FAULT DIAGNOSIS METHOD FOR INTRA-CLASS SELF-ADAPTIVE BEARING UNDER VARIABLE WORKING CONDITIONS

摘要

A fault diagnosis method for a rolling bearing under variable working conditions, solving a problem that the universality of a deep learning model becomes poor caused by processing complex and variable working conditions of mechanical equipment by combining with a transfer learning algorithm on the basis of employing a convolutional neural network learning model. First, data acquired under different working conditions is cut to classify samples; the samples are preprocessed by means of FFT; then low-level features of the samples are extracted by means of improved ResNet-50; then a multi-scale feature extractor analyzes the low-level features from different perspectives to obtain high-level features to serve as input of a classifier. In the training process, high-level features of a training sample and a test sample are extracted at the same time, the conditional distribution distance between the training sample and the test sample is calculated and serves as one part of back propagation of a target function to achieve intra-class self-adaption, the influence of domain drift is reduced, and a deep learning model can be better qualified for a fault diagnosis task under the variable working conditions.