Research On Methods Of Fault Diagnosis For Nonlinear Systems Based On Analytical Models

Along with enlargement in the scale of production and increasing in the complexity of technological systems,how to improve the safety and reliability of control system has already aroused great concern.Fault diagnosis technology is an effective mean for the issue.Research on fault diagnosis has both theoretical and practical utility.The fault diagnosis method based on analytical models is the most in-depth and has excellent effect.At present,this method is mainly aimed at the linear system.However,the nonlinearity exists in practical plant,which makes fault diagnosis of nonlinear system based on analytical models as a main topic.According to the position of fault,system fault can be divided into component fault,actuator fault and sensor fault.They cover all possible faults in system.Thus,this paper presents fault diagnosis methods of these faults in nonlinear system based on analytical models.And a novel approach to fault diagnosis for multiple fault of actuator and sensor in nonlinear system is proposed.To improve the reliability of fault diagnosis,it analyzes the robustness.Moreover,the application of these methods to spacecraft control system is preliminarily explored in this paper.To fault diagnosis for component of nonlinear system,a method based on extended state observer(ESO)is proposed.Firstly,the nonlinear systems are transformed into the systems with triangular form.An ESO is designed to jointly estimate the states and the combination of uncertainty and nonlinear function of systems.Then,linearization of dynamic compensation is employed to linearize the systems via estimated value.Eventually the complex fault diagnosis problem of nonlinear system is simplified as one of linear system.An observer-based residual generator is designed on the basis of the linearized model for fault diagnosis,and threshold is committed.The proposed scheme can primely perform the task of fault diagnosis,and also can give a clear relationship between residual and fault.To fault diagnosis for actuator,sensor and multiple fault of actuator and sensor in nonlinear system,a method based on adaptive observer is proposed.Actuator faults and sensor faults are modeled respectively by unknown parameters in the state equation and the output equation in this paper.Firstly,the affine nonlinear systems are transformed into the particular systems via diffeomorphic transformation.Then,based on the techniques of high-gain observer,and defined the linear combinations of the state estimation error and the parameter estimation error,the derivative of parameter estimation error and the output estimation error,an adaptive observer structure is designed with simple method.Also an algorithm of the fault estimation is derived.Finally,the proposed fault diagnosis method is applied to generally affine nonlinear systems directly by the reversibility of aforementioned coordinate transformation.The robust of fault diagnosis based on adaptive observer is presented by threshold treatment technique.Based on the assumption that the interference is bounded,the threshold and diagnostic logic are given.In addition,the approach is applied to spacecraft control system.