AN ADAPTIVE NEURAL-NETWORK TOPOLOGY FOR DEGRADATION COMPENSATION OF THIN-FILM TIN OXIDE GAS SENSORS

A hybrid neural network for gas sensing application is presented, whic h is based on adaptive resonance theory. The network may use as an inp ut one or more gas sensors. The basic feature of the proposed topology is its ability to learn a new pattern or form a new pattern category at any point of its operation. At the same time it retains knowledge o f previously learned patterns or pattern categories. This adaptation a bility helps the network to solve many of the problems encountered wit h tin oxide gas sensors, like instabilities and degradation. The funct ionality of the network is presented in the two cases of one and four input providing gas sensors. The experimental results show that the ef fect of sensor degradation maybe compensated by the proposed network t opology. (C) 1997 Elsevier Science S.A. All rights reserved.