Rolling bearing fault diagnosis method based on multi-branch multi-scale convolutional neural network

Abstract

The invention discloses a rolling bearing fault diagnosis method based on a multi-branch multi-scale convolutional neural network. The rolling bearing fault diagnosis method comprises the steps of: acquiring acceleration vibration signals of rolling bearings without faults and with different faults in different operating states at first; setting a fault state label according to the fault state corresponding to each acceleration vibration signal; standardizing each acceleration vibration signal; training a multi-branch multi-scale convolutional neural network model by taking the standardized acceleration vibration signals as a training sample, wherein the multi-branch multi-scale convolutional neural network model comprises a low-frequency branch convolutional network, an identity mapping branch convolutional network, a denoising branch convolutional network, a feature fusion layer, a global average pooling layer and a Softmax layer; and then acquiring current acceleration vibration signals of the rolling bearings, and sending the current acceleration vibration signals into the multi-branch multi-scale convolutional neural network model for fault diagnosis. By adopting the multi-branch multi-scale convolutional neural network model, the rolling bearing fault diagnosis method can effectively improve the fault diagnosis performance of the rolling bearing in a strong noise environment and under a variable load working condition.

Description

technical field [0001] The invention belongs to the technical field of rolling bearing fault diagnosis, and more specifically relates to a rolling bearing fault diagnosis method based on a multi-branch and multi-scale convolutional neural network. Background technique [0002] Rolling bearings are important components in industrial application systems, and the failure caused by rolling bearings is an important reason for the failure of machinery and equipment, especially for rolling bearings under high-speed and heavy-load conditions, due to the long-term repeated action of contact stress, it is easy to cause fatigue and cracks , denudation and other faults, these faults will reduce the rotation accuracy of the bearing, generate vibration and noise, increase the resistance of the bearing rotation, and eventually cause the bearing to be blocked and stuck, resulting in the failure of the entire mechanical system, so the fault detection of the bearing Very important. [0003] ...

AI Technical Summary

Problems solved by technology

The traditional intelligent fault diagnosis method has the following disadvantages: 1) The vibration signal of the rolling bearing is affected by other moving parts and structures, and its vibration characteristics are very complex, and the manually extracted features cannot fully characterize the complex dynamic characteristics of the rolling bearing

2) In a strong noise environment, the signal features related to the fault are completely submerged by the noise; under variable load conditions, the fault features are distributed in different feature intervals, so the manually extracted features cannot truly reflect the inherent characteristics of bearing faults characteristic

3) These machine learning classification algorithms are shallow models, it is difficult to learn the complex nonlinear relationship of vibration signals, and it is easy to cause misjudgment

However, under strong noise environment and variable load conditions, the fault feature extraction of rolling bearing vibration signals is more challenging, and the above methods perform poorly in rolling bearing fault diagnosis tasks under strong noise and variable load conditions

Images