Graph matching as a means to energy-visualisation of a counter-flow heat exchanger

Abstract

The objective of this study is to develop an energy visualisation of a counter-flow heat exchanger by making use of graph matching. The energy visualisation developed should be suitable for the purpose of fault diagnosis. Since energy is a multi-domain parameter, it is considered as an ideal approach for fault diagnosis. The heat exchanger model used in this study is based on the gas cooler of an operating CO2 heat pump test bench at the North-West University. A simplified model is developed in Flownexr; a simulation environment that can be used to simulate flow and heat transfer systems. Experimental data have been used to validate the components available in Flownexr. The model is adjusted to incorporate faults by adding the necessary Flownexr components. The faults concerned include a fluid leak, heat leakage and fouling. Energy can be seen as a unifying concept convenient to address multi-domain systems. System information can be reduced to the essentials if energy is used as modelling parameter. The concepts of exergy and energy flow rate are used to fully represent the energy of the heat exchanger. Energy information is related to attributed linear graphs and graph matching is applied on these graphs. The resulting outputs include a permutation matrix, a cost matrix and a distance parameter. Graph matching is a technique that describes how similar two graphs are. Therefore, the three outputs can be viewed as a description of how similar the energy information contained in two graphs are. Visualisation is achieved by computing and plotting the eigenvalues of the cost matrix. Using this visual presentation, a procedure to identify which of the three faults has occurred, is developed. The procedure is successful in identifying a fluid leak, heat leakage and fouling. This study confirms that the energy visualisation of a counter-flow heat exchanger can be achieved by making use of graph matching. The use of energy reduces the number of parameters considered and makes it possible to treat the fluid and thermal domains found in a heat exchanger in a similar fashion. This study shows that energy visualisation and graph matching are suitable for the purpose of fault diagnosis in a practical system