An intelligent approach for engine fault diagnosis based on wavelet pre-processing neural network model

Based on the sound intensity analysis, discrete wavelet transform (WT) and the neural network (NN) technique, a combined intelligent method for engine fault diagnosis (EFD), the so-called wavelet pre-processing neural network (WT-NN), is presented in this paper. Based on the measured multi-condition engine noise signals, a wavelet-based 21-point model for feature extraction of engine noise is established, as is a multi-layered NN model for fault pattern identification. To verify the proposed intelligent method, as an example, the WT-NN models are built and performed for recognizing eight common faults of the 2VQS type of EFI engine. The results suggest that the WT-NN models are effective and feasible for engine fault diagnosis. Due to its outstanding time-frequency characteristics, the WT-NN model can be used to deal with stationary, nonstationary and transient signals. The WT-NN technique is suggested not only to detect the engine faults, shorten maintenance time, but also to apply to other sound-related detection fields in engineering.