Research On Flywheel System Health State Assessment Method Based On Interpretable Belief Rule Base

Flywheel as a key executive element of spacecraft attitude control,its performance and reliability directly affect the life and function of the spacecraft.If the flywheel loses balance during operation,it needs to be corrected immediately,otherwise the engine will be subjected to more and more serious vibration during operation,which will lead to a "shaft-burning" safety accident and even break related spacecraft components,threatening space safety.For this reason,it is of great significance for the spacecraft to evaluate the health status of the flywheel in a timely and effective manner.In the actual flywheel health assessment process,the interpretability of the assessment model should be ensured.This is because interpretable assessment analysis helps to guide the characterization engineering of the flywheel system,increase the transparency and credibility of the model,way and identify deviations,etc.,while avoiding many high-risk pitfalls.Therefore,this interpretable flywheel health assessment model study is of great theoretical and practical value.This research is based on belief rule base(BRB)as the theoretical foundation and focuses on three main aspects.For the problems of difficult construction of expert knowledge in flywheel health assessment model,difficult establishment of assessment indexes,and unclear logical relationships among indexes,a BRB model construction method based on fuzzy fault tree analysis(FFTA)is proposed.The method uses FFTA to analyze the mechanism of modeling,and uses FFTA to enrich the expert knowledge base in BRB to ensure the interpretability of the modeling process of flywheel system health state assessment model.The experimental part of the case study proves that the accuracy of the method reaches more than 95% through the comparison of different methods,which verifies the validity,feasibility and accuracy of this part of the study.For the problem of combinatorial explosion of flywheel health assessment model rules and failure to consider the reliability of rules,the BRB model inference method of combining intervals and considering the reliability of rules is proposed.The method replaces the traditional reference values of BRB with reference intervals,which makes the expert knowledge be expressed more completely.It replaces the rule fusion process in the form of Cartesian product combination rules with the form of interval addition combination rules,which solves the rule explosion problem caused by the exponential increase of reference values;it introduces the ER rule algorithm into the flywheel assessment model newly,replacing the original ER parsing algorithm,and further The reliability of the rules is considered.The accuracy of the method is proved to be 98.57%by the flywheel experimental case,and the comparison with other data sets proves that the method is general and has higher accuracy compared with other methods.In response to the problem that the parameters of the flywheel health assessment model based on BRB violate the actual physical meaning and lose interpretability in the optimization process,an interpretable constrained BRB model optimization approach is proposed.The method adds the interpretability constraint to the traditional optimization algorithm that makes the confidence level recognized by each expert and ensures that the optimized confidence distribution matches the actual system,which solves the problem of uninterpretability of the parameters after the inference process and ensures the interpretability of the optimization process of the flywheel system health state assessment model.The experiments prove that the mean square error value of the method is small compared with other methods,and the trend of the confidence distribution is well fitted,which indicates the feasibility and interpretability of the method.In summary,different countermeasures are proposed for the interpretability problems in traditional BRB from three perspectives,including modeling,inference,and optimization,respectively.From the theoretical point of view,the interpretability of the complete flywheel health condition assessment model is guaranteed,and the rationality of the scheme has been verified by experiments on flywheel health assessment.The experimental results show that the proposed model is feasible,accurate,and valid.