Flow past a cylinder/bluff body
Hello,
Wondering if anyone has performed single phase simulations using dualSPHysics; specifically flow past a cylinder or a bluff body at different Re numbers capturing von Karman vortex sheet?
Vinay
Wondering if anyone has performed single phase simulations using dualSPHysics; specifically flow past a cylinder or a bluff body at different Re numbers capturing von Karman vortex sheet?
Vinay
Comments
If I was to give an immediate response then my answer would be, no DualSPHysics can't capture this effect as long as you are using the standard boundary condition.
There is, of course, more to the response though. Firstly, the weakly-compressible SPH method employed in DualSPHysics can capture these effects, for an example see the paper by Vacondio et al. ("Variable resolution for SPH: A dynamic particle coalescing and splitting scheme") where they use the WC-SPH version of the FORTRAN SPHysics code. However, it can be noted that they had to use a different boundary method. It can also be noted that in gravity-free cases like this , WC-SPH will likely produce artificial voids behind the cylinder/body, especially if you are moving it rather than using an inlet and outlet boundary condition to produce a flow in the fluid (and as DualSPHysics only employs open periodic boundaries, this is what you would have to do). This behaviour is entirely expected but ruins the kinematics of the simulation, so in order to fix this the SPH community typically introduces a small background pressure to all particles in the simulation (magnitude is case dependant) or uses some sort of particle shifting routine to renormalise the distribution throughout the domain. By default, DualSPHysics does not provide either of these mechanisms.
Effectively though, the WC-SPH method and both artificial and laminar+SPS viscosity schemes can capture the effect, but the boundary condition is not sophisticated enough for these cases without some extra treatment.
Saying that, the case should be very simple to set up and numerically speaking I have been able to perform cases at high Reynolds numbers of 10,000+ in terms of stability, so if you fancy giving it a go and seeing what you get then do report back and let me know! I can almost guarantee you'll need a background pressure though, this can be easily introduced by simply adding a small value (Say 100 Pascals) to the value calculated in JSphGpu_ker.cu (float press=...).
Best of luck and let me know how you get on!
Best regards.
Vinay