e3 13 0
e1 23 0 e2 23 27.14 e3 23 0
Dielectric Properties (C V-1 m-1)
11 1.51e-8 22 1.51e-8 33 1.30e-8

Table 2. Composite plate properties

Elastic Properties
Stiffness Matrix Components (GPa)
C11 286.0 C22 286.0 C33 269.5
C13 170.5 C23 170.5 C12 173.0
C44 45.3 C55 45.3 C66 56.5
Density (kg m-3) 1545

The physical model for the experimental validation simulations was developed to resemble the experimental plate.
The piezoelectric actuation and subsequent wave propagation was modeled using the presented derivation. A two-
dimensional Gaussian window was applied to the actuation voltage to accurately simulate the circular PZTs used in the
experiments. In addition to simulating Lamb wave propagation in a pristine composite plate, models were developed to
quantify the effects of various sizes and types of delamination and damage embedded within the composite laminate.
Utilizing the novel electro-magneto-mechanical coupled LISA/SIM framework, piezomagnetic elements were also
incorporated in the model. The material properties (elastic and magnetic) for the piezomagnetic material (CoFe2O4) are
presented in Table 3. Numerical stability was ensured and pulse and amplitude distortion mitigated through satisfaction
of the Courant Friedrich Lewy (CFL) number, Eq. (54).

1 1 1
CFL  cmax t   1, (54)
x 2
y 2
z 2

Proc. of SPIE Vol. 9064 90641B-10

where cmax is the maximum wave speed (i.e., longitudinal wave speed), ∆t is the time step (i.e., sampling period), and ∆x,
∆y, ∆z are the grid spacings for the cuboid elements.42 In addition, at least eight elements per minimum wavelength were
utilized in order to prevent amplitude distortion.43 The grid spacings (i.e., ∆x, ∆y, and ∆z) and time step (i.e., ∆t) for the
studies presented in this paper were chosen to ensure convergence while minimizing numerical error and computational
costs. The grid spacings in the plane of the plate were held at 1 mm while the through-thickness grid spacing was
determined by laminae thickness. Furthermore, the time step was adjusted to satisfy the CFL criterion.

Table 3. Piezomagnetic material (CoFe2O4) properties

Elastic Properties (GPa)
C11 286.0 C22 286.0 C33 269.5
C13 170.5 C23 170.5 C12 173.0
C44 45.3 C55 45.3 C66 56.5
Density (kg m-3) 5300
Piezomagnetic Properties (N A-1 m-1)
q1 11 0 q2 11 0 q3 11 580.3
q1 22 0 q2 22 0 q3 22 580.3
q1 33 0 q2 33 0 q3 33 699.7
q1 12 0 q2 12 0 q3 12 0
q1 13 550.0 q2 13 0 q3 13 0
q1 23 0 q2 23 550.0 q3 23 0
Magnetic Permeability Properties (N s2 C-2)
11 -590e-6 22 -590e-6 33 157e-6

3.2 Experimental Validation

The composite numerical model was experimentally validated using the setup described in Section 3.1. The
utilization of collocated actuators for the selective excitation of Lamb wave modes facilitated the comparison between
the simulated and experimental results for a laminated composite plate over a range of frequencies commonly utilized
for SHM. It should be noted that because of the difficulties associated with perfectly aligning the actuators, inherent
differences in the piezoelectric properties of the PZTs, and through-thickness variability in the composite plate, complete
success was not achieved in maximizing the suppression of Lamb wave modes. However, sufficient suppression was
achieved in order to extract the wave speeds for the three fundamental GW modes (S0, SH0, and A0) propagating in the
composite plate at the range of frequency-thickness products of interest. The experimental and simulation results are
presented in Figure 2 for the frequency-thickness range of fb/2=175 kHz-mm to 875 kHz-mm. It can be observed that
the three fundamental GW modes are accurately represented by the numerical model. The maximum error in the
simulation results was approximately 13% while the model tends to under-predict the group velocity for all fb/2 values.
This discrepancy is likely caused by mesh convergence however the current results are deemed sufficient for further
studies since the overall dispersive trend of each of the GW modes is accurately captured.

8000
Simulation
So t Experiment
a
6000

E 4000
U °' SHo = a
i I
2000
Ao , I I I

00 100 200 300 400 500 600 700 800 900

fb/2 (kHz-mm)
Figure 2. Simulated and experimental dispersion curve comparison for range of frequency-thickness products commonly used for
damage detection in composites

Proc. of SPIE Vol. 9064 90641B-11

3.3 Computational Efficiency
The LISA/SIM solution methodology was originally formulated to run on connection machines (i.e., massively
parallel computers) in a parallel processing environment where each node in the model is provided its own processor.
Even on traditional parallel computer architectures, the computational efficiency of this model offers key advantages
over other wave propagation models. In Ref. 25 the improved computational efficiency of the LISA/SIM framework
with piezoelectric coupling, coded in Fortran 90 and utilizing OpenMP parallelization, over the commercial finite
element software44 was demonstrated. A 247 mm x 247 mm x 4 mm aluminum plate with one actuator and one sensor
was simulated and both models were run in a parallel computing environment on eight Harpertown 2.66 GHz, 8
MB/Cache, 16 GB memory processors. Each model was run in double precision for 1000 iterations with a time step of
9.5e-8 s. In the original demonstration, the computational efficiency of the electromagnetic coupled LISA/SIM based
model was shown to be over 170 times faster than the finite element based model. After improvement of the code for
efficiency and optimization of the parallelization routine, further increases in speed were achieved. The final
computation results are shown in Table 4. With the improvements in computational efficiency, the optimized model had
a wallclock time of 108 s, which is more than 350 times faster than the finite element model presented previously.

Table 4. Computational efficiency comparison between FEM and

current model
Solver Method # of elements Wallclock time (s)
FEM 244,038 40,855

Current Model 567,009 108

3.4 Lamb Wave Propagation in a Laminated Composite Plate

In Ref. 25, the authors illustrated the distinct displacement signatures of the two fundamental Lamb wave modes
(symmetric and antisymmetric) using displacement contours and vector plots of the GW propagating through an
isotropic aluminum plate. The contour and vector plots confirmed that the developed model was capable of accurately
simulating the symmetric profile of the out-of-plane displacement and antisymmetric in-plane displacement profile for
the case of the symmetric Lamb wave mode and the inverse for the antisymmetric fundamental Lamb wave mode. The
technique of mode suppression through selective poling of symmetrically collocated PZT actuators is demonstrated in
Figure 3. In order to verify the model’s ability to capture this phenomenon in an orthotropic composite plate, the GW
propagation in a laminated composite with a [90/0/90/0]s layup and excited with symmetrically collocated piezoelectric
actuators was simulated. The contours of the in-plane displacement (u2) and out-of-plane displacement (u3) and the
through-thickness vector plot are presented in Figure 4 and Figure 5 for the A0 and S0 selective excitation respectively.
Comparison of the contour results obtain from the composite plate with those presented in Ref. 25 reveal the greater
complexity of wave propagation in the composite plate. The increased complexity in wave behavior is due to the
directional dependence of wave speed, reflections between and within laminae, and presence of additional GW modes
such as the shear horizontal mode (SH0).

4
EMI / \ /
Seno

Symmetric modes
u / \ _/
/ //
So So
\ /
\
Actuato

Actuato Seesur
u3
Antisymmetric modes
u,
u3

Actualoc Actuato

Figure 3. Selective Lamb wave mode suppression illustration via relative voltage poling of symmetrically collocated actuators

Proc. of SPIE Vol. 9064 90641B-12

 Propagation direction
1 pm

0 d
E
6

a) Through-thickness in-plane displacement (u2) contour of A0 Lamb wave mode

Propagation direction

b) Through-thickness out-of-plane displacement (u3) contour of A0 Lamb wave mode

Propagation direction

c) Through-thickness vector plot (u2,u3) of A0 Lamb wave mode

Figure 4. Through-thickness plots of (a) in-plane displacement, (b) out-of-plane displacement, and (c) vector field for A0 Lamb
wave mode at t=45.60 s for fb/2=525 kHz-mm

Propagation direction

a) Through-thickness in-plane displacement (u2) contour of S0 Lamb wave mode

Propagation direction
x10"

b) Through-thickness out-of-plane displacement (u3) contour of S0 Lamb wave mode

Propagation direction

c) Through-thickness vector plot (u2,u3) of S0 Lamb wave mode

Figure 5. Through-thickness plots of (a) in-plane displacement, (b) out-of-plane displacement, and (c) vector field for S0 Lamb
wave mode at t=45.60 s for fb/2=525 kHz-mm

Proc. of SPIE Vol. 9064 90641B-13

3.5 Damage Detection Capabilities of the Developed Model
A key advantage of an accurate and computationally efficient wave propagation model with electro-magneto-
mechanical coupling is the ability to investigate SHM damage detection and quantification techniques in a systematic
manner for the purpose of optimizing the techniques for various geometries, layups, damage type and size, and actuation
and sensing techniques. The SIM incorporated in the LISA framework allows sharp material property boundaries, such
as cracks or delaminations, to be simulated without the model incurring significant numerical instabilities or error. This
ability provides an advantage over other numerical modeling schemes and expands the damage quantification scenarios
that can be investigated. In the event of a low velocity impact such as a tool drop, barely visible damage can form below
the composite surface. SHM techniques utilizing GWs have been shown capable of detecting this type of damage.45-46 In
the case of a low velocity impact, matrix cracking is often present surrounding the delaminations.47 The microcracks
distributed throughout the polymeric matrix cause a localized reduction in stiffness that perturbs the GW signal in the
vicinity of the damage. In addition to the Lamb wave mode conversion occurring due to the change in plate thickness
above and below the delamination, the stiffness change surrounding the delamination in the midplane of the composite
will further alter the wave speed and amplitude.
To illustrate the developed model’s capability in detecting and quantifying damage, delamination and damage
(matrix cracking) were embedded in the midplane of the composite plate modeled previously. The delamination was
modeled as a separation of adjacent plies through the application of the material properties of air to the cells in the area
of the delamination. For the present study, six delamination diameters ranging from 3 to 13 mm were embedded within
the composite laminate. For the consideration of matrix microcracking surrounding the delamination, the homogenized
composite laminae properties were computed using reduced matrix elastic properties. The microcracking was assumed to
extend a distance equal to the delamination radius while its severity linearly decreases as a function of the distance from
the delamination center. A contour illustrating the gradient in material stiffness progressing from total reduction in
stiffness (i.e., delamination in blue) to pristine (i.e., undamaged composite plate in dark red) is shown in Figure 6.
Relative 561Poess

Figure 6. Illustration of damage (matrix cracking) surrounding delamination

The sensor signals for three of the simulated delamination and delamination/damage cases each are presented in
Figure 7. It is apparent that the presence of delamination and matrix damage alters the sensor signal; however further
analysis of the signal is necessary to quantify the effect of delamination and damage size. The relative phase shift and
amplitude change of the three fundamental GWs (S0, SH0, and S0) were computed for each of the delamination and
damage sizes and compared to the results from a pristine specimen. The results from this analysis are presented in Figure
8 where the relative ToF is plotted with respect to delamination diameter in Figure 8(a) and Figure 8(c) and the relative
amplitude change with respect to delamination diameter is plotted in Figure 8(b) and Figure 8(c) for the case of
delamination only and delamination and damage respectively. From the plots of relative ToF vs. delamination size, the
A0 mode is most sensitive to small diameter delamination but the S 0 mode shows the greatest relative change for larger
delaminations. Regarding the effect on the maximum amplitude of each fundamental mode, the A 0 mode demonstrates
the greatest sensitivity for small scale damage and relative amplitude change for large diameter damage. These results
are in agreement with experiments indicating that because of its shorter wavelength, the A0 mode is more capable of
detecting small scale damage such as cracks or delaminations.48

Proc. of SPIE Vol. 9064 90641B-14

 0.04 0.04
-0 mm delam. -0 mm delam.
0.03- 7 mm delam. 0.03- 7 mm delam. -

13 mm delam. 13 mm delam.
0.02 - - 0.02-
>
m 0.01 á, 0.01
ca

> 0

v°> -0.01 -
0
>
°C -0.01 -
0
a>

( I ) -0.02 - u) -0.02

-0.03 - -0.03

-0.04 -0.04
40 60 80 100 120 40 60 80 100 120
Time (µs) Time Gis)
a)
Sensor voltage for three delamination sizes b) Sensor voltage for three delamination sizes
(delamination only) (delamination and matrix damage)
Figure 7. Sensor voltage time signature demonstrating the phase shift and amplitude change as a result of delamination and damage

Comparing the relative phase shift plots for the delamination only and delamination/damage cases, it can be
observed that the sensitivity to small scale damage increases for the SH 0 and S0 modes while the maximum phase shift
for the large diameter delaminations increases due to the presence of damage for all three fundamental modes. A similar
comparison for the relative amplitude indicates a different trend. Instead of the maximum relative amplitude change
increasing due to the damage, the maximum relative amplitude of the A 0 mode decreases. It is hypothesized that this
reduction in maximum relative amplitude is caused by wave dampening due to the reduced stiffness surrounding the
delamination. Further investigations are necessary to fully test this hypothesis.
04 20
-a-So So
.... SHo .... SHo
^0.3 Ao Ao
LL

;o
0.2-
fr

fX 0.1
.'' ` ,

a g
2 4 6 8 10 12 -50 2 4 6 8 10 12
Delamination Size (mm) Delamination Size (mm)
a) Relative ToF vs. delamination size (delamination only) b) Relative amplitude vs. delamination size (delamination only)
08 15
tSo ,So
.... SHo .... SHo
0.6- A0
0 Ao
u_
o
~m O.
>
c
m
0.2 -

2 4 6 8 10 12 2 4 6 8 10 12
Delamination Size (mm) Delamination Size (mm)
c) Relative ToF vs. delamination size d) Relative amplitude vs. delamination size
(delamination and matrix damage) (delamination and matrix damage)
Figure 8. Relative ToF and peak amplitude for three fundamental GW modes (A0, SH0, S0) in composite plate with fb/2=525 kHz-mm
for varying delamination and damage (matrix cracking) sizes

Proc. of SPIE Vol. 9064 90641B-15

 Utilizing the newly formulated electro-magneto-mechanical coupled LISA/SIM, a composite model was simulated
with an embedded piezomagnetic sensor for the same delamination and delamination/damage cases previously
investigated. Since piezomagnetic sensors provide the capability of noncontact sensing and actuation, composite
structures can be manufactured with these materials embedded below the surface for the purpose of GW excitation and
sensing without the need for wires to supply/measure the energy to/from the transducer. Since the A0 mode was shown
to be more sensitive to small scale damage in Figure 8, the relative ToF and amplitude of the antisymmetric Lamb wave
mode are presented for the delamination and delamination/damage cases in Figure 9. The overall trend of ToF vs.
delamination size differs from that witnessed in Figure 8(a) and Figure 8(c). In addition, the sensitivity to small scale
damage and small changes in delamination diameter appears to improve. In regards to the delamination effect on relative
amplitude of the A0 mode, the embedded sensor results indicate an increase in maximum relative amplitude for large
delamination. However, the trend of a decrease in maximum relative amplitude for the case of delamination and damage
was observed similar to that seen in the case of a surface-bonded PZT sensor.

04 30
De lam. Only Delam. Only
Delam. & Damage \ 25 Delam. & Damage
0.3
-a 20
'Q 15
E
< 10
m
>
5
m
Ct 0.
-0.10
2 4 6 8 10 12 -50 2 4 6 8 10 12
Delamination Size (mm) Delamination Size (mm)
a) Relative ToF vs. delamination size c) Relative amplitude vs. delamination size
(delamination only) (delamination and matrix damage)
Figure 9. Relative ToF and peak amplitude for A0 Lamb wave mode in composite plate with fb/2=525 kHz-mm for varying
delamination and damage (matrix cracking) sizes

4. CONCLUSION
A fully coupled electro-magneto-mechanical elastodynamic model for wave propagation in a heterogeneous,
anisotropic material system was developed to simulate the GW propagation in a composite plate for piezoelectric and
piezomagnetic actuation and sensing. The model was shown to accurately predict the experimentally determined group
velocity of the three fundamental GWs (S0, SH0, and A0) over a range of frequency-thickness products commonly used
for damage detection in composites. Once the wave propagation model was experimentally validated, the characteristic
displacement profile of Lamb waves was demonstrated in a composite plate cross section. Comparison with similar
displacement profiles for an isotropic aluminum plate reveals the greater complexity in the wave propagation in
composites due in part to wave speed angular dependence and reflections between individual laminae. The improved
efficiency of the developed model over commercial finite element packages was presented and damage (delamination
and matrix cracking) was introduced to study its effects on the three fundamental GW modes. The relative ToF and
mode amplitude were compared for the three fundamental modes over a range of delamination diameters. It was
concluded that the A0 mode provides the best sensitivity to small scale damage and demonstrates the greatest change in
amplitude over the range of delamination diameters. Finally, embedded piezomagnetic sensors were modeled for the
purpose of studying whether improved damage detection capabilities could be achieved with this type of sensing. The
results indicate that improvement in the sensitivity and magnitude of response can be achieved depending on the type
and size of the damage. The developed model has been proven accurate and efficient for the simulation of GW
propagation in laminated composite specimens and used to study the effect of damage on the wave propagation
behavior. With its improved actuation modeling capabilities, the developed model can serve as a valuable tool in a
virtual sensing framework to assess damage quantification and localization techniques.

Proc. of SPIE Vol. 9064 90641B-16

 5. ACKNOWLEDGEMENTS

This work is supported in part by the National Science Foundation Graduate Research Fellowship under Grant No.
(2011124478) and Army Research Office under Grant No. (60766-EG); Program Manager Dr. Larry Russell.

REFERENCES
[1] Raghavan, A. and Cesnik, C. E. S., “Review of Guided-wave Structural Health Monitoring,” Shock and Vibration
Digest 39(2), 91-116 (2007).
[2] Giurgiutiu, V., [Structural Health Monitoring with Piezoelectric Wafer Active Sensors], Academic Press, Boston
(2008).
[3] Andrews, J. P., Palazotto, A. N., DeSimio, M. P., and Olson S. E., “Lamb Wave Propagation in Varying Isothermal
Environments,” Structural Health Monitoring 7(3), 265-270 (2008).
[4] Alleyne, D. N. and Cawley, P., “The Interaction of Lamb Waves with Defects,” IEEE Transactions on Ultrasonics,
Ferroelectrics and Frequency Control 39(3), 381–397 (1992).
[5] Staszewski, W. J., Pierce, S. G., Worden, K., Philp, W. R., Tomlinson, G. R., and Culshaw, B., “Wavelet Signal
Processing for Enhanced Lamb-Wave Defect Detection in Composite Plates Using Optical Fiber Detection,” Optical
Engineering 36(7), 1877–88 (1997).
[6] Jha, R. and Watkins, R., "Lamb Wave Based Diagnostics of Composite Plates Using a Modified Time Reversal
Method," Proc. of 17th AIAA/ASME/AHS Adaptive Structures Conference (2009)
[7] Liu, Y., Mohanty, S., Chattopadhyay, A., “Condition Based Structural Health Monitoring and Prognosis of
Composite Structures under Uniaxial and Biaxial Loading,” Journal of Nondestructive Evaluation 29(3), 181-88
(2010).
[8] Liu, Y., Kim, S.B., Chattopadhyay, A., Doyle, D., “Application of System Identification Techniques to Health
Monitoring of On-Orbit Satellite Boom Structures,” Journal of Spacecraft and Rockets 48(4), 589-98 (2011).
[9] Soni, S., Das S., and Chattopadhyay, A., “Simulation of Damage Features in a Lug Joint using Guided Waves,”
Journal of Intelligent Material Systems and Structures 20(12), 1451–64 (2009).
[10] Chattopadhyay, A., Peralta, P., Papandreou-Suppappola, A., and Kovvali, N., “A Multidisciplinary Approach to
Structural Health Monitoring and Damage Prognosis of Aerospace Hotspots,” Journal of Royal Aeronautical
Society 113(1150), 799-810 (2009).
[11] Talbot, R., and Przemieniecki, J. S., “Finite element analysis of frequency spectra for elastic waveguides,”
International Journal of Solids and Structures 11(1), 115-138 (1975).
[12] Zienkiewicz, O. C., [The Finite Element Method 4th edn.], McGraw-Hill, New York (1989).
[13] Koshiba, M., Karakida, S. and Suzuki, M., “Finite Element Analysis of Lamb Waves Scattering in an Elastic Plate
Waveguide,” IEEE Transactions on Sonics and Ultrasonics 31(1), 18–25 (1984).
[14] Cheung, Y. K., [Finite Strip Method in Structural Analysis], Pergamon, Oxford (1976).
[15] Liu, G. and Achenbach, J. D., “Strip Element Method to Analyze Wave Scattering by Cracks in Anisotropic
Laminated Plates,” Journal of Applied Mechanics 62(3), 607–13 (1995).
[16] Liu, G., Xi, Z., Lam, K. Y., and Shang, H., “A Strip Element Method for Analysing Wave Scattering by a Crack in
an Immersed Composite Laminate,” Journal of Applied Mechanics 66(4), 898-903 (1999).
[17] Yamawaki, H. and Saito, T., “Numerical Calculation of Surface Waves using New Nodal Equation,” Nondestructive
Testing and Evaluation 8(9), 379–89 (1992).
[18] Cho, Y. and Rose, J. L., “A Boundary Element Solution for Mode Conversion Study of the Edge Reflection of
Lamb Waves,” Journal of the Acoustical Society of America 99(4), 2097–2109 (1996).
[19] Fornberg, B., [A Practical Guide to Pseudospectral Methods], Cambridge University Press, Cambridge, UK (1998).
[20] Krawczuk, M. and Ostachowicz, W., “Spectral Finite Element and Genetic Algorithm for Crack Detection in
Cantilever Rod,” Proc. 4th International Conference on Damage Assessment of Structures (2001).
[21] Hu, N., Fukunaga, H., Kameyama, M., Roy Mahapatra, D., and Gopalakrishnan S., “Analysis of Wave Propagation
in Beams with Transverse and Lateral Cracks using a Weakly Rormulated Spectral Method,” Journal of Applied
Mechanics (ASME) 74(1), 119–127 (2007).
[22] Delsanto, P. P., Whitcombe, T., Chaskelis, H. H., and Mignogna, R. B., “Connection Machine Simulation of
Ultrasonics Wave Propagation in Materials I: One-Dimensional Case,” Wave Motion 16(1), 65–80 (1992).
[23] Delsanto, P. P., Schechter, R. S., Chaskelis, H. H., Mignogna, R. B., and Kline, R., “Connection Machine
Simulation of Ultrasonics Wave Propagation in Materials II: Two-Dimensional Case,” Wave Motion 20(4), 295–
314 (1994).

Proc. of SPIE Vol. 9064 90641B-17

[24] Delsanto, P. P., Schechter, R. S., and Mignogna, R. B., “Connection Machine Simulation of Ultrasonics Wave
Propagation in Materials III: Three-Dimensional Case,” Wave Motion 26(4), 329–39 (1997).
[25] Borkowski, L., Liu, K., and Chattopadhyay, A., “Fully Coupled Electromechanical Elastodynamic Model for
Guided Wave Propagation Analysis,” Journal of Intelligent Material Systems and Structures, 24(13), 1647-1663
(2013).
[26] Ludwig, R., You, Z., and Palanisamy, R., “Numerical Simulations of an Electromagnetic Acoustic Transducer-
Receiver System for NDT Applications,” Magnetics, IEEE Transactions on 29(3), 2081-2089 (1993).
[27] Achenbach, J. D., “Quantitative Nondestructive Evaluation,” International Journal of Solids and Structures 37, 13–
27 (2000).
[28] Wilcox, P., Lowe, M. J. S., and Cawley, P., “An EMAT Array for the Rapid Inspection of Large Structures using
Guided Waves,” Proc. AIP Conference A, 814-821 (2003).
[29] Jian, X., Dixon, S., Grattan, K. T. V., and Edwards, R. S., “A Model for Pulsed Rayleigh Wave and Optimal EMAT
Design,” Sensors and Actuators A: Physical 128(2), 296-304 (2006).
[30] Mariani, S., Nguyen, T., Phillips, R. R., Kijanka, P., di Scalea, F. L., Staszewski, W. J., Fateh, M., and Carr, G.,
“Noncontact Ultrasonic Guided Wave Inspection of Rails,” Structural Health Monitoring, 12(5-6), 539-548 (2013).
[31] Li, J. Y., and Dunn, M. L., “Micromechanics of Magnetoelectroelastic Composite Materials: Average Fields and
Effective Behavior,” Journal of Intelligent Material Systems and Structures, 9(6), 404-416 (1998).
[32] Aboudi, J., “Micromechanical Analysis of Fully Coupled Electro-magneto-thermo-elastic Multiphase Composites,”
Smart Materials and Structures 10(5), 867 (2001).
[33] Lee, J., Boyd IV, J. G., and Lagoudas, D. C., “Effective Properties of Three-phase Electro-magneto-elastic
Composites,” International Journal of Engineering Science, 43(10), 790-825 (2005).
[34] Zhou, Z. G., Wu, L. Z., and Wang, B., “The Behavior of a Crack in Functionally Graded
Piezoelectric/Piezomagnetic Materials under Anti-plane Shear Loading,” Archive of Applied Mechanics, 74(8),
526-535 (2005).
[35] Chen, J., Pan, E., and Chen, H., “Wave Propagation in Magneto-electro-elastic Multilayered Plates,” International
Journal of Solids and Structures, 44(3), 1073-1085 (2007).
[36] Pang, Y., Liu, J., Wang, Y., and Fang, D. “Wave Propagation in Piezoelectric/Piezomagnetic Layered Periodic
Composites,” Acta Mechanica Solida Sinica 21(6), 483-490 (2008).
[37] Wu, T. L. and Huang, J. H., “Closed-form Solutions for the Magnetoelectric Coupling Coefficients in Fibrous
Composites with Piezoelectric and Piezomagnetic Phases,” International Journal of Solids and Structures 37, 2981–
3009 (2000).
[38] ANSI/IEEE. Standard 176, Piezoelectricity. IEEE, New York (1987).
[39] IEEE Standard 319-1990, IEEE Standard on Magnetostrictive Materials: Piezomagnetic Nomenclature (1990).
[40] Sundararaman, S., “Numerical and Experimental Investigations of Practical Issues in the use of Wave Propagation
for Damage Identification,” Ph.D. Dissertation, Department of Mechanical Engineering, Purdue University (2007).
[41] Liu, K. C., Chattopadhyay, A., Bednarcyk, B., and Arnold, S. M., “Efficient Multiscale Modeling Framework for
Triaxially Braided Composites using Generalized Method of Cells,” Journal of Aerospace Engineering 24(2), 162-
169 (2011).
[42] Virieux, J., “P-SV Wave Propagation in Heterogeneous Media: Velocity Stress Finite Difference Method,”
Geophysics 51(4), 889-901 (1986).
[43] Balasubramanyam, R., Quinney, D., Challis, R. E., and Todd, C. P. D., “A Finite-Difference Simulation of
Ultrasonic Lamb Waves in Metal Sheets with Experimental Verification,” Journal of Physics D: Applied Physics
29(1), 147–155 (1996).
[44] Abaqus Version 6.7, Abaqus/CAE and Abaqus/Explicit. Simulia World Headquarters, Providence (2007).
[45] Guo, N. and Cawley, P., “The Interaction of Lamb Waves with Delaminations in Composite Laminates,” Journal of
the Acoustical Society of America 94(4), 2240–46 (1993).
[46] Liu, Y., Fard, M. Y., Chattopadhyay, A., and Doyle, D., “Damage Assessment of CFRP Composites using a Time–
Frequency Approach,” Journal of Intelligent Material Systems and Structures 23(4), 397-413 (2012).
[47] Liu, Y. and Chattopadhyay, A., “Low Velocity Impact Damage Monitoring of a Sandwich Composite Wing,”
Journal of Intelligent Material Systems and Structures 24(17), 2074-2083 (2013).
[48] Su, Z., Ye, L., and Lu, Y., “Guided Lamb Waves for Identification of Damage in Composite Structures: A
Review,” Journal of Sound and Vibration 295(3), 753-780 (2006).

Proc. of SPIE Vol. 9064 90641B-18