Pages

Friday, December 23, 2011

Permanent and Non-Permanent Ineffective Flow Areas

Written by Chris Goodell, P.E., D. WRE | WEST Consultants
Copyright © RASModel.com. 2011. All rights reserved.

The traditional way of changing non-permanent ineffective flow areas to permanent involves going through each cross section, one at a time, in the cross section editor, selecting Options…Ineffective Flow Areas, and then manually checking which ineffective flow areas to make permanent. 
image
This can be long and tedious if you have a lot of cross sections to change.  HEC-RAS has an easy-to-use method for changing ineffective flow areas from non-permanent to permanent for multiple cross sections.  In the geometry editor, select Tools…Ineffective Areas…Set to Permanent Mode.  
image
Here you select the cross sections with which you want to change the non-permanent ineffective flow areas to permanent, and RAS will do that for you.  If you have a lot of cross sections where you want to do this, this utility is much faster than the traditional way of selecting each individual cross section in the cross section editor and manually changing the status.

There are a couple of disadvantages to the former method:  First, with the traditional method, if you select a cross section to have its ineffective flow areas changed to non-permanent, all of the ineffective flow areas will be changed for that cross section.  You cannot pick and choose within one cross section.  Second, once you have changed a cross section’s ineffective flow areas to permanent, there is not a similar multiple cross section selection utility to go back to non-permanent. 

Dr. Ray Walton, of WEST Consultants passed along a very easy way around this second disadvantage.  First, open up the geometry file (*.g##) that you want to edit in a text editor.  Wordpad, Notepad, WORD will all do.  Go to the search tool and search on the string “       T”.  That’s 7 spaces followed by a “T” which means “true”.  This string is used in the text editor to indicate an ineffective flow area is permanent.  Simply replace that text with “       F” (seven spaces followed by an “F”). 

image

A simple “search and replace” action will allow you to make this change for a number of cross sections very quickly.  If you have specific cross sections you wish to do this to, first search on the River Station.  In the example above, the River Station is “11576”.  Then scroll down to “       T” and change it to “       F”. 

As always, be very careful when editing a RAS file directly in a text editor.  It's very easy to corrupt the file if you make a small mis-type (i.e. you put in 6 spaces instead of 7 before the "F").  I always suggest making a copy of the original file for safe-keeping, before you make your edits. 

Monday, December 19, 2011

Theta Implicit Weighting Factor and its Effect on Sample Datasets

Written by Aaron A. Lee   | WEST Consultants
Copyright © RASModel.com. 2011. All rights reserved.
Adding to the previous topic on the Theta Implicit Weighting Factor (Theta), this post takes an objective look at how the unsteady-flow option affects model output. Theta is a weighting factor for the spatial derivative used in solving the finite difference forms of the St. Venant equations. Adjusting Theta can improve model stability or increase the accuracy of the output. In a practical sense, how much is Theta really changing the solution? This post observes the influence of Theta by running the 21 installed (sample) projects in HEC-RAS version 4.1.

Theta can be found by navigating to Calculation Options and Tolerances under the Unsteady Flow Analysis Options menu. The default value is 1.0, but the user can define a value of Theta anywhere between 1.0 and 0.6. A value of 0.5 represents a half weighting explicit to the previous time step’s known solution, and a half weighting implicit to the current time step’s unknown. A value of 1.0 gives a fully implicit formula that is highly diffusive. In theory, a higher value will improve model stability but is less accurate in the solution. The opposite is true for lower values of Theta, which can make the model more sensitive to errors and lead to oscillations.

The table below summarizes the results for the sample projects included in the experiment. Water surface elevations (WSEL) are compared at each river station between the current plan and the default plan, Theta = 1.0. The values in the table are the largest maximum differences in WSEL for the entire reach. The cells in red are the plans that failed.

image

Apart from the three crashed runs, the difference in the solutions is very small. The results demonstrate that for these simulations, the model is not highly sensitive to changes in Theta. Keep in mind that these models are relatively simple (shorter reach lengths, plain structures, uniform geometry) when compared to other unique project situations.

Changing Theta has a direct effect on how the solution is solved, but other factors may have more of an effect on stability and accuracy. The Hydraulic Reference Manual notes that factors such as cross-sectional properties, abrupt slope changes, flood wave characteristics, and complex hydraulic structures often overwhelm any stability considerations associated with Theta. When testing a model, pay special attention to the stability considerations listed above before laboring over Theta. While lowering Theta will yield (technically) more accurate results, it can also propagate errors where other factors may be causing problems. The User’s Manual suggests making sure that the computation interval is accurately defined, and that the maximum number of iterations is reasonable.

The HEC-RAS User’s Manual (page 8-32) suggests starting out with a Theta value of 1.0. Paraphrasing from the User’s Manual, page 8-32: “Once the model is up and running, the user should experiment with changing Theta towards a value of 0.6. If the model remains stable, then a value of 0.6 should be used. In many cases, there may not be an appreciable difference in the results when changing Theta from 1.0 to 0.6. However, every simulation is different, so you must experiment with your model to find the most appropriate value.”

The results of adjusting Theta for the 21 sample projects validates the approach suggested in page 8-32 of the HEC-RAS User’s Manual.

Wednesday, November 30, 2011

Initial Conditions Trick

Written by Chris Goodell, P.E., D. WRE | WEST Consultants
Copyright © RASModel.com. 2011. All rights reserved.

I’ve mentioned many times on this blog and in class about the importance of having your initial conditions flow match your first time step flow, when running an unsteady flow model.  As with any “rule” in RAS modeling, there are exceptions to this, but generally speaking, if your first timestep flow is “X”, then your initial conditions flow should also be “X”. 
image
This is generally very easy to do-if you have a single reach.  The hard part is just remembering to do it.  So, if you know you’re going to be changing your initial flow a lot, maybe while testing different hydrologic scenarios, or just trying to stabilize your model, you can save yourself some time and extra mouse clicks by leaving the initial flow cell blank.  By leaving it blank, you’re telling RAS to just use the first timestep flow provided in the flow hydrograph.  Give it a try, it works great.
image
This works perfectly for single reach systems.  For more complex systems, there is a catch.  For dendritic systems, this trick only applies to the upstream end of Rivers, not at junctions.  For looped systems, this trick only applies to the upstream end of Rivers that do not originate as distributaries from a junction.  Don’t worry, if you fail to enter an initial flow value where one is required, RAS will tell you about this when you try to compute. 

Monday, October 31, 2011

Breaching of Condit Dam

Written by Chris Goodell, P.E., D. WRE | WEST Consultants
Copyright © RASModel.com. 2011. All rights reserved.

Last Friday, Condit Dam, on the White Salmon River was intentionally breached by blowing a hole in the spillway as part of a plan to remove the dam and restore the river to its natural state.  Although the actual breaching of the dam was not all that spectacular (i.e. the dam itself wasn't destroyed-only a relatively small hole was blown through the base of the spillway), the drawdown and subsequent erosion of the reservoir was very interesting.  It was all captured in "time-lapse" in a video posted at the following location:


Please enjoy this video.  I'm sure there will be more to come later.

Tuesday, September 13, 2011

Dam Failure of a Coal Slurry Impoundment

Written by Chris Goodell, P.E., D. WRE | WEST Consultants
Copyright © RASModel.com. 2011. All rights reserved.

I was recently asked my opinion on a good way to model the following dam breach event:
I …recently completed three consulting projects where I simulated the breach of three proposed coal slurry impoundments.  The permitting agency required a RAS model of an “instantaneous” hypothetical breach (over full depth, almost 80 ft for one of the impoundments).  I was able to achieve a stable model with brief breach formation time and satisfy the permitting agency.  The client (a coal company) considered the results to be unrealistic due to the rapid failure time and the fact that much of the impoundment is very viscous slurry; they have asked me to revisit the problem.  They asked me to model a partial breach of the top 10 ft, which they estimate is the distance from the top of slurry to the top of impoundment and occupied by water for the failure scenario, followed by the viscous slurry.  I was wondering if HEC-RAS could model such a complex situation.  I was thinking it might be modeled using the sediment transport capabilities within RAS.  I do have properties of the slurry, including particle size distribution, etc.  I suspect a more complex model is needed, but wanted to get your opinion, since I frequent your blog and have seen many complex issues addressed with RAS.
Thanks to Jason Hill, Ph.D., P.E. for sending in this interesting problem.  I don’t know if it is ultimately the best solution, but one that I think may work is as follows:
to model the breach of a partial water, partial slurry impoundment, you’re going to have to get creative.  First of all, RAS technically cannot model highly viscous fluids, like mud or slurry flows.  Really your only option for a “RAS-Only” model is to bump up Manning’s n values to account for the highly viscous flow.  Without a means of calibrating these high n values, you really are just guessing when you increase them. 
Here’s my suggestion:  Not sure if this would work, but what I would explore is the use of a combination of HEC-RAS, NWS BREACH, and FLO2D.  First, assume the first “pulse” of flow (water flow) will be separate and distinguishable from the second pulse (slurry flow).  The initial (water) part of the breach and the first pulse can be modeled and mapped using HEC-RAS exclusively.  For the second pulse of flow, I would model the remainder of the breach using NWS BREACH.  This model will simulate the breaching process and will generate a breach outflow hydrograph for you.  An advantage of NWS BREACH over RAS is that it provides an input for sediment concentration of the breach flow.  Once BREACH has provided you with a breach outflow hydrograph, use that as the inflow to a FLO2D model.  I say FLO2D only because I’m familiar with it and it can model highly concentrated mudflows.  But any model that you can find that models mudflows will work in this case. 
In summary, you’ll end up with two hydrographs to route downstream and to map independently:  the water hydrograph, and the slurry hydrograph.  The “water” breach will be modeled, and the water hydrograph will be routed using HEC-RAS.  The “slurry” breach will be modeled with NWS-BREACH, and the slurry hydrograph will be routed using FLO-2D (or other model capable of simulating mudflow).”
Although I know a little bit about NWS BREACH and FLO2D, I freely admit I haven’t tried this before. I think it can be made to work but I can also foresee a few hurdles.  Namely, what happens when/if the slurry flow and the water flow ultimately mix together somewhere downstream?  How do you map that condition?    If any of you out there have other suggestions, please feel free to comment to this post.  

Friday, September 2, 2011

Overflow Gates

Written by Brian Wahlin, Ph.D.,  P.E., D. WRE | WEST Consultants
Copyright © RASModel.com. 2011. All rights reserved.
In modeling irrigation canals in HEC-RAS, a typical structure that is encountered is a check gate. Check gates are designed to back the water up behind them in an effort to keep the water level immediately upstream of these gates at a constant level. As the flow rates change in the canal, the gate openings on the check gates are adjusted (i.e. opened or closed) in order to pass the new flow rate while maintaining the water level upstream of the check at the desired elevation. Why go through this effort? Farmers usually get their water from these irrigation canals via orifices just upstream of the check gates. The flow through these orifices is dependent on the head (or the water surface elevation) acting on it. Since farmers want a constant flow rate delivered to their fields, the check gates in the main canal are adjusted when the flow changes to make sure the water level upstream of the check (and hence the rate of flow delivered to the farmers) remains constant.
Check gates can take a wide variety of forms, but typically they fall into two categories: undershot gates and overshot gates. Undershot gates are things like radial gates and sluice gates. As the name implies, water shoots “under” these gates. Inline weirs are typical examples of overshot gates. For these structures, water does not pass “under” the gate but instead flows over the top of the structure. Since many irrigation districts operate on limited budgets, it’s not uncommon to see an overshot gate made simply of several 2x4 wood boards that slide into groves in the canal walls. The irrigation canal operator “opens” these types of gates by manually removing one or more of the wood boards. In a similar manner, these gates are “closed” by adding one or more wood boards.
Modeling undershot gates has been straightforward in HEC-RAS for many years. You simply select whether you have a radial gate or a sluice gate and then enter the appropriate input data for the gate. The gate openings are then set through the flow editor-either steady state or unsteady state-depending on your situation.
Modeling overshot gates in HEC-RAS has been a little more challenging. Unlike the undershot gates, there really wasn’t a gate type that allowed water to flow over the top like a weir. Thus, you had to model overshot check gates using the geometry of the inline structure. An example is shown in the figure below. The four long white rectangles in the middle of the structure are undershot types of gates (in this particular case, they happened to be sluice gates). The two short white rectangles at the far right and the two at the far left (the ones with the open top) are overshot gates. Because in previous version of HEC-RAS, there wasn’t a particular “overshot” type of gate, you were stuck with coding these gates using the weir/embankment button. While this is perfectly fine for steady state mode where you only have one weir height setting, it becomes problematic if you run the model in unsteady mode and the weir height changes during the simulation. Because you have to model the weir as part of the geometry, there is no easy way to adjust the weir height as a function of time in unsteady mode.
clip_image001
Starting with HEC-RAS version 4.0, there is a new type of gate called an “overflow gate.” As the name implies, this type of gate allows water to flow over the top of the gate as shown in the schematic below. There are two types of overflow gates in RAS: open air and closed top. Both of these gates allow water to flow over the top of the structure. The difference is that the closed top gate is kind of like an elevated orifice. At some water levels, water will flow over a closed top gate like a weir. At higher water levels, the closed top gate will function as an orifice.
clip_image002
For irrigation canals, open air overflow gates are exactly what we need to model overshot check structures. To use this option, enter the inline structure information exactly as before. Now, select “Overflow (open air)” as the gate type. There are three types of weir shape methods: Broad Crested, Sharp Crested, and Ogee. For modeling irrigation checks, using the Sharp Crested option is probably most appropriate. With this option, there are three ways to enter the discharge coefficient for the weir equation: User entered coefficient, Rehbock equation, or Kindsvarter-Carter equation.
clip_image004
The gate opening is still set through the flow editor (either steady state or unsteady state). For overflow gates, the gate opening is now from the top of the gate rather than the bottom. The figure below shows a check structure modeled with open air overflow gates. As can be seen, the gate opening on the far left is 1.56 feet. But this distance is measured from the top of the gate frame rather than the bottom. Now, if you are running this model in unsteady mode and the gate opening changes (that is, another weir board is added or removed), you can reflect this in the gate opening boundary condition.image

Tuesday, June 21, 2011

Hotstarts and HTab Parameters at Bridges

Written by Chris Goodell, P.E., D. WRE | WEST Consultants
Copyright © RASModel.com. 2011. All rights reserved.


An advantage of running a step-down scheme hotstart run is the ability to spatially evaluate stability issues with difficult reaches. One of the elemental features of the step-down scheme is the artificially raising of the downstream boundary during the hotstart simulation to “drown-out” the reach and effectively create a very stable environment. During the hotstart simulation period, problem areas will identify themselves as the water surface elevation slowly lowers itself into a realistic solution. This is a great way to diagnose instability issues. However, if you have bridges or culverts, drowning-out the reach creates water surface elevations that are much higher than the normal water surface range you’d expect at those bridges, and a normal set of HTab parameters may not work well during the initial period of your hot start simulation. I’m speaking specifically about the “Head water maximum elevation” HTab parameter that is required at bridges and culverts. The figure below shows the problem that occurs with a normal headwater maximum elevation during a step-down hotstart run. Notice the flat pool downstream of the 3rd bridge, followed by a severe drop in water surface elevation. In a good hotstart simulation, the downed-out reach should show a consistently level water surface elevation. Ultimately this hotstart simulation crashed.

image

At first, it seems like an easy fix: Simply increase the head water maximum elevation for the affected bridge to an elevation around the “drowned-out” condition. In this case, I increased the head water maximum elevation to 350 ft, which is equal to the initial drowned-out elevation set at the downstream boundary. I make this fix and the profile looks good and the hotstart simulation runs to completion without errors. The figure below shows 6 profiles using the hotstart simulation with a step-down scheme. Notice the level pools as the water surface steps down to the true initial conditions.

image

Now that we have a good, stable hotstart run, switching to the real plan should be seamless, right? Actually, now that I’ve expanded the range of computation points for the rating curve at the 3rd bridge by using an artificially high head water maximum elevation, I’ve lost a lot of resolution in my rating curve for my real plan-particularly down in the range of real water surface elevations. Notice in the figure below that my expanded HTab curves go up to elevation 350 ft. Since I have a finite number of submerged curves, and points on the surbmerged and free flow curves, I have a loss in resolution in the range of realistic solutions (namely down in the 250 ft range).

image The real plan then crashes when the front end of flood wave reaches the 3rd bridge. The figure below shows error at the upstream end of the bridge that leads to the instability-and ultimately the crash.

image

There is actually a very easy fix. Simply change the headwater maximum for the “real” plan to a realistic maximum water surface elevation for that bridge. In my case, I reduced it from 350 ft to 270 ft. If you read through the hotstart posts in this blog, you’ll notice that I say any change in geometry will require a re-run of the hotstart plan, before running the real plan. To stabilize the hotstart plan, you’d have to put the headwater maximum elevation back to 350 ft, and then we’d be back to where we started. However, this is a rare exception to the rule. Luckily for us, changing the headwater maximum does not prompt HEC-RAS to want to re-run the hotstart plan. So we are free to change the HTab parameters at bridges for the real plan. Just make sure that if you DO re-run the hotstart plan, you change the headwater elevation back to the “drowned-out” condition (350 ft). Then “re-change” it to the realistic condition for your real plan (270 ft). It also helps to maximize the number of submerged curves, and the number of points on the free flow and submerged curves. This too provides more resolution. The figures below show the real plan solution with the more refined HTab parameters and the resulting profile plot. Notice the HTab curves are squeezed to a narrower range, providing more resolution. The difference in the solution on the profile plot is subtle, but makes all the difference between a stable and unstable solution at this bridge.

image

image

Monday, May 23, 2011

Modeling Junctions for Unsteady Flow Analysis

Written by Aaron A. Lee   | WEST Consultants
Copyright © RASModel.com. 2011. All rights reserved.

In the current version of HEC-RAS (v 4.1.0) there are two methods of modeling the hydraulics at a junction for unsteady flow. By default RAS selects the Force Equal WS Elevations (Forced) method, which forces the upstream bounding cross-sections’ water surface equal to the downstream water surface. This method may be adequate for some situations like high depths and shallow bed slopes, but can also cause major instabilities in your model if depths are too low and/or bed slopes are too steep. The alternative is the Energy Balance (Energy) method, which uses the energy equation across the junction to solve for WS elevations. The model presented in this post is part of a dam breach simulation and will demonstrate that there can be significant differences between the two methods. This simulation is a hotstart run which seeks to identify stability issues by starting the downstream stage artificially high, and slowly lowering it to the true solution over the run time. The river system in this model has a normal flow combining junction with a steep transition. Special attention will be paid to the steep transition, especially at low flow conditions. The Figure 1 below shows the 3D view of the model extents, which includes the Middle Reach, Tributary C and Lower Reach.


image Figure 1


For a normal flow-combining junction, the water surface elevations at the upstream bounding cross-sections are based on the computed downstream WS elevation. Longer lengths between the bounding cross-sections will generally make your results less accurate and less stable. By looking closely at the above figure you can see that the bounding cross-sections are spaced far apart, which corresponds to long junction lengths. The results for both methods are shown below in a series of profile plots. Figures 2a and 2b show the junction approximately halfway through the simulation. Figure 2a shows the Energy Balance method and Figure 2b shows the Forced Equal Water Surface method. The water surface is high enough that there are no differences between the two methods. For reference, Tributary C is the steeper of the upstream reaches.


image Figure 2a


image Figure 2b


Significant differences develop in Figures 3a and 3b as the downstream stage is lowered. At the same time-step, the two profiles are dramatically different. The Forced method produces a large drop at the junction (Figure 3b), while the Energy method produces only a minor instability (Figure 3a). The large drop (shown in Fig. 3b) occurs because RAS must balance the momentum equation from the upstream bounding cross-section at the junction (an unrealistically low water surface) to the cross-section immediately upstream. The only way to provide a balance is to overestimate the upstream WS elevation, which is why the profile for 3b is much higher than 3a. Notice the spike in the energy grade line. The same problem occurs for the Energy method, but at a much smaller degree.


image Figure 3a

image Figure 3b


Figures 4a and 4b show Tributary C only, just prior to the model crashing for the Forced Equal Water Surface method. There are obvious oscillations in the profile plot, which indicates a very unstable solution. As the stage is lowered downstream, the WS elevation at the junction also becomes lower. At a certain point the WS elevation at the junction approaches the invert for the upstream cross-section; and the model crashes. Figure 4b shows a zoomed in view of the WS elevation relative to the invert of the channel as the channel runs dry.

image Figure 4a

image Figure 4b


The best solution is to shorten the junction lengths as much as possible, which is done by adding cross-sections closer to the junction. By adding additional cross-sections you are decreasing the length over which RAS makes its calculations, which helps to remove the problems with low water surface elevations over a junction. If surveyed data is unavailable, then start by copying the most downstream cross-sections of the upstream reaches to a location closer to the junction. The positioning of these new cross-sections will be based on the judgment of the modeler, who should know the actual conditions of the river system. Make sure to adjust the downstream reach lengths and junction lengths accordingly.


In this example, cross-sections were placed within 20 ft of the junction. Junction lengths were changed from 573’ and 534’ to 35’ and 28’ for Tributary C and Middle Reach, respectively. In addition, cross-sections were added every 40’ on the steep section of Tributary C by interpolation. Figures 5a and 5b each show the profile plots for both the Energy and Forced method at the junction of Tributary C and Lower Reach.

image Figure 5a

image Figure 5b


By redefining the geometry around the junction the error is significantly reduced for both methods and the results appear stable. Both profiles are very similar in this case, showing only a slight difference in WS elevations. The dotted line-type represents the profile for the Forced method. It might not always be possible, or realistic, to place new cross-sections close to the junction. The Energy method allows this model to run to completion without the addition of new cross-sections, though the results appear to not be as good. The table below lists the WS elevations at the bounding cross-sections for each of the different plans: the initial Energy method, and the Forced and Energy method after adding additional cross-sections.













The initial plan has the geometry with the long junction lengths, which consistently calculates lower WS elevations than the plans with shorter junction lengths. Although the elevations were underestimated in the initial runs, they are still within 1 ft of the new profiles. For this model, the Energy method provides a stable solution at the junction without having to modify the geometry. However, given the steepness of Tributary C, the addition of cross-sections near the junction improved the accuracy and stability of the model output. Therefore, even though the Energy method can produce stable results for long junctions in steep reaches, adding more cross sections will improve the results.

Friday, April 1, 2011

Mixed Flow Regime Options – LPI Method

Written by Aaron A. Lee | WEST Consultants
Copyright © RASModel.com. 2011. All rights reserved.
By using the Mixed Flow Regime option for Unsteady Flow Analysis, RAS can better handle transitions from subcritical to supercritical flow. This option should be utilized only after determining that a mixed flow situation exists, which requires judgment from the modeler. One application where this could be particularly useful is dam breach modeling, or any other extreme and flashy flood event. Even though a model is stable there may still be small errors in the solution (caused by max. iterations). The Local Partial Inertia (LPI) factor may eliminate or reduce these errors, particularly if they occur when the Froude number is near 1. Figure 1 shows the Unsteady Flow Analysis window with the Mixed Flow Regime option selected. This post will focus on the LPI Filter, which is enabled when Mixed Flow Regime is selected by the modeler.


clip_image003


Figure 1. Unsteady Flow Analysis Window


Once the Mixed Flow Regime option is selected, additional settings can be adjusted to help stabilize the model. Navigate to Options, Mixed Flow Options. This window, shown in Figure 2, allows the user to adjust two inputs for the LPI factor.


clip_image007


Figure 2. Mixed Flow Options Window


For the unsteady flow computation scheme, RAS accounts for a local acceleration and convective acceleration (inertial terms) through the St. Venant equation of Conservation of Momentum. The St. Venant equations, and by extension, HEC-RAS, are designed to work best in gradually varied flow. Transitions from supercritical to subcritical flow (hydraulic jump), and to a lesser extent subcritical flow to supercritical flow, are rapidly varied flow situations. These are not gradual changes, in the hydraulic sense. Near critical depth (Froude number approaching 1) the convective acceleration terms can change very rapidly over a short distance (think of a hydraulic jump) and can lead to oscillations in the solution. These oscillations tend to grow larger until the solution goes completely unstable (HEC, 2010). The LPI factor systematically reduces these inertial terms to dampen the oscillations, helping to stabilize the model. The user can influence the magnitude of reduction by varying the two inputs in Figure 2.


The first input, m, is the exponent for Froude number reduction factor. Its default value is 10 and ranges from 1 to 128. Adjusting m will change the shape of the curve on Figure 2, thus influencing the rate of reduction of the inertial terms. You can see that by making m smaller there is an earlier and more direct reduction in the inertial terms, with respect to the Froude number. Increasing m can make the model more accurate but increases the likelihood of numerical instability.


The second input, FT, is the Froude number threshold at which the LPI factor is set to zero. In other words, if the calculated Froude number at the current cross-section is larger than FT the inertial terms will be eliminated from the computations at that cross-section for the current computational time step. The default value is 1. Making FT smaller will improve the stability of the model, but will also reduce the accuracy. A larger FT can make the model more accurate, but increases the likelihood of numerical instability as the inertial terms will be more sensitive to fluctuations in Froude number.


A good place to start is to run the simulation with the default values




and see what the profile looks like. For this flume example, the model ran without reporting any maximum water surface errors. The profile for the default LPI inputs is shown in Figure 3.


clip_image010


Figure 3. Profile Plot, Default Values


Next, a value of 1.6 was chosen for FT. This simulation yielded small maximum water surface errors, but had maximum iterations at various locations. The value of m was left unchanged. Figure 4 shows the results.


clip_image013


Figure 4. Profile Plot, Increased Froude Number Elimination Threshold


Even though the errors were small, instabilities could be seen in the downstream end. Notice the instabilities around the transitions between the flow regimes. The value of m was reduced from the default of 10 to 7 in order to improve the stability of the model. Figure 5 shows the profile for reduced m and increased FT. The modeler should choose the largest values of m and FT that produce a stable model. However, check the results to make sure that the output is reasonable. Notice how the transitions between flow regimes are much better defined in Figure 5 then the default setup shown in Figure 3. That’s because the default LPI parameters (m = 10 and FT = 1) provide dampening of the results. Though Figure 3 looks very stable (and it is), Figure 5 (m = 7 and FT = 1.6) is both stable and (by my engineering judgment) more accurate. Also, notice how the slight increase in energy (green dashed line) is less in Figure 5 versus Figure 3. An increase in the energy elevation in the direction of flow is an indication of error in most cases. Further adjustment of the LPI parameters may help to eliminate the error in the energy grade line, while still producing a stable solution.


clip_image015


Figure 5. Profile Plot, Increased Froude Number Elimination Threshold and Decreased Exponent ,m

Wednesday, March 9, 2011

More on HTab Parameters

Written by Chris Goodell, P.E., D. WRE | WEST Consultants
Copyright © RASModel.com. 2011. All rights reserved.

I’ve mentioned this a lot in this blog, but I’m finding more and more that a good, solid definition of your HTab parameters up front will go a long way in helping construct a nice stable unsteady flow model in HEC-RAS.

First, let me recap how they work. In unsteady flow, HEC-RAS will convert the geometry of cross sections into a set of curves defining relationships between hydraulic parameters and stage (it does the same thing for bridges and culverts, but we’ll save that for another post). These hydraulic parameters include conveyance, flow area, storage area, and top width. This is done for the main channel, overbanks, and for the total cross section. Storage area in this case represents any ineffective flow areas in the given cross section. The figure below shows the conveyance HTab curve for a cross section.

image

These curves (actually the paired data that creates them) are used by HEC-RAS during the unsteady flow computations. Rather than re-computing the hydraulic parameters at every time step, RAS can simply pick the value off the chart. Because these curves are defined by a number of discrete points, RAS usually must interpolate to grab a value in between points. RAS does this linearly. Herein lies the problem that can lead your model to errors and instabilities.

Notice how at the lower stages, there is a significant curvature to the conveyance relationship in the above figure? A linear interpolation in this range can be quite wrong if the resolution of points is too coarse. Notice how when I zoom in to a rather coarse Conveyance HTab curve, it becomes obvious that in between points, linear interpolation is going to give me a bad answer. This is why a good definition of HTab points is particularly advantageous when running at low stages.

image

Start by maximizing the number of HTab points for you cross sections. HEC-RAS allows up to 100. The grey horizontal lines in the figure below represent computation points at different stages. I know what you’re thinking. “Won’t that many points on every cross section really slow down the computations?” Well, that may have been the case years ago, but computers are so fast now, that you’ll probably never notice the difference. Furthermore, the development of these curves is only done once, during the preprocessing part of the computations. As long as you don’t change the geometry, RAS won’t have to recompute the HTab curves.

image

Then make the computation increment as small as possible to squeeze all the points together. You only need to extend your HTab curves to contain the maximum computed water surface.

image



Finally, go to the Stating El. column and click the button “Copy Invert”. Notice in the figure below there is a gap between the invert of the channel and the first computation point which is set 1 ft above the invert of the main channel. image This is the default starting computation point in RAS (well, not completely true). If you check the 2nd figure up above, you’ll see that there actually was a computation point at the 0 depth point (the invert). That’s because RAS will still compute the 0 depth point, but then the next computation point, by default, is 1 ft (0.3 meters in SI Units) above the 0 depth point. From there on up, RAS will space the computation points based on the increment you define. To get more points between the invert of the main channel and 1 ft (0.3 m) above the invert, you have to set your Starting Elevation to the invert. That’s why it’s always a good idea to click “Copy Invert” and make sure that your Starting Elevation is the same as your “Chan Min” value. That way, the small computation increment is started from the channel invert, not 1 ft (0.3 m) above the channel invert.

image



I was recently informed that in the next release of HEC-RAS (version 4.2), the default starting point for HTab computations will be 0.5 ft above the invert (0.15 m?). This will improve things somewhat, but it still may be necessary to “copy the invert”, particularly if you have very low stages in your simulation.