open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

We have modified the subroutine TFOND in order to correct the high shear velocity in the shallow water. But we don't know it for sure that whether the velocity will be recalculate after calling TFOND or not? Because the shear velocity will affect the calculation of the velocity field. Does the subroutine TFOND use the velocity from previous time step or current time step?

Thanks in advance!
Qilong

Hello,

Actually tfond.f uses the velocity of the previous time step, that will be used to get the friction in the new velocity equation. In other terms the U**2 in friction terms is semi-implicited in the form U(time n) * U(time n+1).

Regards,

JMH

Hello,

I found the code for calculation of UETCAR in tfond.f was

DO N=1,NPOIN
UETCAR(N)=(U2D(N)**2+V2D(N)**2)*0.5D0*CF(N)
ENDDO

Do you mean the U2D and V2D are from previous time step? Is there any other subroutine that can be used to calculated UETCAR? Is it possible to calculate UETCAR based on the velocity from current time step and update the velocity field afterwards?

Is the semi-implicit U**2 you mentioned above the velocity?

Regards,
Qilong

Hello,

You are right but the implicit friction coefficient AUBOR is UETCAR**2 divided by a VNORM which is also explicit. Yes we could take the new velocity, which would require at least 2 sub-iterations to be effective. I did not do that because I want the U**2 to be semi-implicit. Normally it should be ((U(n)+U(n+1)/2)**2, which is not linear in U(n+1) (here n means old time and n+1 new time), so I choose to take U**2 as U(n)*U(n+1). This is what is done and why it is done, but of course you can try other possibilities. The only, but strong, requirement is that friction must remain a fully implicit term (i.e. a term -k*U(n+1)), otherwise you will get sometimes inversion of velocity due to high friction, then there is not so many possibilities.

With best regards,

Jean-Michel Hervouet