open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi there,
I am trying to run a TELEMAC-3D simulation of a very simple case of flow in a straight compound channel. The configuration is a straight trapezoidal channel with floodplains.
Boundary conditions: upstream and downstream elevations.
Initial condition: average of u/s and d/s elevations.

When using a constant viscosity model for the horizontal turbulence and mixing length model for the vertical, steady state is achieved (Run1).
However when attempting to use a constant viscosity (CV) model for both vertical and horizontal turbulence, I had difficulties reaching a steady state (for which I watch the mass-balance in the sortie file). However the change in mass-balance takes place as very gradual small increments, but still no steady state. Have tried reducing timestep, increasing duration of run, using the 3D file of Run1 as a previous computation file to speed up the computation, no luck.
Also when plotting the velocity vertical profile (CV both vertical and horizontal) I noticed that there is very little variation over the vertical (0.02 m/s), does this suggest that the vertical velocity profile hasn’t developed yet?
How can I achieve a (1) a steady state case (2) proper velocity profile using a constant viscosity model?
Any suggestions,
Thanks

Hello,

Depending on the constant you use for diffusion on the vertical, the velocity profile on the vertical may take a long distance to establish. We are not familiar with a constant diffusion on the vertical and favor the mixing-length model in most cases.

When a canal is not long enough, it is sometimes necessary to prescribe a logarithmic velocity profile at the entrance. This is obtained by keyword

VELOCITY VERTICAL PROFILES : 2;2;2... (as many as liquid boundaries)

and the subroutine vel_prof_z.f is called and may be amended.

Regards,

Jean-Michel Hervouet

Hello,
Thanks very much for the response.
I am hoping that using the keyword will help fix the profile. Will let you know of the outcome
Thanks again,
Zeinab

Hi again,
An update: I assigned the keyword VELOCITY VERTICAL PROFILES the values 2;2 (since I have two liquid boundaries). In the TEL3D dictionary the keyword can take only 3 values, 0, 1, and 2. I experimented a bit with it but I think there is still something wrong, because when I change the value to say 4 I don't get prompted that its not an option.
Also when I chose 0 (for which I should define my own profile) I don't get prompted to modify a subroutine.
Tried running the model with it and no change at all. So it must have had no effect on the vertical profile. I seem to get no variation at all over the vertical!!! can't be right.
Am I missing something?
Best Regards

Zeinab Bedri

Hello,

Yes, you must probably specify the keyword : PRESCRIBED FLOWRATES (one per liquid boundary, e.g. 50.;75.), to activate the profiles, otherwise the values in the boundary conditions file are taken and copied on the vertical. You can see this in subroutine BORD3D, NDEBIT is the number of prescribed flowrates given in the parameter file, if you give none, the section where the profile function vel_prof_z.f is called is bypassed. Sorry this is complicated because we mix a number of possibilities, files, user functions, profiles...

I hope it works,

Jean-Michel Hervouet

Hello again,
So I did a few more tests on the constant viscosity model:
(1) I ran a constant viscosity model (both horizontal and vertical) using an inflow boundary condition (Q) and downstream fixed elevation. VERTICAL VELOCITY profile keyword set to 2 for both boundaries. Result: While a logarithmic velocity profile was produced at the inflow boundary, this has not translated into the channel.

(2) Ran a simulation in which the Prandtl mixing length model was used for vertical turbulence (horizontal turbulence= constant viscosity). Model had upstream and downstream fixed elevations as boundary conditions, the model reached a steady state and produced a very reasonable profile, used this simulation as an intial condition for the constant viscosity model by just changing the vertical turbulence model to constant viscosity and continuing the computation. Result: vertical profile changed gradually during the simulation to a constant vertical velocity!!!

(3) Generally I spend a long time to get 3D simulations to reach a steady state. I watch the mass-blance i.e. previous and present mass, I notice it gets very close to converging and then keeps moving on and on even though I reduce timestep and/or increase duration of simulation. Have tried both SUPG and characteristics method for advection. Any tricks or tips, any rule of thumb in timestep/meshsize/horizontal diffusion coefficient ratio? I find so far it is a matter of trial and error. Also if I continue a computation (using a previous 3D file) but use a smaller timestep it converges which is a very arbitrary thing.

Can anyone shed some light here?
Thanks alot,
Zeinab

Hello Zeinab,

On points 1 and 2, I do not really understand what you want to do. The logarithmic velocity profile is obtained in theory with a mixing-length turbulence model, this is what you find. Now if you put another turbulence model the velocity profile is no longer logarithmic and so it changes even if you prescribe it at the entrance. So what you say in points 1 and 2 is quite logical and normal, it is not for me a problem.

As for point 3, there must be something wrong in what you do. In any test in a canal with constant discharge at the entrance and fixed elevation at the exit, the mass in the domain remains bounded. If you don't find what happens send your case and we'll have a look at it.

Generally speaking I prefer the methods of characteristics to SUPG for the advection of velocities. For the constant horizontal diffusion it must not be too large (think that when there is a slope in the bottom it may interfere with the vertical diffusion).

With best regards,

Jean-Michel Hervouet

Hello Jean-Michel,
Thank you very much for taking the time to respond to my queries. Regars the constant viscosity model, I was under the impression that this is not the right (Expected profile) I thought there is something wrong with it. Glad to know it is fine.

Regards point 3, I found that for the case I have (Straight channel), a steady state is reached quicker when using fixed elevations upstream and downstream as boundary conditions than when using inflow upstream and fixed level downstream.
Please find attached the case that I am struggling with at the moment.
Thank you very much again for your response.

Zeinab

Hello,

There is a problem in your steering file : the name of the steering file should be series2_8_3d4.str, and the value indicated is run1600_3d.str. Maybe you do not run the case you think.

I strongly advise (message to everyone !) to never give the name of the steering file in the steering file ! Unfortunately some old examples are given like this. It is not useful and in case of mistake you do not look at the right file and it can drive you mad.

Zeinab you should just cancel the corresponding line and check if it works. If not I'll look again.

Best regards,

Jean-Michel Hervouet

Hi Jean-Michel,
I did the test you requested i.e. removed the keyword “steering file” from the steering file and re-ran the test case. The same results were produced.

However, I did some more experiments on the case. I used a simulation of the same configuration with a reduced number of horizontal planes (8 planes), ran it and reached a steady state.

I used the 3D file of the above computation as a previous computation file for a setting in which the number of planes were increased to 20.
I experimented with 3 timesteps and used the same number of timesteps in each computation.
I noticed in one of the computations that the mass balance hasn’t changed at all, but was iterating and has resulted in an improved vertical velocity profile. The other two simulations kept iterating and converging but never reached a steady state.

Please find attached a summary results file for the tests I did.

Thank you very much for your tireless support and patience.
Regards,
Zeinab