DETECTION OF COMBUSTIBLE MATERIAL WITH PI EZOELECTRIC CRYSTAL SENSOR ARRAY USING PATTERN-RECOGNITION TECHNIQUES

A gas sensors array with seven piezoelectric crystals each coated with a different partially selective coating material was constructed to i dentify four kinds of combustible material which generate smoke contai ning different components, The signals from the sensors were analyzed with both conventional multivariate analysis, stepwise discriminant an alysis (SDA), and artificial neural networks (ANN), The results show t hat the predictions were even better with ANN models, In our experimen t, we have reported a new method for training data selection, ''stepwi se expanding training set method'' to solve the problem that the netwo rk can not converge at the beginning of training.