To solve the problem of parameter selection of variational mode decomposition and the difficulty of determining the network structure of sparse autoencoder in fault diagnosis of rolling bearings in rotating machinery

a fault diagnosis model combining variational mode decomposition and sparse self-adaptive autoencoder is proposed in this paper. Firstly

the envelope entropy is calculated to determine the decomposition layers and modal components of the variational modal algorithm. Then

the optimal component selection is achieved by signal decomposition and noise reduction. And the envelope spectrum of the best component is calculated and used as the input of the sparse autoencoder. Particle swarm optimization algorithm is introduced to optimize the network structure of the sparse autoencoder to obtain the optimal feature representation ability of automatically extracting vibration data

which greatly enhances the adaptability of the model on the premise of satisfying the better feature learning ability of the model. The simulation results of fault type identification of bearing and variable speed bearing data sets of Case Western Reserve University show that the proposed method has strong adaptability and excellent accuracy.