Author(s)

M.A. Golberg

Abstract

We consider the problem of analyzing the convergence rate of projection methods for solving a variety of integral equations which arise from integral reformulations of boundary value problems when both integration and data approximation errors are present. This is done by extending Miel's perturbation theory for split equations of the first kind [6] and generalizes the work of Anselone and Lee [5]. To illustrate the theory we discuss the numerical solution of some boundary integral equations which arise from the conversion of boundary value problems for Laplace's equation in the plane to equivalent boundary integral equations. In particular, we show that the dominant error comes from approximating the kernel and right-hand sides. Such errors appear to have been largely

Keywords