Rolling bearing fault diagnosis method and system based on relative entropy and k nearest neighbor algorithm

Abstract

The invention discloses a rolling bearing fault diagnosis method based on relative entropy and K nearest neighbor algorithm, which includes obtaining vibration data generated by the operation of the bearing in various fault states and vibration data generated by operation in a healthy state; The relative entropy vector sequence between the vibration data and the vibration signal generated by the operation in the fault state; the fault type is used as the training sample to obtain the trained classification model; the vibration data generated by the operation in the unknown state is obtained, and the operation generated in the fault state The relative entropy vector between the vibration signals; the obtained relative entropy vector is used as the test sample of the classification model, and the classification model is used to test the test sample, and then continue to diagnose the rolling bearing fault, and obtain the diagnosis result. The present invention adopts relative entropy to measure the difference of vibration signals between bearings in different states, does not need to calculate and optimize the combination of different characteristic indexes, and directly utilizes the distribution of original vibration signals.

Description

technical field [0001] The invention relates to the technical field of fault diagnosis, in particular to a method and system for fault diagnosis of rolling bearings based on relative entropy and K nearest neighbor algorithm. Background technique [0002] Rolling bearings are called "mechanical joints" and are one of the most critical structures in rotating machinery. They are widely used in aerospace, machinery manufacturing, automobiles and ships, and their operating status is often decisive for the safe and stable operation of these mechanical equipment. effect. According to statistics, one of the main failure causes of mechanical equipment is rolling bearing failure. Failure of rolling bearings directly causes equipment downtime, and if there is no good monitoring and diagnosis method, it may cause more serious accidents. Therefore, it is very important to carry out fault diagnosis on rolling bearings, which can be used not only to troubleshoot and locate faulty equipme...

AI Technical Summary

Problems solved by technology

However, the spectral analysis method based on Fourier transform also has great limitations: it is only suitable for steady-state signals

In fact, the original signals generated by rotating machinery often have very strong unsteady characteristics. For unsteady signals, the effect of this method is greatly reduced, and even wrong results; profound theoretical knowledge is required

In actual use, if you want to achieve the ideal effect, you need to make a lot of judgments when truncating the original signal to ensure the relative stability of the obtained time domain signal. The judgment algorithm increases the complexity of the algorithm and requires a very deep theoretical foundation and expert knowledge; Extensive expert experience required

Converting to frequency domain signals is very good, but due to the complexity of the bearing structure, the identification of characteristic frequencies in the frequency domain also depends on expert experience

[0005] Although the Fourier transform is very mature, the Fourier transform is only suitable for time-invariant signals, and in practical application scenarios, effective fault diagnosis frequency components are often buried in a large amount of noise and other useless vibration signals. Reliable diagnostic conclusions can only be obtained by analyzing the frequency spectrum combined with expert experience

Images