open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello everyone!!
I am testing the model with impermanent flow in a section of the Paraná river between the cities of Puerto San Martin and Rosario. The upstream boundary condition was represented with a inflow hydrograph [Q(t)]. Downstream was defined a curve representing discharge versus water level [H(Q)].

I test the model with different time step; 10 seconds, 5 seconds and 1 second. The roughness employed is n = 0.023. I attached an image with the comparison of the water levels calculated for the three time step for two cross sections.

I notice too much difference between the values that take the water levels for the different simulations, is it possible that the time step influences in this way?

Hello

It is normal that any time discretiaztion (explit, implicit or semièimplicit), the results are dependant of the time step. This dependance is also dependant of the space discretisation (finite element and finite volumes do not behave in the same way).
Moreover, the depensance will change with the model size, the total simulation time and many other parameters
For Telemac 2D with finite elements, recommanded time step is the one that doesn't make suffer the solver. In other words, when running your simulation, you can see in the listing informations about the linear solver. The smart time step is the one that makes the solver converge with 20 to 40 iterations more or less

I hope that this helps

with my best regards

Riadh

Hello

as Riadh said, there is some dependency on the discretization and the numerical scheme/numerical flux. Choosing a small time steps does not mean to be more accurate.

Do you have values of the measured water depth at the outflow boundary? Maybe it can help to switch the boundary at the outflow to a given water depth. As a result, the water depth should be equal for all three time steps at the outflow. Doing so, you can compare the differences/influence of your model on the time step.


Best regards

Leo

thank you!! Now I'm testing with a more discreet mesh, the results are closer than expected!