A Novel Machine Learning Based Fault Diagnosis Method for All Gas-Path Components of Heavy Duty Gas Turbines With the Aid of Thermodynamic Model
作者
Jingchao Li,Yulong Ying
出处
期刊:IEEE Transactions on Reliability [Institute of Electrical and Electronics Engineers] 日期:2024-04-15卷期号:73 (4): 1805-1818被引量:14
标识
DOI:10.1109/tr.2024.3383922
摘要
Heavy-duty gas turbines are key engines for clean energy utilization and efficient conversion in natural gas power plants. Gas-path components are the components with the highest failure rate in gas turbines, and their faults are highly hidden and destructive. In response to the shortcomings of existing gas-path diagnostic methods, a machine-learning-based diagnostic method for all gas-path components with the aid of thermodynamic model was proposed for the first time. A comprehensive rule base was established for the relationship between the internal fault modes of gas-path components and the external fault symptoms of gas-path measurable parameters. A mathematical model for all gas-path component fault diagnosis suitable for machine learning framework was established. The proposed method can be used to comprehensively diagnose the different types and severity of faults in all gas-path components under various operating conditions after grid connection. Case analysis shows that the proposed method can achieve a success rate of 100% for diagnosing different types of faults and can achieve an overall success rate of over 97% for diagnosing the types and severity of faults under a few base sample conditions. and the accuracy of fault diagnosis has increased at least by 3.4%. The proposed approach has excellent diagnostic accuracy and real-time performance.