Research On Fault Diagnosis Method Of Bearing Based On Linear Discriminant Analysis Theory

Rolling bearing is an important component of rotating machinery.The status of bearing affects the normal operation of mechanical equipment.Therefore,the monitoring and diagnosis of bearing running conditions are extremely crucial to ensure the normal running of mechanical equipment.This study takes train bearings and rolling bearings as the research objects,proposing an improved linear discriminant analysis(ILDA)feature extraction method and an adaptive linear discriminant analysis(ALDA)feature extraction method.These two methods research how to extract effective status features from different perspectives,and combine machine learning theory to realize bearing fault diagnosis.In the traditional bearing status feature extraction methods,the linear discriminant analysis method pays too much attention to the samples with large between-class dispersion,and does not consider those with small between-class dispersion.In response to these problems,this paper proposes ILDA to extract the status features of train bearing wayside distortion correction acoustic signals.The algorithm uses the distance function to adjust the between-class divergence matrix of the LDA algorithm,increasing the weight of samples with crosses or overlaps between different types of classes,reducing the weight of samples without crosses or overlaps between different types of classes.Firstly,train bearing wayside distortion acoustic signals are collected and corrected.Then,the fusion status features of the train bearing wayside correction acoustic signals are extracted.Finally,the Extreme Learning Machine(ELM)classifier is used to identify the running conditions of the train bearing.The experimental results show that the status features extracted by this method have a higher recognition rate than the status features extracted by traditional methods.In addition,a bearing status feature extraction method based on ALDA is proposed for rolling bearing fault diagnosis.The sample clustering evaluation index(SI)is used to adjust the weight of the within-class divergence matrix of the LDA algorithm,reducing the cross or overlap among different types of samples,especially for the marginal samples.Firstly,the vibration signal of the rolling bearing is collected.Then,the fusion features reflecting rolling bearing running status are calculated by ALDA.Finally,the Support Vector Machine(SVM)is used to identify the rolling bearing working conditions.The experimental results show that the status features extracted by ALDA compared with traditional algorithms can be effectively used for rolling bearing fault diagnosis.Firstly,this paper verifies the effectiveness of ILDA based on the train bearing simulation data and experimental data.Then,ALDA is verified to be effective based on bearing experimental data of CWRU.Finally,the effectiveness of ILDA and ALDA is verified based on bearing fatigue experimental data.The results show that the proposed two methods are effective in extracting fusion features reflecting bearing running conditions,and the recognition accuracy of status features is significantly higher than other traditional methods.The research results of this paper are significant for bearing fault diagnosis.