Wind turbine bearing fault diagnosis method based on rapid multi-separation dictionary learning

Abstract

The invention discloses a wind turbine bearing fault diagnosis method based on rapid multi-separation dictionary learning, and the method comprises the following steps: obtaining a wind turbine bearing signal, reconstructing a one-dimensional wind turbine fault signal to a high-dimensional phase space through phase space reconstruction, obtaining a matrix of a reconstructed signal, and forming a bearing fault diagnosis original feature set Y; partitioning the sample matrix, and dividing the sample matrix into L classes according to different classes through a K _ means classifier; optimizing and updating the ith class of matrix signals through separable dictionary learning, and outputting sparse coefficient separable dictionaries Ai and Bi; repeating the above steps to obtain a set of allsparse coefficients and dictionary matrixes, and performing reconstruction through a 2D _ OMP algorithm; performing integration processing on the signal Yi to obtain Y ', and outputting a signal Y' reflecting a fault characteristic so as to obtain non-stationary fault information of the wind turbine bearing; according to the method, the local spatial correlation in the high-dimensional signal is maintained, and the reconstruction efficiency is greatly improved.

Description

technical field [0001] The invention relates to the field of fault diagnosis of bearings, in particular to a fault diagnosis method for wind turbine bearings based on fast multi-separation dictionary learning. Background technique [0002] At present, rolling bearings have penetrated into all aspects of mechanized production. Once a failure occurs and the failure is not discovered in time, it may not only cause new failures, but also damage equipment and cause huge economic losses. Therefore, the research on early fault diagnosis methods for bearings is of great significance to ensure the safe operation of equipment. [0003] The traditional bearing fault diagnosis technology is based on the time-domain or frequency-domain feature extraction of vibration signals for fault identification. The effective extraction of fault features directly affects the accuracy of fault diagnosis. Especially when the bearing hides early weak faults and the fault information is overwhelmed by ...

AI Technical Summary

Problems solved by technology

However, the above algorithm has obvious shortcomings in the application. When the signal is relatively long, the atomic weight of the overcomplete dictionary is extremely large, contains a large amount of redundant information, and the operation speed is slow; when the signal noise is relatively large, K_SVD is affected by the signal phase. Larger, the signal features cannot be extracted well, and the reconstruction accuracy and efficiency will decrease

Images