open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I have a T3D model that I intend to use for modelling a sediment plume using an inflow source with sediment concentration.

The study area is complex with shallow and deep areas and with several underwater strucutres (tunnel, bridges, locks), hence the need of a 3D model. The model is also characterized by locally small elements (1 m) and high velocities (up to ca 4 m/s).
The aim of my sediment modelling is to assess environmental risks due to sediment spill in case of excavation works.

I use version 6.2 (to be able to use Postel-3D directly) and T3D in non-hydrostatic mode. I have 10 vertical levels on my 3D grid.

The 3D hydrodynamics results are fine, with only small and local stability issues, mainly downstream of bridge piers. I use a time step of 0,5 s.

As mentioned above, inflow is applied via a source on an internal node. No sediment is taken into account on liquid boundaries, and I set the critical shear stress for erosion sufficiently large to ensure no erosion of the existing river bed.

I first used MOC for advection of tracers (sediment) and time step of 0,5 s, but the results were showing poor mass balance with a relatively high error, about 10-30% of the inflow rate. After about 500 s of simulating time, mass balance started to diverge significantly. I tried with a shorter time step of 0,05 s but the results were unfortunately the same, with the divergence occuring somewhat later (1200 s). The log of this simulation is attached.

I tried to change the advection scheme to 5 and 14, which are mass conservative, but I then have convergence issues using the SOLVER 1 (exceeding maximum number of iterations). I tried to use other solvers (2, 7), but I still got convergence problems, and with much longer computation times.

I also attach my steering file.

Would someone have useful tips regarding my numerical set up / sediment mass balance issues?
I will try to run in hydrostatic mode in the meanwhile and see if results can get better.

Thank you in advance for your help!

Best regards
PL

Hello,

It would be important to know what raises a problem. If it is the sediment (probably if hydrodynamics works), the refinement on the bottom is of course important but you should not have very small layers, because the CFL number of the settling velocity then becomes very high. As new developments were done in the last version 7.0 which is already released in version R0, it would be interesting to know if it works better (and if the problem with Postel-3D is still there we definitely shall have to solve it !). For example our colleagues at Wallingford have developed a new option for the settling velocity which seems more robust.

With best regards,

Jean-Michel Hervouet