A coupling diagnosis method of sensors faults in gas turbine control system. (15th August 2020)

• Record Type: Journal Article
• Authors: Sun, Rongzhuo; Shi, Licheng; Yang, Xilian; Wang, Yuzhang; Zhao, Qunfei
• Is Part Of: Energy. Volume 205(2020)
• Publication Date: 2020-08-15
• Subjects: Gas turbine -- Engine health management -- Sensor fault diagnosis -- Wavelet energy entropy -- Support vector regression; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2020.117999
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: Gas turbines usually operate under complex conditions, such as frequent start-stop, complex environment (dust, salt fog). There are many sensors equipped in a gas turbine for the sake of monitoring and control. The sensors may fail to output normal signals since working continuously for a long time and in the harsh conditions. To avoid misjudgment of gas turbine control system due to sensors' failures, it's necessary to diagnose the sensors faults from the output signals beforehand. In this paper, a coupling diagnosis method of sensors faults in gas turbine control system based on machine learning was proposed. We coupled the wavelet energy entropy (WEE) and support vector regression (SVR) for sensor fault diagnosis where WEE was used to extract the signals features and SVR was used to classify the types of faults. A sensors faults database with five typical types was built by using the experimental data of a 7000 kW gas turbine under different operating conditions to verify the accuracy and effectiveness of the proposed coupling method. The results show that the accuracy of the coupling method is more than 90% with a shorter diagnosis time. Highlights: A coupling fault diagnosis method of sensors for gas turbine is presented. Wavelet analysis is used to extract the signals features. Support vector regression is used to diagnosis signals. The accuracy of this coupling method is more than 90%. With the same accuracy, this method can greatly shorten the diagnosis time.