Research On Key Technologies Of Gas Turbine Prognostics And Health Management

In order to realize the maintenance mode of gas turbine transform from regular maintenance to conditional maintenance,and to ensure the safety and reliability of the operation process of gas turbine system,it is necessary to realize health monitoring during the whole life of gas turbine.Therefore,it is necessary to study gas turbine prognostic and health management technology.Based on the research results of domestic and foreign scholars,this paper studies the key technologies involved in the gas turbine prognostic and health management system.Firstly,the overall structure of the gas turbine prognostic and health management system was studied.The basic concept and functional architecture of PHM system are studied.The abnormal working mode of gas turbine and the decomposition method of complex system according to function and structure are studied.The overall structure of gas turbine prediction and health management system is designed and its main components are analyzed.Secondly,the data-driven kernel principal component analysis method is studied.The steps of fault detection and isolation using KPCA are summarized.The typical fault diagnosis and isolation of gas turbine sensors are realized by KPCA.Matlab/Simulink simulation results show that the KPCA method can effectively detect the sensor constant deviation fault,drift fault and intermittent fault.Based on the model-based fault diagnosis method,the gas turbine gas path fault diagnosis system model is constructed,and the overall structure and design method of gas turbine gas path fault diagnosis system are proposed.The fault detection and faults of neural network and multi-hypothesis test are introduced in detail.The identification method and simulation verify the effectiveness of the two algorithms.Based on the analysis of the advantages and disadvantages of the two algorithms,an algorithm fusion method is designed to make up for the shortcomings of the above two diagnostic methods.The simulation shows that the accuracy of fault diagnosis of the diagnostic system is improved.Next,the measurement parameter trend analysis method is studied.The specific implementation steps of the parameter trend analysis method,the single-parameter trend analysis and the trend analysis of the relationship between parameters,the abnormal mode of the parameter trend and the application of the fingerprint analysis method for fault diagnosis are studied,and the advantages and disadvantages of the parameter trend analysis method are summarized.Finally,aiming at the health management simulation platform built,a unified standard for evaluating fault diagnosis algorithm is proposed.Based on the performance of the algorithm,the performance evaluation criteria of the fault diagnosis algorithm are proposed,the relationship between the above performance evaluation index and the threshold is analyzed,the calculation method of the evaluation criterion is given,and the accuracy confidence intervals of the above evaluation indexes are given.