Neural network methods for boundary value problems defined in arbitrarily shaped domains.pdfVIP

IEEE TRANSACTIONS ON NEURAL NETWORKS, VOL. 11, NO. 5, SEPTEMBER 2000 1041 Neural-Network Methods for Boundary Value Problems with Irregular Boundaries Isaac Elias Lagaris, Aristidis C. Likas, Member, IEEE, and Dimitrios G. Papageorgiou Abstract—Partial differential equations (PDEs) with boundary conditions (Dirichlet or Neumann) defined on boundaries with simple geometry have been successfully treated using sigmoidal multilayer perceptrons in previous works. This article deals with the case of complex boundary geometry, where the boundary is determined by a number of points that belong to it and are closely located, so as to offer a reasonable representation. Two networks are employed: a multilayer perceptron and a radial basis function network. The later is used to account for the exact satisfaction of the boundary conditions. The method has been successfully tested on two-dimensional and three-dimensional PDEs and has yielded accurate results. Index Terms—Boundary value problems, engineering problems, irregular boundaries, neural networks, partial differential equa- tions (PDEs), penalty method, multilayer perceptron, radial basis function (RBF) networks. I. INTRODUCTION N EURAL networks have been employed before to solve boundary and initial value problems [1] as well as eigen- value problems [2]. The cases treated in the above mentioned articles were for simple orthogonal box boundaries either finite or extended to infinity. However, when one deals with realistic problems, such as the human head-neck system [3] or the flow and mass transfer in chemical vapor deposition reactors [4], the boundary is highly irregular and cannot be described in terms of simple geometrical shapes, that in turn would have allowed for a relatively simple modeling scheme. In this article we propose a method capable of dealing with such kind of irregular boundaries (with either Dirichlet or Neu- mann boundary conditions). As before [1], [2], our approach is based on the use of feedforwa