Engine Multi-attribute Decision-making Evaluation And Application Research Under Imperfect Maintenance

As the main power source of civil aircraft,aero-engine is a complex machine with high requirements on safety and reliability,and its components are complex and interact with each other.In the current research on engine health management,there is a lack of methods to influence the periodic maintenance behavior during engine degradation.Meanwhile,the low utilization rate of multi-source monitoring data of engines will directly affect the scientific decision of aviation operators on the health management of engine fleet.Aiming at the above problems and requirements,the research objective of this paper is to establish an engine degradation model considering imperfect maintenance based on the multi-source monitoring data of the engine.Aiming at the actual demand of multi-source performance parameters and periodic imperfect maintenance behavior,this paper firstly observes the data characteristics in the engine's historical trajectory,and proposes the engine performance degradation model based on wiener process with drift points in order to guarantee the randomness of the model and the maintenance behavior.According to the characteristics of the engine low utilization rate of multi-source data,this paper puts forward a kind of hierarchical variable weight engine performance evaluation method: this method is first based on clustering algorithm on engine performance degradation stage,then based on the deep belief network DBN solving weights of each parameter,compared with other methods of multiple attribute assessment validation: principal component analysis method and ReliefF-PCA method.Finally,based on the analysis of the results of the example comparison,it is found that the evaluation method based on the hierarchical variable weight proposed in this paper can represent the health state of the engine to the greatest extent and get a better prediction result.Finally further build better fit the engine performance degradation model,factors affecting maintenance activities,based on Shared LSTM network layer model to modify engine maintenance factors of degradation model,and through the evaluation method based on hierarchical variable weight to data fusion of the performance parameters,get the revised engine performance degradation model;Through comparative analysis and verification,it is shown that the maintenance factor modified based on the Shared LSTM network layer model can improve the predicted value of the model.The research in this paper provides theoretical basis and technical support for enriching the engine fleet management and improving the scientific maintenance decision of aviation operators.Meanwhile,the established theoretical method is of great research value for the optimization of maintenance decision.