Workbench meshing - 1
Posted June 27, 2013 at 09:16 by kartrmswy
So I modeled (with some difficulty) a concentric heat exchanger pipe on ANSYS WB Design modeler, and found that it's not that difficult to start using. However, the main challenge encountered was in the mesh, wherein the automatically generated mesh proved to be too extensive (even with a coarse relevance center) and the total number of nodes was in the order of 20,00,000 - this was going to take very long even on the 8-core machine I was working on.
I need to change the mesh so that I'm able to get a more suitable mesh - finer radially, since heat transfer coefficient changes are more in the radial direction, whereas the axial mesh could be coarse. Moreover, the auto-generated mesh was a tet-mesh, which meant a large number of elements. A structured Quad-Mesh would be better for solving.
This led to a potential solution: a mapped surface mesh that could be extended to the entire volume. An alternative is to use a sweep method. You can change the mesh shape to all quad/tri-quad/all tri. In addition, an inflation layer needs to be added near the walls to improve precision near the walls. It is important to note that this could affect the turbulence model used for the run, so reconsidering it is essential.
Finally managed to add inflation to my mesh - after selecting the swept method, right click to add inflation. Small catch in that: you can't select the boundary without selecting first layer thickness - if you go in for smooth transition, it just ignores your selection. The best thing to do in this case is to key in a value for first layer thickness. [Note: The selector sometimes selects faces instead of edges - the best way to get over that is to tinker around with the selections for a bit and it resolves itself.] I had no idea about mine so I used the y+ calculator on CFD online (http://www.cfd-online.com/Tools/yplus.php). The y+ value was coming out to be around 2.2 mm for the water-pipe interface.
A similar calculation for oil was made inflation values were keyed into the sweep grid parameter tab.
The mesh generated was not structured but could be changed to get it as closed to structured as possible. Another thing that can be done to improve quality is the refinement option - you can refine the mesh upto a factor of 3 - which makes it finer with each increasing step.
I tried using the mapped mesh option today, but there were issues with the grid size - increasing the sizing didn't really give me the results I wanted, so I stuck to the Sweep mesh.
Managed to get an optimum mesh size for the computation, though there are still a large number of elements for such problem.
I need to change the mesh so that I'm able to get a more suitable mesh - finer radially, since heat transfer coefficient changes are more in the radial direction, whereas the axial mesh could be coarse. Moreover, the auto-generated mesh was a tet-mesh, which meant a large number of elements. A structured Quad-Mesh would be better for solving.
This led to a potential solution: a mapped surface mesh that could be extended to the entire volume. An alternative is to use a sweep method. You can change the mesh shape to all quad/tri-quad/all tri. In addition, an inflation layer needs to be added near the walls to improve precision near the walls. It is important to note that this could affect the turbulence model used for the run, so reconsidering it is essential.
Finally managed to add inflation to my mesh - after selecting the swept method, right click to add inflation. Small catch in that: you can't select the boundary without selecting first layer thickness - if you go in for smooth transition, it just ignores your selection. The best thing to do in this case is to key in a value for first layer thickness. [Note: The selector sometimes selects faces instead of edges - the best way to get over that is to tinker around with the selections for a bit and it resolves itself.] I had no idea about mine so I used the y+ calculator on CFD online (http://www.cfd-online.com/Tools/yplus.php). The y+ value was coming out to be around 2.2 mm for the water-pipe interface.
A similar calculation for oil was made inflation values were keyed into the sweep grid parameter tab.
The mesh generated was not structured but could be changed to get it as closed to structured as possible. Another thing that can be done to improve quality is the refinement option - you can refine the mesh upto a factor of 3 - which makes it finer with each increasing step.
I tried using the mapped mesh option today, but there were issues with the grid size - increasing the sizing didn't really give me the results I wanted, so I stuck to the Sweep mesh.
Managed to get an optimum mesh size for the computation, though there are still a large number of elements for such problem.
Total Comments 0
