A vision of energy-based visualisation of large scale industrial systems for the purposes of condition monitoring

Abstract

Most large scale industrial systems can be viewed as processes that convert energy from one form to another. Energy is a universal quality of which the distribution holds meaning across physical domains i.e. thermal-fluid, mechanical, electrical and chemical. This paper presents a vision of using energy information of a process to visualise or characterise what is happening in the process with regards to energy. This energy information is then represented in a form that retains the structural information, making it possible to relate patterns in the representation to specific locations in the process. The energy attributes considered include energy flow rates between components of the system and the change in exergy flow rates across components. An energy representation of a process can form the basis for fault detection and diagnosis (FDD) or optimal control through the definition of a reference energy representation. The use of energy information linked to specific locations in the system, classifies the approach to FDD in the hybrid class which represents a hybrid between a pure data-driven and model-based approach. Attributed graphs are considered in this work as a structure that can describe the system’s energy attributes while retaining structural information. The usefulness of the energy representations for the purpose of condition monitoring is illustrated through a few case studies. The case studies include a heated two tank system, a gas to liquids process and a Brayton cycle. The results illustrate that energy-visualisation as a means to condition monitoring shows promise in terms of detecting and diagnosing typical faults in large scale industrial processes where energy conversion is the main concern