Active Fault Detection?Fault Isolation And Fault Tolerant Control For Non-gaussian Stochastic Distribution Control Systems

Since the 1970 s,automation becomes popular on the development of complex system control and advanced intelligent control.It is focused on how to conduct fault diagnosis and fault-tolerant control for the system early and accurately.At the same time,the rapid development of the intelligent control theory provides a strong theoretical support for fault diagnosis and fault-tolerant control of complex systems.As a result,the actual controlled object is often affected by the outside world stochastic disturbance and stochastic parameter perturbation,so the hypothesis of Gaussian distribution for the controlled object is restricted.Therefore,the research on fault diagnosis and fault-tolerant control for the non-Gaussian stochastic distribution control system has important theoretical significance and wide practical prospect.Serious nonlinear,coupling and uncertainty exist in most of the industrial processes which make it difficult to establish accurate mathematical model of control systems.As T-S fuzzy model combined with the expert experience has very strong approximation ability,which has great superiority in dealing with complex nonlinear function.T-S fuzzy model can approximate the dynamic behaviors of nonlinear systems,extending the application of modeling and control methods for linear dynamic systems to complex nonlinear systems.In the coexistence condition of disturbance and multiple fault,the performance of control systems faces more problems and challenges.In this case fault detection and isolation is particularly important.In order to realize the detection of small fault and improve the accuracy of fault detection,the exploration and research on active fault detection method for the stochastic distribution control system is conducted.The chemical reaction process is conducted in this thesis.The RBF neural network is used to approximate the output of the probability density function(PDF)of the non-Gaussian stochastic distribution control system.In order to meet higher performance requirements and study more comprehensively for stochastic distribution control systems,the research on active fault detection,fault isolation,and the modelpredictive fault-tolerant control for non-Gaussian stochastic distribution control system based on T-S fuzzy model are given respectively.The concrete contents are shown as follows:(1)Active fault detection is conducted for the non-Gaussian SDC systems.The potential fault is motivated by the designed auxiliary signal so as to improve the fault detection quality.The active fault detection strategy is that the appropriate auxiliary signal is added into the system during the given test cycle.The system output residual is described by ellipsoid set using set-membership estimation method.By judging whether the output ellipsoid set intersection of the normal and fault system is null for fault detection.The influence of auxiliary input signal to the normal system output is reduced by the design of minimum energy auxiliary signal.The set-membership estimation method doesn't need the distribution information of the disturbance.It is assumed that the disturbance is unknown but bounded,which meets the case that non-Gaussian SDC systems don't need to obey the Gaussian distribution assumption.Considering the biggest uncertainties of the system,the ellipsoid set is updated,and then active fault detection is conducted by substituting the solved optimal auxiliary signal into the system.Finally,the computer simulation verifies the effectiveness of the proposed algorithm.(2)A T-S fuzzy model is applied to approximate the nonlinear dynamics of SDC systems,in which RBF is adopted to approximate the output PDF of non-Gaussian SDC systems.Considering the situation that disturbance and multiple actuator fault may occur at the same time,fault detection,isolation and estimation is conducted.The fault not to be isolated is regarded as disturbance,forming a augmented disturbance vector.The decoupling is realized by using non-singular linear coordinate transformation.In order to realize the decoupling,multiple transformation matrices whose number is the same as that of actuators,are designed respectively to divide the system into two subsystems.Thus one of the subsystems contains at most one actuator fault,and it is convenient for the fault isolation and estimation later.The earliest fault happened time and fault location are given.The adaptive observers for fault estimation are given respectively.Finally,computer simulation results show the effectiveness of the proposed control algorithm.(3)The model predictive fault-tolerant control is conducted for the SDC systems described by T-S fuzzy model.For a post-fault system,the quality of the products will be influenced and huge property losses will be caused without timely fault tolerant control.Thus,fault diagnosis and fault-tolerant control for the SDC systems is very necessary.Fault diagnosis is realized by designing residual and adjusting the observer gain and fault adaptive regulation rate.Based on the fault estimation information,the fault tolerant control strategy combined with model predictive control(MPC)algorithm is given for SDC systems.The system weight vector is predicted in the predictive time domain.By choosing proper objective function the optimal control input increment is solved in the control time domain.Thus the active fault-tolerant controller is constructed.Finally,computer simulation verifies the effectiveness of the proposed control algorithm.