Research On On-Line Estimation Methods For The Health State Of Aircraft Electromechanical System Components

The aircraft electromechanical system is one of the key systems of aircraft,and its safe operation is essential to ensure the normal flight of the aircraft and the safety of the crew.An important prerequisite for its safe operation is that the components of aircraft electromechanical system(hereinafter referred to as "aircraft electromechanical system components")are in healthy state.However,aircraft electromechanical system components have the characteristics of complex structure,frequent use,and time-varying operating conditions.When their health state degrades to a certain extent,it will lead to more serious damage to the aircraft,which will increase the operation and maintenance cost of the aircraft,or even lead to aircraft accidents.On-line estimation of health state is an effective way to evaluate whether the health state of aircraft electromechanical system components can meet the requirements of flight mission support,and to provide decision-making references for the condition-based maintenance of the aircraft electromechanical system components.It refers to the process of studying the mapping relationship between the state monitoring data and the health state characterization parameters,and the evolution law of the health state characterization parameters,so as to realize on-line estimation of the health state characterization parameters.During the on-line operation of aircraft electromechanical system components,the degradation process of their health state is often more complicated,which is mainly reflected in three aspects.Firstly,the health state characterization parameters are difficult to directly on-line monitor.Secondly,the operating conditions are time-varying.Thirdly,there is multiple failure mode coupling.The complicated degradation process makes it difficult to meet the requirements of flight mission support and condition-based maintenance of aircraft electromechanical system components.Therefore,this dissertation focuses on the research on on-line estimation methods for the health state of aircraft electromechanical system components.The main contributions are shown as follows.(1)Under the condition that the mechanism model can reflect the mapping relationship between the indirect monitoring data and the characterizing parameters of the component health state,there is instability of independent variable coefficients in the on-line indirect health indicator construction method based on multiple linear regression.To address this issue,a mechanism model-based feature fusion on-line indirect health indicator construction method is proposed.On the basis of extracting multiple performance degradation features from indirect monitoring data by using the method of mechanism model analysis,the elastic network regression method is utilized to realize the fusion of performance degradation features.Through its regularization process,the coefficients of independent variables are shrunk.And the influence of strong correlation between performance degradation features and health state characterization parameters on the coefficients of independent variables is reduced.Then the performance of the on-line indirect health indicators of the aircraft electromechanical system components under such condition can be improved.Experimental results show that the proposed method can realize the construction of on-line indirect health indicators with stable independent variable coefficients.The ability to characterize the health state of aircraft electromechanical system components under such condition is better than similar methods,and has better interpretability and accuracy performance.(2)Under the condition that the mechanism model cannot reflect the mapping relationship between the indirect monitoring data and the characterizing parameters of the component health state,it is difficult to determine the optimal subset of independent variables in the on-line indirect health indicator construction method based on multiple linear regression.To address this issue,a data-driven-based feature fusion on-line indirect health indicator construction method is proposed.On the basis of extracting multiple performance degradation features from indirect monitoring data by using data-driven method,the stepwise linear regression method is utilized to realize the fusion of performance degradation features.Through its statistical testing and screening process,the selection of the optimal subset of independent variables is realized.And the influence of correlation uncertainty between performance degradation features and health state characterization parameters on the independent variable set is reduced.Then the performance of the on-line indirect health indicators of the aircraft electromechanical system components under such condition can be improved.Experimental results show that the proposed method can realize the construction of on-line indirect health indicators with the optimal subset of independent variables.The ability to characterize the health state of aircraft electromechanical system components under such condition is better than similar methods,and has better interpretability and accuracy performance.(3)It is difficult to match the estimation model and indirect health indicators with sudden change in the health state on-line estimation method of strong tracking filter adaptive Wiener process for time-varying operating conditions.To address this issue,an health state on-line estimation method with the suppression ability of the influence of sudden change in operating conditions is proposed.Firstly,an adaptive health state on-line estimation model based on state space model is designed.The historical health indicator data are integrated into the on-line estimation model by using the adaptive Wiener model.On this basis,through the proposed saturation tracking filter method,the parameter sudden change of the health state on-line estimation model caused by the sudden change of operating conditions is suppressed.And the matching characteristic of the adaptive on-line estimation model and the indirect health indicators with sudden change can be maintained.Then,the performance of the adaptive health state on-line estimation can be improved.Experimental results show that the proposed method can realize the construction of an adaptive health state on-line estimation model with the ability to suppress the influence of sudden change in operating conditions.Compared with similar health state on-line estimation methods,it has the strong ability to suppress the influence of sudden change in operating conditions and small estimation errors.(4)It is difficult to match the estimation model with the slowly varying indirect health indicators in the health state on-line estimation method of Copula function for multiple failure mode coupling.To address this issue,an health state on-line estimation method with the suppression ability of the influence of slow change in multiple failure mode coupling is proposed.Firstly,a fusion health state on-line estimation model based on joint probability distribution is designed.The joint probability distribution function of indirect health indicators for multiple failure modes is solved by using the probability fusion ability of Copula function.On this basis,through the proposed linear scaling factor,the parameter deviation of on-line estimation model caused by the slow change of coupling relationship is suppressed.And the matching characteristic of the fusion on-line estimation model and the indirect health indicators with slow change can be maintained.Then,the performance of the fusion health state on-line estimation can be improved.Experimental results show that the proposed method can realize the construction of a fusion health state on-line estimation model with the ability to suppress the influence of slow change in multiple failure mode coupling.Compared with similar health state on-line estimation methods,it has the strong ability to suppress the influence of slow change in multiple failure mode coupling and small estimation errors.