Reflection on the energy graph-based visualisation approach to FDI of large-scale industrial systems

Abstract

Modern industrial systems are complex and large-scale, constituting multiple components and interconnections. These components interact through material and energy interfaces. Energy can be considered as a universal quality that allows system characterisation across different physical domains (thermal-fluid, mechanical, electrical and chemical). Therefore, an energy representation of a system can be used as a basis for fault detection and diagnosis. In order to cope with representing complex interconnections of systems, graphs can assist in presenting the structure and attributes of the system components. This paper reviews FDI applications using energy and graph-theoretical methods and then focuses on a unique approach that combines the strengths of both approaches. It furthermore specifically exploits the attributed graph as mathematical formalism for characterising an industrial process in terms of energy. The graph formalism allows capturing energy characteristics while retaining the structural information, i.e. linking the energy attribute to a physical location in the process. This approach is named the energy graph-based visualisation approach. The contribution of this paper lies in the portrayal of the full research endeavour for the purposes of comparison and future reference. Two applications are presented to illustrate the usefulness of the approach namely a gas-to-liquids process and a practical heated two-tank system.