Optimization of a Compact Manifold Heatsink in External Airflow for Power Devices Using a Multiobjective Genetic Algorithm

Abstract

This article presents a robust multiobjective genetic algorithm (MOGA) optimization of the novel converging–diverging (CONDIV) manifold heatsink for external airflow environments. The CONDIV manifold is designed to protect internal heatsinks from debris-laden flows while maintaining high thermal performance. Previous studies relied on simple parametric sweeps for only a limited exploration of the design space of external manifolds, limiting the discovery of optimal designs. Here, a comprehensive MOGA framework is implemented, combining 3-D computational fluid dynamics (CFD) simulations validated against experiments. Seven geometric parameters are optimized to simultaneously minimize thermal resistance and pumping power, maximizing the coefficient of performance (COP). The design of experiments (DOE) is generated via Latin hypercube sampling (LHS), and a response surface model is constructed to approximate heatsink performance. The MOGA efficiently explores the design space, yielding a Pareto front of optimal solutions. Best compromise solutions (BCSs) are selected and verified through CFD. Results show that MOGA-optimized designs consistently outperform previous CONDIV and baseline plate-fin heatsinks, achieving up to 33% higher COP and 4.3% greater heat flux than previous CONDIV-H design and 42.4% higher mean COP than the baseline plate-fin heatsink. Detailed flow analyzes reveal that optimized BCSs improve airflow into internal channels and improve flow uniformity, resulting in superior thermal management. This demonstrates the strong potential of MOGA-driven optimization for advanced manifold heatsink applications in electrified transport.