Motor fault diagnosis method and system based on improved stack type sparse auto-encoder

Abstract

The invention discloses a motor fault diagnosis method and system based on an improved stack type sparse auto-encoder, and belongs to the technical field of computers. The method comprises the following steps: S1, collecting a training sample, and initializing a network; s2, training a sparse self-encoding layer; s3, training a Softmax classification layer; s4, adjusting the deep learning network; and S5, carrying out motor fault diagnosis. According to the method, the improved sparse auto-encoder which performs non-negative weight constraint by using LP and L2 regularization is adopted, and the Softmax classifier is combined, so that the error of reconstruction by using the features learned by the sparse auto-encoder can be reduced, the sparsity of the learned features is improved, and the method has the excellent characteristics of high applicability and high stability and is worthy of being popularized and used.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a motor fault diagnosis method and system based on an improved stack sparse autoencoder. Background technique [0002] The traditional motor fault diagnosis method has certain limitations. It cannot automatically extract features and needs to manually extract features. The sparse autoencoder solves the problem of feature self-learning and has the advantage of automatic feature extraction. The error of sample reconstruction of features and the sparsity of learning features are improved, and a method using L P and L 2 The improved sparse autoencoder with non-negative weight constraints is regularized. In order to achieve accurate motor fault diagnosis, a motor fault diagnosis method and system based on an optimized SAE stacked sparse autoencoder are designed in combination with the Softmax classifier. Contents of the invention [0003] The technical problem to be solved by th...

AI Technical Summary

Problems solved by technology

[0002] The traditional motor fault diagnosis method has certain limitations. It cannot automatically extract features and needs to manually extract features. The sparse autoencoder solves the problem of feature self-learning and has the advantage of automatic feature extraction. The error of sample reconstruction of features and the sparsity of learning features are improved, and a method using L P and L 2 The improved sparse autoencoder with non-negative weight constraints is regularized. In order to achieve accurate motor fault diagnosis, a motor fault diagnosis method and system based on an optimized SAE stacked sparse autoencoder is designed in combination with the Softmax classifier.

Images