open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi everybody,

I have tried to simulate lock exchange problem using the distributive scheme for the advection of velocity and tracer. I noticed that when I use the option of NUMBER OF SUB ITERATIONS FOR NON LINEARITIES to be 4 iterations, there is no diffusion at all compared with not using this option where there is obvious diffusion for the tracer.

is the distributive scheme is not suitable for advection of velocities? Actually in the validation case I found in Telemac3d, the distributive scheme is used only for advection of the tracer but not in the advection of velocity.

Has anyone an answer for that

Hello,

Actually you can use any distributive scheme for advection of velocity also, but it might be more diffusive than the method of characteristics. The sub-iterations should not change the result regarding diffusion (and should not change a lot in the case of the lock-exchange), so if you confirm changes that cannot be explained I'll have to look at this, and the problem would certainly be on the sub-iterations side, not on the distributive scheme side.

JMH

I have used the sub-iterations with the method of characterstics and I did not find any change, however when I am using distributive scheme the results are totaly different, where the dense fluid did not move at all which is not logic.

attached here are the files of the case

Hello,

OK, I'll look at this as soon as possible (but am leaving now for SimHydro 2012 till the end of the week). In the while you can rely on the distributive schemes and not on sub-iterations (where the problem probably is, as there is only one test-case with sub-iterations and many with distributive schemes).

With best regards,

Jean-Michel Hervouet

It is ok, and thanks in advance

Hello,

Thanks for pointing out this problem, which was really a mistake. The combination : source terms + distributive scheme + iterations for non linearities (the latter seldom used, hence we had no test case with this combination) forgot the source terms from the second iteration on. Enclosed are 3 files with the correction. The real correction is in CVDF3D, but then we must be more careful in wave_equation.f for the cases where the source term S0U must be applied. In the interface the only change is that S0F in CVDF3D must now have a target property.

With best regards,

Jean-Michel Hervouet

Hello,

Now the wave_equation.f file in previous post is also wrong and takes the source terms into account twice ! Hopefully correct version enclosed, with modification dated 19/09/2012.

JMH

Thank you