Research On Life-cycle Maintenance Strategy Optimization Of Civil Aeroengine Fleet

The civil aero-engine,which is a multi-component power machine with high safety requirements,provides thrust and power for aircrafts.Currently,the single-centered maintenance policy is still adopted for most of civil aero-engines.The aero-engine maintenance optimization methods that consider the life-cycle maintenance strategy and fleet optimization deserve further studies.However,the complexity of the optimization problem would increase dramatically,when the maintenance optimization extend from single-centered to fleet-centered.To address the aforementioned problems,the maintenance optimization methods for civil aero-engine fleet are studied in this paper.The key technologies in single aero-engine and single-time maintenance optimization,the life-cycle maintenance policy optimizaiton method,and the maintenance policy optimization method for civil aero-engine fleet are studied.Firstly,the gas-path performance prediction model and fault diagnosis method for civil aero-engine are studied,to provide supports for the single-centered maintenance optimization of civil aero-engine.To address the time series characteristics and operating condition information,a gas-path performance prediction model is proposed based on the long short term memory(LSTM)deep learning network.To address the problems of small sample size and individual differences,a fault diagnosis method is proposed based on performance deviation model.The case studies show that the proposed gas-path performance prediction model obtains the better prediction accuracy,and the proposed fault diagnosis method is able to diagnose the civil aero-engine fault under the small sample condition.The after-maintenance performance prediction model of civil aero-engine provides the infrastructural support for maintenance optimization.The aero-engine after-maintenance performance is affected by the before-maintenance performance and maintenance levels.The aero-engine performance is represented by the time series;however,the maintenance level data are regarded as the discrete variables.For these characteristics,a simplified structure identification and mixed variables Takagi-Sugeno(SMTS)method is proposed,and an after-maintenance performance prediction model of civil aero-engine is proposed based on SMTS method.The proposed prediction model is able to forecast the after-maintenance performance time series under small sample condition.The case study shows that the SMTS model has the distinct advantages of prediction accuracy and stability.In engineering,the hybrid maintenance strategy is adopted to improve the aero-engine operational reliability.Thus,the long service-life and hybrid maintenance strategy should be considered synchronously in civil aero-engine maintenance policy optimization.This paper proposes an aero-engine life-cycle maintenance policy optimization approach that synchronously addresses the long service-life and hybrid maintenance strategy.The reinforcement learning method is adopted to illustrate the civil aero-engine optimization method.The Gauss-Seidel value iteration algorithm is adopted to optimize the maintenance policy.Compared with traditional aero-engine maintenance policy optimization methods,the long service-life and hybrid maintenance strategy can be addressed synchronously by the proposed approach.Compared with the single-centered aero-engine maintenance optimization method,the fleet resources and collaborative optimization of aero-engine maintenance policies should be addressed in fleet maintenance optimization method.As the spare aero-engine is the key fleet resource,a spare aero-engine demand forecasting model is proposed based on deep Croston method.The Croston framework is adopted to address the intermittent demand characteristic.The LSTM deep learning network is adopted in spare aero-engine demand interval forecasting and demand amount forecasting.Accompany with the proposed deep Croston model,a comprehensive intermittent demand forecasting evaluation method for spare aero-engine is proposed.The real-life spare aero-engine demand data of an airliner company are adopted to validate the proposed spare aero-engine demand forecasting model.The case study shows that the proposed deep Croston model achieves evident advantage.The fleet maintenance optimization method is more effective in rational allocation of the entire fleet resources and reducing the fleet operation and maintenance costs.In the civil aero-engine fleet,the maintenance plan of the single aero-engine has a certain impact on the maintenance policy of the entire fleet and directly determines the fleet operation and maintenance costs.As the civil aero-engine is with a long service life,it is necessary for the optimization method to address the long-term collaborative maintenance policy optimization.However,the civil aero-engine fleet maintenance policy optimization is characterized by high state space dimension,complex action set and multiple restrictions.To address the aforementioned characteristics,a civil aero-engine fleet maintenance policy optimization method is proposed based on multi-agent deep reinforcement learning approach.In the proposed optimization method,the convolutional neural network is adopted to perceive the fleet state and decide the maintenance action of the corresponding aero-engine.And the aero-engine fleet maintenance policy is obtained by the multi-agent framework.It is validated by the comparative simulations that the aero-engine fleet maintenance policy optimization method based on multi-agent deep reinforcement learning approach is able to obtain the superior maintenance policy.The research in this paper has theoretical significances and practical application values for enriching the civil aero-engine maintenance decision-making optimization technical system.The related maintenance optimization methods have a good reference value for the maintenance policy optimization of other complex equipment.