Early Fault Diagnosis of Rolling Bearings Based on RSSD and CNNSE-BiLSTM
The early fault of rolling bearing has a strong noise background and the signal strength is weak,which leads to low diag-nostic accuracy.In order to solve the problem,a fault diagnosis method based on resonance sparse signal decomposition(RSSD)was proposed,which improved 1D convolution and bidirectional long short-term memory.The 3σ principle was used to determine the early degradation starting point of the bearing throughout its entire life cycle,and RSSD denoising was applied to the time-domain signal of the starting point to improve the resolution of early weak faults.The preprocessed signal was input into the convolutional neural network adding SE attention mechanism to extract key local features,and the feature extraction of the current and before and after time series in-formation was performed by its output and input bidirectional long short-term memory neural network(BiLSTM).Finally,the early multi fault classification was carried out through the fully connected layer and Softmax layer.The above method was used for the entire life cy-cle fault signal experiment of XJTU-SY bearing.The results show that the proposed method has a higher recognition rate for early weak fault signals,with a diagnostic accuracy of 99.75%,which is superior to other diagnostic methods.
resonance sparse signal decompositionconvolutional neural networkattention mechanismbidirectional long short term neural networkearly fault diagnosis
万方数据
维普
