Particle Filtering Based Fault Diagnosis Of Electro Hydraulic Position Servo System

Electro hydraulic servo system is a typical complex automatic control systemprocessing mechatronics and hydraulics integration, which is applied in variousoccasions, especially in large power to weight ratio, and irreplaceable in fast responsespeed conditions. Because the complex，high precision and highly integrated of thecomponents resulting in processing difficulty and poor ability in resisting pollution,consequently, the fault is inevitable, which enable its fault diagnosis very importantsignificance to conduct research on. The fault diagnosis method based on the analysismodel has been widely used in the fault diagnosis problem of electro-hydraulic servosystem, but at present this kind of method is mostly linearized the near operating pointof the system to realize the fault diagnosis. As the electro-hydraulic servo system is astrong nonlinear system in essence, inevitably, this method will affect the accuracy offault detection and diagnosis. Considering the deficiencies of existing methods and thesuperiority of particle filtering method in the treatment of non-Gauss and nonlinearproblem, this paper proposed the fault diagnosis method that based on particle filteringto the electrohydraulic servo system. This paper focused on how to accomplish themethod of fault diagnosis of electro-hydraulic servo system based on particle filtering,the main research contents and conclusions are as follows.Firstly, the existing method of fault diagnosis of electro-hydraulic servo systemand its strengths and weaknesses were described, meanwhile, the research status ofthe improved particle filter algorithm and its application in fault diagnosis weresummarized;Secondly, the basic principle of particle filtering was discussed,then, acomparative study of filtering estimation performance of the standard particle filtermethod and the Extended Kalman filter, Unscented Kalman filter was carried out. Theresults suggest that, regardless of the nonlinear Gauss model and nonlinear non Gaussmodel, filtering accuracy of the particle filtering method was higher than that of thelatter two traditional filtering methods;Thirdly, the results was concluded that fault detection method based on stateestimation and residual smoothing is superior to fault detection method based on thelikelihood function by a contrastive study of the performance of two kinds of detectionmethod based on particle filtering; At the same time, the fault identification method based on information divergence was presented to resolve the problem that fault typeis tough to identify only though residual error,and the validity of this method wasverified by the simulation results;Fourthly, the electro-hydraulic position servo system as the research object wastreated, and its nonlinear model was established; Additionally, research on the faultdetection method based on particle filtering for state estimation and residualsmoothing and recognition method based on information divergence were conductedto solve the detection and identification of the typical fault system; Similarly, thesimulation results verified that the two methods can respectively detect the fault andidentify fault type accurately and timely;Lastly, experiments were conducted to study the hydraulic cylinder leakage faultby fault detection method based on particle filter and fault identification method basedon information divergence; The experiments results verified that the two methods arepractical and effective.