An automated OpenFOAM-based optimization of triply periodic minimal surface heat sinks in forced convection

With new advances in AI, multiphysics simulations, exascale computing, and the digitization of industries, there exists a growing need for compute resources and correspondingly, an equally growing need for advanced cooling options for these electronics. Advanced data centers use liquid or two-phase cooling, yet there exists a need across the heat rejection technology spectrum for continued improvement. One such area is in the advancement of forced convection heat sink design. Much research has taken place in using lattice structures for heat transfer improvement. Here, we expand the design space of triply periodic minimal surfaces (TPMS) to act as heat sinks and extend the state-of-the-art to include coordinate system variation, functional grading, and variable lattice thickness all within a custom developed OpenFOAM simulation and optimization framework capable of greatly expanding the design space of these lattices. The optimization framework allows heat sink designers to select a desired flow rate and pressure drop and determine a unique class of TPMS heat sink structures that has a reduced temperature while meeting the desired pressure drop. This was explored for cylindrical and rectangular coordinate systems and the TPMS thickness and spatial grading was allowed to vary parametrically within the optimization algorithm. Three cases were created in this study with six TPMS heat sink geometries (Diamond, Gyroid, Lidinoid, Neovius, Schwarz, and SplitP) and include: case I optimized for low thermal resistance, case II optimized for low mass, and case III a hybrid of both. These designs were compared to a standard plate fin design and all designs but one were printed in AlSi10Mg using an EOS M290-2 PBF-LB machine. Experimental testing was conducted for each heat sink and designed to match the simulation conditions for comparison. The cylindrical Lidinoid design was found to have the lowest thermal resistance at 0.78 K/W representing a 15.2% decrease over the plate fin design. Additionally, the mass normalized thermal resistance was reduced by up to 72% using these TPMS structures which holds potential application in mass sensitive transportation sectors. The OpenFOAM models were validated against experimental data suggesting the automated heat sink design process presented here can improve the design cycle and performance of generating advanced TPMS-based heat sinks for cooling applications.