Robustness and Accuracy of the Particle Swarm Optimisation Method in the Solution of Inverse Heat Transfer Problems with Phase Change

Abstract

Some engineering heat transfer problems cannot be solved directly as certain parameters influencing heat transfer are unknown. This is where the solution of inverse heat transfer problems might be utilized. The paper explores application of the Particle Swarm Optimisation (PSO) method to heat transfer problems with phase change. In order to assess the robustness and accuracy of the method, the study was conducted on simulation results. The explored case is an air-PCM heat exchanger. In the first step, the simulations with various values of parameters influencing heat transfer were performed. All cases were simulated as transient with time-dependent boundary conditions. In the second step, some of these parameters (phase change temperature, density, thermal conductivity, etc.) were assumed to be unknown and they were obtained from the simulation results with the use of the PSO method. The cost function was defined in the form of a root mean square error between the simulation results (the outlet air temperature from the heat exchanger) and the results of the optimized scenario. The number of unknown parameters varied from one to seven (in case of the description of the relationship of effective heat capacity on temperature, which was parameterized in the form of an asymmetric Gaussian function). The results have shown that the PSO is a robust and accurate tool with only up to 2 % difference in the enthalpy of fusion and no significant discrepancies during the higher dimension optimisation.