A Method Of Rolling Bearing Life Prediction Based On Multi-Strategy Coebolution Particle Filtering

It is very important to study the life prediction technology of rolling bearing to ensure the safety and maintenance decision of mechanical equipment.For researching the life prediction method of rolling bearing which is working in the actual working conditions,some in-depth research has been make in three aspects,such as,signal blind source separation,feature extraction and life prediction.A method of multi-strategy coevolution particle filtering algorithm has been put forward by me,and it is applied to rolling bearing life prediction.It is proved by experiment that this method can improve the precision and accuracy of bearing life prediction.The main contents of this paper are as follows:1.The importance of signal blind source separation are introduced in the actual working conditions.Using Blind Source Separation(BSS)algorithm based on ASTFA to separate the experimental signals.2.A new characteristic index named Weighted Fusion Index(WFI)is proposed.It is made from two procedures.One is evaluating the tracking abilit y of every original characteristic index.Another is clustering the chosen characteristics to one characteristic index according to their weight.So this new characteristic has the good performance of tracking bearing decline trend.3.WFI is put in Paris-Erdogan model.First of all,initializing the parametersvof state model,then using particle filter algorithm updates the model in real time,finally according to the degradation of system state achieves predicting residual life of the bearing.The effect of particle filter on the life prediction of rolling bearing is verified by experiment.4.An improved particle filter algorithm named a method of multi-strategy coevolution particle filtering algorithm is proposed.Aiming at the problems of particle degradation and low diversity in the particle filter algorithm,I improve the resampling process,and put forward an adaptive fitness function which can guide the resampling particle samples to the position where the after posterior probability density has high value.At the same time,this algorithm can make the particles evolve through several different strategies,then it can improve the weights of particles and particle diversity,so it can improve the accuracy of the algorithm.Finally,it is verified by experimentthe improved particle filter algorithm has better precision and accuracy than particle filter algorithm.The study on rolling bearing condition monitoring and fault diagnosis techniques is of crucial importance for assuring the operation safety of mechanical equipment.The vibration signals of rolling bearing are non-stationary and non-linear.Meanwhile,the weak fault feature of the rolling bearing can easily be submerged by the noise,especially in the case of intensive background noise.