[Scott Miller]

What do you have as a tailwater connection type?

[Scott Miller]

This sounds fairly straight forward. Good description. I would use a lateral structure with an “out of system” tailwater.

[Scott Miller]

When a bridge is included in geometry, whether 1D or 2D, the geometry preprocessor calculates a set of curves that relate flow rate past the bridge as a function of headwater and tailwater. It is different from modeling the crossing with 2D equations. The water surface ought to be comparable between bridge and no-bridge scenarios, but it likely won’t be the same.

My preference, if the water surface never reaches the lower rung of the bridge, is to model reaches with bridges as open channel. In design of crossings in jurisdictions I work in, the bridge would have to be one or two feet higher than the 100-year flow. Which clearance depends on the scale of the flow. Piers and abutments can be modeled as topography.

[Scott Miller]

Nice! My new personal build does 5.5 GHz with the intel 13900k, and it is much faster than my usual workstations.

There are so many factors to account for in comparing where the error comes from in these machines. I’m running a 100-year balanced hydrographs through a low gradient valley with a couple of culverts and foot bridges in the flow. The solution time on Xeon Gold 6242s (2.79 GHz) configured as a 16 core vm are takes 20% to 40% longer than an 18 core Xeon @-2195 (2.3 GHz), varying the number of sockets those cores go through, and has 10x greater volume error. The Core i9 runs in about 60% of the time, but also has about 10x the volume error (1.0% instead of 0.1%). I assume the difference in error on the Core i9 is exactly because it is not a Xeon, but why would the vm have so much more error?

If you know of any guidelines for configuring vms for computation heavy processing, my IT people could benefit.

[Scott Miller]

Point of clarification: What format is the survey? Easting/northing/elevation or distance off baseline station?

[Scott Miller]

An approach to extend 1D cross-sections within Mapper would be to use the survey to interpolate a channel, then export it as a surface. Bring that surface into a new terrain with lidar below it. Copy the surveyed cross-sections to a new geometry associated with the hybrid terrain, extend them, and use the terrain to cut new elevations.

You could, alternatively, generate a survey surface in CADD. Check with survey or CADD staff.

[Scott Miller]

QUESTION regarding how flow through gated structure is calculated. I modeled the gap under part of the fence using a gate – fully open when the water is high enough to flow through it. The gate is rectangular, but the topography is trapezoidal. About a third of the gate is obstructed toward the corners. Does RAS take the land surface into account, or is flow through the gate calculated as though the full gate rectangle has flow – based just upon the head and tail waters?

[Scott Miller]

Vineela – Have you turned on iterations for 1D/2D flow computation?

Unsteady Flow Analysis Plan – Options > Computation Options and Tolerances > 1D/2D Options tab. The default maximum iterations between 1D and 2D is 0. It does not iterate unless to tell it to. Even if there is a difference in water surface elevation over a lateral structure, the simulation holds each solution moving into the next time step. In a single time step, for the entire model, 1D/2D water surface elevation can iterate up to 20 times. Iteration of 1D/2D water surface elevations within a single time step can help the difference you see converge.

On the General tab – 1D2D Unsteady Flow Options you can also adjust the lateral structure flow stability factor. Also look into setting flow tolerances. In a 1D/2D model with multiple lateral structures, I believe it may be helpful to equalize the flow through each (more or less), such that the flow through one does not control all the iteration.

Unsteady Flow Analysis Plan – Options > Computation Options and Tolerances > View Computation Log File. Verify that the Volume Accounting is reasonable.

—

Unfortunately there is no longer image upload on the forum. You can put an image somewhere on the web, like imgur, and include a link in your post.

[Scott Miller]

It is difficult to know what is happening with the weir in your model without comparing it to the connected grid cells and terrain elevations. But easy to caution not to read too much into the particle tracing. That is only a graphic effect, and it is imperfect.

[Scott Miller]

Odd – the link does not appear to be coming through. Check for the post titled “Optimizing Your Computer for Fast HEC-RAS Modeling” in the RAS Solution. See the comments there.

• This reply was modified 2 years, 2 months ago by Scott Miller.

[Scott Miller]

The best way to answer your question is to establish a benchmark and compare the performance of different computers. Be sure to check for differences in volume error. HEC-RAS also allows you to limit the number cores used during simulation on a single computer. In the 2D models I’ve run I find that increasing the number of cores HEC-RAS is using consistently increases simulation speed, albeit with diminishing returns.

There are two more things to consider when trying to get a 2D model to run as fast as possible. 1. Pay attention to what your model is doing. If particular cells are iterating excessively to converge to a solution, or dominating courant condition time step, they may cause a simulation to take hours or days longer. 2. 2D models may take a very long time anyway.

There are quite a few comments on Chris’s post. Be sure to take a look through them.

• This reply was modified 2 years, 2 months ago by Scott Miller.

[Scott Miller]

Hello William – The error refers to weir stationing. It may help if you look at a graphic of the weir. Lateral structure stationing is peculiar in that it is measured from upstream to downstream, whereas cross-section stationing is measured from downstream to upstream. In 1D, the upstream end of the lateral weir, 0+00, should be a distance measured downstream from a model cross-section. Check the stationing to make sure each station increases (or is the unchanged) in a downstream direction.

[Scott Miller]

Thank you, Luis. Good point about perpendicular flow. The fence is oblique across the floodplain. Overbank flow is partially deflected, and the deflection would vary with stage. I did not give thought to flow direction with regards to a cell face. It seems full momentum ought to help with regard to vectors at the cell face, but, for flood plain modeling, diffusion wave ought to be sufficient. I am not familiar enough with 2D at a code level to knowledgeably switch the equation sets for this situation. What difference might a full momentum might make? How much of a problem would non-perpendicular flow be? Perpendicular flow would be an assumption, and without gauging specific to this fence – it is what it is. Reasonable?

[Scott Miller]

Check to see if you have a water surface elevation set as an initial condition for the downstream 2D flow area. If that is the case, the initial condition shouldn’t matter if it is not significant compared to the flood wave.

[Scott Miller]

If you’re looking at design alternatives, start with the culvert inlet. Wingwalls and a headwall would get more flow into the culvert. Another remedy would be to upsize the culvert. Whatever is going on at that constriction will affect how the bed adjust to stream flows.

If you’re asking about changing the modeled bed to reduce modeled overtopping, it is possible to design a transition that minimizes head loss at the inlet. But that is design for concrete form work that might be applied to a low loss aqueduct ($$), not natural bed and bank materials.

[Author]

Posts