Design of Liquid Cold Plates for Thermal Management of DC-DC Converters in Aerospace Applications

Abstract

Due to increasing power demands and decreasing component size, thermal management has become the bottleneck for many power electronic applications. The aerospace industry has focused on reducing weight, operating temperature, and pumping power of power converters since these will limit an aircrafts' range and load carrying capacity. This paper outlines a tool created in MATLAB to automate the cold plate design process for DC-DC converters (or similar applications). The tool incorporates a genetic algorithm to fi nd the optimal aligned or staggered pin fi n confi guration that maintains the devices below their critical junction temperature while reducing the system's overall weight and pressure drop. Utilizing this MATLAB design tool, a cold plate was designed, manufactured, and tested. The convection coefficient calculated within MATLAB (via empirical correlations) was veri fed using simplifi ed CFD simulations within 5% of each other. The same CFD setup, boundary condition types, and methodology are then applied for the full-sized prototype cold plate simulations. These simulations were then validated using the experimental results. For all cases, the percentage error between the simulated convection coefficient values (CFD) and the experimental results was less than 12%. The experiments' measured surface temperature and pressure drop errors were less than 8% of the predicted CFD results. Therefore, the MATLAB tool and its correlations/calculations could be veri fied (via CFD) and validated (experimentally) based on good agreement between the CFD and the experimental results. This three-pronged approach (analytical calculations, CFD simulations, and experimental validation) is an effective and robust method to solve heat transfer problems. Overall, with the framework outlined in this thesis, a complete cold plate design can now be completed in weeks instead of months. This streamlined approach will save companies signifi cant time and money in the design and simulation phases, making this tool a valuable addition to the current literature available.