Skip to Main Content
In the present study, an expert system is developed to identify and classify fault condition for diesel engine. Vibration signals are collected on a diesel engine test platform. Wavelet packet analysis (WPA) coefficients of vibration signals are used for evaluating their Shannon entropy and treated as the features to identify the fault conditions of diesel engine in the preprocessing. A back-propagation neural network (BPNN) is used to classify the fault condition. To improve the convergence of BPNN, a hybrid particle swarm optimization (PSO) with a differential operator named PSO-DV is used to adjust the weights and threshold of BPNN in fault diagnosis of diesel engine. To verify the proposed PSO-DV hybrid method has the better convergence, a classical PSO based BPNN is compared with a PSO-DV based BPNN in fault classification of diesel engine. The experimental results showed the proposed hybrid intelligent PSO-DV method not only achieved classification for diesel engine, but also can escape from local optima, so has better convergence than classical PSO.