Description

Making Your HEC-RAS Model Run Faster

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

Related Documents

Share

Transcript

1
Making Your HEC-RAS Model Run Faster
Gary W. Brunner, HEC
Introduction
The purpose of this document is to provide guidance on model modifications that can be made to an HEC-RAS unsteady flow model (1D and/or 2D) in order to improve computational speed, while maintaining model accuracy. Some of the subjects in this document are specific to 1D modeling, some are specific to 2D modeling, and others will affect both types of modeling. The following list are the items contained within this document, and what the user should focus on to improve computational speed and reduce run times: 1.
1D Model Stability (this is the #1 problem with most models) 2.
2D Flow Area Stability 3.
Dam and Levee Breaching 4.
Boundary Conditions 5.
Variable Time Step 6.
Too Many Points in Structure Station/Elevation Definitions 7.
Too Many 2D cells 8.
Pardiso Solver for Large Complex 1D Models 9.
Re-Ordering the Reaches for Large Complex 1D Models Each of these issues are discussed in detail below. Any or all of these issues could be causing your model to be computationally inefficient and the cause of long run times. Working on these issues will improve computational efficiency and reduce model runs times.
1D Model Stability
Model stability problems are often the number one source of long model run times. Improving model stability can dramatically reduce the computational time to run a model. For example, if you have a model that is going to the maximum number of iterations (or even a high number of iterations) for a significant amount of time steps, even if the errors are small, this is significantly slowing down the model computations. Improving the model in the areas where it is having trouble solving the equations (cross sections, 2D cells, storage areas, or hydraulic structures) can dramatically speed up the computations and reduce run times. Chapter 8 of the HEC-RAS User’s Manual has a very detailed discussion on model stability issues for 1D models. If you are having 1D modeling issues, please review that section of the User’s Manual very closely, in addition to what is discussed in this document. This document will discuss some of the most common model stability problems that arise for both
2 1D and 2D models which end up increasing run times. If your model is repeatedly going to the maximum number of iterations, please review these items to help reduce the number of iterations being used in the model.
Cross Section Spacing, Cell size, and Time step
1D cross section spacing, velocity of the flood wave, and computational time step go hand in hand. If you have too large of a time step for the cross section spacing and the flood wave velocity, the model will iterate more. Shown in Figure 1 is a hydrograph plot for an unstable 1D model with too larger of a time step for the cross section spacing and flood wave speed being modelled.
Figure 1. Hydrograph plot for unstable 1D model.
The same thing is true for 2D models. In general, if the time step is too large for the cell size and floodwave speeds, then the model will iterate more or possibly go unstable. Sometimes reducing the time step can speed up the model, even though it has to solve for more time steps (that is as long as it improves the model stability). In general users should follow the Courant condition criteria for selecting a time step. Meaning you should try to keep the Courant number around 1.0 or lower during the run, if you are having model stability issues. The general Courant condition is:
0.1
≤∆∆=
xt V C
wr
And therefore:
240006001200180024000600120018002400060012001800240010Feb199911Feb199912Feb1999-100000100002000030000400005000060000
Plan: Unstead lat River: Beaver Creek Reach: Kentwood RS: 5.97
Time
F l o w ( c f s )
Legend
Flow
3
w
V xt
∆≤∆
Where: V
w
= Flood wave speed, which is normally greater than the average velocity. C
r
= Courant Number. A value = 1.0 is optimal.
Δx
= Average distance between cross sections, or average cell spacing.
Δt
= Computational time step. If a model is well structured and stable, users can get away with selecting a larger time step that will produce Courant numbers greater than one, as long as the model is stable and still producing accurate answers. Additionally, if the flood wave rises slowly and falls slowly, then larger time steps can be used because the water surface, depth, and velocity are changing slowly with respect to time. Cross section spacing, cell size, and the time step should be tested for sensitivity, stability, and accuracy.
Bad Low Flow Channel Data
Many models do not have good information for the low-flow portion of the channel (i.e. 2-yr flow and lower). Poor channel data, (erratic bed profile, or channel area changing dramatically from one cross section to the next) can cause model instability or significant model iterations at low to moderate flows. It is very common to have missing channel data when using terrain data derived from LiDAR. This is due to the fact that (typically) the LiDAR cannot penetrate through the water to obtain accurate information about the terrain below the water. An example cross section, having the tell-tale “flat” channel bottom, extracted from a LiDAR-based terrain model is shown in Figure 2. When you have poor low-flow channel data, first try to verify the accuracy of the low flow channel. If it is deemed that the low flow channel is not accurate, consider using the HEC-RAS channel modification routines to cut in a trapezoidal channel that is reasonable, based on all available data sources. Possibilities of source data include: old studies (either HEC-RAS models or even HEC-2 models); FEMA studies; cross sections from old reports; channel invert profiles from older studies; using Google Maps/Earth to make measurements of channel width at the top of the channel and water line; and going out into the field and taking measurements of channel width and depth.
4
Figure 2. Cross section Extracted from LIDAR based Terrain data.
1D Initial Conditions
The initial conditions for a model are often a source of model stability issues right at the beginning of a run, but can also cause stability problems to persist for some time during the run. For 1D models, this is often a problem with not having the correct flows set up for the 1D reaches, or not having enough base flow. If the initial flows are too low, then pools and riffles will form. Where the model goes from a pool to a riffle, it may rapidly pass through critical depth. This rapid change in water surface, depth, and velocity can cause model instability and/or excessive iterations. To prevent this from happening, increase the initial conditions flows to a high enough flow to drown out these rapid transitions. An additional problem is that when there are lateral structures connected to the 1D river reaches, and the computed water surface in the river reaches is higher than the lateral structures, the model may end up computing very larger flows going across the lateral structures during the very first actual unsteady-flow time step. This is due to the fact that the program does a steady-flow backwater computation for all of the 1D rivers, for the flows it has been given as the initial conditions. If the computed water surface is higher than the Lateral
100012001400160018002000650655660665
PMF Event with Breach of Dam
Geom: Existing GIS Data Nov 2006 Flow: Station (ft)
E l e v a t i o n ( f t )
Legend
WS Max WSGroundBank Sta.08.04.08
5 Structure, thenHEC-RAS wants to send flow over the lateral structure right at the first time step. This problem is made even worse, when the lateral structure is connected to a storage area, and the initial conditions of the storage area is either dry, or at an elevation much lower than the computed water surface in the 1D river on the other side of the Lateral Structure that is connecting them (see Figure 3). The solution to this problem is to either reduce the initial flows, so the computed water surface is below the lateral structures, or set the starting water surface of the storage areas to the same elevation as the water surface in the river on the other side of the Lateral Structure.
Figure 3. Example of Initial Conditions with Lateral and Inline Structures in the model.
Other things that can help with this problem are to turn on the model warmup option and to create a restart file. You can set model warmup for a specific number of times steps (e.g. 100), and you can also set the computation interval during model warmup to something lower than your base computation interval, which will give the model a better chance of solving when there are many flow and stage discontinuities at the beginning of the simulation. A restart file is the solution to your model for a specific time step that is written to a file which can then be used to set the initial conditions for the entire model. The benefits of a restart file are good stable model conditions that can be used at the beginning of other model runs/events. However, if you change anything in your model, such as adding cross sections, removing cross sections, or adding or deleting 2D cells, then previously created restart files are
70000750008000085000580600620640660680
Bald Eagle Creek Example Dam Break Study Plan: PMF - Froehlich Breach Parameters 10/28/2008
Main Channel Distance (ft)
E l e v a t i o n ( f t )
Legend
EG 03JAN1999 0300WS 03JAN1999 0300Crit 03JAN1999 0300GroundBald Eagle Cr. Lock Haven

We Need Your Support

Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...Sign Now!

We are very appreciated for your Prompt Action!

x