open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi all !

I have to model a surge wave (backwater effect) upstream of a dam corresponding to the following conditions:
1st phase: permanent flow through the dam, with constant reservoir elevation
2nd phase: flow through the dam set to zero, thus initiating a surge wave propagating in the reservoir.

I guess it is possible to define such BC using a fortran file, but as i face some problems with such procedure on my machine (W7 64 bit), and as I am not skillful in programming, I had the idea of modelling this case as follows:
1st computation: reaching a permanent state in the model ; the DS boundary would be located at the dam and the BC would be a rating curve ; the computation stops when a steady state is reached
2nd computation: the IC used are taken from the first calculation; in this 2nd calculation, the downstream liquid BC is deleted in order to model the shifting from Q = constant to Q = 0 (solid boundary).

But I don't know if it is possible to use IC from a result file runned on a different model (same geometry but different boundary conditions).

If no, does anyone have other tricks ?

Thank you in advance for your help !
PL

Hello,

Yes you can change the boundary conditions before a restart, but the initial velocities of a previously open boundary may go accross a wall, so the first time steps may be a little hectic. Another solution would be to have the dam treated as a weir (this would require some tuning to get the correct discharge, and then to brutally change the coefficient mu of the weir, setting it to 0 (this is just an idea, not guaranteed, not tested).

Wit hbest regards,

Jean-Michel Hervouet

Hi,

Thank you for your reply. Yes, i also thought of using either a weir or culvert object to model the dam, but in that case i would need to model the river downstream of the dam.

Furthermore, without using a fortran file (in which the mu coefficient would be modified), the procedure will probably remain the same, with the same problem at the 1st time step (initial velocities through a impervious structure) ?

Regards,
PL

Another method that worked for me was to perform a restart simulation but this time with a discharge boundary condition at the dam (replace your initial level type 5 4 4 to the discharge type 4 5 5 in the BC file).

In the restart simulation, impose the exact same discharge that was exiting the boundary initially and have it suddenly go to zero. You can simply put a hydrograph in the FICHIER DES FRONTIERES LIQUIDES. No Fortran necessary. The discharge value should be negative since water exits the boundary.

In this way you can accurately control the shutdown time of your power house, spillway gate, etc.

Regards

Jeremie

Hi all,

I have another question regarding this topic : we expect the surge wave to propagate towards the upstream boundary, where will be introduced the inflow (liquid).
How do behave liquid boundaries in such case ? Will the flow brought by the wave "leave" the model by the boundary or simply being reflected on it ?

Thank you in advance for your help,
PL

Hi
By default, as you impose the discharge, you will have a reflection on this boundary.
Maybe but i'm not sure, you could use thompson method.

Hope this helps

Hi,

I think ideally you'd want to extend your model far enough upstream to include a hydraulic discontinuity (critical depth at rapids, another powerplant, etc.). In this way you wouldn't have to worry about reflexion off a discharge boundary. Of course, this is only valid if it is technically possible for you to run a larger model.

I put an example screenshot where the inflow boundary (right end) is upstream of a rapid. The water surface is colored with the Froude number (red being Fr>=1). This case is about 6,5 km long. A surge wave originating from the left-end boundary does not travel past the rapids, so there is no relexion off the inflow boundary.




Cheers

j

Thank you Christophe and Jérémie,

Yes, I thought about the possibility to extend the model upstream, but I am not sure if we will get enough data for that !
One of the issues was also to try to limit the model size in order to be able to have detailed results (with both small elements and short time steps). I think a compromise will have to be found...

Thanks for you help and best regards,
PL

Hi again !

I created my model and launched a first calculation. The method suggested by Jeremie works perfectly !
I have calibration data I can use in order to compare with T2D results. The data consists of water surface elevation measured over time after the shutting thanks to a pressure sensor located in depth.

I have prepared a graph, but the upload does not work... sorry.

Anyway, the main differences are:
- the computed waves' period is longer,
- the computed waves' amplitude attenuates more quickly.
On the other hand, the general trend is well reproduced.

Does anyone have an idea to explain these differences ?

I think that the following might affect the results:
- there are possibly 3D effects both near the surface that accuentuate the waves movement (can the element size have a significant role ?) and/or in depth (local increase of pressure that could explain the data but without effect near the surface)
- the turbulence model ? I used nr 1
- less significantly, the effect of wind and/or residual waves ; i don't consider their role as important (calm reservoir).

Thank you in advance for your comments !
PL

I will investigate your problem with uploading images (maybe the size could be an issue ?)