Rolling bearing fault diagnosis method based on improved SSA optimized VMD and CNN parameters

Abstract

The invention discloses a rolling bearing fault diagnosis method based on improved SSA optimized VMD and CNN parameters. Comprising the following steps: inputting a rolling bearing data set; optimizing the decomposition number and the penalty factor of the VMD algorithm based on an improved sparrow search algorithm SSA to obtain an optimal parameter combination of the decomposition number and the penalty factor of the VMD; decomposing the vibration signal into a plurality of signal components containing fault information by using the parameter-improved VMD; optimizing hyper-parameters of the CNN based on an improved sparrow search algorithm SSA, wherein the hyper-parameters comprise a learning rate, a training frequency, a convolution kernel size of each convolution layer and the like; and fault feature extraction and fault diagnosis are realized by using the CNN after parameter improvement. According to the method, a good fault classification and diagnosis effect is obtained on a rolling bearing public data set, the problem that a traditional VMD algorithm is limited by penalty factors and decomposition numbers is solved, accurate extraction of fault feature information is realized, the problem that a traditional CNN network is subjected to repeated test selection of hyper-parameters, and consequently a large amount of experiment cost and time are consumed is solved, and the method is suitable for popularization and application. The diagnosis precision of the CNN network is improved.

Description

Technical field [0001] The invention is a mechanical failure diagnosis field, especially in a rolling bearing fault diagnosis method based on improving SSA optimization of VMD and CNN parameters. Background technique [0002] The manufacturing industry is the foundation of the country and the foundation of a strong country. With the rapid development of modern technology industry technology, especially information technology, engineering systems in various fields such as aviation, communications, and industrial applications are becoming more and more complicated. The chance of failure and functional failure has gradually increased, so the diagnosis of faults has gradually become the focus of researchers' attention. [0003] Regarding the classification of fault diagnosis methods, different research institutions and organizations at home and abroad are not consistent. From the perspective of current mainstream technology and application research work, it can be divided into: model...

AI Technical Summary

Problems solved by technology

[0006] In the 1980s and 1990s, due to the limitations of computer level and other technologies, deep learning did not show excellent performance

For a given problem, it is difficult to know the best combination of hyperparameters, but replicating values ​​​​used on other problems by past experience, or finding the best value through trial and error will consume a lot of experimental cost and time, and may not be the optimal solution

Images