Abstract: Agas sensors array with seven piezoelectric crystals each coated with a different partially selective coating material was constructed to identify four kinds of combustible material which generate smoke contammg different components.The signals from the sensors were analyzed with both conventional multivariate analysis, stepwise discriminant analysis (SDA), and artificial neural network_s (ANN).The results show that the predictions were even better with ANNmodels.In our experiment, we have reported a new method for training data selection," stepwise expanding training set method" to solve the problem that the network can not converge at the beginning of training.

Key words: Artificial neural network_s (ANN), Piezoelectric crystal sensor array, Stepwise discriminant analysis(SDA)