Multi-resolution Simulation and Design Tool For Thermal Management of Electronics Packages in Complex Systems

Optimizing the placement of electronics packages to minimize thermal loads in large systems (e.g., power electronics modules in vehicles), and prediction of overall system thermal performance, is typically a very complex problem. Physical prototyping of each permutation during design cycles is time-intensive and expensive. Use of full-scale computational fluid dynamics (CFD) solvers may be impractical for system-level analyses because of: (a) typically large manual efforts needed for grid generation; and (b) long simulation times. There is a need, therefore, for simulation and design tools based on reduced models that will enable rapid, system-level, thermal-fluidic analyses. In this paper, we describe a multi-resolution flow-thermal simulation approach and prototype software tool for thermal management of electronics packages in large systems. This method offers higher efficiency than CFD analyses in terms of computational resources and time. Our demonstration example is related to flow-thermal analysis of a simple vehicle model containing multiple electronics packages. Artificial neural networks (ANNs) are used for rapid modeling of the electronics components and subsystems. The tool is capable of performing fully-resolved (CFD), mixed-resolution (CFD-ANN), and reduced-model (ANN) flow-thermal simulations. Coupled CFD-ANN simulations are a powerful method for design and thermal management of electronics packages and other electrical-mechanical components in large multi-component systems