Research On Remaining Service Life Prediction Method Based On Performance Monitoring Data Under Multiple Operating Conditions

Fault prediction and health management technology integrates safety,reliability,fault management,testability and cost analysis into a unified framework,which is of great significance to the maintenance of complex systems and equipment maintenance costs,and to improve the operation and maintenance efficiency.The remaining service life prediction is an important part of fault prediction and health management.Accurate remaining service life prediction plays an important role in preventing and reducing catastrophic accidents,improving equipment reliability and safety,guiding maintenance decisions and reducing costs.In this paper,the data-driven residual service life prediction method based on similarity is studied.After summarizing and analyzing the existing residual service life prediction methods,the improvement of similarity model and the establishment of similarity model under multi working conditions are studied respectively.Aiming at the complex equipment with multi condition performance monitoring data,the degradation process modeling method,the influence of similarity calculation,similarity matching and condition switching on the remaining service life are studied,and the condition information is integrated into the similar remaining service life prediction framework.(1)The definition of remaining service life is briefly introduced.This paper introduces the definition of remaining service life,the classification of remaining life prediction methods,common remaining life prediction methods and evaluation indexes,and expounds the basic principle and implementation method of remaining service life prediction method based on similarity model in detail.The principle of residual service life prediction method based on similarity model is described.(2)Processing of performance monitoring data.The performance monitoring data generated by the sensor are preprocessed by noise reduction and impurity removal.The sensor data series under different working conditions are extracted and standardized by Z-score.The differences caused by the running state are eliminated,which makes the sensor data comparable in different time periods.(3)An improved method of remaining service life prediction based on similarity model is proposed.After analyzing the principle of traditional prediction method based on similarity model,the idea of sliding matching is integrated into the life prediction of the method,which greatly improves the utilization of information and the accuracy of prediction.(4)This paper proposes a method to establish the residual service life prediction model based on similarity under multiple working conditions.This method assumes that the degradation process of equipment is composed of different working conditions,processes the data under different working conditions,and establishes the linear regression model under different working conditions.In the same data set,this method is compared with the traditional similarity method to verify the importance of the working condition in the prediction,which shows the effectiveness of the proposed modeling method.