Rolling bearing fault identification method based on AVMD ASWT PCNN

DOI:	10.11809/bqzbgcxb2024.09.031
Keywords:	rolling bearing; adaptive variational mode decomposition; adaptive synchrosqueezing wavelet transform; convolutional neural network; fault identification
Abstract:	Aiming at the problems of inadequate noise reduction caused by the direct discarding of the high frequency component of the signals using with the traditional method and poor signal characterization in the time and frequency domains, a noise less time frequency image generation method based on Adaptive Variational Modal Decomposition fused with Adaptive Synchrosqueezing Wavelet Transform (AVMD ASWT) is proposed, based on which a Convolutional Neural Network combined with Particle Swarm Optimization is used to realize the identification of rolling bearing faults. The AVMD ASWT algorithm is used for processing of bearing vibration signals, and the mutual information entropy correlation coefficient criterion is also introduced, which can obtain high resolution time frequency images with less noise.The less noisy time frequency image is used as the input of the network model for fault identification, while the dynamic inertia weight particle swarm optimization algorithm (PSO) is used to optimize the parameters of the convolutional neural network model (CNN), which can solve the problem of difficult to determine the structure of the model, and the correct rate of the model identification and the identification speed have been significantly improved. Engineering examples show that the time frequency image obtained by using the AVMD ASWT method has higher resolution, significantly reduces the influence of noise in the signal, furthermore the correct rate of bearing fault identification has reached more than 99%.
Published:	2024-09-30
PDF