Monday, August 3, 2009

Problems with bridges and culverts.

Written by Chris Goodell, P.E., D. WRE
Copyright © 2009. All rights reserved.

(Thanks Eric for the topic). It's very common for HEC-RAS models to show inconsistencies around crossings (bridges and culverts). Usually, 1 of 3 things is going on here: 1. Bad geometry-either incorrectly entered, or poorly defined. 2. Numerical errors. 3. The results are actually correct and can be explained.
For the first case, "bad geometry", here's a technique that can be used to help spot sources of problems. Create and evaluate the hydraulic property plots for the crossing. If you are running unsteady flow, this is done for you during the geometry preprocessing task. If you are running a steady flow model, you can create an unsteady flow plan and just run the geometry pre-processing task (you don’t need to run the computations or the post-processing. Once that’s done, on the main RAS window, go to View…Hydraulic Property Plots. Click Type…Internal Boundaries, and you’ll see the family of rating curves for your crossing. Here you’ll want to examine the curves and look for any abrupt changes, or discontinuities, particularly in the range of flows/depths where you are seeing the discrepancy. Typically you see problem areas where RAS changes equations (i.e. going from low flow to pressure flow, or pressure flow to pressure and weir flow), or when ineffective flow triggers turn off/on. Also, keep in mind that the equations for culverts are very different from those used for briges in HEC-RAS.

Take the following example, in the figure below. First of all, I always like to open up the bridge plot along side its htab plot (make sure the vertical axis is consistent) so that I can graphically explain any discontinuities in the htab curves. This example shows a significant discontinuity at around 10,000 cfs (you can click on the figure above to get a better view). It's very obvious from looking at the plot that this is the range at which the flow transitions from low flow to pressure flow and then on to pressure and weir flow. Also, notice that the ineffective flow triggers turn off in this range. It appears that the creater of this model tried to lessen the impact of the ineffective flow areas instantaneously turning effective by significantly raising up the n-values in the overbank. Not a bad technique, but obviously didn't completely solve the problem.
Things you can tweak that may provide more sensible results and a better set of Htab curves are:

-Coefficients (bridge and culvert coefficients).
-Ineffective flow areas upstream and downstream of the crossing.
-Bridge modeling approach.
-Placement of cross sections. Sometimes if they are too far from the crossing, or to sparsely spaced leading up to the crossing, it can cause these types of problems.
-Consider modeling a bridge as a culvert, particularly if it has a very deep deck and small relative opening. Likewise, consider modeling a culvert as a bridge, particularly if it has a very large opening, relative to the deck thickness (conspan culverts are good examples).

If the problem is only a very small discrepancy in upstream head levels, a refinement of the computation tolerances might yield better results. For example, let's say you are trying to provide a "no-rise" condition, and you feel that there should be no rise (i.e. new bridge opening is bigger than old bridge opening with a higher low chord). However, you're results are showing a 1 to 2 hundreths of a foot of rise for the new bridge. This is most likely numerical issue and a refinement of the computation tolerances might yield better results.