Rolling Bearing Fault Diagnosis Method Study Based On Image Processing

Rolling bearing is one of the commonly used rotating machinery in industry,which has a wide range of applications in many fields.The normal operation of rolling bearing is related to the safety of the whole industrial equipment,so it is very important to monitor and diagnose the rolling bearing in real time.Feature extraction is the key point in the whole process of fault diagnosis.The vibration signal of the bearing is collected and analyzed effectively,so as to achieve the purpose of fault diagnosis.The vibration signals of bearings are generally nonlinear and non-stationary.It is difficult to extract effective fault features by traditional timefrequency analysis technology.In this paper,the original signal is transformed into symmetrical point image(SDP)which is easy to extract fault features.Because its shape is similar to snowflake,it is also called snowflake image.Then,improved Manhattan,improved Chebyshev and convolutional neural network(CNN)are used for fault feature extraction and fault recognition.The main research work of this paper is as follows:(1)A rolling bearing fault diagnosis method based on EMD and improved Manhattan distance is proposed.In this method,the collected vibration signals are divided into 10 equal parts,and the equal parts are decomposed into several intrinsic mode function(IMF)components by empirical mode decomposition(EMD).The first five IMF components are transformed into five symmetrical point images in polar coordinates.Then each image is processed by image binarization,image segmentation and image denoising to get a clearer local image.The average image of 10 local images of each IMF component is calculated,and the maximum eigenvalue of the average matrix is calculated.The Manhattan distance between each local image and its average image is extracted,and the maximum feature value of the average matrix is added to the original Manhattan distance to obtain the improved Manhattan distance.Finally,according to this feature,the fault diagnosis and classification of rolling bearing are carried out.(2)A bearing fault diagnosis method based on EMD and improved chebyshev distance is researched.Firstly,the original data of rolling bearing is divided into 10 equal parts.The original signal of rolling bearing is decomposed into several intrinsic mode functions(IMF)by empirical mode decomposition(EMD),and each IMF component is transformed into snowflake image by the basic principle of symmetrical point image.Then the local image is extracted by image binarization,image denoising and image segmentation,and the average image of 10 local images and the maximum eigenvalue of its matrix are calculated.The Chebyshev distance between each local image and its average image is calculated.In order to improve the accuracy of diagnosis,a new improved Chebyshev distance is obtained by adding the maximum eigenvalue of the mean matrix to the original Chebyshev distance.This method has higher accuracy than the original Chebyshev distance in fault diagnosis.(3)A method of rolling bearing fault diagnosis based on convolution neural network is studied.Firstly,the collected vibration signals are divided into training set and test set,and all signals are transformed by SDP to get the symmetric image in polar coordinates.Then,the SDP images in the training set are input to the input layer of the convolutional neural network model,and the fault types are automatically diagnosed through the convolution layer,pooling layer,full connection layer and output layer.The number of convolution layers and the size of convolution kernel in CNN model are adjusted continuously,and a convolution neural network model is selected according to the new index.Finally,the effectiveness of the selected CNN model is verified by the test set images.Finally,a summary of the research contents is presented.Moreover,the further research object and the target are pointed out.