Evaluation Method Of Rolling Bearing Quality Based On Vibration Signal And Its Transform Image

Rolling bearings are important components of rotating machinery.The quality classification of rolling bearing is beneficial to the application of bearing classification.Traditional rolling bearing product quality inspection relies on the number of acceptances and rejections in the sampling inspection scheme to determine whether the batch of bearings are qualified.This method is complicated and inaccurate.Based on this,a rolling bearing quality grade evaluation method based on vibration signals and its transform images is proposed.The specific research methods are as follows:Aiming at the difficulty of extracting the vibration signal features and the low accuracy of pattern recognition methods in the process of quality grade evaluation of rolling bearings,a method for evaluating the quality of rolling bearings based on variational mode decomposition(VMD)and support vector machines(SVM)was proposed.Through the analysis of the simulation signals by the VMD method,the effectiveness of the method for the separation of mixed signals was proved.Firstly,the vibration signals of bearings of superior products,first-class products and qualified products were decomposed into 6 intrinsic mode functions(IMF)by VMD method.Secondly,the best IMF component number was optimized using the mutual information determination method,and the singular value decomposition(SVD)was carried out to obtain the singular value characteristics.Finally,singular value characteristics were used as inputs of SVM to establish models for quality grade prediction and evaluation.The experimental results show that compared with the EMD-SVD-SVM model,the VMD-SVD-SVM bearing quality evaluation model is better,and the average recognition rate of the prediction set is increased from 87.67% to 90.33%.VMD method is beneficial to the extraction of bearing vibration signal detail feature information and can effectively improve the resolution of vibration signal.In order to solve the problem of low recognition rate caused by insufficient extraction of vibration signal features of rolling bearing,a multi-domain feature rolling bearing quality evaluation method was proposed.Statistical analysis,Fast Fourier Transform(FFT)and VMD methods were used to extract the original vibration signal characteristics from multi-domain(time domain,frequency domain and time frequency domain),which can fully mine the stateinformation and inherent characteristics of the original vibration signal.Secondly,the Laplacian Score(LS)algorithm was introduced to select sensitive features according to the importance of each feature,so as to remove redundant information and improve the calculation efficiency.Finally,the selected characteristics were used as inputs of SVM to establish models for quality grade prediction and evaluation.The experimental results show that the multi-domain analysis of vibration signals can fully extract the features and improve the recognition rate of quality grade evaluation of the rolling bearing,and the average recognition rate of the predictive set can reach 92.50%.In view of the non-stationarity of the vibration signal of rolling bearing and the advantage of the intuitiveness of the image processing technology,a method for the quality evaluation of rolling bearing was proposed,which converts the vibration signal into vibration image.Firstly,the one-dimensional vibration signal was transformed into a two-dimensional image according to certain laws.Secondly,Histogram of Oriented Gridients(HOG)algorithm was introduced to extract features from 2D vibration image.Finally,the HOG features were used as the input variable of SVM to build the prediction and evaluation model of bearing quality grade,and the common data sets of bearing faults and the experimental data sets of bearing quality grading were used for verification..The experimental results show that: for the common vibration signal data sets of different fault types of bearings,the vibration image-HOG-SVM diagnosis model can obtain better analysis results than the direct use of vibration signal to identify the SVM,and the average recognition rate of the prediction set is increased from 95.62% to 99.63%.For the experimental bearing data sets with different quality levels,the vibration image-HOG-SVM classification model is better than the SVM classification model directly using vibration signals,and the average recognition rate of the prediction set is increased from 84.83% to 95.83%.It shows that the quality evaluation of rolling bearing can be realized by converting vibration signal into vibration image and extracting feature.Aiming at the problem that traditional machine learning algorithms cannot achieve deep-level feature extraction,a rolling bearing quality evaluation method based on vibration image with deep learning was proposed.Firstly,the one-dimensional vibration signal was transformed into 2D image according to certain laws.Secondly,the convolutional neural network(CNN)was used to extract the features of 2D images and then achieve quality classification.The common data set of bearing faults and the experimental data set of bearing quality grading were used for verification.The experimental results show that the average recognition rate of the prediction sets can reach 100.00% for the common fault diagnosis data sets.For the experimental data set of bearing quality classification,the average recognitionrate of the prediction set is 98.16%.The results are slightly better than the vibration image-HOG-SVM model.Therefore,the high precision evaluation of the quality grade of rolling bearing can be realized by transforming the vibration signal into two-dimensional vibration image and combining with the deep learning method,indicating that deep learning can realize deep feature extraction.