Bearing fault diagnosis method based on random search and convolutional neural network

Abstract

The invention relates to the field of bearing fault prediction and diagnosis and discloses a bearing fault diagnosis method based on random search and a convolutional neural network, which optimizes hyper-parameters in combination with a random search algorithm and establishes a convolutional neural network model for intelligent diagnosis of rolling bearing faults; problems that a traditional method in the prior art is not enough in accuracy and manual parameter adjustment in an intelligent method is tedious and time-consuming are solved. The method comprises the following steps of (1) initializing a hyper-parameter combination; (2) configuring a distribution function of random search; (3) continuously updating the distribution function of random search; and (4) selecting optimal hyper-parameter configuration; and training to obtain a final bearing intelligent diagnosis network model. The method is advantaged in that through an alternate connection structure of two convolution layers and a single pooling layer, the convolution layer performs convolution operation and data feature learning on input data, the pooling layer is designed to be maximum pooling, and pooling kernel operation of the maximum pooling layer can enhance data features obtained by learning of the convolution layer.

Description

technical field [0001] The invention relates to the field of bearing fault prediction and diagnosis, in particular to a bearing fault diagnosis method based on random search and convolutional neural network. Background technique [0002] At present, the diagnosis schemes of rolling bearing faults are divided into two categories. One is the traditional method, which uses signal analysis and feature extraction to extract features from the original data and then uses pattern recognition to diagnose rolling bearing faults. The other is the intelligent diagnosis method, which uses machine learning to directly learn the original vibration data to diagnose rolling bearing faults. The traditional method is time-consuming and laborious, and the manual extraction of features is easy to lose the original information, and the accuracy of the diagnosis is not enough. Compared with the traditional method, the intelligent diagnosis method is more accurate, but there are also problems such...

AI Technical Summary

Problems solved by technology

[0003] In order to overcome the lack of accuracy of the existing traditional methods in the fault diagnosis of rolling bearings and the problem that the artificial parameter adjustment in the intelligent method cannot optimize the network well, the present invention designs a random search algorithm based

Images