Regular Papers

International Journal of Control, Automation and Systems 2022; 20(8): 2759-2767

Published online July 12, 2022

https://doi.org/10.1007/s12555-021-0061-9

© The International Journal of Control, Automation, and Systems

Fault Detection Based on Graph Model for Dead Zone of Steam Turbine Control Valve

Yi-Jing Zhang, Ye-Yuan, and Li-Sheng Hu*

Shanghai Jiao Tong University

Abstract

The abnormal control valve dead zone in the steam turbine causes system oscillation severely, affecting the stable power generation of the power plant. In this study, a graph model is developed, and it can not only detect multiple faults at a time but also reduce the dependence on statistics. The graph topology of the steam turbine is built, including the process nodes and fault node indicating process variables and valve dead zone. Moreover, the graph convolution operation is conducted to classify the faulty and normal conditions. As a result, the simulation and experimental examples demonstrate that the accuracy of dead zone detection reaches 90% and 89% respectively, surpassing Multilayer Perceptron and Principal Component Analysis. The high-precision dead zone detection rate ensures that the method can be effectively applied to industry and improve economic benefits and operation safety.

Keywords Control valve dead zone, graph convolution, graph model, hidden fault node, steam turbine.

Article

Regular Papers

International Journal of Control, Automation and Systems 2022; 20(8): 2759-2767

Published online August 1, 2022 https://doi.org/10.1007/s12555-021-0061-9

Copyright © The International Journal of Control, Automation, and Systems.

Fault Detection Based on Graph Model for Dead Zone of Steam Turbine Control Valve

Yi-Jing Zhang, Ye-Yuan, and Li-Sheng Hu*

Shanghai Jiao Tong University

Abstract

The abnormal control valve dead zone in the steam turbine causes system oscillation severely, affecting the stable power generation of the power plant. In this study, a graph model is developed, and it can not only detect multiple faults at a time but also reduce the dependence on statistics. The graph topology of the steam turbine is built, including the process nodes and fault node indicating process variables and valve dead zone. Moreover, the graph convolution operation is conducted to classify the faulty and normal conditions. As a result, the simulation and experimental examples demonstrate that the accuracy of dead zone detection reaches 90% and 89% respectively, surpassing Multilayer Perceptron and Principal Component Analysis. The high-precision dead zone detection rate ensures that the method can be effectively applied to industry and improve economic benefits and operation safety.

Keywords: Control valve dead zone, graph convolution, graph model, hidden fault node, steam turbine.

IJCAS
June 2024

Vol. 22, No. 6, pp. 1761~2054

Stats or Metrics

Share this article on

  • line

IJCAS

eISSN 2005-4092
pISSN 1598-6446