Motor Fault Diagnosis Based On Wavelet Analysis And CPSO-NP Optimized SVM

As a driving device of emancipating manpower,asynchronous motors have a wide range of applications in fields such as industry and agriculture because of their simple structure,good stability,and moderate price advantages.If a motor has bad,it will cause damage to the motor and affect the production schedule.It will jeopardize the lives of operators and the safety of corporate assets.Therefore,how to finish rapidly and achieve diagnosis of electric motors and avoid early detection of faults have important practical significance.The article takes the asynchronous motor as the research object,firstly it introduces its working principle,and summarizes the fault type and diagnosis technology of the motor,and studies the frequency characteristics of the rotor broken bar fault,and the stator fault and the rolling bearing fault,and provides the theory for the next fault diagnosis.Secondly,the article will be divided into three major fault diagnosis systems:signal acquisition module,signal processing module and pattern recognition module,because the article uses Case Western Reserve University bearing data sets,so the latter two modules are articles Content of the study;In the signal processing module of the article,an improved wavelet denoising method and wavelet packet feature extraction are proposed.Considering the inevitable noise in the collected signal and affecting the extraction of fault signals,this paper proposes wavelet denoising method using neighboring coefficients based on a multi-threshold new function(NCMN);NCMN contains neighboring hard threshold value,neighboring window threshold value and neighboring expansion threshold value,In order to fully reflect true information of the signal,the wavelet coefficients are preserved or shrinked in the light of different threshold values in the same window;The novel thresholding formula which well reflects the relationship with the filtered noise is used to decrease the effect of noise;Then,by combining with revised universal threshold,NCMN makes use of chaotic particle swarm optimization algorithm finding the optimal values of parameter and which respectively come from neighboring expansion threshold and universal threshold,furthermore,the processed optimal wavelet coefficients are reconstruct the original signal;Finally,Compared with other threshold de-noising methods,the simulation results show that NCMN can improve the signal to noise ration,reduce the useful signal disortion and eliminate noise effectively.secondly,the multi-resolution characteristics of wavelet packets are used to decompose thedenoised signals and use the energy of different frequency bands as feature vectors.Simulation results show that the extracted feature vectors can represent different fault information.;In the pattern recognition module of the article,a fault diagnosis method based on improved particle swarm optimization support vector machine is proposed.In order to effectively balance the exploration and exploitation's capabilities and overcome the shortcomings of local minima and low convergence in Particle Swarm Optimization(PSO),a Chaotic PSO algorithm with Natural selection and Predatory search(CPSO-NP)was proposed.The evolution mechanism of survival of the fittest in natural selection was employed to improve the rate of algorithm's convergence;predatory search can balance the global search and local search though adjust the level of restriction so as to optimize the search performance;The results of test functions and fault diagnosis of bearing show that CPSO-NP algorithm can search the parameter optimization of SVM accurately with high precision and fast convergence.It raises the accuracy of fault diagnosis.