Application Of Improved Particle Swarm Optimization And Neural Network In Rolling Bearing Fault Diagnosis

The precision bearing is a very important component in modern machinery and equipment,and easy to break down in a non-normal conditions.Because the bearing failure will directly cause vibration,it has been widely used in Bearing Fault Diagnosis Based on Vibration Signal.Bearing vibration signal is used in this paper and use wavelet soft threshold method to denoise for sampling data.The signal preprocessing model based on wavelet packet extracting energy of vibration signal and neural network fault diagnosis model is set up and combine with the two models to study bearing fault vibration signals.Sinusoidal signal superimposed Gaussian noise is used to analog bearing fault vibration signal in this paper and then use wavelet soft threshold method to denoise vibration signal.After the vibration signal is denoised,use wavelet packet to extract energy of vibration signal.The fault signal preprocessing model based on wavelet packet extracting energy of vibration signal is established.The basic principles of standard PSO algorithm is researched and find the inadequate of standard PSO algorithm is not only the slow speed of convergence early in search,but also easily local optimum late in search and non-linear decreasing weights strategy is used to improve the standard PSO.Using four classic test functions to verify the improved PSO algorithm convergence,convergence of iterations and the convergence time shows that improved PSO is superior to standard PSO.The improved optimization algorithm model is established.Aiming BP neural network is easy to fall into local minimum,slow convergence and other shortcomings,this paper will optimize BP neural network by using the improved particle swarm optimization algorithm.Make use of a three-output function as bearing fault signal energy features,results show that iterative of the BP neural network improved by IMPSO is fewer and the prediction accuracy of neural network bearing failure is improved.Fault bearing diagnosis model based on neural network is established.The paper combine with the signal preprocessing model based on wavelet packet extracting energy of vibration signal and neural network fault diagnosis model to finish earing fault diagnosis.Calculations show that the optimized BP neural network can improve the accuracy of fault diagnosis and comprehensive performance of the algorithm is better than the standard BP neural network.Research of this paper provides new ideas and methods for bearing fault diagnosis using neural network.