Experimental Validation of Indirect Liquid Cooling Computational Thermal Modeling

Abstract

Indirect dry-interface liquid cooling for electronics offers a modular and lightweight approach to thermal management. As part of development for the Navy’s integrated Power and Energy Corridor (NiPEC), multiple computational thermal analyses of this cooling method have been conducted but not yet tested against experimental results. NiPEC involves the deployment of converter units referred to as integrated Power Electronics Building Blocks (iPEBBs), which experience heat generation from four rows of MOSFET switches and a transformer. The goal of thermal simulations is to predict the liquid flow rate through a cold plate given a heat flux which will keep the iPEBBs below 100◦C. Experimental data on indirect liquid cooling of a layered substrate was collected with varied power, temperatures and flow rates. The data was verified through a STAR-CCM+ CFD simulation by David Hernandez, then used as a training and test set for a physics-informed neural net- work (PINN) model by Aniruddha Bora. The PINNs simulations were shown to improve in consistency and accuracy when trained on data gathered in the experimental trials. The collected experimental data set is available to the public for training and testing computational models.