I am trying to simulate a kayak moving through water at a speed of 4ms-1 by using multiphase flow. The goal is to extract the drag coefficients and study the kelvin wave patterns
In my project, I have followed the boundary conditions and initial setups as prescribed here
However, my simulation always diverges within less than 10 iterations. I suspect its a setup problem and have tried hex and tet meshes but both simulation equally blows up.
I am really grateful is someone could help me out, thanks!
I’m not used to multiphase transient simulations and moreover I prefere Stand Up Paddle.
However I had a shallow look at your BCs setting and it seems that Symmetry BC was applied to too many faces. I made a copy of your project (Kayak Client) and did run simulation for 10 core-hours. It works, as far as I can see from the residual plots:
Well, I’ve tried Stand Up Paddle before but the speed and exhilaration from a sprint kayak still takes the cake
Thank you so much for troubleshooting the B.Cs, I wouldn’t even think of that and definitely will give it go.
Just a question for fundamental’s sake. Can you speculate why symmetric boundary conditions fail?
The reason for symmetric B.Cs around the sides and base of the domain is to ensure no mass/momentum flow in/out. It seems that this implement led to unphysical results which ultimately led to a diverged simulation. I would speculate that my domain is too small to ensure that mass/momentum does not moves out of the domain.
This is a ‘tunnel’ definition. In that case you could apply Wall > Slip BC.
Symmetry BC is definitely for real symmetry plane (of you full geometry): however I do not know why it should be avoided in far fields definitions…
Personally I prefer more ‘elastic’ BC on far field. I use Custom > Zero gradient for all fields, but your Custom In / Out BC should be doing similar trick. Your domain is not too small: low Reynolds numbers are in play, turbulence is minimal.
Beware that transient simulation, even for relatively small mesh, with your parameters, will consume about 1000 core-hours. I suggest to observe residuals and stop simulation when no more progress is observed.