An Innovative Modeling Approach to Thermal Management using Variable Fidelity Flow Network Models Imbedded in a 3D Analysis

Speed and accuracy are the critical needs in software for the modeling and simulation of vehicle cooling systems. Currently, there are two approaches used in commercially available thermal analysis software packages: 1) detailed modeling using complex and sophisticated three-dimensional (3D) heat transfer and computational fluid dynamics, and 2) rough modeling using one-dimensional (1D) simplistic network solvers (flow and thermal) for quick prediction of flow and thermal fields. The first approach offers accuracy at the cost of speed, while the second approach provides the simulation speed, sacrificing accuracy and can possibly lead to oversimplification. Therefore, the analyst is often forced to make a choice between the two approaches, or find a way to link or couple the two methods. The linking between one-dimensional and three-dimensional models using separate software packages has been attempted and successfully accomplished for a number of years. However, this coupling procedure involves a very tedious and time-consuming task of interfacing between the two packages made more difficult by the lack of access to the source code. Furthermore, there may be issues relating to overall convergence, as well as the convergence of each solver, which can lead to compromised accuracy. For a truly coupled approach modifications to the source code of the solvers are required.

This paper introduces a complete thermal modeling and simulation methodology that offers a variety of approaches for modeling complex systems and components, avoiding any of the unnecessary “overhead” associated with a single modeling approach type. As an example, by allowing the combination of a full three-dimensional analysis in the radiator and the use of a network-based approach for pipes and pumps, the analyst can study the effects of orientation, fouling and related environmental conditions within the radiator and its effect on the complete system. The ability to combine these modeling approaches within a single system model allows one to employ the right tool for the job and avoid any unnecessary overhead associated with a single modeling approach or linking different packages. This greatly reduces the model development / analysis time allowing for the timely generation of results from which informed design decisions can be inferred.

A new software package, ADFlo, under development with the support of the US Army is used to demonstrate these benefits. This approach is applied to the thermal management of a liquid cooled Electronics Rack to demonstrate the ability to solve complex systems. The rack, boxes and the heat exchanger are all modeled in three-dimensions, with a one-dimensional flow network running through them. The flow network is complete with a pump and is tied to a one-dimensional airside model. The model is validated by experimental data.