Trouble converting from 2D to 3D
Hi, I changed the case of the flow around a cylinder in the XML file located at DualSPHysics_v5.0\examples\inletoutlet\01_FlowCylinder from a 2D case to a 3D case but I'm having doubts. In a normal CFD simulation the spanwise boundaries will have a symmetric boundary condition but to my understanding this option is not available yet. As of now I have walls on the sides of the cylinder but I'm unsure if I should allocate for both walls a velocity like what was done for the 2D case. If I do not give the sidewalls a velocity, the fluid particles will enter the solid boundaries. How can I fix the simulation? I've attached my XML file for reference.
Comments
Hi Ityap,
To the best of my knowledge, periodic boundary conditions are not compatible with open boundaries in DualSPHysics. That means that, for example, if you have flow through the x-direction, you cannot enforce y- or z- periodicity. I may be wrong and this may have been updated in latest releases, but I doubt it.
Back when the validation cases were made for I/O conditions, the idea was to have walls moving at the free stream velocity, so to generate the minimum amount of disturbance between the particles entering from the inlet (and in the vicinity of the wall) and the wall particles theirselves. This worked out well so I would recommend doing the same. As an alternative, you could apply a velocity a profile at the inlet that approaches 0 at the wall, this way you do not have to move the walls. However, keep in mind that the flow past these stationary walls will generate boundary layers that will interact with the shear layer produced by the fluid past the cylinder, which is an unwanted effect. You can obviate to this by putting the walls away from the cylinder (there are papers that study the effect of the wall distance on the solution), however this will cost you a lot especially in 3-D.
Best,
Angelo
Hi @taffo , thank you for the reply. I re-ran my simulation using your recommendation and got a stable simulation, although I do not get vortex shedding at Re200 like in the 2D case, which is something I have to look into. Regarding applying a velocity to the side walls, how much of an impact will it have on the final solution? Would you recommend making sure that the cylinder is long enough in the spanwise direction? At the moment, my cylinder has a diameter of 0.1m and a spanwise length of 0.1m. I've attached my updated xml file for reference. Thank you.
Just an update, I increased the spanwise length of the cylinder and I do get better force coefficients now compared to experimental results plus I have shedding which is nice, but I still have an issue with the simulation. As seen in the screenshot of my simulation below, I have some fluid particles entering my solid cylinder, which should not be happening. Do you have any suggestions on what could be the cause? I've attached my updated XML file too for reference if need be.