7  Concluding Discussion

It is suggested in Kress [13] and Amini [15] that the weighting parameter should be chosen as a function of k and S only. For Neumann problems, it has been shown in this paper that the value chosen for the weighting parameter cannot be independent of the number of elements if the resulting numerical method is to be consistent. Figures 4 and 5 show that for the particular test problem used in this work, choosing a value of parameter that balances the contribution from the component matrices gives rise to a well-conditioned composite matrix and generally smaller error.
It is clear from the results from the test problems and the analysis that the value assigned to the parameter has a major influence on the resulting error. However, the error in the test problems is not critically dependent on l. Thus, given the trade off between the computational cost of choosing a value for the parameter and its beneficial effect on the numerical solution, the most efficient route is likely to be that of making only a rough estimate of its optimal value.
For the exterior Helmholtz equation with a Robin boundary condition (1), an indirect BEM can be derived in the standard way from the following integral equation:
f(p) = a(p) æ
è
a{Lk s}S (p) + b{ Mk s} S (p) + 1

2
bs(p) ö
ø

+ b(p) æ
è
a{ Mkt s}S (p) + b{ Nk s}S (p) - 1

2
as(p) ö
ø
    for p Î S
which follows from the substitution of (2), (3) into (1). Derivation of the direct BEM for the Robin problem is more awkward. Nevertheless, both methods require solution over the Nk integral operator and hence the outcome of the analysis in this paper for the Neumann problem will also apply to the more general Robin problem.

Acknowledgement. The author is grateful to Dr D J Henwood of the Department of Mathematical Sciences, Brighton Polytechnic for his support in this work and to Dr G T Symm and Mr G F Miller of the National Physical Laboratory, Teddington for their advice on the work in this paper.