Neural network applications in multisensor systems

Abstract

Reviews the suitability of different neural network architectures for use with typical multisensor systems required by their increasing use in complex engineering applications. Outlines the learning mechanisms that are required [to generate the transformation between the data at the input and the corresponding output] involving back‐propagation networks and self‐organising map networks. Looks at the three main problem areas of classification, quantification and descriptions and uses the case study of an electronic nose as a system which encounters each of these problems. Concludes that the combination of artificial neural networking tools with mutisensors is becoming more widely accepted and defines the need for the investigation of alternative supervised and unsupervised architecture if the true potential of multisensor systems is to be realized.