Research On Filtering Fault Diagnosis Based Spatial Affine Contraction

In the study of filtering fault diagnosis methods,with the superiority of the set membership filtering method,scholars have been studying the filtering methods more and more deeply.In the fault diagnosis of systems with unknown but bounded noise,in order to accurately realize the system parameter estimation and state estimation,reduce the amount of calculation,and improve the efficiency of fault diagnosis,this paper studies the diagnosis methods of abrupt and slowly varying faults based on different spatial affine methods.It has important theoretical significance and remarkable application value for enriching and developing the filtering fault diagnosis method.The specific work of this thesis includes the following four aspects.1.For a system with unknown but bounded noise,a finite data window parameter estimation algorithm based on weighted ellipsoids is first studied.In order to reduce the amount of calculation,the idea of a limited data window is adopted,and a rolling data matrix is formed by the limited data window,which effectively reduces the calculation amount.Improve the data utilization rate and parameter estimation accuracy,and verify the effectiveness of the algorithm under different noise intervals and data window lengths through simulation.Then,we adopt the fault detection by detecting whether the feasible set of parameters is empty or not.The model matching algorithm is used to complete the weighted ellipsoid contraction based fault identification.2.Aiming at the problem of low convergence of the ellipsoid algorithm,a filtering fault detection algorithm based on zonotopes contraction was derived.By solving the intersection of zonotopes and bands,the fault detection problem is transformed into a computational problem of the outer boundary of the set membership.The method of minimizing the volume is used to calculate and update the zonotopes,and the smallest zonotopes is selected as the approximate outer boundary,to improve the convergence of the algorithm.By detecting whether the state feasible parameter set is empty,judging whether the system has a fault.3.For fault diagnosis of linear systems with unknown but bounded noise,a filtering fault diagnosis method based on orthotopic is proposed.This method uses linear programming equations to describe constraints in recursive operations.During the recursion,the spatial expression of the orthotopic is updated.The feasible form of the parameter is described by the spatial form of the orthotopic.Whether the system is empty determines whether the system is faulty or not,and adopts the method of model matching to realize fault diagnosis.Aiming at fault libraries with more fault types,the calculation amount based on the model matching method is reduced,and the fault diagnosis efficiency is improved,a fault diagnosis method based on the fault library layering is presented.4.For the problem of slowly varying parameters of input nonlinear system,a nonlinear fault diagnosis method for diagnosing the slowly varying parameters is proposed.The idea of linear programming constraints is used to recursively calculate the orthotopic to approximate the feasible parameter set.A fault diagnosis algorithm for global extended filtering of parameters is proposed.In addition,in order to reduce the amount of calculation and improve the efficiency of fault diagnosis,a parameter-oriented extended filtering fault diagnosis algorithm based on spatial dimensionality reduction is proposed.In summary,this paper mainly studies the problem of filtering fault diagnosis problem based on spatial affine contraction.The simulations verify the effectiveness and feasibility of these proposed algorithms.