Deep learning rolling bearing fault diagnosis method based on feature fusion and hybrid enhancement

Abstract

The invention discloses a deep learning rolling bearing fault diagnosis method based on feature fusion and hybrid enhancement, which adds feature engineering on the basis of deep learning, and comprehensively considers the time domain feature, frequency domain feature, working condition feature and time difference feature of an original signal, the features and original signals are fused into new one-dimensional signals, the new one-dimensional signals are converted into a two-dimensional image format, virtual samples and labels are constructed in a linear interpolation mode through hybrid enhancement, and the virtual samples and labels are input into a ResNet18 network for training by means of the powerful feature extraction capacity of a two-dimensional convolutional neural network. According to the method, potential features of original data are comprehensively considered, data distribution is disturbed, and the generalization ability of the model is improved. The rolling bearing fault diagnosis method not only improves the precision of rolling bearing fault diagnosis, but also has good domain adaptability and is suitable for fault diagnosis under various working conditions.

Description

technical field [0001] The invention relates to the field of deep learning technology, in particular to a method for deep learning rolling bearing fault diagnosis based on feature fusion and mixed class enhancement. Background technique [0002] In recent years, with the national industrial upgrading, energy conservation and emission reduction, it is necessary to realize real-time fault diagnosis for large mechanical systems such as aviation generators, rail transit equipment, and agricultural equipment, and develop management systems with health monitoring, fault diagnosis detection and diagnosis, and remaining life prediction. Functional advanced health monitoring system. Rolling bearings are important parts in rotating machinery. Among the faults of rotating machinery, the faults caused by bearing damage account for about 30%. Therefore, the fault diagnosis of rolling bearings is of great significance to the condition monitoring and maintenance of rotating machinery equ...

AI Technical Summary

Problems solved by technology

[0009] 1. The generated new samples belong to the same category;

[0010] 2. The relationship between different samples of different categories has not been modeled;

[0011] 3. Overfitting occurs

Images