open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi All,

I've invoked the culvert model (OPTIONS FOR CULVERTS = 2) to simulate flow through a short (~10 m) box culvert. On the upstream side of the culvert is an open drainage canal that dead ends at the box culvert, on the downstream side is a rectangular concrete channel. The two waterways are perpendicular, and the drainage canal is a tributary to the concrete channel (see attached images).

The culvert model seems to work great and agrees with hand-calculated flow rates. However, my problem is that the water surface elevation spikes at the culvert exit node and is depressed at the culvert entrance node. These perturbations effectively reduce, or all together cancel out the head difference across the culvert. This condition leads to too little flow getting through the culvert.

My Courant number at the exit seems fine at ~0.5. Also, "IMPLICITATION FOR VELOCITY = 1" and "FREE SURFACE GRADIENT COMPATIBILITY = 0" help, but do not completely resolve the issue (see attached case file for other settings). I'm not sure what else may be at play, but this is a big problem affecting the accuracy of my model.

Any help toward resolving it would be greatly appreciated!!


Thanks,
Ben

Here is a plan view image of the mesh and water surface elevations (see attached).

I should also note that the culvert exit node is located on a short, relatively steep section of the concrete channel where the flow is supercritical. I've noticed elsewhere that steep slopes/bathymetry can cause surface oscillations - perhaps that steep slope is causing the issue here?

Thanks,
Ben

Hi
It's a possibility...
Try to move the output node to another point to see effect.
Another possibility is also to reduce time step...

regards

Thanks for the feedback. I had the same thoughts. Moving the exit node off of the steep section and onto a flatter reach did not fix the problem unfortunately (see attached image). Nor did cutting the timestep in half. Reducing the timestep had no effect.

I'm wondering if a work-around may be to distribute the culvert flow across several adjacent nodes. I.e., represent the culvert as several "sub-culverts" so as to distribute the outflow over more than a single point. At least this might attenuate the error if not eliminate it. Thoughts?

Thanks!
Ben

To follow up on this thread, I have not had much progress finding a solution to this problem. It seems no matter what I try, the culvert entrance node will have a depressed free surface and the exit node with have a super-elevated free surface. The phenomenon causes the culvert flow to be self-limited, biasing the culvert flow low - significantly in some cases. The culvert flow can be more sensitive to this error than even to the assigned loss coefficients!

The problem is more noticeable the higher the flow rate is. Spreading the flow across several nodes (i.e., turning one culvert into several sub-culverts) helps a bit, but is awkward and doesn't totally fix the problem.

Currently, my next approach will be to work-around the problem by having the culvert flow equations use the surface elevation from a neighboring node (i.e., a node where the free surface is not perturbed by the presence of the culvert) rather than the node assigned to the culvert entrance/exit. Does anyone have guidance on how to modify the BUSE.F subroutine to achieve this goal?

I see that the subroutine defines the free surface elevations at the entrance (S1) and exit (S2) using the P_DMAX and P_DMIN variables, but I'm unclear on how:

S1=P_DMAX(S1)+P_DMIN(S1)
S2=P_DMAX(S2)+P_DMIN(S2)

If someone could explain to me what's going on here, perhaps I could hack the subroutine to have it retrieve elevations from manually-specified nodes (for example by hard-coding in the neighboring node's global index).

Thanks so much for any input,
Ben