A Study Of Condition Assessment And Prognosis Based On Synthesized Health Indicators For Marine Power System Failure

With the development of modern technology and the improvement of ship automation and intelligence,the operation and maintenance(O&M)process of marine systems and equipment are facing new challenges.Ship management means relying on human experience and knowledge could no longer meet the O&M needs of intelligent ships.Based on the comprehensive equipment monitoring,evaluation,diagnosis,and prediction technology,health management based on a more active maintenance strategy has gradually become a new trend of intelligent operation and maintenance.Therefore,as an essential means of health management,the research on health status evaluation and life prediction of the multi-state system has important theoretical significance and practical application value.Focusing on the practical engineering problems and internal scientific problems of ship multi-component state systems,such as performance state degradation and uncertainty in the process of equipment operation,combined with the theory of modern evaluation and prediction methods,this thesis researches system-level health condition assessment and prediction,to complete the application research of health condition identification and prediction in the whole life cycle and realize the health management of the system.The main contents of the dissertation are as follows:(1)Analysis of influencing factors of system structure,function,and health conditionFirstly,the characteristics of ship system structure and classification are analyzed.Secondly,the internal and external uncertainties affecting the health condition are studied.Then the function analysis method is proposed to realize the analysis of system function.Finally,the fuel oil supply system is selected as the research object of this thesis,and the system principle and state parameters are introduced.(2)The adaptive threshold and baseline models of system parameters are establishedAffected by the uncertainty in the whole life cycle,the health condition of the system changes gradually,and the reference standard of parameters(baseline and threshold)should vary accordingly.To address the problem,adaptive baseline and threshold models based on data-driven methods are adapted to provide more accurate references for evaluation and prediction.(3)System health condition assessment based on cloud barycenter modelThe connotation and concept of health indicator(HI)for system health condition assessment are proposed,and the indicator system is established.The potential law and knowledge of data,human subjective factors,and mechanism analysis are fused by utilizing the improving cloud barycenter model.The advantages of qualitative knowledge and quantitative data are integrated,and the characteristics of the correlation and influence between multidimensional factors and components of the system are considered.The dynamic update of each indicator weight is accomplished by using the comprehensive weight method,and the changes in system health conditions are effectively tracked.(4)System failure prediction based on synthesized health indicator(HI)The connotation of system failure prognosis is expounded,and the system failure HI is constructed.Combining the fractal theory and dynamic principal component analysis(DKPCA)method,the thesis puts forward the fusion method based on adaptive extraction of system features.The stacked autoencoder(SAE)method is applied to obtain the mapping relationship from multi-dimensional features to one-dimensional,to retain the health status information of the system as much as possible.Aiming at the phenomena of global degradation,local selfhealing,and local interference in the ship system,the variational mode decomposition method is proposed to decompose the health indicator.According to the different characteristics of the decomposed signal,the hybrid failure prediction that can improve the prediction accuracy is adopted to track the health condition,which solves the problem that different prediction models are suitable for diversedata characteristics.(5)System application case studyBased on the above research results,taking the ship fuel oil supply system as the object,the application verification of system health condition assessment and prediction is carried out.Combined with the verification results,we find that the continuous dynamic assessment and prediction can better quantify the health condition of the system,especially for the ship system with obvious self-healing function.The theories and methods studied in this thesis supply a quantitative benchmark for the decision-making for ship managers,provide an effective solution for ship intelligent operation and maintenance,and have engineering application value.