My simulation crashes frequently when I use very small grids ( less than 0.05 mm). It doesn't for grids > 0.1 mm. Always there is an error related to partitioning. When I decrease the number of processors involved in parallel computing, I have found that the crash gets delayed (few hundreds of extra time steps)
I am doing micro-combustion in suddenly converging 2D tubes with bigger diameter = 5.67 mm and smaller diameter as 2.83 mm. I am using detailed chemistry (methane) . My mesh is uniform with 0.2 mm size and with adaptive gradient refining, refines to 0. 025 mm. Here the crash starts. I am using 1st order upwind scheme and even increased relaxation parameters but the crash happens. The crash doesn't happen at grids of 0.0625 mm or higher.
Please suggest me some partitioning methods which would allow higher order of refining. I am using Ansys Fluent.
Dear Aditya: I think you could use your ICEM, however the problem could also be in the way your particionas your model, I recommend you do not use auto partition and do it manually.
Please check this document
@Hennis Meyerhenke, I have used the default Metis partitioning method. It is auto-partitioning method. I don't have much idea on the manual partitioning methods and how to choose them. Any method which helps me get through the entire solution without crashing, would do the job. The region which has the highest number of cells moves as the flame carries it with them. The solver usually crashes just before mesh is refined dynamically.
As my experience, METIS (a graph partitioner) crashes when the number of vertices of the graph to be partitioned is smaller than the number of partitions (so, maybe you are using too many cores), or the weight of any edge of the graph to be partitioned is 0. If you can control the partitioning algorithm and the elements/vertices to be partitioned have coordinates information, try just cutting the simulation space into simple stripes.
- Badges
- Science topic