This paper describes an approach to fault detection and classification usin
g neural networks and methods of combinatorial optimisation. Given the exis
tence of an effective fault detection procedure, the problem arises as to h
ow the sensors should be placed for optimal efficiency of the detector. In
this paper, a neural network is used to locate and classify faults and a nu
mber of different methods are applied to determine an optimal (or near opti
mal) sensor distribution. A survey of recent work on sensor placement is gi
ven. (C) 2001 Elsevier Science Ltd. All rights reserved.