Research On Fault Diagnosis Of Motor Rolling Bearing Based On Vibration Signal

At present,the rapid development of my country’s manufacturing industry— "Made In China" has gone to the world.Agricultural rotating machinery equipment such as tractors and combine harvesters are an indispensable member of the manufacturing industry.The continuous development of society makes it more and more demanding,and the monitoring of its health status has gradually become particularly important.As the power source of intelligent agricultural machinery equipment,the core component of the motor is the bearing.When the bearing fails during the operation of agricultural machinery equipment,its internal motor cannot operate normally,which leads to irreparable and major losses in the entire system.Therefore,it is very important to carry out research on the health detection and failure diagnosis of bearings of agricultural machinery.Vibration signals are the most common detection method for bearing failure diagnosis.Therefore,in this paper,we focus on rolling bearings of agricultural machinery and equipment motors,and use the rolling bearing vibration signals as detection signals for noise reduction pretreatment,feature extraction,and failure identification of this type of vibration signal.The specific research contents are as follows:(1)Analysis of the fault mechanism of the rolling bearing of the motor.This paper expounds the basic mechanical structure,fault types and causes of motor rolling bearings in agricultural machinery and equipment,and gives the corresponding fault frequencies of motor rolling bearings in various states,which lays a theoretical foundation for the following fault diagnosis.(2)Noise reduction and preprocessing of rolling bearing vibration signals.In this paper,the vibration signals of rolling bearings are collected in four bearing states,namely,normal,ball failure,outer ring,and inner ring failure.The working environment of rolling bearing in running state is very complex,so the vibration signal collected is often mixed with a lot of nonlinear noise.In view of this situation,the mode variant VMD algorithm is used to indicate the aforementioned vibration signals in this paper,and the effectiveness of the VMD algorithm for signal de-noise is verified through simulation experiments.(3)Feature extraction of rolling bearing vibration signals.The feature extraction of vibration signal of rolling bearing after noise reduction was carried out.In this paper,the statistical features,the energy features of wavelet packet and the entropy features of vibration signals are extracted respectively,so as to analyze the state characteristics of vibration signals from three perspectives.Due to the different extraction methods,these three kinds of features can be effectively complementary,so as to more comprehensively reflect the feature information of rolling bearings in different states.Characteristics of three types of formation of the combination of high dimensional feature vector is easy to increase the computational load of the SVM recognizer and often contain some redundant information,so this article introduces a local linear embedding in the manifold learning algorithm LLE combination of high-dimensional feature dimension reduction for effective integration,thus completed the features of three kinds of rolling bearing vibration signals have the effect of fusion,make such a low dimensional feature contains more category information,for the following fault pattern recognizer SVM provides effective input feature vectors.(4)Fault identification of rolling bearing vibration signals.A motor rolling bearing fault pattern recognition model based on support vector machine(SVM)theory is established.Since the parameters of SVM model usually depend on manual experience selection,therefore,this document uses a genetic particle optimization algorithm and a whale hybrid improvement algorithm to optimize the parameters of the SVM model.Through experiments,it can be known that the error detection rates of the SVM models optimized by the two algorithms have reached 99.22% and 100%,respectively.The effectiveness of the proposed method is verified by comparative analysis.