A Spectrum-Based Bearing Fault Classification Method and System

Abstract

The invention discloses a frequency spectrum-based bearing fault classification method and system. The method includes S1, calculating the frequency spectrum of the vibration signal of each state of the bearing; S2, calculating the frequency difference between each state of the bearing; S3, calculating the ratio K of each difference between each state spectrum; S4, selecting the ratio K Greater than the frequency of the frequency amplitude of the set value; S5, set the parameter R according to the following formula, R=R1 / R2, R1 is the common frequency amplitude number that needs to analyze the training set frequency, and R2 is the total frequency amplitude number of all training set frequencies ; S6. Select the amplitude of the common frequency from each state to form R eigenvectors, and input the calculated eigenvectors into K-adjacent value classification, K-means clustering, and linear support vector machine for classification. The invention realizes the selection of feature vectors through fewer FFTs and common frequency selections, has small calculation amount, and is convenient for engineering realization.

Description

technical field [0001] The present invention relates to the field of bearing fault classification, and more specifically relates to a frequency spectrum-based bearing fault classification method and system. Background technique [0002] Equipment tends to be larger and larger, and more and more researches in the field of fault diagnosis and fault detection have shown that health detection and fault diagnosis play a vital role in the safe operation of equipment. Bearings are important parts connecting rotating parts and supporting parts in equipment, and there are already a large number of fault detection methods. [0003] At present, the methods applied to bearing fault classification include eigenfrequency analysis methods and envelope analysis methods in the frequency domain, statistical analysis methods in the time domain, and neural network analysis methods. Among them, the eigenfrequency analysis method is the simplest analysis method, but when the noise is large and t...

AI Technical Summary

Problems solved by technology

Among them, the eigenfrequency analysis method is the simplest analysis method, but when the noise is large and the fault is early, it cannot analyze the fault in a timely and effective manner; the neural network method can distinguish the fault more accurately, but requires accurate characteristic input or large amounts of data

This leads to the general complexity of neural network methods

Images