Multi-objective optimization of a microchannel heat sink with a novel channel arrangement using artificial neural network and genetic algorithm

Abstract

Effective cooling solutions become increasingly crucial as micro-electrical devices advance in complexity and miniaturization. Microchannel heat sinks (MCHSs) are one of the solutions, and researchers are continuously exploring innovative designs and optimization techniques to ensure that these cooling solutions meet the evolving demands of modern electronics. In the current research, a novel disk-shaped MCHS with strategically positioned baffles was selected to be investigated. Total thermal resistance (Rt) and pressure drop (ΔP) were scrutinized to assess the role of the baffle’s structural parameters (length of baffles in the first row (L), with the values of 500 μm–1500 μm, space between the baffles (S), with the values of 1700 μm–2500 μm, and angle of baffles (θ), with the values of 45°–90°). Firstly, various combinations of structural parameters of the baffles were selected as study designs, and each was numerically simulated. Artificial neural network (ANN) models and a genetic algorithm were employed to predict the data and identify the optimal designs. Based on the obtained results, higher R2 values (0.9940 for Rt Model and 0.9996 for ΔP model) indicated the accurate performance of both ANN models. Two optimal designs of optimal thermal performance (OTP) and optimal hydraulic performance (OHP) with separate objectives of minimizing Rt and minimizing ΔP were recommended. The utilization of baffles with the dimensions of L = 1500 μm, S = 2500 μm, and θ = 90° in the disk-shaped MCHS led to the attainment of the OTP. While to acquire the OHP, it was necessary to use the baffles with dimensions of L = 500 μm, S = 1700 μm, and θ = 81.8325°. In the OTP design, Rt was reduced by about 49.47 %, while in the OHP design, it decreased by approximately 30.36 % compared to the reference design.