Motor bearing fault recognition method

Abstract

The invention provides a motor bearing fault recognition method, which comprises the following steps of S1, collecting a vibration signal of a motor bearing, and preprocessing the vibration signal; s2, performing feature extraction on the preprocessed vibration signal to obtain a feature vector of the motor bearing vibration signal; and S3, inputting the motor vibration signal feature vector obtained after preprocessing into a pre-trained fault classification model based on a convolutional neural network, and carrying out fault classification to obtain a bearing vibration signal fault classification result. According to the method, superposition and offset of white noise parts are carried out on the parts obtained after EEMD decomposition, so that reconstruction errors generated by noise disappear in the decomposition process, the decomposition precision is guaranteed, and the bearing fault recognition rate is increased.

Description

technical field [0001] The invention relates to the field of bearing fault identification, in particular to a motor bearing fault identification method. Background technique [0002] Bearing is an important rotating part in electromechanical equipment. Its failure often affects the normal operation of electromechanical equipment. Seriously, it may cause failure of electromechanical equipment and bring huge economic losses. Therefore, it is of great significance to ensure the normal operation of bearings. [0003] When a bearing fails, a vibration signal containing a large amount of impact noise will be generated. The generated vibration signal has a significant time-varying characteristic, thus showing an obvious nonlinear behavior. If the fault feature can be accurately extracted, the bearing can be effectively fault identification. [0004] The signal processing method can effectively reveal the nonlinear characteristics of the vibration signal, and comprehensively reflec...

AI Technical Summary

Problems solved by technology

Commonly used fault feature extraction, wavelet analysis, Fourier transform, and empirical mode decomposition, but wavelet transform requires artificial selection of wavelet bases, which has great difficulties, and threshold selection has great uncertainty, and EMD method cannot Effectively retain useful information, and there is modal aliasing and empirical mode decomposition method introduces white noise to cause signal distortion

Therefore, traditional fault feature extraction methods are not enough to mine features sensitive to all fault types.

Images