Research On Fault Detection Methods Based On State Observer For Nonlinear Systems

With the development of science and technology,the control system becomes more and morecomplicated,and then inevitably there may be faults.In the event of a fault,it will affect the quality of the product,reduce the production efficiency,and even cause property damage and life-threatening.Therefore,fault detection technology emerges as the times require.Through the accurate fault detection method,the purpose of detecting sudden faults in time is of great significance for timely stop loss.With the deepening of research by experts and scholars at home and abroad,the fault detection method for linear systems has approached maturity.However,in the actual industrial production process,most control systems exhibit different degrees of nonlinearity.At this time,some fault detection methods of linear systems are no longer applicable,which forces us to study the fault detection for nonlinear systems.With the research of scholars for many years,the fault diagnosis technology of nonlinear systems has made great progress.However,for different degrees of nonlinear systems,for example,discrete nonlinear systems with bounded disturbances and sensor faults,there is no specific analytical solution for fault detection and fault estimation based on state estimation observers,therefore,they cannot achieve accurate fault detection and estimation;For nonlinear systems with unknown inputs and constant actuator faults,the construction of fault estimation observers and the simplification of parameter solving are all difficult issues to be solved;For a class of nonlinear unstructured model uncertain systems with sensor faults,considering the existence of both system inputs and state variables,and partial state-measurable nonlinear cases,the research of fault detection methods becomes extremely challenging.In this paper,the nonlinear control system is taken as the research object,and the main tools and means such as state estimation observer,unknown input observer and linear matrix inequality are used to design the sensor or actuator fault detection method,and the effectiveness of the method is verified by simulation experiments.The main research contents of this paper are as follows:?1?For discrete nonlinear systems with bounded disturbances and sensor faults,a fault detection and estimation method based on state estimation observer is proposed.We construct a suitable state estimation observer.When building the system model with sensor faults,the sensor fault is transferred to the state equation,and then the H₂ optimization technique is used to reduce the influence of the interference input on the residual.The parameters of the observer can be solved by the method.By designing the residual generator,the residual signal and the sensor fault are in one-to-one correspondence,thereby detecting the sensor fault.?2?For nonlinear systems with unknown input and constant actuator failures,a fault estimation method is proposed by constructing an unknown input observer.For a Lipschitz nonlinear system with unknown input and constant actuator fault,we construct a full-dimensional unknown input observer,and give a sufficient condition for the existence of nonlinear unknown input observer?UIO?,and then by means of LMI,the solving problem of nonlinear matrix inequality is transformed into the solving problem of LMI.The LMI toolbox is used to solve the nonlinear UIO unknown gain matrix,which simplifies the computational complexity.Combined with the parameter estimation method,a parameter estimation observer is established to estimate the constant actuator fault in the system.?3?For a class of nonlinear uncertain systems with sensor faults,considering the nonlinear cases of system input and state variables and partial state measurable,a fault detection method based on adaptive threshold is proposed.Firstly,in order to solve the partial state measurable problem in the system,a nonlinear estimator with sensor fault is designed by introducing linear transformation.Secondly,an additional nonlinear term is introduced in the process of modeling.And then the corresponding residual generator is designed according to the system model,and the residual generator generates residuals for detecting sensor faults.Finally,the adaptive threshold selection strategy is designed,and the sensor fault is detected by the generated residual signal.