Research On Fault Prediction And Remaining Useful Life Analysis Methods For Turbofan Engines

Turbofan engines are the most commonly used engines in jet aircraft,featuring high efficiency,reliability,and energy-saving characteristics.They are one of the key technologies in modern aviation.During long-term operation,the engine may experience varying degrees of wear or aging due to high temperature,high pressure,high altitude,and high humidity.At the same time,the increasing complexity of turbofan engine structures resulting from advancements in the aviation industry leads to many uncertainties in manufacturing,operation,and maintenance.Currently,complex system modeling of turbofan engines is based on physical models,data-driven,and hybrid-driven methods.However,the modeling process does not take into account the correlation between monitoring data and the fusion of multivariate information,making it difficult to effectively reflect the credibility and accuracy of the data,leading to reduced interpretability.Therefore,it is necessary to adopt a method with better interpretability for turbofan engine performance evaluation and to estimate the performance degradation and remaining useful life(RUL)in the next stage to take appropriate fault repair or maintenance measures.Fault prediction and remaining life analysis of turbofan engines,as an important part of prognostics and health management(PHM),are key technologies for transitioning from preventive maintenance to predictive maintenance.To address these issues,this study uses the C-MAPSS turbofan engine degradation simulation dataset provided by NASA and employs evidence theory for fault prediction and remaining life analysis of turbofan engines.The research mainly includes three aspects:1.In complex system modeling of turbofan engines,there are usually problems with integrating qualitative knowledge and quantitative data,as well as correlations between variables from different sensors.This paper uses the Evidential Reasoning rule(ER rule)to fuse multivariate information,describing evidence correlation from the perspective of mutual information and quantifying evidence correlation using Copula entropy.This allows for performance evaluation of the turbofan engine and the construction of a one-dimensional composite health index.The Analytic Hierarchy Process(AHP)is used to demonstrate the rationality of the model,and Pearson correlation coefficient,Spearman correlation coefficient,and Kendall rank correlation coefficient are used to compare the correlation between the one-dimensional composite health index and the remaining life.Comparative experiments show that the health assessment model based on the ER rule has good adaptability and effectiveness.2.To address the fuzzy uncertainty and probability uncertainty issues in expert knowledge for turbofan engine fault prediction,a Belief Rule Base(BRB)based on fuzzy set theory and knowledge base for reasoning and the optimization algorithm with constrained optimization are used to deal with the probability uncertainty,fuzzy uncertainty and completeness of one-dimensional composite health index.The BRB follows Bayesian rules for evidence logic inference,ensuring the effectiveness and better interpretability of the turbofan engine fault prediction results.Expert knowledge is used to define the initial parameters and premise attributes of the BRB model,and the Projection Covariance Matrix Adaptation Evolutionary Strategies(P-CMA-ES)is used to optimize the reference values of the reference points,achieving more accurate prediction results.3.Regarding the issue of mutual influence of degradation features in the RUL prediction process of turbofan engines,this paper uses the evidence correlation ER rule to establish a one-dimensional composite health index.By analyzing the characteristics of the composite health index and combining it with the linear Wiener process,the remaining life prediction of the turbofan engine is achieved under the first passage time concept.Comparative experiments show that the proposed Wiener process-based turbofan engine remaining life analysis model has good adaptability and effectiveness.In summary,this paper addresses the issue of evidence correlation in traditional ER rules by using Copula entropy to quantify evidence correlation.The rationality of the proposed method was theoretically analyzed from the perspective of mutual information,and its effectiveness was verified in turbofan engine fault prediction and RUL analysis.