Title: Bond graph model based design of supervision algorithm for distributed fault tolerant control systems

Authors: A.K. Samantaray, Sanjoy K. Ghoshal, Saurav Chakraborty

Addresses: Department of Mechanical Engineering, Institute of Technology (IIT), Kharagpur, India. ' Department of Mechanical Engineering, Asonsol Engineering College, West Bengal, India. ' Tata consultancy Services (TCS), 54B Hadapsar Industrial Estate, Pune-411013, India

Abstract: For distributed analytical model based fault detection, some additional signals besides those required for distributed process control are required to evaluate residuals. This paper presents a graphical means of identifying those signals for design of a supervision system by analysing the causal structure in the bond graph model of a process. The global fault isolation scheme is initially implemented at a central monitoring system based on the alarm states generated at distributed controlling units. In safety critical systems, to minimise dependence on network communication, causality analysis is used to identify appropriate sensors that lead to information decoupling between various localised units of the process. Consequently, each controlling unit, called a smart station, is able to perform its own fault isolation and Fault Tolerant Control (FTC). The central monitoring system is used only for supervision of interfaces between different smart stations and operator mode management.

Keywords: bond graphs; distributed control systems; DCS; fault detection; fault isolation; FDI; smart station; service; operating mode; fault tolerant control; FTC; analytical redundancy relations; ARR; residuals; model based design; supervision; monitoring; safety critical systems; causality analysis.

DOI: 10.1504/IJAAC.2007.013296

International Journal of Automation and Control, 2007 Vol.1 No.1, pp.28 - 47

Published online: 19 Apr 2007 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article