Research Based On Intelligent Particle Filter For Nonlinear System Fault Detection

With the rapid development of modern industry,the development direction of machine production gradually towards large-scale,automated and highly complex,which leads to a situation that there is more and more complex nonlinear system existing in the industrial field.In the meanwhile,the precision of the modern industrial filter requirements are constantly improved.Because the emergence of more and more nonlinear system models in industrial production process,the Kalman filter algorithm,which plays an important role in the state estimation of linear systems,can no longer meet the development trend of modern industry.As a new filtering technique of nonlinear system,particle filter has a good application prospect in solving nonlinear and non Gauss system problems.However,the traditional particle filter technology has its own shortcomings: particle degeneration and loss of particle diversity,which leads to inevitable deviation in the nonlinear system state estimation.As a result,the traditional particle filter algorithm can not meet their needs in some high precision and high requirements of industrial application.So how to mortify the traditional particle filter algorithm and make it colud satisfy the needs of modern industrial production has become a research trend.In this paper,at very first a fault detection method based on the traditional particle filter algorithm is proposed,and the fault monitoring of the nonlinear system is realized.Then in order to compensate for the above shortcomings existing in the traditional particle filter algorithm,this paper proposed the intelligent particle filter algorithm,the intelligent particle fi lter algorithm needs to verify its reliability of the estimation of the system state,and also to test its diagnosis effect.The proposed intelligent particle filter algorithm is inspired by the genetic algorithm.Because the traditional particle filter algorithm suffers from the particle degeneration and loss of particle diversity and the intelligent particle filter algorithm is combined by the traditional particle filter algorithm and genetic algorithm to alleviate the loss of particle diversity problem an d could improve the particle degradation.In the intelligent particle filter algorithm,we introduce the selection,crossover and mutation operations from genetic algorithm before the particle filter resampling process to avoid getting rid of those particles with smaller weight directly,which could maintain the diversity of particles.It could help improve the accurate estimation of system state and fault diagnosis.What's more,in order to verify the reliability of intelligent particle filter algorithm,thi s paper will apply these two kinds of fault detection methods into the numerical simulation cases and three tanks model,and compare the results of two kinds of filtering algorithm from different simulation test.