[kkpal]

Dear all,
In the simulation of vortex induced vibration of a flexibly mounted rigid circular cylinder, I modified the case from the wingMotion tutorial and tried to change the ddtScheme from Euler, which is first order accuracy, to backward, which is second order one. The case with Euler went well, however, with backward scheme the computation diverged very quickly!

With a fixed cylinder simulation where there is no mesh deformation, both euler and backward worked well.

Have someone else encountered this problem?

[kkpal]

I attach my case file here. Hope some one could test it and give me some hints.

https://drive.google.com/open?id=0By...0N6anFZTGNpRDA

[saneku]

Hi! Did you try to reduce timestep?

[kkpal]

Yes, I tried reducing time step 10 times smaller, the courant number is around 0.2! But the problem is still there.

Originally Posted by saneku

Hi! Did you try to reduce timestep?

[saneku]

have you look here http://cfd.direct/openfoam/user-guide/fvschemes/ ?
section 4.4.6

[kkpal]

Originally Posted by saneku

have you look here http://cfd.direct/openfoam/user-guide/fvschemes/ ?
section 4.4.6

Dear Saneku!
Among the ddtschemes listed in section 4.4.6, I tested Euler, backward and CrankNicolson.

Euler worked well;
backward diverged;
CrankNicolson with values smaller than 0.4 worked well, with values larger than 0.5 diverged.

[kkpal]

hi, guys.
The reason for the blow up with second order schemes is that i was using a very small mass ratio for my cylinder. It seems that the weak coupling fsi algorithm does not work well for low mass ratio.
As I increase the mass ratio, this problem is resolved.

Best regards,