An artificial neural network model for predicting frictional pressure drop in micro-pin fin heat sink

Abstract

This study is part of two studies conducted for developing artificial neural-network-based tools for predicting the thermal and hydraulic performance of micro-pin fin heat sinks used for high-heat-flux electronic devices. The thermal design of high-heat-flux electronics requires a strong understanding of both pressure drop and heat transfer coefficient. Increasing the pressure drop in a cooling system significantly increases the required pumping power and decreases the system energy efficiency, in addition to considerably increasing temperature nonuniformity and causing reliability issues. Micro-pin fin heat sinks can help in the thermal management of high-heat-flux electronic systems owing to their effective heat transfer characteristics, namely, a large surface area and passage flow turbulence generation, and the requirement of lower pumping power compared with the microchannel heat sink. Studies conducted over the past decade have revealed that the thermal and hydraulic performance of micro-pin fin heat sinks are highly dependent on their geometric and operational parameters. However, a universal approach to predicting the frictional pressure drop, which influences the amount of power required, in pin fin arrays for various operating conditions and geometric shapes has not been developed so far. In this study, a trained artificial neural network (ANN) was used to develop a universal model for predicting the friction factor of micro-pin fin arrays. The friction factor correlation was predicted from 1,651 experimentally determined friction factor data points obtained from 22 studies. The relationship between a wide range of geometric and operating conditions and the hydraulic performance was investigated for accurately training the ANN. Furthermore, the universal model was analyzed by comparing values predicted by it with values obtained in other experimental studies. The model was found to show superior performance compared with other regression-method-based correlations.