6  Boundary Element Method Solution through Frequency Interpolation

The concept of frequency interpolation in the BEM was first introduced with a view to speeding up the numerical solution of boundary value problems [32-33]. The use of frequency interpolation as a means to solving the eigenvalue problem was first considered in Kirkup and Amini []. A similar technique for the modal analysis of coupled fluid-structure problems is considered in Kirkup and Amini [34].
The method employed for solving (23) requires that in an interval [kA, kB] of values of k the matrix Ak is approximated by a matrix polynomial in k
Ak » A[0] + k A[1] + ... + km A[N]   for k real.
(39)
The non-linear eigenvalue problem (23) can be replaced with the following eigenvalue problem
[A[0] + k A[1] + ... + km A[N]]
~
j
 
= 0.
(40)
The solutions of (40) are the same as those of the following generalised linear eigenvalue problem

é
ê
ê
ê
ê
ê
ê
ë
A[0]
A[1]
A[2]
.
.
A[N-2]
A[N-1]
0
I
0
.
.
0
0
:
:
:
:
:
:
:
0
0
0
.
.
I
0
0
0
0
.
.
0
I
ù
ú
ú
ú
ú
ú
ú
û
é
ê
ê
ê
ê
ê
ê
ê
ê
ê
ë
~
j
 
~
k
 
~
j
 
:
~
k
 
m-2
 
~
j
 
~
k
 
m-1
 
~
j
 
ù
ú
ú
ú
ú
ú
ú
ú
ú
ú
û
(41)

               =   
~
k
 
  é
ê
ê
ê
ê
ê
ê
ë
0
0
0
.
.
0
-A[N]
I
0
0
.
.
0
0
:
:
:
:
:
:
:
0
0
0
.
.
0
0
0
0
0
.
.
I
0
ù
ú
ú
ú
ú
ú
ú
û
é
ê
ê
ê
ê
ê
ê
ê
ê
ê
ë
~
j
 
~
k
 
~
j
 
:
~
k
 
N-2
 
~
j
 
~
k
 
N-1
 
~
j
 
ù
ú
ú
ú
ú
ú
ú
ú
ú
ú
û
  .
Equation (41) is amenable to solution by the QZ algorithm. Methods for solving problems of the form (40) are considered in Kublanovskaya [26,35-38]. On solution, the eigenvalues [k\tilde] are the approximations to the eigenfrequencies k* of the Helmholtz problem. The approximation [(j)\tilde](p) to the eigenfunctions j*(p)  (p Î D ÈS) can be recovered through the substitution of the approximations into (18),
~
j
 
(p) = - n
å
i=1 
~
j
 

i 
{M[k\tilde] zi }(p)    (p Î D)    and
(42)

~
j
 
(p) = - 1

c(p)
n
å
i=1 
~
j
 

i 
{M[k\tilde] zi }(p)    (p Î S) .
The full set of solutions will contain approximations to the true eigenvalues of the underlying Helmholtz problem. Spurious solutions are produced as a result of the collocation method and approximation (39), but they can be easily excluded.
In this paper we consider interpolants with N=2 only. In this case equation (41) takes the form
é
ê
ê
ë
A[0]
A[1]
0
I
ù
ú
ú
û
é
ê
ê
ê
ë
~
j
 
~
k
 
~
j
 
ù
ú
ú
ú
û
=
~
k
 
é
ê
ê
ë
0
-A[2]
I
0
ù
ú
ú
û
é
ê
ê
ê
ë
~
j
 
~
k
 
~
j
 
ù
ú
ú
ú
û
.
(43)
where Ak is computed at the three Chebyshev (¥ norm) interpolation points for any selected range [kA, kB]. The coefficient matrices A[0], A[1] and A[2] in (39) are obtained through Newton's divided differences. The generalised eigenvalue problem (41) is then solved through invoking NAG routine F02GJF [39]. Further detail on the development of the method is given in Kirkup and Amini [18].