Optimal designs of hybrid jet impingement microchannel heat sink across various pumping powers: A deep-learning approach

Abstract

Hybrid jet impingement/microchannel heat sinks (HJIMCHS) are an effective cooling technique for high heat flux electronics, providing efficient heat dissipation and enhanced temperature uniformity. To expand the applicability of HJIMCHS to a broader range of cooling challenges, it is crucial to optimize their heat transfer performance across a wide range of pumping powers ($\Pi$), rather than focusing on a single operating point. This study numerically solves the fluid flow and heat transfer characteristics of a silicon-based HJIMCHS with water as the coolant, and then the geometric variables of the HJIMCHS, including jet slot length ($L_{\mathrm{jet}}$), jet slot and channel width ($W_{\mathrm{jet}}$), jet slot height ($H_{\mathrm{jet}}$), and the number of microchannels (N), are simultaneously optimized over a broad range of $\Pi$ ($0.01W\le \Pi \le 0.15W$). The optimization process employs the non-dominated sorting genetic algorithm II (NSGA-II), powered by a surrogate artificial neural network model, and objective functions are total thermal resistance ($R_T$), maximum temperature difference on the bottom wall ($\mathrm{TD}$), and entropy generation rate (${\dot{S}}_{\mathrm{gen}}$). Optimal trade-offs between pumping power and the objective functions are explored by optimizing the HJIMCHS design at each pumping power. Results indicate that using ${\dot{S}}_{\mathrm{gen}}$ as an objective function across all $\Pi$ values may lead to inferior heat transfer performance, necessitating careful consideration. Additionally, with a heat flux of $200W/{\mathrm{cm}}^2$, for $\Pi =0.057W$, the optimal compromise solution results in $R_T=0.0963K/W$ and $\mathrm{TD}={2.9}^{°}C$, similarly for $\Pi =0.15W$ the results are $R_T=0.079K/W$ and $\mathrm{TD}={2.3}^{°}C$.