Application Of Wavelet Packet Combined With SVM And Neural Network In Fault Diagnosis Of Rolling Bearing

The rolling bearing is a key component of rotating machinery,its health conditions will directly affect the running state of the whole equipment.With the development of monitoring technology,the amount of data collected related to the running state of rolling bearings increases exponentially,which enables the concept of big data for fault diagnosis.Therefore,extract useful features from big data by advanced technology and identify the health conditions of roll bearings accurately has been a new research topic.In this paper,the vibration signal of rolling bearing running parameters is selected as the research object,and the intelligent fault diagnosis of rolling bearing is studied by wavelet packet transform,support vector machine(SVM)and BP neural network.The main contents and conclusions of this paper are as follows:1.The shortcomings of Fourier transform and Windowed Fourier transform in timefrequency domain analysis are mentioned,and wavelet packet transform is selected for feature extraction of vibration signals of rolling bearings.Wavelet packet transform decomposes the original signal into different frequency bands and reconstructs the signal of each frequency band to generate energy feature vectors.The research shows that energy feature vectors contains abundant fault information,and can be used for intelligent classification methods as sample data.2.The process and method of intelligent fault diagnosis model of rolling bearing based on support vector machine and BP neural network are studied respectively.Support vector machine(SVM)is a supervised learning method,its goal is to obtain hyperplane equations of sample data based on support vector.And classification can be carried out by these equations.Essentially,the learning process of BP neural network is to adjust the connection weight between each neuron and the threshold value of each neuron according to the sample data.The SVM and BP neural network both have advantages and disadvantages,and the method should be selected according to the actual situation of sample data.3.Four kinds of rolling bearing faults(normal bearing,outer ring fault bearing,inner ring fault bearing and ball fault bearing)were tested in the experiment.In order to better evaluate the performance of the model,"large sample" and "small sample" sets were generated from the experimental data,and SVM model and BP neural network model were established by Matlab.The data processing results show that the accuracy of SVM model is higher than that of BP neural network model in the classification of small samples.When classifying large samples,the classification accuracy of both models is high enough,but the SVM model will cost more computing resources and time.Theoretical knowledge and experiments prove that wavelet packet transform can effectively extract fault feature information of rolling bearing.In the intelligent classification method,SVM model is suitable for small sample classification,and BP neural network is suitable for large sample classification.