Aimed at the common gas sensor fault

the approach of fault diagnosis based on wavelet packet and neural network is discussed.The output of gas sensor by three-layer wavelet packet is decomposed to achieve eight signals of different frequency bands.An 8-dimensional eigenvector of vibrating signals was constructed for training 3-layer neural network.The fault type classifier is established for the gas sensor fault diagnosis.The simulation result shows that the proposed method can detect faults accurately.