I have been able to find very little guidance on appropriate values for entrance (ENTK) and exit (EXITK) loss coefficients for EXTRAN. The Users Guide barely mentions it as does the SWMM 4.4 documentation. Can anyone point me in the direction of a good resource on this issue?

I realize this is a relatively new part of SWMM but I am still surprised at the lack of documentation on what seems to be an extremely important parameter.

You're right, there is very little out there, and much of what is there was intended for water distribution or open channel systems, not storm drains and manholes. I've been doing a literary search and looking for good coefficients for a couple of years, but the final write-up hasn't gotten off the back burner.

There have been a couple of good studies in the past 20 years. Jiri Marselek's work for APWA was probably the one that gave the best numbers for design or modeling work. He did flume studies, but concentrated on full-flow measurements. There's still not much experimental data on open channel flows through manholes. We need some basic research. Anyone with a flume interested?

I will try to light a fire under myself and get the results of my literary search written up this year.

Here is some additional information on the effect of entrance, exit and other losses in the Extran hydraulics routing in SWMM. As far as other information or other sources, I find that text book estimates, FHWA culvert nomograph estimates, HEC-2 and HEC-RAS recommendations, etc. work comparably in SWMM.

Minor losses such as entrance and exit losses through the manholes of a SWMM hydraulic network can be simulated in the Extran layer by assuming a loss coefficient for the entrance, exit or "other loss" categories. Examples of "other losses" would be bend losses. The table below shows the head loss in feet for various combinations for the loss coefficient and modeled velocities using the loss equation K*V2/2g as discussed in the paragraph below the table.

Table Relationship between the minor loss coefficient, velocity & head loss in a conduit.

K factor

Velocity

Loss

K factor

Velocity

Loss

(dimensionless)

(ft/s)

(ft)

(dimensionless)

(ft/s)

(ft)

0.05
1
0.001
0.05
6
0.03
0.10
1
0.002
0.10
6
0.06
0.25
1
0.004
0.25
6
0.14
0.50
1
0.008
0.50
6
0.28
1.00
1
0.016
1.00
6
0.56
0.05
2
0.003
0.05
8
0.05
0.10
2
0.006
0.10
8
0.10
0.25
2
0.016
0.25
8
0.25
0.50
2
0.031
0.50
8
0.50
1.00
2
0.062
1.00
8
0.99
0.05
4
0.012
0.05
10
0.08
0.10
4
0.025
0.10
10
0.16
0.25
4
0.062
0.25
10
0.39
0.50
4
0.124
0.50
10
0.78
1.00
4
0.248
1.00
10
1.55

The entrance/exit/other loss term in the momentum equation has a slope (See) defined by:

which fits in the St. Venant equation similarly to the friction slope and bed slope term:

This loss is calculated at each time step (if you use the right NEQUAL parameter) based on the current upstream, midpoint or downstream velocity. The important suggestion is that you can use other literature sources for the values of the coefficients and Extran will apply those unmodified coefficients in the proper manner in the St. Venant equation.

I'd recommend looking at FHWA's Hydraulic Engineering Circular No. 22 - Urban Drainage Design Manual. There is a section on Energy Losses that discusses losses due to transitions, junctions, & manholes. The MH sections discuss deflection, benching influence, plunging flow, etc. The reference may also lead you to other good publications.

Entrance and Exit losses seem pretty straight forward to me in EXTRAN, but how about Junction/Momentum losses due to lateral flow coming into a manhole from another conduit and\or from surface flow plunging down into the manhole? Do the St. Venant equations handle this at all?

Yes, I'm well aware of HEC-22. My question though is with EXTRAN.

If the algorithms involving the St. Venant equations do not account for these losses, then is work going to be done to incorporate this into the new versions of SWMM?

Hmm - must have done something wrong. Thanks Bill!

Let me add another good reference - Larry Mays "Stormwater Collection System Design Handbook", again a chapter by Ben Chie Yen, more extensive than in the Hydraulic Design Handbook.

BTW, I've looked at most of these references ...

Sangster's results can be converted to head loss coefficients (see below) and they match pretty well with Marsalek's and others' measurements of full flow loss coefficents once you've done that. Sangster did a lot of very good research, then presented the results in a form which, as you mention, is difficult to use in SWMM.

I started with ASCE MOP 72 and found that most of the reported coefficients are either are not backed by any research I can find, or can be traced to water distribution or open channel research. That's what got me started on this.

I've also looked at a lot of State DOT Design Criteria that give loss factors with no research backing them up whatsoever, which is quite disappointing, because the method of applying them is usually very straigtforward. I get nervous when adjacent states have methods that give loss coefficients different by a large percentage and there's no way of figuring out why.

The approach I've come around to uses reported flume research on sewers or storm drains to derive a loss coefficient from manhole geometry (e.g. ratio of upstream/downstream pipe diameters, ratio of manhole diameter to pipe diameter, and invert benching). That's for straight full flow through a junction. Once that calc is done, I'd like to apply factors for open channel flow and bends. And then apportion it to entrance and exit losses. And then deal with the problem of two or more flows into a junction, which are much more difficult.

Gotta go look for those matches. Even a 3/4 complete writeup sounds like it would be worthwhile. :-)

I'd like to directly address Fred's question about whether the St. Venant equations as implemented in SWMM (either 4.x or 5) incorporate momentum losses due to lateral inflows or from plunging flows. The answer is no. In fact, the Extran code doesn't even account for the momentum flux term (dQv/dx) in the momentum equation the way most other solutions do. The latter methods, like the 4-point implicit scheme, represent this term in the equation for flow in link i with 2 terms:

[v(i+1)Q(i+1) - v(i)Q(i)] / dx

where there is influence on the downstream link i+1 from the upstream link i. In Extran, the momentum equation for link i contains only variables that are local to link i. That's one of the reasons why in the SWMM 5 Redevelopment Plan we mention the possibility of adding additional dynamic wave solution methods into the code (at some later date).

Another factor to consider when looking at tabulated loss coefficients is whether they apply to (or were derived from) steady flow conditions and if so, how applicable would they be to conditions being modeled by dynamic wave routing.

Thank you Mr. Rossman for the clarification. This is where I always get hung up on "selling" the model to local regulators and colleagues not familiar with EXTRAN in terms of HGL\EGL capabilities. I usually end up offering the red herring about how the rigorous dynamics and routing more than make up for not accounting for these losses...

Bill - I totally agree - 3/4 would be better than what we have..

I have a reservation on that formula. If Dd=Du the formula defaults to Ke=Kp, and the Sangster and Wood graphs range between -6 and 3 or more. such values are well out of reasonable range. Also, Kp has different meanings in some cases. I must admit my head was swirling after reading "Head Losses in Storm Drain Junction Boxes" HRB Proc.v35, '56, by Wood, one of Sangster's collegues. They collaborated with others in ASCE HYD 6/59, and that is not any easier to use, at least for SWMM. What one wants of course is the head loss coefficiient to be applied to the velocity head. Some of there graphs relate the pressure coefficient upstream to the velocity head downstream, for example.

What might be a good sub-project for that book of yours would be to get all of Sangster & Wood's data again, put it on a spreadsheet and convert it to standard head loss coefficients. And who knows - there might even be some spare bucks somewhere inside the beltway to pay for it. Maybe if they could just restrain themselves a bit on such things as cow-flatulence studies...

I've been following this excellent discussion all day and now I can't resist putting in my two cents worth.

I haven't been using the entrance and exit coefficients feature at all in any of my studies. Instead I've been estimating these losses and ensuring that they are covered within the friction losses by cooking Manning's roughness.

The reason is that Extran, up to version 4.4x, doesn't compute either HGL or EGL and thus I don't exactly how the energy losses computed by these coefficients would be applied. I suspect the coefficient times the velocity head would be a loss added to the computed water level for the node in question. My concern is that there is almost always a difference in velocity through a node and thus the hydraulic grade line and the energy grade line are not the same for both pipes but Extran uses a common value to represent a hydraulic grade line (quasi-HGL) for both. So, if I don't know just what the real HGL or EGL is then what do I really have when I add some more loss to this water level? Does anyone know?

It is most likely that the results would be the same if I used a Manning 'n' that didn't include a miscellaneous loss component and let the entrance and exit loss feature jack the computed water levels up the same amount. Assuming only an exit loss coefficient were applied then the upstream pipe would flow against a water level equal to the computed level but this would be an internal value that would not be output. Thus the user would see a profile showing the nodal water level which would not be the computed depth for that end of the pipe.

A few years ago I made a four pipe test file with steady state flows and no exit or entrance loss features. I manually computed and plotted the HGL and EGL and added the Extran water levels. There was a significant increase in velocity at each node. I found that Extran levels were at the HGL only for the first node from the outfall and that all other Extran levels were below the HGL by an amount equal to the sum of the pressure drops along the route from the outfall. This is logical - if the pressure changes are not considered then the induced error must equal the sum of what is ignored. I suppose that I could repeat my test file with the coefficients turned on and determine just where the Extran levels would plot but I haven't done that as yet.

I once did some research on friction factors in concrete pipes and found an old Bureau report that had measured flows and losses for several dozen conduits with sizes from 30-inches to quite large (forget the number but larger than 10 ft) and all sorts of physical condition from excellent to fairly rough. ("Friction Factors for large Conduits Flowing Full", US Bureau of Reclamation, Engineering Monograph No.7, 1965)

They presented friction factors for Darcy-Weisbach so I converted them to Manning values. There was data for 57 sites of which 18 did not raise concerns about questionable flow measurement, bends, neglect of outlet velocity heads, very high velocities or abnormally rough pipe surfaces. Of these 18 the corresponding Manning 'n' values were;

0.010 - 6 sites
0.011 - 7 sites
0.012 - 4 sites
0.013 - 1 site

There was no correlation to pipe size and, as I recall, to description of pipe surface roughness.

The 0.011 value seemed a good choice for sewer studies. I then applied Marselek's data to typical manholes and spacings - one velocity head would cover manhole losses for a good range of manhole configurations. I found that an 'n' of 0.013 would cover a 0.011 friction factor with allowance for typical manholes at 350 foot spacing.

Since every report or study I've ever seen uses 0.013, or something close, I figure I can use the same value that everyone expects to see and also feel comfortable that I am considering miscellaneous losses as well. If I used 0.013 for a friction factor and added the entrance and exit loss coefficients it would amount to including the manhole losses twice. The preceding considers manholes for small pipes, up to 21-inches. Larger pipes generally use lower loss manhole configurations but there is a junction loss aspect. I used 0.5 foot junction losses for the larger pipes at 600 foot spacing. The 0.013 covers this loss at the longer spacing, too. Thus 0.013 is a good value for all sizes with no need to worry about manholes and junctions. Atypical situations must be individually considered because there are a lot of them that a broad brush rule can't cover.

Egad, another project :-)

I agree that it would be great to get the raw data together from this study (and others) and rework it for head loss. Not sure where the data is, though. I got Sangster's original report on an inter-libary loan a few years back and the data weren't published in it. I'd've made a copy if I could have.

Regarding the formula, I don't have my calcs with me, but I'm pretty sure that whatever I did normalized the values to a reasonable range.

Good comments Reinhard - the quaisi HGL-EGL is probably why one runs into trouble in EXTRAN with high velocity heads, particularly Froude numbers greater than 1. Your experiment description was interesting and if you do run into the original, I suggest you put it up on the web for folks to see.

I go along with your idea of a simple .013 for "long" lines. The form losses become insignificant. But for short lengths they can greatly exceed internal friction effects. An n value of .013 could easily double or more, depending on the length and the magnitude of the K values.

Bill - My experience with the Sangster et al work was not good. When one is in the middle of a project and you want to use experimental data you have to draw a line when something does not feel right and move on. There were some charts that seemed reasonable, but others that seemed absurd; .with the clock ticking I had to find something more in the mainstream. I never could find a copy of the complete report either; too bad the experimental data are apparently not available.