Fault Diagnosis Of Stochastic Switched Nonlinear System

This paper studies the fault detection and fault estimation of a class of stochastic switched nonlinear system.Different filtering methods are used to obtain the state estimation of the system and achieve good results,and based on this,the residual signal is constructed to determine whether a fault has occurred and the fault amplitude is estimated.The main work is as follows:Firstly,the basic concepts of stochastic switched nonlinear system are studied,including: switching rules and subsystem state space equations.Run one of the subsystems at different times.As time goes on,different subsystems are stochastically switched according to the switching rules,and each subsystem is described by a nonlinear state space model.After that,the knowledge about fault diagnosis is studied,and the method based on analytical model is the main research focus.Secondly,the related theory of traditional Kalman filtering is studied.Then,based on its existing problems and limitations,square root unscented Kalman filtering(Square Root Unscented Kalman Filter,SRUKF)and square root cubature Kalman filter(Square Root Cubature Kalman Filter,SRCKF)are studied to make it possible to estimate the state of nonlinear systems.In order to solve the problem that the stochastic switched nonlinear system cannot determine which subsystem to run,interactive multiple model(Interactive Multiple Model,IMM)is studied to improve SRUKF and SRCKF.Interactive multiple model square root unscented Kalman filter(IMM-SRUKF)and interactive multiple model square root cubature Kalman filter(IMM-SRCKF)are proposed to estimate the state of the system.This algorithm can get more accurate system state estimation results,and can greatly reduce the complexity of directly filtering stochastic switched nonlinear system.According to the filtering result,the fault diagnosis of the stochastic switched nonlinear system can detect the fault occurrence and estimate the fault amplitude more accurately.After that,the advantages and disadvantages of the two algorithms are compared by simulation.Finally,because Kalman filter has a poor filtering effect on nonlinear non-Gaussian systems,the particle filter(Particle Filter,PF)method is studied for this kind of constraint condition.This method can achieve better results for nonlinear non-Gaussian systems.In view of the problems of the large number of collected particles and the degradation of particles in the traditional PF,the combination of UKF and PF is introduced.An unscented Kalman particle filter(Unscented Particle Filter,UPF)is proposed,which optimizes the selection of the importance density function,reduces the number of sampling points,and reduces the computational complexity.Next,for the problem of stochastic switched nonlinear system filtering,the IMM-UPF filtering algorithm is proposed by using IMM to improve the UPF algorithm,and the better state estimation results can be obtained.On this basis,fault diagnosis of the system under non-Gaussian noise can accurately detect the fault occurrence and estimate the fault amplitude.