Research On Fault Diagnosis Method Of Rolling Bearing Based On PSO-BP Neural Network And Improved PSO-SVM

The role of rolling bearings in actual industrial production is crucial.In industrial production,the normal operation of rolling bearings directly affects whether the enterprise can perform normal production.At present,ordinary fault diagnosis companies mainly adopt some relatively single methods,such as time domain or frequency domain analysis.These methods can only roughly diagnose the location of the fault,and they are not efficient.Until 2016,with the development of artificial intelligence,researchers began to try to combine artificial intelligence with fault diagnosis,and gradually worked out accurate and efficient intelligent diagnostic methods.In this thesis,we will use wavelet packet transformation algorithms to separately compare with BP neural network,particle swarm optimized BP neural network(PSO-BP),particle swarm optimized support vector machine(PSO-SVM),and particle swarm optimization with improved acceleration factors.A comparison experiment of four intelligent diagnostic methods combined with the past support vector machine(PSO-SVM)is carried out to obtain a more efficient intelligent diagnostic method.This thesis takes rolling bearings as the research object,and designs a set of schemes starting from signal acquisition until identifying the type of failure.First of all,for the signal acquisition system,starting from the hardware and software design,implement a system to collect vibration signals,and realize the signal sharing through cloud computing technology.Then,wavelet noise reduction is used to remove the interference of the noise signal.Secondly,the wavelet packet transform is used to decompose the signal into frequency bands with the same width,and then extract the energy from the decomposed frequency band.Parameters,through the analysis of these parameters can get different fault locations.Finally,these feature vectors are combined into a training sample set for training.The four methods of BP neural network,PSO-BP neural network,PSO-SVM,and improved PSO-SVM can effectively improve the diagnosis speed and accuracy.The four methods are verified through experiments,and it is found that they can effectively perform fault diagnosis.Comparing the accuracy of the four methods and the efficiency of fault diagnosis,it is finally verified that the improved PSO-SVM has better diagnostic performance.it is good.