| Aero-engine is an important part of the aircraft,whose health state directly affects the normal operation of the aircraft and the safety of passengers’ life.The failure probability of gas system is the highest among the many subsystems of aero-engine.The health state prediction of aero-engine gas system plays an important role in the aero-engine health management.There are many difficulties during the prediction of gas system’s health state such as poor effective fault data because of the high reliability of aero-engine and the influence of monitoring environment.This paper proposes a health state prediction model of aero-engine gas system based on BRB to solve the problem that lacks of effective feature data in health state prediction,which can combine the expert knowledge and limited monitoring dataFirstly,the health states evaluation system of gas system is established based on the analyzing of gas system’s fault mechanism and working mechanism.The relationship between different health states is established based on MF,which can solve the problem that the attrbution of aero-engine gas system when it is in the critical state.Secondly,the CMA-ES optimization algorithm is used to optimize the parameters given by experts,which can overcome the subjectivity of expert knowledge.The working condition of aero-engine is complex.The heath state prediction model of aero-engine gas system under the single working condition cannot meet the accurate prediction of health state under multiple working conditions.The health state prediction model of aero-engine gas system considering working condition is proposed based on time domain feature analysis and BRB,which can improve the accuracy of the health state prediction model of aero-engine gas system under the condition switching.The simulation experiment is carried out to verify the rationality of the health state prediction model of aero-engine gas system.The software of aero-engine gas system health state prediction is designed and developed based on the health state prediction method proposed in this paper. |