Fault Diagnosis of Axial Piston Pump Based on CNN-SE-LSTM and Multi-sensor Data
Axial piston pump is the key component of the hydraulic system,its condition monitoring and fault diagnosis are the key to ensure the safe and reliable operation of the entire hydraulic system.However,due to the complex structure and bad working environ-ment of axial piston pump,the collected signal is often mixed with strong noise,using single sensor data to monitor its health status often does not achieve the expected effect.Therefore,an axial piston pump diagnosis method based on convolutional neural networks and long short-term memory networks(CNN-LSTM)with channel attention mechanism and multi-sensor data(MSD)was proposed.The size of convolution kernel in CNN was improved to optimize the structural parameters of CNN-LSTM to improve the anti-noise performance of the model,and the channel attention mechanism SENet block was introduced to enhance the model's representation ability.Then the data of two vibration sensors at different positions were fused as input.Finally,the fused data were input into the improved CNN-SE-LSTM and the diagnosis results were output through the Softmax layer.The experimental results show that the fault diagnosis accuracy of the proposed method is 100%without adding noise.The proposed method has good diagnostic accuracy and rapidity.Compared with multilay-er perceptron(MLP),deep convolutional neural networks with wide first-layer kernel(WDCNN)and other models,the proposed method has better fault diagnosis accuracy and robustness under noise interference with different signal-to-noise ratios.
万方数据
维普
