TY - JOUR
T1 - Investigation on the thermal control and performance of PCM–porous media-integrated heat sink systems
T2 - Deep neural network modelling employing experimental correlations
AU - Rehman, Tauseef ur
AU - Sajjad, Uzair
AU - Lamrani, Bilal
AU - Shahsavar, Amin
AU - Ali, Hafiz Muhammad
AU - Yan, Wei Mon
AU - Park, Cheol Woo
N1 - Publisher Copyright:
© 2023 Elsevier Ltd
PY - 2024/1
Y1 - 2024/1
N2 - Phase change material (PCM)-based heat sinks can offer reliable and effective thermal management (TM) solutions for increasingly sophisticated applications. A critical aspect of such heat sinks is determining how long it takes them to reach a set-point temperature. However, no generalised method exists in the literature that can predict and interpret the thermal performance of a wide range of PCM–porous media-integrated heat sinks. In this regard, this study examines the heat transfer characteristics of PCM-based heat sinks integrated with various metallic foams through experimental and deep learning (DL) techniques. The experiments are performed for transient TM analysis of various PCM-based heat sinks. Diverse variables, including foam porosity (0.95–0.97), PCM fraction (0.6–0.8), heat flux (0.8–2.4 kW/m2), foam materials (Fe–Ni alloy, Ni and copper) and PCM type (RT-35HC, RT-44HC, RT-54HC and paraffin wax), are investigated in this study. The experimental data are fed to the optimal DL model using the Bayesian surrogate model-tuned hyperparameters. Utilising a correlation analysis, as exemplified by the heat map and correlation plot, in conjunction with explainable artificial intelligence, it has been deduced that the thermal performance of the heat sink is principally influenced by factors such as PCM type, PCM fraction, foam material, foam porosity, and heat flux. Comparing the model's predicted data with the empirical findings, a good agreement was observed. Specifically, the mean absolute error (MAE) for the anticipated temperature and gradient registered at 0.0438 and 0.0054, whilst the mean square error (MSE) manifested values of 0.0579 and 0.0087, respectively. The proposed model can accurately assess the heat sink's thermal performance (correlation coefficient, R2 = 0.99) for various PCM types, fractions, foam materials, applied heat flux and foam porosity.
AB - Phase change material (PCM)-based heat sinks can offer reliable and effective thermal management (TM) solutions for increasingly sophisticated applications. A critical aspect of such heat sinks is determining how long it takes them to reach a set-point temperature. However, no generalised method exists in the literature that can predict and interpret the thermal performance of a wide range of PCM–porous media-integrated heat sinks. In this regard, this study examines the heat transfer characteristics of PCM-based heat sinks integrated with various metallic foams through experimental and deep learning (DL) techniques. The experiments are performed for transient TM analysis of various PCM-based heat sinks. Diverse variables, including foam porosity (0.95–0.97), PCM fraction (0.6–0.8), heat flux (0.8–2.4 kW/m2), foam materials (Fe–Ni alloy, Ni and copper) and PCM type (RT-35HC, RT-44HC, RT-54HC and paraffin wax), are investigated in this study. The experimental data are fed to the optimal DL model using the Bayesian surrogate model-tuned hyperparameters. Utilising a correlation analysis, as exemplified by the heat map and correlation plot, in conjunction with explainable artificial intelligence, it has been deduced that the thermal performance of the heat sink is principally influenced by factors such as PCM type, PCM fraction, foam material, foam porosity, and heat flux. Comparing the model's predicted data with the empirical findings, a good agreement was observed. Specifically, the mean absolute error (MAE) for the anticipated temperature and gradient registered at 0.0438 and 0.0054, whilst the mean square error (MSE) manifested values of 0.0579 and 0.0087, respectively. The proposed model can accurately assess the heat sink's thermal performance (correlation coefficient, R2 = 0.99) for various PCM types, fractions, foam materials, applied heat flux and foam porosity.
KW - Bayesian optimisation
KW - Deep learning
KW - Heat sink
KW - Phase change materials
KW - Thermal management
UR - http://www.scopus.com/inward/record.url?scp=85178155749&partnerID=8YFLogxK
U2 - 10.1016/j.renene.2023.119719
DO - 10.1016/j.renene.2023.119719
M3 - Article
AN - SCOPUS:85178155749
SN - 0960-1481
VL - 220
JO - Renewable Energy
JF - Renewable Energy
M1 - 119719
ER -