A chromosome inspired fin structure for microchannel heat sinks: CFD driven XGBR, ANN, and MLR models for thermal and fluid flow prediction

Abstract

Microchannel heat sinks are eminent for thermal management in microscale systems. The performance of these systems is greatly enhanced through geometric optimization and advanced nanofluids. Using water and two nanofluids (SWCNT and SWCNT-Cu) at volume fractions of 0.01 and 0.02, this work examines four new chromosome-shaped pin fin topologies, both perforated and non-perforated, embedded within a microchannel. Nu, pressure drop, friction factor and overall thermal performance (OTP) are assessed through numerical simulations across a range of Reynolds numbers, from 150 to 350. Results show that perforated fins perform better than non-perforated ones. The top-view perforated chromosome fin utilizing SWCNT (φ = 0.02) improves its Nu by 25 % and significantly enhances thermal uniformity. Additionally, the nanofluids' enhanced flow distribution through perforations results in a pressure drop reduction of up to 39.79 %. Perforated designs demonstrate more uniform cooling and lower peak temperatures, according to thermal studies. Results indicate that the friction factor decreases with Reynolds number and remains largely unaffected by nanofluid type. It is influenced by fin geometry rather than perforations. Three ML models are trained using simulation data to forecast performance measures. Among them, the MLR model performs exceptionally well in estimating both pressures drop (R² = 0.9999, rRMSE = 0.74 %) and Nusselt number (R² = 0.9999, rRMSE = 0.26 %). These outcomes support the integration of ML based prediction and geometry optimization for effective and scalable MCHS design. Additionally, KFold cross validation has been conducted to find the mean CV score (k = 5). To confirm the performance evaluation and cross validation, hyperparameter sensitivity analysis has been executed as well.