Section 13.
2 389
Premrov, M.; Spacapan, I.
An Iterative FEM for Solving Wave Problems of a Halfspace
An iterative finite element method for solving wave problems of a halfspace is presented in this paper. The halfspace
is first truncated by introducing a fictive finite boundary on which some fictive boundary conditions must be imposed.
A finite computational domain is in each iteration subjected to actual boundary conditions on real boundary and to
fictive Dirichlet or Neumann boundary conditions on the fictive boundary. The radiation condition is satisfied by
using DtN operator. The DtN operator is not introduce in the finite element formulation on the fictive boundary so
any finite elements can be used. The method is simple and specially useful for computing higher harmonics.
1. Introduction
By solving wave problems of a halfspace the most important boundary condition is a radiation condition. This asserts
that at infinity all waves are outgoing and there are no reflecting waves. Radiation condition is satisfied automatically
as a part of the fundamental solution in the boundary element method. Unfortunately, the fundamental solution is
not always available. More flexible and useful is a finite element method. To obtain a finite computational domain
the infinite domain should first be truncated by introducing a fictive finite boundary (β). As a consequence spurious
reflection of waves from β is obtained. In order to diminish this reflection suitable boundary conditions on the fictive
boundary should be introduced.
2. Computational procedure of the method
In the presented method an iterative solution for solving wave equation in infinite domain is obtained. The infinite
domain is first truncated by introducing a fictive finite boundary (β). So the finite computational domain Ωf
bounded with actual boundary Γ and fictive boundary β is obtained. The radiation condition in the halfspace
outside of the fictive boundary can be represented with DtN (Dirichlet to Neumann) operator (SN ) in the form
∂u
= SN · u ; on β (1)
∂n
DtN operator can be in a non-local or in more simple local form and can be obtained from the references [1 − 2].
Figure 1: Figure 1: Iterative procedure of the method
The finite computational domain Ωf is in each iteration subjected to actual boundary conditions on Γ and to different
fictive Dirichlet (DBC) or Neumann (NBC) boundary conditions on β. The wave equation for the out-of-plane wave
motion is in each iteration first in the form
∇2 u + k 2 u = 0 ; u = u0 on Γ , u = DBC on β (2)
Tractions on β as a result of Dirichlet boundary conditions now must be computed. Analogously displacements due
to prescribed fictive Neumann boundary conditions can be obtained. The line y(i) which connects the two obtained
values is then projected to the DtN operator (SN ). The new Dirichlet and Neumann fictive boundary conditions on
390 PAMM, Proc. Appl. Math. Mech. 1 (2002)
β for the next iteration are so obtained - Figure 1. It is very important that DtN operator is not introduced in the
finite element formulation on the fictive boundary, so any finite elements can be used.
3. Numerical example
Consider a nonsymmetric halfspace problem with a hole of radius (a = 1), with the considered wave number (k)
and the maximal number of considered harmonics (N ). The fictive boundary is located on the distance (R) from
the origin of waving. The prescribed actual boundary conditions at the hole (r = a) are
u(a, θ) = cos(nϕ), n = 0, 1, .., N − 1 (3)
The exact solution is in the form
Hn(2)(kr)
u(r, θ) = (2)
· cos(nϕ), n = 0, 1, ..., N − 1 (4)
Hn (ka)
By solving the problem with the presented method and by using linear isoparametric finite elements and S2 DtN
operator a high accuracy in the results is easily achieved. Solutions for diplacements on the fictive boundary for the
numerical example with R = 2.0, k = 1.0 and by using first two harmonics are presented in Figure 2. The problem
was solved by using 48 finite elements in the tangential and 6 elements in the radial direction.
Figure 2: Figure 2: Solution for real and imaginary part of diplacements after the second iteration (N=2)
By using more harmonics the solution is normally not so accurate. The solutions for the same problem as before
but by using N = 4 are presented in Figure 3.
Figure 3: Figure 3: Solution for real and imaginary part of diplacements after the second iteration (N=4)
4. Conclusions
The presented method is based on an iterative variation of boundary conditions on a fictive boundary. The DtN
operator is not introduced in a finite element equation, thus we do not have problems with continious conditions
between finite elements. The method is so applicable also for using higher harmonics.
5. References
1 Givoli, D.; Keller, J.B.: Non-reflecting boundary conditions for elastic waves; Wave motion 12, (1990).
2 Pinsky, P.M.; Thompson, L.L.: Local high-order radiation boundary conditions for the two-dimensional time-dependent
structural acoustic problem; Journal for Acoustical Society of America 91 (3), (1992).
Ph.D. Miroslav Premrov, Ph.D. Igor Spacapan, Faculty of Civil Engineering, Smetanova 17, University of
Maribor, SI-2000 Maribor, Slovenia.