A Fast Algorithm For Thermal Analysis On A Dense Mesh

The objective of this program is to reduce the analysis cycle time for full vehicle thermal models. Cycle times include geometry and mesh generation, property attribution, solution, and post-processing. Of these, mesh generation can take the most time, particularly if separate meshes are required for a CFD tool and a thermal tool. If a conventional thermal solver is employed on a high-density CFD mesh, the time for a solution becomes prohibitive, and if existing thermal solvers are ported for use on an HPC or GPGPU and then applied to high-density meshes, convergence issues will cause computational times to be excessive. Our objective is to build upon work started in Phase I and modify the thermal solver in MuSES so that solutions on high-density meshes are feasible. The work will be broken into a series of specific tasks that include developing a solution technique for the coupled conduction and radiation problems; developing a method to automatically coarsen the mesh for the radiation problem; and implementing appropriate advanced solvers. The algorithms and techniques resulting from these individual tasks will be integrated together to form an advanced tool capable of achieving multiple orders of magnitude reduction in analysis time for full-vehicle thermal analysis.