Transient heat transfer analysis of a phase change material heat sink using experimental data and artificial neural network

Abstract

Transient thermal management of electronics is vital to increase their reliability. In the present study, heat transfer characteristics of a phase change material (PCM) heat sink was analyzed experimentally. Then, an artificial neural network (ANN) based on feedforward back-propagation multilayered perceptron (MLP), was utilized to predict transient heat transfer coefficient. Experiments were conducted at three different powers applied to the heat sink in the absence and presence of PCM. It was found that the use of PCM reduces the transient temperature while increasing the time to reach a steady-state temperature. Transient heat transfer during melting at a constant heat flux was characterized by dimensionless numbers including Rayleigh, Fourier and Stefan numbers. An optimal structure MLP network with 15 neurons in the hidden layer was obtained with trial and error method to predict the Nusselt number (Nu) during melting. The correlation analysis results based on ANN for predicting the Nu during PCM melting indicated the high accuracy of the neural network.