Dam Failure Crashing

[longislandeng]

Greetings. I have a nice and stable unsteady flow model I am using for a Dam Failure Analysis.
BUT when I breach the dam. The model becomes unstable and crashes.
I have ran through the usual remedies. Changing computational interval and interpolating cross sections.
I’ve ran through Chris’ excellent guide. http://hecrasmodel.blogspot.com/2013/10/stabilizing-dynamic-unsteady-hec-ras.html

But I am still snagged.

My water surface error jumps to 45 in one location.
I also have two RS that extrapolated beyond the rating curve for a bridge and culvert.

The problem appears to be at downstream bridges. That is where I have a large WS spike and “things go wrong”.

Does anyone have any tips on trouble shooting instabilities causes by bridges? I think I need some guidance on Htabs for bridges.

Thank you!
Todd

[Anonymous]

When you do a dam breach, you need to make sure that the bridge and rating curves go high enough. There is a user entered maximum head water elevation and an optional maximum flow. These values may need to be adjusted upwards for the higher flows and WSEs from the breach.

When you get a message that RAS extrapolated beyond the rating curve, it could mean that the curves don’t go high enough. Or, it could be that RAS went unstable and got a ridiculous ‘bad’ WSE/flow and the fact that that is beyond the curve isn’t the real problem.

As an intermediate step, try removing the bridge and culvert. Do you get good results or is it still going unstable?

[Anonymous]

HI Jarvus. Thank you for the post.
I did adjust the max wse up and the max flow up. This did not have the effect I hoped for. The calculated WSE are way off the chart.

What I ended up doing was trying the less is more approach. The model reacted poorly if I decreased the computational time lower and increased the the number of interpolated cross sections. This seems counter intuitive to me. But as I removed interpolated XS and increased the time step, model stability increased.

I eventually found a sweet spot and was able to stabilize the model. And the results make sense.

For now I think I am in the clear!

[Anonymous]

Generally, reducing the time step makes things better. But I have seen situations with dam breaks where longer time steps and fewer cross sections work better. Dam breaks can have a really sharp rise in flow and WSE at the leading edge. The longer, fewer approach maybe can smooth things out over a longer distance, time step so there is less of a shock.

The 2D code seems to handle this better, but that can be a lot of work. There is a ‘mixed flow’ option for 1D. Even if your model does not go supercritical, turning it on may help with the dam break runs.

[Author]

Posts