Parallel simulation of radiative heat transfer using an unstructured finite volume method
Radiative heat transfer is a major mode of heat transfer and it is usually strongly coupled with fluid dynamics in many high-temperature systems such as boilers, furnaces, aircraft engines, etc. Here, a spatial domain-based parallel algorithm is developed for simulating radiative heat transfer in a distributed computing environment. The radiative transfer equation is solved using an unstructured finite volume method which is applicable for any 2D planar, axisymmetric, and 3D problems with structured, unstructured, or hybrid grids. The domain decomposition is carried out by equally partitioning the spatial domain into many sub-domains along the longer geometrical dimension. Communication between the sub-domains on each processor is performed through a message passing interface library. In order to examine the parallel performance of the unstructured radiation code, two benchmark problems are investigated for different absorption coefficients, scattering coefficients and grid sizes in a parallel computer. To help one understand the change of parallel performance, a new parameter, the total inner iteration number, is introduced to analyze the results. For all the cases examined, as expected, the parallel performance is seen to degrade rapidly with an increase of the processor number. However, in contrast with other studies, the parallel performance is found to degrade with an increase of absorption coefficient for a temperature prescribed problem. Also, the global iteration number is found not necessarily to be independent of the grid size.
- Research Organization:
- Engineering Sciences, Inc., Huntsville, AL (US)
- OSTI ID:
- 20002479
- Report Number(s):
- CONF-990805--
- Country of Publication:
- United States
- Language:
- English
Similar Records
Spatial domain-based parallelism in large scale, participating-media, radiative transport applications
A paradigm for parallel unstructured grid generation