Learning Forum

[Lucas Dewaele]

Hi Everyone and thank you in advance for your time.

I have red multiple topics regarding similar issues but I wasn't able to find anything to solve my issue.

I'm tying to simulate the sinking of an object in a volume of water to have data regarding the speed and pressure aroung the geometry. It's a very basic geometry with two disks, a larger one over a smaller one, with a hole (). I am running a transient 3D simulation, using VOF model for multiphase (air and water) and k-omega SST for viscous effects. I have set gravitationnal force (-9.81 in the y-axis); the top of the domain is a pressure outlet, the contour are symmetry type and the bottom wall. My mesh seems to be very good (orthogonal quality minimum 0.5 and skewness maximum 0.5). I have created an overset mesh around the geometry and created it accordingly in fluent. The dynamic mesh uses an udf file which only give the mass and moments of inertia for Ixx, Iyy and Izz. 

For the methods, I have tried multiple options (piso and coupled scheme), pressure on Presto and upwind for the rest and set the advanced stabilization for the vof. In the initialization tab, i set the turbulent kinetic energy to 0.01, added the patch for the water volume and i'm running the calculation for 2000 time steps, with a time step size of 0.001, and max iterations of 20. My goal would be to simulate at least 2 seconds to start. 

My problem is around the residuals. As i'm in a transient simulation, the residuals are "periodic", so the converge then get back to high values. During the 100 first steps, everything converges well (10-3 which is acceptable for me). But after that, the high value at each time step for the vf-phase increases. When i'm around 1000 iterations, i'm at e1 values for the peak, then get back to low values.

I am scared that this behavior lead to bad results, do you know how i can fix that issue ? I've tried many things but nothing works, i'm kind of desperate...

[Prashanth]

Hi: In that case, use adaptive time stepping based on a conservative global cfl estimate value

[Lucas Dewaele]

Hi ! 

Thank you for your reply, I will try that ! I didn't think about an adaptative time stepping.