open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

Recently I'm testing my 3D hydrodynamics model with salinity.I also have very large ERRORS ON THE MASS during computation. I have already modified the steering file according to the advice found in the previous post.

Here is the information ptinout on the screen:

MASS BALANCE

WATER
MASS AT THE PREVIOUS TIME STEP : 0.5361970E+11
MASS AT THE PRESENT TIME STEP : 0.5360712E+11
MASS LEAVING THE DOMAIN DURING THIS TIME STEP : 0.1256341E+08
ERROR ON THE MASS DURING THIS TIME STEP : 9741.434
FLUX BOUNDARY 1: -5319.107 M3/S ( >0 : ENTERING <0 : EXITING )
FLUX BOUNDARY 2: -2507363. M3/S ( >0 : ENTERING <0 : EXITING )

TRACER 1
ADVECTIVE FLUX THROUGH BOUNDARIES OR SOURCES : 0.3737988E+08
DIFFUSIVE FLUX THROUGH THE BOUNDARIES : 0.000000
MASS AT THE PREVIOUS TIME STEP : 0.8065292E+12
MASS AT THE PRESENT TIME STEP : 0.8063433E+12
MASS EXITING (BOUNDARIES OR SOURCE) : 0.1868994E+09
ERROR ON THE MASS DURING THIS TIME STEP : -973401.9

Besides, after several time steps, I saw the following messages on the screen:

GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 200 RELATIVE PRECISION: NaN
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 200 RELATIVE PRECISION: NaN
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 200 RELATIVE PRECISION: NaN
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 200 RELATIVE PRECISION: NaN
......

Could anyone help me to improve this? Attached is my steering file.
Any help will be appreciated.

Hello,

a quick look into your steering file:
you specified a Nikuradse roughness height of 60 m. This value seems very large and nonphysical.
In general I would start with the default values for the implicitation coefficients and for the mass lumping coefficients except for the mass lumping for depth if you use treatment of negative depths = 2. However these settings should not be the reason for the large mass errors.

Best regards,
Clemens

Thanks Clemens!
I've changed LAW OF FRICTION ON LATERAL BOUNDARIES to ZERO as well as mass lumping coefficients for velocities and diffusion. I still got large mass errors.
There was no errors like GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 200 RELATIVE PRECISION: NaN at first but it appeared again after several time steps.

Hello,
I don't know your model but what you can do:
Inspect your intermediate result file, there you will probably see where the problem is.
Check the boundary conditions, if you have strong wiggles or velocity gradients at your imposed liquid boundaries.
Estimate roughly the Courant number for your case and adjust the time step, maybe you have a too large time step.
Try first a simpler turbulence model, sometimes I had wiggles and no proper convergence with the k-e model, however reducing the time step especially when using the k-e model helped a lot.

Best regards,
Clemens

Thanks for your advice, Clemens.

I have tried to run the simulation with constant viscosity instead of k-e model but it still became unstable after 2h45m. Then I tried it again without salinity and it was OK. So I think the problem is related to the way dealing with the salinity, especially in the shallow water areas.