基于熵选择小波包分量和PSO-BP神经网络的发动机故障诊断

Engine Fault Diagnosis Based on Entropy Selective Wavelet Packet and PSO-BP Neural Network

为实现发动机活塞组故障的快速诊断,对活塞销不同工况下的振动信号进行小波包分解并重构,求出各分量的信息熵,根据信息熵的大小选取出合适的小波包分量进行分析。将选取出的各分量的能量归一化后作为特征参数,输入到PSO-BP神经网络模型中进行训练和预测识别。试验结果表明:基于熵选择小波包分量和PSO-BP神经网络的方法,能够对活塞销不同程度的故障进行分类识别。

To quickly diagnose the faults of engine piston groups,the paper firstly decomposes and reconstructs the vibration signals of the piston pins under different working conditions with wavelet packet method. Then,it computes the entropy of wavelet packet components and selects suitable components by the values. Lastly,it normalizes the energy of the chosen ones and inputs them as characteristic parameters into the PSO-BP neural network model for training and prediction recognition. The experiment results show that this method combining entropy selective wavelet packet with PSO-BP neural network can effectively classify and identify different levels of faults of the piston pins.