Session: 05-05 Data-Driven Methods and AI for Diagnostics

Paper Number: 123564

123564 - Gas Turbine Gas-Path Fault Diagnosis Based on Decision Fusion of Model-Based and Data-Driven Methods 

Model-based methods and data-driven methods are currently the primary approaches for gas turbine gas-path fault diagnosis. However, the accuracy of model-based methods can be compromised due to low-quality models, while data-driven methods struggle to handle untrained fault types. To solve this problem, this study proposes a fusion framework based on Bayesian network for decision fusion of model-based and data-driven methods, aiming to enhance the reliability of the gas-path fault diagnosis system. Initially, a model-based diagnostic system is constructed using the component-level model of the gas turbine and the unscented Kalman filter. Following this, a data-driven diagnostic system is developed using historical gas turbine operational data and corresponding machine learning algorithms. The outputs of each diagnostic system are then unified into a common output format. In the next step, reasonable decision objectives are defined and the Bayesian network parameters are solved based on these objectives, enabling the construction of the decision fusion system for gas-path fault diagnosis. Finally, the fusion diagnostic results are compared with those obtained from the individual diagnostic systems. The results indicate that decision fusion effectively combines the strengths of each method, leading to improved reliability in gas turbine gas-path fault diagnosis. This research has significant importance for the performance monitoring and maintenance of gas turbines.

Presenting Author: Xianda Cheng Shanghai Jiaotong University

Presenting Author Biography: Xianda Cheng received the B.S. degree in Mechanical Engineering from Shanghai Jiao Tong University, Shanghai, China, in 2020.He is currently working toward the Ph.D. degree in Engineering Thermophysics with the Department/School of Shanghai Jiao Tong University, Shanghai, China. His research interests include engine-level performance simulation and gas path fault diagnosis.

Authors:

Xianda Cheng Shanghai Jiaotong University
Haoran Zheng Shanghai Jiao Tong University
Wei Dong Shanghai Jiao Tong University