Unsteady Model Troubleshooting (Dam Breach)

[tduffy83]

So I’m newish to unsteady flow modeling in RAS and I’m struggling to troubleshoot the inevitable instabilities in the models I produce. I have taken an ASCE webinar, but it was not the most detailed.

My model has a steeper area at the beginning that I hypothesized was supercritical, which can cause instabilities. My thought was to run the model in steady state and use the modified puls method at every cross section to determine where the model was supercritical.

The result is:

Which is what I would expect to see for an unsteady model. As I thought, the model does go supercritical at the steep beginning, so my thought was to keep using modified puls at the beginning but let it switch to the full st venant equations (which are more accurate) for the later part.

Unfortunately, this causes the model to go unstable:

With the hydrograph and rating curve of a downstream cross section completely haywire:

This is versus the hydrograph and rating curve of when it was still modified puls:

My model can be downloaded from here: https://drive.google.com/open?id=0Bzzp7FkGAwnAOGlDbXhqWGxNOGs

Any help would be greatly appreciated.

[cameron]

Are the downstream cross-sections from the dam a canyon or wide open? Did you attempt to increase roughness values to 0.1 or so to help stabilize the model? Did you try to add interpolated cross-sections?

[cameron]

A few additional thoughts:

I would switch to a 2D model if you have the option, far more stable.

You need to update the HTAB data for the cross-sections to be every .25 feet or less.

Bump up the roughness to 0.1 or higher just downstream of the dam as 0.032 is low for all of the debris that would be coming down.

I would add ineffective areas so that the active top widths don’t change much if at all downstream of the dam.

You may need to interpolate xs.

You set initial conditions to 0 cfs in the main channel, the model does not like this as you can’t have zero flow in a cross-section or HEC-RAS will blow up.

You timestep of 2 minutes is probably too high and should be lower around 15 to 30 seconds maybe less.

[Author]

Posts