Applied Mathematics, 2013, 4, 963-967

http://dx.doi.org/10.4236/am.2013.46132 Published Online June 2013 (http://www.scirp.org/journal/am)

Simulations of Rayleigh’s Wave on Curved Surface

Shih-Yu Shen1*, Chin-Yu Wang2
1
Institute of Applied Mathematics, National Cheng-Kung University, Taiwan
2
Department of Mechanical Engineering, Cheng Shiu University, Taiwan
Email: *shen@mail.ncku.edu.tw

Received March 15, 2013; revised April 11, 2013; accepted April 21, 2013

Copyright © 2013 Shih-Yu Shen, Chin-Yu Wang. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

ABSTRACT
Impulsive line load in a half-space (Lamb’s problem) can be solved with a closed form solution. This solution is helpful
for understanding the phenomenon of Rayleigh’s waves. In this article, we use a boundary element method to simulate
the solution of an elastic solid with a curved free surface under impact loading. This problem is considered difficult for
numerical methods. Lamb’s problem is calculated first to verify the method. Then the method is applied on the prob-
lems with different surface curvatures. The method simulates the phenomenon of Rayleigh’s wave propagating on a
curved surface very well. The results are shown in figures.

Keywords: Elastodynamics; Lamb’s Problem; Boundary Element Method; Rayleigh’s Wave

1. Introduction 2. The Problem

The phenomenon of surface wave is interesting and im- Because our goal is to simulate the Rayleigh wave on a
portant for many engineers and scientists. Impulsive line smooth surface near the loading point, the boundary of
load in a half-space can be solved by analytic methods the 2D domain  is modeled as a arc with curvature
[1]. Herewith scientists and engineers may get under- 1
standing with Rayleigh’s wave. When the free surface is . The loading is a single line impact at the surface.
R
not flat the phenomenon of surface wave should be simi- The schematic diagram is shown in Figure 1.
lar, but some detail can be different. On the case of curved Boundary value problems with zero initial conditions
surface, analytic solutions are no longer available. There- and absent body force for 2D elastodynamics in plane
fore numerical methods become an alternative way to
understand the phenomena. These problems are consid-
ered difficult for numerical methods. To the best of our
knowledge, there are no related reports on this problem.
Simulating transient wave in elastodynamics needs
mass computing time and huge memories. Boundary ele-
ment methods are efficient for simulating elastodynamics
[2,3]. Recently personal computers (PC) become much
more powerful and are equipped with more rams then
ever. Practical problems can be simulated with a PC pre-
cisely.
In this article, we use a boundary element method to
solve two dimensional elastodynamic problems with
curved surfaces. The curved boundary is assumed to be
an arc. The loading is an impulse. The numerical method
is implemented with fortran programs and a PC. In Sec-
tion 2, the mathematical problem is described. In the fol-
lowing section, we formulate the numerical method
briefly. The results are shown in Section 4. Figure 1. The schematic diagram of the geometry of the
*
Corresponding author. problem.

Copyright © 2013 SciRes. AM

964 S.-Y. SHEN, C.-Y. WANG

strain condition are considered. The material is homo- we have

geneous, isotropic and linear elastic. m 1 2 N
0
Therefore, the displacement u  x ,t  fulfills Navier’s   x aik  x Bik n   y aik  y Bik n     s, t   1  ,
(7)
k 0 i 1  
equation; i.e.
When a collocation method apply on Equation (7), the
 2u coefficients Bik  x j , tl  n are as many as m 2  2 N  . It
2
 2 u           u    , (1)
t 2 is difficult to calculate a precise elastodynamic solution
where  and  are the Lame constants and  is on a personal computer with this formulation. Therefore
the mass density of the elastic material. we take the advantage of the symmetry of the boundary.
The boundary conditions are Let
0 s s
  x  s  , t  n  s     s, t    , aik  cos i y aik  sin i x aik
n
(2) (8)
R R
1
s s
where  is the stress tensor, n  s  is the outward nor- s
aik   sin i y aik  cos i x aik (9)
mal vector on  , s is the coordinate on the boundary, R R
  s, t  is the Dirac delta function, and x  s  presents and
n  s s
the map from coordinate s to 2D spatial coordinates Bik  cos i y Bik  sin i x Bik (10)
 R R
 s 
 R cos    s  s s
 x  R  Bik   sin i y Bik  cos i x Bik (11)
x    . Here, x   . The time-varied dis- R R
 y   s 
 R sin   s s
Bik  cos i y Biku  sin i x Biku
n u
(12)
  R R R
placements on the curved surface are Rayleigh’s waves. s s
Bik   sin i y Biku  cos i x Biku
s u
(13)
R R
3. The Method
Substituting Equations (8) and (9) into (7), we have
We use a boundary element method to calculate the sur- m 1 2 N
0
face displacements. The boundary  is approximated   n aik  n Bik n   s aik  s Bik n     s, t   1  , (14)
as a polygon. k  0 i 1  
There are two families of particular solutions of Na- Then, we decomposite the traction into normal and
vier’s equation, x Biku  x, t  and y Biku  x, t  . tangent directions.
Then, the approximated displacements field are
  n aik  n Bik n  s    n  s   s aik  s Bik n  s    n  s 
m 2N
m 1 2 N
u  x , t      x aik x Biku  x, t   y aik y Biku  x,t  . (3) k  0 i 1
k  0 i 1
(15)
0
Using Hooke’s law, we have stress bases, x
Bik  x , t     s, t     n  s 
1
and y Bik  x, t  i.e.
    n aik  n Bik n  s    t  s   s aik  s Bik n  s    t  s 
m 2N
x
Bik  x , t      x Biku  x , t  I
k  0 i 1

 
(4) (16)
   x Biku  x, t    x Biku  x, t 
T
0
   s, t     t  s 
1
y

Bik  x , t      y Biku  x, t  I 
where n  s  and t  s  are normal and tangent vectors
 
(5)
   y Biku  x , t    y Biku  x, t  at s respectively, and  denote the inner product of two
T

u 
vectors.
Note that B are vector fields and Bik are tensor
ik Let si  is and tk  k t . Then apply the semicol-
fields. location method [4] to Equations (15) and (16). Rear-
The bases x Biku  x , t  , y Biku  x, t  , x Bik  x , t  and ranging the equations for n aik and s aik , we have the
Bik  x , t  have been derived in a close form [4]. The
y 
stepping equations.
the coefficients, x aik and y aik , will determined with  1 2 N
1
the boundary condition (2).    2  n a j   j    n aik biknn,j  s aik biktn,j (17)
Then, the approximated stress field is 2 k 0 i 0
m 1 2 N  1 2 N
c
   x , t      x aik x Bik  x, t   y aik y Bik  x ,t   (6)    2  s a j   j    n aik biknt,j  s aik biktt ,j (18)
k  0 i 1 2 k 0 i 0

Substituting Equation (6) into boundary condition (2), where

Copyright © 2013 SciRes. AM

 S.-Y. SHEN, C.-Y. WANG 965

1   s j  s j 1      s j  s j 1    
biknn, j   n Bik  x 
3  

   
   
 , t  n s j  Bik x  s j  , t n s j  Bik  x 
n  n  


   

 , t  n s j   n s j ,
  2     2  
1   s j  s j 1      s j  s j 1    
biknt, j   n Bik  x 
3  

   
   
 , t  n s j  Bik x  s j  , t n s j  Bik  x 
n  n 
   

 , t  n s j   t s j ,
*


  2     2  
(19)
1   s j  s j 1      s j  s j 1    
btn
ik , j    s Bik  x 
3  
 s 
   
 s 
  
 , t  n s j  Bik x  s j  , t n s j  Bik  x 



   

 , t  n s j   n s j ,
  2     2  
1   s j  s j 1      s j  s j 1    
biktt , j   s Bik
3 
 x 

   
  
 , t  n s j  Bik x  s j  , t n s j  Bik  x 
s   s 
   




 , t  n s j   t s j ,
  2     2  

and treatment with integral transform technique was given by

De Hoop, but the results were in complex function form.
 1
 if j  0 and   0 Nevertheless explicit form for surface displacements are
   s t
n
j
available. The analytic solution for surface displacement
0 else (20)
can be found in page. 614-626 of [1].
 j  0.
t
The schematic diagram of the geometry of the problem
is shown in Figure 1.
The   s, t  is approximated by In this example Poisson’s ratio   0.25 . Therefore
1  s   s 
H  s   H   H  t  tk  H  tk 1  t  where H the shear and Rayleigh’s wave speeds are 0.57735cd
s t  2   2  and 0.5308cd respectively, where cd is the dilatation
is the Heaviside step function. wave speed.
In formulae (17) and (18), the coefficients bik ,j take Figure 2 shows the displacements with vertical
the form of b i  j  k  ,00 . The usage of computer memory loading when R  1000 . The dashed line shows exact
is enormously deduced to 2Nm. displacement when R   . There is a Dirac delta func-
 
When k  l , Bi x  s j  , t n  s j   0 , if i  j . Thus tion at the Rayleigh wave front in the exact solution. The
vertical line indicate the arriving time of Rayleigh’s
Equations (17) and (18) are uncoupled. Thus, the time- 1
stepping scheme is explicit. wave. For this figure, s  . Even though this pro-
600
Then the time-stepping technique is applied on Equ-
ations (17) and (18) to solve n ai and s ai on  th blem is considered difficult for numerical methods, our
step. results are precise. This example shows the method is
After solving the coefficients, x aik and y aik , the applicable for impact problems.
numerical displacement may be calculated by When R  5 , the results are shown in Figure 3. In
this case tangential displacement is almost the same with
 
m 2N
u  x, t     aik x Biku  x, t   y aik y Biku  x ,t 
x Lamb’s problem, but the normal displacement is changed
(2.11)
k  0 i 1 much. The normal displacement has a strong and short
peak at the shear wave front and the head wave is
and the numerical interior stress by enlarged. The head wave is the wave beyond the shear
m 2N wave front [5]. The dashed line is the displacement for
   x , t  =    x aik x Bik  x , t   y aik y Bik  x , t  . (21) R   . Figure 4 shows the displacements for R  3 .
k  0 i 1
The phenomenon is similar to the case of R  5 .
4. The Results 5. Conclusion
In order to verify this method, we calculate the problem
We use the boundary element method and write a fortran
with very large R first. For R   , the problem be-
program running on a personal computer. In order to
comes Lamb’s problem in which the exact surface dis-
simulate the phenomenon of Rayleigh’s wave propagat-
placements are available. The elastic half space  y  0 
ing on a curved surface, the free surface is assumed to be
0
is loaded at t  0 by a unit line impulse at x    . a constant,
1
. The method is verified by the problem of
0 R
The solutions for problems of a line impulse load on R  1000 . Two examples, R = 5 and R = 3, calculated to
an elastic half plane were derived by Lamb. A modern demonstrate the phenomenon. On the shear wave front,

Copyright © 2013 SciRes. AM

966 S.-Y. SHEN, C.-Y. WANG

Figure 2. The displacements with vertical loading when R = 1000. The dashed line shows exact displacement when R = ∞. The
vertical red lines indicate the arrival time of Rayleigh’s wave. There is a Dirac delta function at the Rayleigh wave front in
1
the exact solution. In this figure, s  is used.
600

Figure 3. The displacements when R = 5. The dashed line shows exact displacement when R = ∞. The vertical red lines
1
indicate the arrival time of Rayleigh’s wave. In this figure, s  is used.
400

Copyright © 2013 SciRes. AM

 S.-Y. SHEN, C.-Y. WANG 967

Figure 4. The displacements when R = 3. The red line shows exact displacement when R = ∞. The vertical dashed lines
1
indicate the arrival time of Rayleigh’s wave. In this figure, s  is used.
400

the wave is very different from that of Lamb’s problem.
REFERENCES
[1] A. C. Eringen and E. S. Suhubi. “Elastodynamics, Vol-
ume II,” Academic Press, New York, 1975.
[2] D. E. Beskos, “Boundary Element Method in Dynamic
Analysis,” Applied Mechanics Review, Vol. 40, No. 1,
1987, pp. 1-23. doi:10.1115/1.3149529
[3] D. E. Beskos, “Boundary Element Method in Dynamic
Analysis: Part II (1986-1996),” Applied Mechanics Re-
view, Vol. 50, No. 3, 1997, pp. 149-197.
doi:10.1115/1.3101695
[4] S. Y. Shen, “An Indirect Elastodynamics Boundary Ele-
ment Method with Analytic Bases,” International Journal
for Numerical Methods in Engineering, Vol. 57, No. 6,
2003, pp. 767-794. doi:10.1002/nme.702
[5] J. D. Achenbach, “Wave Propagation in Elastic Solid,”
Noth-Holland, Amsterdam, 1976.

Copyright © 2013 SciRes. AM