Stuck in EPANET 9

[boedekek]

I am a GIS, NOT engineering intern for the City of Bloomington Utilities Department and have been given the tall task of creating a water quality model for all of the water pipes within the water system having a diameter greater than or equal to 12 inches. I have been working on this model for over a year using the users manual and various engineering forums as guidance. Unfortunately, I have had no values to reference my numbers (ie time pattern, pump curves, total head at reservoir etc.) and because of my relative unfamiliarity with the program am not sure if these values are even in the acceptable range of values to have a successful run. All the engineers and my supervisor at the city are also unfamiliar with the this program and it is solely up to me to create a successful water quality model. I know everyone is busy with their own jobs, but I was wondering if someone could look at my .net file and offer me suggestions that could help fix the many errors I encounter when I try and run the program. I really can't figure out the errors unless I have an EPANET expert look at them as I have little experience. Additionally, does anyone know of any EPANET experts in Bloomington Indiana that I could contact for help? I will email you the .net file from my google drive account.

Thanks
Kent

 

[fel3]

I mean no disrespect to you, but you're not the right person to complete this task (I think you know it) and it isn't your fault. Frankly, I can't understand why your boss assigned you to build AND RUN a water model when you don't have a background in civil engineering hydraulics and specifically water distribution systems. Building the basic piping network is pretty straightforward and a GIS intern like yourself should be able to handle it. However, when it comes to adding all the little details to the model (control valves, tanks, pumps, etc--if any) and dealing with water demands and demand patterns, that is best handled by an engineer with the requisite experience. It appears that you have learned a lot on your own and kudos for tackling this with only the users manual for help, but you can only take this so far and you are there.

I have to believe that at least some of the engineering firms in your area that do municipal civil engineering will have people well versed in EPANET. I don't know if your boss would accept a suggestion to bring in a local firm to either help or take over the task, but that's what I think should happen (I'm in California, so I have no dog in this hunt). In the meantime, if you can post your model I am willing to take a look at it and give you some suggestions.

==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[boedekek]

Hello fel3,

Thanks for the response. I completely agree with you. I really wanted to try and figure this out on my own and got pretty far doing it, but I don't have the skills to complete this task. Building the piping network was indeed very easy but once I had to put in all the curves and patterns and user-based controls, I was unsure whether these values were correct because of the lack of experience in water distribution systems and civil/mechanical engineering. As you can see if you pull up my model, when I run it, the reservoir is emptying even before the time series begins and I have no idea why. I believe this could be due to me assuming that the demand patterns for each specific zone directly correspond to the pumping rates within each distinct zone (which the engineer working with me who had no prior experience in EPANET advised me to use). Let me try and be more clear with the previous statement. I assumed that because a certain pump in the South zone pumps water into the South tank at higher rates (in order for the tank to remain within its optimal water elevation/pressure), the demand for water at each node in the South zone is higher because a larger magnitude of water is being drawn out of the tank when compared with other zones having pumps with lower pump rates. Does that make any sense?
Another problem was that although I did obtain the pump curves for each pump in my water system, the range in flow (GPM) was usually different for each pump. Ie WBoost1 goes from 0-3600 GPM flow while Redbud1 pump curve only goes from 0-250 GPM flow. Lastly the gate, butterfly, and altitude valves in my system don't appear to follow the user-based rules I put into EPANET.
My boss was aware of consulting firms that do this kind of stuff but she wanted to see if it could be done by an intern using EPANET in order to save some money. Do you think someone with experience in EPANET could use my .net file with help from our engineering department to build off of my work and fix all my problems so that the EPANET model could indeed work instead of having to start all over with an engineering firm?

 

[fel3]

I just started looking at your model. I will have more comments next week, but here are some to start with. By examining Google Earth, I was able to determine that you are in Bloomington, Indiana, not Bloomington, Illinois. I did not know there was one of each.

[1] The RESERVOIR elevation of 3390 ft is too high by an order of magnitude. Per Google Earth, the surface of Lake Monroe is about 540 ft. Also, your model has the RESERVOIR connected directly to the distribution system, which cannot be correct. A water treatment plant (WTP) would be required to treat the surface water and Google Earth shows one. Since modeling the hydraulics of a WTP is beyond the ability of EPANET, I suggest eliminating the RESERVOIR and using TANK MONROE as the main system feed in the southeast. If there are pumps that feed TANK MONROE based on tank level and this needs to be modeled, then I would use the finishing tanks (?) in the WTP as your reservoir and model the pumps between there and the tank.
[2] The diameter and base elevations seem reasonable for the following TANKS: MONROE, GENTRY, SOUTHWEST, WEST, REDBUD, and DYER.
[3] The dimensions provided for TANK 4MIL indicate a volume of only about 900,000 gallons, but name implies a capacity of 4 million gallons. This leads me to believe that TANK 4MIL is actually the larger tank on this site.
[4] I was not able to locate TANK HARRELL on Google Earth, which by your model should be in the hills east of S. Harrell Road, perhaps near the upper end of Stirling Ave. Regardless, the dimensions provided are unreasonable. First, the base elevation should be around 760 ft, not 100 ft. Second, a maximum level of 500 ft and a diameter of 300 ft equates to about 264 million gallons. The largest steel water tank in the USA is about 1/10 this size (

http://www.waterworld.com/articles/...ater-tank-in-us-gets-new-coat-of-paint.html).

[5] Under the menu PROJECT|ANALYSIS OPTIONS, I see that maximum number of trials is set to 5. While this is adequate for simple systems, you have lots of pumps and rules and I suggest increasing this to maybe 25.
[6] I didn't try to understand all of the rules because I don't know your system, but it occurs to me that some rules may be working against each other. For example, RULE 13 and RULE 17 both turn off the same pump but using water levels in two different tanks. I suspect that the pump actually turns off only when the second tank reaches it's set point and the first tank may be protected from overfilling by an altitude valve. If this is the case, then i would combine these rules with an IF statement and include the altitude valves. This is where an engineer (or maintenance person) who really knows the system would be a big help. This would apply to RULES 14 & 18, RULES 15 & 19, and RULES 16 & 20. It may also apply to some of the rules above these.
[7] It appears that the model is focused on the big pipes in the system and that the neighborhoods are mostly ignored. This creates a conservative model with respect to the ability of the system to move large amounts of water from one area to the other, but it also misses some important elements of the system. First, the smaller neighborhood pipes would add a small to medium but certainly significant amount of overall system capacity. Second, time period simulations based on this skeletonized system MAY produce results that don't tell you the real story and may even give you a unrealistic result (the better you know how a system actually functions, the easier it is to prepare a valid skeletonized model). Third, it is likely that some of your critical fire flow locations are inside the neighborhoods, but without the neighborhoods in the model it is impossible to tell. This has happened to me twice, where I have taken skeleton model prepared by other engineers, who proclaimed the system fully functional for fire protection, yet when I added the neighborhoods and ran fires there found that pressures were too low.
[8] It appears that you have included a few fire hydrants in your model (the short, dead-end 6-inch pipes). That is something most modelers do not include, but which I think is a good idea, at least in critical areas and/or for very long hydrant runs. It is very easy to lose several psi in a hydrant lateral and several more in the hydrant head (I have tested fire hydrants for a manufacturer: e.g. 2.25 psi @ 1000 gpm and 6.3 psi @ 1500 gpm). The other option is to delete all the hydrants and laterals and shoot for a minimum of 30 psi at the fire flow node rather than the 20 psi (?) required at the hydrant head by the fire department.

I will have more later.



==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[fel3]

Some more thoughts before I go to Yosemite to do an astronomy program tomorrow night and Monday night:

[1] The skeletonized model will be unable to tell you if there are water quality problems within the neighborhoods and, in your system, the neighborhoods are the most likely places for such problems to occur. It should be self-evident that it will take longer for the water to travel into the middle of the neighborhoods than it does for the water to get around the large pipes surrounding the neighborhoods. Dead ends are particularly prone to water quality problems. The term we use for the last places to receive water is "hydraulically remote." Hydraulically remote locations are also critical locations for testing fire flows, as are the highest points in the system.
[2] Water tanks with single inlet/outlet pipes and/or tanks whose levels don't fluctuate much are also potential sources of water quality problems. The water in these tanks may not be turned over fast enough to prevent stagnation and loss of chlorine. You should find out from your operations people how they operate these tanks. In some systems with multiple tanks, they regularly rotate drawing down one tank at a time to make sure they don't have "old water."
[3] Diurnal water demands rarely, if ever, correspond to pumping rates. Residential areas typically see demand spikes in the morning before work/school and in the evening after work/school. There may also be a small peak around lunch time. Commercial areas typically experience the bulk of their demand during the day/early evening, with little at night unless there are extensive areas to irrigate (in one very small system I know, the day/night demand ratio for commercial is about 40:1, with the peak daytime demand being even higher). On the other hand, parks and golf courses usually irrigate at night and daytime demands are much less. There are small day-to-day differences in the actual diurnal demands, but the seasonal differences are large. In addition, the weekend/holiday demand patterns are going to be somewhat different than the weekday demand patterns. However, rather than get lost in all the possibilities, we usually settle on a couple of extreme cases so that all other operational conditions are between these two extremes. Generally, water demand is highest in the summer and lowest in the winter. I would shoot for a maximum day/peak hour scenario for one extreme (this will govern for fire flows) and minimum day/nighttime for the other (this will likely govern for water quality).
[4] Overall diurnal demand patterns can be determined by doing a mass balance for water pumped and water going into and out of the tanks. It's basically calculus without the equations. In one mostly residential system I worked on extensively years ago, we further simplified the overall diurnal demands determined by mass balance into a simple average daytime demand and an average nighttime demand. It worked like this:
[a] We started by taking the average annual water demand (say acre-feet per year or million gallons per year) and converting it to gallons per minute (gpm) for ease of use.
Based on the mass balance, we determined that the daytime demand was ~1.5x the average for 14 hours and the nighttime demand was ~0.3x the average for 10 hours.
[c] Based on daily water demand records, we determined that the Maximum Day was ~2.1x the Average Day (IIRC) and the Peak Hour was ~3.0x the Average Day (IIRC).
[d] Based on the average annual water demand and population figures, we determined the average per capita water demand. We also used meter records to determine average demands for commercial, industrial, etc land uses. Back in the day, these demands were distributed manually using tract maps and assessor's maps to assigned parcels to the nearest node. Now, this can all be done with GIS.

==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[fel3]

Here's some more food for thought:

It is good to see that you are using the Hazen-Williams Formula instead of EPANET’s other two options (Darcy-Weisbach and Chezy-Manning). The most rational* equation for this type of hydraulic calculation is the Darcy-Weisbach Equation. However, it is rarely used for modeling distribution systems because it requires iterating the Colebrook-White Equation to solve for the pipe friction factors, while simultaneously iterating the entire network. Not only that, but the key to properly implementing the Colebrook-White Equation is knowledge of the pipes’ relative roughness values and good data is not as easy to find as for Hazen-Williams or Manning. There are several direct solution approximations for the Colebrook-White Equation that are OK, but not great. Even so, normal “industry practice” is to instead use the empirical Hazen-Williams Formula for modeling water distribution systems as you are doing. Hazen-Williams is faster and easier to apply, though it is less accurate, and doesn’t require any iteration (the network itself must still be solved by iteration). Regardless, the error bars on most of the data in a water model are large enough** that, in the hands of an experienced hydraulic modeler, the differences between Darcy-Weisbach and Hazen-Williams are easily accounted for.

Determining appropriate values for the Hazen-Williams Roughness Coefficient (C) is slightly an art form with a foundation in laboratory testing. C varies with the size, type, and condition of the pipes. That being said, it is not uncommon to use a single value of C for all the pipes in a model (I have done this myself). It is also common (and usually better practice) to use a handful of C’s and apply them judiciously based on size, type, and age of pipe (which is a reasonable proxy for condition, especially for cast iron). We often “derate” the C’s a little bit to account for differences between nominal and actual pipe diameters and minor losses at valves, bends, and other fittings in the distribution system. It’s much more time consuming to include all minor losses in a distribution system and you gain little by doing it. My practice is to still account for specific minor losses in pumping stations and other facilities where the minor losses are many and the pipe lengths are short.***

I have tables that list estimated C values from about 20 (e.g. 100-year-old, 3-inch-diameter, unlined cast iron with severe tuburculation) to about 155 (e.g. 48-inch-diameter PVC). However, because most pipes in modern distribution systems are in the 110 to 150 range (not counting minor losses), I can bet that C=100 is far too low for your system. Absent more detailed information about type and condition of the pipes in your system, I suspect that an average value in 120 to 135 range would be appropriate and would account for incidental minor losses and differences between nominal and actual diameters. There are ways to measure (actually back calculate) C values in place if you have a meter and two pressure gauges, but that is likely well beyond the scope of your work. It is also recommended (but not always done) to calibrate the model with actual operation of the system. That is another time consuming task that is also likely well beyond the scope of your work.

* Here, “rational” means that the variables and units attached to those variables accurately reflect the physics of fluid flow. The same cannot be said for the Hazen-Williams Formula.

** For example, the actual diameter for 12” DR 18 PVC pipe is 11.65”; for DR 14 it is 11.20”. That’s one reason why, if you use the nominal diameter in the model instead of the actual, it is imperative to “derate” the Hazen-Williams C. Here’s the comparison, based on a published C=147 for a “true” 12” diameter PVC pipe (in a table I have): the DR 18 pipe has an effective C=136 (using nominal d=12”, not actual d=11.65”) and the DR 14 pipe has an effective C=123. I calculated this using a Hazen-Williams “pipe slide rule” program I wrote more than 30 years ago for my HP-41 calculator and later ported to my daily driver, the HP-42S calculator. By combining the Hazen-Williams Formula, the Continuity Equation (Q=A*V), making substitutions for hydraulic radius and hydraulic slope, and including conversion factors for my preferred units, I am left with an equation that is tailor made for doing “what if” calculations for water systems. I can provide additional info on this if you need. Here’s the basic procedure I used: Set L=1000’, C=147, d=12.00”, & Q=1000 gpm (L & Q are simply arbitrary numbers), then solve for HL=2.02’; change d from nominal to actual, so d=11.65”, then solve for Q=925 gpm (this is what the DR18 can actually deliver versus a “true” 12” PVC pipe); change d back to nominal (d=12.00”) and solve for the effective roughness, C=136.

*** Years ago, I evaluated a pumping system at a farm that wasn’t functioning correctly. The system was simple: a wetwell collected drainage from excess irrigation via a subsurface drain tile system and a small pump sent it about a mile to a discharge pond. It turns out that the design engineer had neglected to include the minor losses in the pumping station analysis. I discovered that the pipe losses and minor losses in the 4” pump discharge piping were about twice as large as the pipe losses in the mile-long 12” inch pipe to the pond. This caused the pump to underperform, which was exacerbated by having several air & vacuum valves positioned above the actual hydraulic gradient. This caused a surging flow as air was frequently introduced and expelled.


==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[fel3]

Here are some thoughts from perusing the pipe data:

[1] Out of the nearly 5,700 pipes in the model, there are approximately 1,400 whose length is less than 10 ft, another 900 or so between 10 ft and 20 ft, and hundreds more that are only slightly longer. In fact, I estimate that about 3,000 pipes (53%) are less than 50 ft long. Short pipes like this (and the nodes between them) very often don’t serve any useful purpose and just “clog up” a model by increasing model setup time, increasing computational needs, and making it more time consuming to review the results. Where possible (and if I had the time), I would combine pipes to simplify the model (e.g. 22892+26253+22907, 22879+22880, etc). By doing this, you should be able to shave between 2,000 and 3,000 pipes from the model. I suggest looking at every pipe shorter than about 50 ft and many of the pipes between 50 ft and 100 ft to determine if these are candidates for combining with other pipes. However, some short pipes will still be required, including short fire hydrant laterals at critical locations, short connecting pipes such as your 17990 & 17991 (though I think only one pipe is needed here instead of two), etc. If you have two tees very near each other, it is often OK to combine them into one node (set between the two tees) to further simplify the model. On the other hand, with modern computers and modeling software, this is not as important as it once was. Many years ago I had two large, complex models that each sometimes took 30 minutes or more to solve via a remote teletype connection to a mini mainframe. This was due partly to the model, partly to the software, partly to the connection speed, and partly to having only a mid-level priority for computer core time. I have taken models that required 30 to 50 iterations using old software and solved then in less than 10 iterations with modern software).

[2] Quite a few pipes have lengths expressed to 2 decimal places, which is unnecessary. Rounding lengths to the nearest foot, or even the nearest 5 ft or 10 ft is usually accurate enough. When you build a network by tracing over an imported map, it is usually best to keep the lengths EPANET calculates and not waste any time “cleaning up” the data. However, Project|Defaults|Properties shows that Auto Length is OFF, so it appears the model was traced, but the lengths were added manually. This probably explains why 10 pipes still have the default length and diameter (Pipes 31, 35, 313, 333, 334, 989, 1013, 1015, 1027, & 1059 are all listed as d=12” & L=1000’). All ten should be checked against the system maps. Some of these 10 pipes are drawn VERY short (e.g. 989 & 1015 are probably 6” fire hydrant laterals); Pipe 1059 is in line between two 36” pipes and is probably also 36”; and so one.

[3] Miscellaneous Pipe Stuff
[a] Pipe 1011 & 1012 are relatively long for being only 6”. The lengths makes sense based on the map, so I would verify the diameter.
Pipe 1032 is 24”, but Valve 26 and Pipe 333 are 12”. I suspect all three should be the same diameter.
[c] Pipes 53, 7060, 7059, & 1062 are 24” but on either side are 36” pipes. Should these four pipes also be 36”?
[d] Pipes 357, 442, & 709 and Valve 366 are closed. Should they be?
[e] At Junc 529, the pipe size changes from 36” (Pipe 828) to 24” (Pipe 809). Is this really the case (it certainly could be)?

This is far from an exhaustive look at the pipes, but it should get you going and give you some ideas for additional things to check. One thing that I have found helpful over the years for debugging water models is to use the graphical capabilities of the program to show you what is in the model. For example, the way I found the pipe diameter issues was to use the Browser to turn on the color coding for pipe diameters, then I looked for suspicious changes in pipe size, etc. You can also go to Report|Table to bring up your input data, then sort the columns and look at the extreme values (that’s how I found the short pipes and did my estimate of the numbers of short pipes).


==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[fel3]

More Pipe Stuff

PIPES 28216, 28218, 28225, 28227, 28230, & 29004 have a diameter of 1”; PIPES 651, 698, 699, 700, 702, 10481, 10580, & 24741 have a diameter of 2”; and, PIPES 210, 211, 703, & 4133 have a diameter of 4”. All of these pipes are smaller than a typical fire hydrant lateral* and should be verified. These pipes are also much smaller than stated in your analysis criteria. For large distribution systems, I almost never include 1” and 2” pipes in the model. Sometime I include 4” pipes…it depends on the system.
All of the 1” pipes are very short laterals off the most northerly reach of pipes in the systems. The 2” pipes are found in several locations. PIPE 10580 is the last pipe in a string of 12” pipes and maybe should 12” itself, rather than 2” (a typo, perhaps).

* I have seen fire hydrants and especially wharf head hydrants on 4” laterals.


==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[fel3]

If I’m interpreting this model correctly, the Lake Monroe / Water Treatment Plant / TANK MONROE / PS MONROE complex is the primary water source for the entire city. TANK MONROE / PS MONROE feeds TANK 4MIL and TANK HARRELL plus several very small service areas en route. TANK 4MIL and TANK HARRELL in turn (via their pumping stations) feed the bulk of the distribution system, including the other tanks. These other tanks will fill during low demand periods (primarily at night). When demand is high (primarily during the day), these other tanks, via their pumping stations, also feed the distribution system. Downstream of TANK 4MIL and TANK HARRELL are several major loops of large diameter pipe that form the backbone of the distribution system. The system between the Lake Monroe / Water Treatment Plant / TANK MONROE / PS MONROE complex and TANKS 4MIL and HARRELL is thus one pressure zone, while the rest of system appears to be a second, lower pressure zone. Do I have the big picture correct?

JUNCs 358 & 401 still have the default elevation of 100 ft, which is obviously too low. Perusal of online topo maps for Bloomington indicates to me that the range of elevations in your model (575 ft to 958 ft) is reasonable except for these two nodes and TANK HARRELL and the RESERVOIR, which I already covered. However, the difference between high and low ground elevations in the system is 383 feet, which is equivalent to a static pressure difference of 165 psi. I would expect such a large range in elevations to require perhaps several pressure zones, possibly interconnected via pressure reducing valves so that the high zone(s) can feed the lower zone(s), but so far I haven’t found evidence for this except for the functional separation of TANKS MONROE, 4MIL, and HARRELL from the rest of the system. Are there additional pressure zones that I'm not seeing (yet)?


==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[fel3]

Junction Nodes

[1] Only JUNC 949 has a non-zero BASE DEMAND and only 8 nodes show any demand at TIME=0 (which comes from DEMAND CATEGORY 2). At nodes with BASE DEMANDS or additional DEMAND CATEGORIES set to zero, the DEMAND PATTERNS are multiplied by zero, for a net demand of zero. As it sits, the model is only trying to move water between tanks. I suggest developing average water demand rates based on land use and using GIS to apply these demands by land use to each node as I briefly described above. This generally starts with meter records and the City’s General Plan documents. This is a conceptually simple task, but it can be time consuming. As I mentioned earlier, the DEMAND PATTERNS need a complete overhaul. I suggest coming up with different DEMAND PATTERNS and CATEGORIES for residential, commercial, industrials, parks & golf courses, etc.

[2] At TIME=0, some nodes have very low to even negative pressures and some nodes have pressures well in excess of 150 psi. Of course, we don’t have a valid solution yet, so this may be much to do about nothing. However, this is good seque into the topic of acceptable system pressures. In all of the water systems I have worked on (mostly in California), here is how pressures have been evaluated (the rules for your system may be differ somewhat):
[a] Minimum required pressure at a flowing fire hydrant: 20 psi. For me, this is governed by the California Fire Code. I think NFPA has the same requirement. This usually means you need 25-30 psi in the adjacent water main to account for head losses in the hydrant lateral and in the hydrant head. Where the laterals are modeled, you can shoot for about 25 psi at your fire hydrant node.
Absolute minimum system pressure: this used to be 5 psi, but the code in California was changed several years ago to 20 psi. This differs from the fire hydrant requirement because you also need to check high points in pipes where there may not be a fire hydrant (e.g. where a pipe goes over a hill or a ridge in an area with no development). The way to deal with this requirement is to make sure you have a node at the top of the hill or ridge so you can check this.
[c] Minimum required pressure during Peak Hour for normal service: 35 psi, or 40 psi, or 43 psi, or 45 psi (I have encountered each of these standards…BTW, 43 psi = 100 ft of head, and the round number is why that one water district used it).
[d] Maximum pressure for normal service without a pressure regulator ahead of building plumbing: 80 psi. For me, this is governed by the California Plumbing Code and the fact that plumbing fixtures in the US are rated for 80 psi maximum.
[e] Maximum pressure for normal service with a pressure regulator between the water meter and building plumbing (and thus the responsibility of the property owner): 150 psi. This is based on the pressure rating of the water meters being used.
[f] Maximum pressure for normal service with a pressure regulator ahead of the water meter (and thus the responsibility of the water agency): 300 psi. This is based on the pressure rating of the distribution system piping and the pressure regulator. Where system pressures exceed 300 psi, special designs are called for.


==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[fel3]

Some thoughts on node and pipe numbering:

Node and pipe numbering within this model is without an obvious organizational scheme. The numbers bounce all over the place, which makes the model and the tables of results hard to work with. It is better to organize node and pipe numbers geographically, by pressure zone, and/or some other sensible scheme. Here are some suggestions:

[a] Every municipal utility that I have ever worked with has been covered by a master set of atlas maps. Even in the electronic age, most of these atlases are set up to be printed on multiple D- or E-size sheets, typically at scales between 100’=1” and 500’=1”. Most of these atlases also have a superimposed grid system to further break up the sheets. Sometimes a master grid covers the entire utility system (so that grid square A-1 occurs only once) and sometimes the grid is repeated on each page (so that grid square A-1 occurs on every sheet). Either way, this makes it easy to find things. Grid lines are typically 2” to 5” apart. If your atlas doesn’t have a grid, I suggest adding one.

This grid system provides a convenient way to number nodes to correlate with the atlas. For systems with multiple pressure zones, I would prefix the node numbers with the nominal HGL (hydraulic grade line) for its pressure zone. For special nodes, like fire hydrants, I might also add a suffix. Some modelers like to include the names of street intersections, at least for major intersections. The idea here is to build the node numbers left-to-right from the general to the specific. For proper sorting with different numbers of digits, make sure to include leading zeros. Many modelers initially number nodes and pipes in increments of 5s or 10s to provide room in the numbering scheme for future additions.

For example, in a water system with a master grid and two pressure zones (nominal 925-foot HGL and nominal 1050-foot HGL), you might number a particular node 0925-E5-015 (or 0925_E5_015) and a nearby fire hydrant 0925-E5-020-FH. In your upper pressure zone you might have node 1050-G7-005 and 1050-F2-Tank Hilltop. If your grid repeats, then my first example (if on sheet 7) might be 0925-07E5-015 and so on.

Numbering nodes this way makes it easier to sort nodes by pressure zone for reading the results and makes it easier to find nodes in the model on screen using the browser. In the old days, we only had limited-length numbers and some programs assigned these automatically. Now we have long and full alpha-numerics so we should make use of it.

[c] Numbering pipes is not quite as easy because pipes can cross multiple grid squares. In this case, I often use one of the grid squares the pipe crosses (perhaps at one end or in the middle). Include the pressure zone as a prefix and you end up with pipe numbers that are similar to node numbers. Because of this similarity and to avoid potential confusion, some modelers use as the first prefix N for node (or J for junction), P for pipe, T for tank, BP for booster pump, and so on.

For large-diameter transmission mains, I will sometimes number the pipes in order starting with the source. You have two parallel transmission mains between TANK MONROE (nominal HGL=782 ft) and TANKs 4MIL and HARRELL. In this case, I could see including an additional designator for which tank the transmission main feeds. For example, once the pipes branch a little ways downstream of TANK MONROE you could have 782-4MIL-100, 782-4MIL-110, 782-4MIL-120, etc heading to TANK 4MIL and 782-HAR-100, 782-HAR-110, 182-HAR-120, etc heading to TANK HARRELL. Before the branching, you might have 782-010, 782-020, 782-030, etc. If you wanted to insert a middle element to this last set of numbers, then I would add something that would sort ahead of 4MIL, such as 000. If you choose to number transmission mains sequentially and you want to also include the grid square, the grid square needs to be added at the end so that the sequence is not violated.

==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[fel3]

More on Tanks & Appurtenances:

[1] It occurs to me that your RESERVOIR may actually be the finishing tanks at the water treatment plant and not the lake. If this is correct, then it looks to me like your model has PS MONROE pulling water directly out of the finishing tanks and TANK MONROE sitting off to the side as additional storage for high demand periods. I apologize for the brain freeze.

[2] The FLOW CONTROL VALVE feeding TANK 4MIL (VALVE 317) has a flow setting of ZERO, so it won’t let any water into the tank.

[3] TANKs GENTRY, WEST, REDBUD, & DYER have no control valves on their inlet/outlet lines. How are these tanks kept from overfilling? I suggest reexamining your system maps and diagrams to see if there are valves here and also asking your operations people about the operation of these tanks. There are RULES pertaining to TANKs GENTRY, WEST, & REDBUD, but not DYER, and most of these RULES pertain to the operation of pumping stations. This may be enough to prevent overfilling in practice (except for DYER), but if these tanks share a pressure zone (which I think is the case), then there may still be altitude valves on the lower elevation tanks. If not, then any overage hopefully has a physical overflow to go through.

[4] The FLOW CONTROL VALVE on the inlet to TANK SOUTHWEST (VALVE 226) has a flow setting of ZERO, so it won’t let any water into the tank. The PRESSURE REDUCING VALVE on the outlet of TANK SOUTHWEST (VALVE 336) has a pressure setting of zero, so it won’t let any water out of the tank since the downstream pressure should always be higher. In addition, I can’t figure out the purpose of a PRV here. I suggesting verifying the PRV’s existence and, if it does exist, talking to your operations people about what this valve’s purpose is.

More on Pipes

[1] PIPES 12976, 9482, & 9481 are 12”, but feed a dead end 24” (PIPES 9483, 9485, & 521). I have a suspicion that the 24” pipes are actually 12”.

[2] PIPE 677 is 6” but is between two 12” pipes. I suspect it should also be 12”.


==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[fel3]

I hope you are finding this evaluation of your model to be useful. It has been a fun and interesting exercise to me. I am almost done with my evaluation, but not quite.

More Tank Stuff

TANK DYER has no water levels set. I forgot to comment on that the first time through.

Pumping Stations

As I mentioned previously, my usual practice is to model pumping stations in more detail than the distribution system. I include all of the short pipe segments in the suction and discharge headers, add valves (e.g. gate valves, butterfly valves, check valves, control valves) and fittings and their associated minor losses, etc. The pumping stations in this model do not have that level of detail and thus the model will underestimate head losses in the immediate vicinity of the pumps.

[1] PS MONROE has four pumps, each with a duty point of Q=2000 ft, TDH=800 ft. Based on (a) an average water level in TANK MONROE of about 775 ft and (b) the pipes between the tank and pumps being so short and large that head losses can be ignored for the following approximation, I get a discharge HGL for PS MONROE of about 1575 ft. The discharge node (JUNC 7164) has an elevation of 744 ft, which means a discharge pressure of about 360 psi. This is not unheard of,* but it seems awfully high for this system considering that one of the two downstream tanks that receive this water is not that much higher: TANK 4MIL has an average water level of 850, which equates to a static pressure of 314 psi. The other downstream tank (TANK HARRELL) has bad data, so I can’t comment on it further. I suggest verifying the pump curves (and using curves rather than one duty point) and checking with your operations people on the typical discharge pressures at this pumping station.

* (I once worked with a water system that had a pumping station with a discharge pressure of 590 psi, which it needed to pump water WAY up a hill to a small service area on the other side.

[2] As drawn, PS SOUTH CENTRAL is paralleled by a 36” pipe (PIPE 972) that creates an unwanted recirculation loop. PIPE 972 is part of the 36-inch transmission main that fills TANK 4MIL from PS MONROE. By examining the model in detail, I discovered that this transmission main is a “straight” shot between PS MONROE and TANK 4MIL. PIPE 972 isn’t connected to the main distribution network and it only serves a couple of very small areas in route.
Because PIPE 972 is connected to both the suction and discharge sides of the pumping station (in the model, not in real life), when PS SOUTH CENTRAL is in operation, it would simply pump back to its suction side. There are no RULES in the model to close PIPE 972 to prevent recirculation. I suspect, though, that PIPE 972 is not actually connected to the pump discharge. Or, if it is connected, there is a pressure reducing valve between the tank fill line and the discharge side of the pumping station (this would be a useful way to bypass TANK 4MIL and PS SOUTH CENTRAL). If there is no connection to the discharge side of the pumping station or if there is a pressure reducing valve in between, then there is no need for a RULE for PIPE 972.

The curves for SOUTHCENT1, SOUTHCENT2, and SOUTHCENT3 will produce a discharge pressure on the order of 50 psi, which seems reasonable. The curves for SOUTHCENT4 and SOUTHCENT5 indicate pumps that produce a somewhat higher discharge pressure (these two curves are slightly different, but probably not enough to matter). I’m not sure why there are two different sets of curves here. Most pumping stations have pumps with matching performance curves, but not always. If there really are two different sets of pump curves, then I would guess that the first three pumps are for normal usage and the other two are for when a higher pressure is needed (perhaps during an emergency). To me, it doesn’t look like pumps from the first group would be happy operating with pumps from the second group. I suggest verifying the curves and checking with your operations people

[3] PS HARRELL has one pump with a duty point of Q=1900 ft, TDH=845 ft. As with PS MONROE, this seems awfully high, but without accurate data for TANK HARRELL, it’s impossible to tell how high is too high. I suggest entering the complete pump curve rather than relying on just one data point because actual pump curves often deviate significantly from the idealized curve EPANET assumes if you just have one point.

[4] PS TAPP is a “run of the pipe” pumping station that appears to feed TANK SOUTHWEST and a good portion of the distribution system. Being a “run of the pipe” pumping station means that it sees larger fluctuations in suction pressure than if it was adjacent to a water tank. This is not necessarily a problem, just something to be aware of.
The three pump curves provided for PS TAPP are not the same, but they appear to be very close. Were these curves produced from test data rather than a curve in the manufacturer’s catalog?

[5] PS WEST BOOSTER has two pumps (WBoost1 and WBoost2) that appear to have the same curve, but the third pump (WBoost3) has a lower curve and is missing data from 0 to 2000 gpm. Is there a reason that WBoost3 has a different curve and incomplete data? This needs to be checked.
It looked for minute like PS WEST BOOSTER & TANK WEST created a separate pressure zone, but, as drawn, it appears the system connects back to the suction side of the pumping station via the pipe that passes by TANK DYER and possibly elsewhere. Generally we don’t recirculate pumped water back through the distribution system to the suction side. It would make sense for this to be a separate pressure zone and if there are any connections back to a lower pressure zone in the middle of the city, I would expect a pressure reducing valve or normally closed valve between them.

[6] PS GENTRY has two pumps with different duty points and no curves. I suggest entering complete pump curves.
Similar to PS WEST BOOSTER & TANK WEST, I think this is probably a separate pressure zone, but as drawn, the model recirculates pumped water back to pump suction via the loops north (3rd Street) and south (E. Moores Pike) of the PS-to-TANK pipeline.

[7] PS REDBUD has four pumps with curves that won’t work together. I suggest checking these curves.
Similar to the two previous pumping stations, as drawn, PS REDBUD recirculates back to itself.

[8] It seems to me that this system consists of four pressure zones, although the model does not separate out the last three:
[a] The transmission mains (and small service areas) between PS MONROE and TANKs 4MIL and HARRELL.
The central part of the city, which is served by TANK DYER and supplied by PS 4MIL and probably PS HARRELL.
[c] The westerly part of the city, which sits at a higher elevation than the center and which is served by TANKs SOUTHWEST (HWL=1027’) and WEST (HWL 1025’) and supplied by PSs TAPP and WEST BOOSTER.
[d] The easterly part of the city, which sits at a higher elevation than the center and which is served by TANKs GENTRY (HWL 953’) and REDBUD (HWL 951’) and supplied by PSs GENTRY and REDBUD.

I don’t know if the last two are further split into four separate zones, but with the high water levels between the tanks being so close in each zone as I surmise, there would be no reason to split them.


==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[fel3]

With this post, I have reached the end of my review. However, if you have additional questions or think I have skipped something important or that interests you, please don’t hesitate to ask for more help. Either way, I would like to get your feedback. Also feel free to update us on your progress.

The Model

As you have seen, there is still data to correct, details to add, and debugging to perform. You are at or near the place where an engineer with a background in water system modeling and water system operations needs to step in to at least oversee your work if not take it over. As you indicated several weeks ago, that’s what you thought should happen anyway. Regardless, you have tackled a complex problem completely outside your area of education and expertise and in the process you have done a lot more and learned a lot more (on your own, no less) than I would have expected from an intern. So, while the execution isn’t perfect (yet), the effort has been extraordinary.

To the first approximation, I think the model accurately reflects your city’s water distribution system. Where the model fails—as evident in my previous comments—is in the details. In my experience, this is almost always were water models fail. But all this is fixable as I have described. What happens when you build a model from small-scale atlas maps (whether paper or electronic) is that you get the big stuff right (excepting, of course, human error), but you miss all the details such as complex piping configurations at tanks and pumping stations, how interconnections between pressures zones are really handled, and so on. To model these special areas may take digging out the plans and/or visiting the sites, in addition to getting help from your operations people.

Except for the smallest models, I like to make sure the steady state system works before I try an extended period simulation.

Learning More

A good introduction to water system modeling is “Distribution Network Analysis for Water Utilities” (AWWA M32) by the American Water Works Association (1989). This publication has been superseded—apparently several times—by “Computer Modeling of Water Distribution Systems” (AWWA M32; Third Edition 2012). I have the old publication but not the new one. According to the tables of contents, both publications cover the same foundational material, but the newer publication covers the computer modeling side of things, including water quality analysis, in much more detail. You can find used copies of the old publication on Amazon and elsewhere for about $30. The new publication is $74 (members) / $118 (non-members) through AWWA and elsewhere. Even without an engineering background, you should be able to follow the older publication just fine and probably the newer one as well.

In-depth treatments of water system modeling can be found in “Water Distribution Modeling” and “Advanced Water Distribution Modeling and Management.” both by Walski, Chase, & Savic. The first book (which I have; First Edition, 2001) was written when WaterCAD was owned by Haestad Methods and might only be available used (just don’t quote me on that). The second book (which I do not have) is available through Bentley, which now owns WaterCAD, at

http://www.bentley.com/en-US/Training/Products/Resources/Books/AWDM.htm.

These books pre-suppose an engineering background (they’re actually written as engineering textbooks) and include quite a bit more math than AWWA M32. However, they are also well-written and I think you could handle quite a bit of either one without too much difficulty.

Since you’re a GIS intern and not an engineer or engineering intern, if you want to get just one book on the subject, I recommend one of the AWWA M32 manuals as the best place to start.

Water Quality

I don’t have much to say about water quality modeling since I have done very little of it and it wasn’t recent. Most of my experience is with hydraulic modeling. You will probably need some guidance from your staff engineers and/or operations people (or the publication I referenced above) for setting the values related to water quality modeling under Options|Quality, Options|Reactions, and Options|Times.

Your main chemical source is undoubtedly the water treatment plant, but if you have other injection points (probably associated with your tanks and/or booster pumping stations) you need to deal with those nodes as well.

Pipe specific parameters should be keyed to pipe type and condition rather than use single global values.

In fact, the shear size of the task to edit each pipe for water quality parameters should be incentive enough to combine pipes as I suggested above and reduce the number you have to deal with.

Reviewing the Results

When you get the model running properly, pay close attention in your results to things like node pressures (which I covered previously) and pipe flow velocities. Generally, flow velocities during a Peak Hour Demand scenario should be less than about 5 feet/second; during a Maximum Day Plus Fire Flow scenario you might see velocities upwards of 10 fps (especially in fire hydrant laterals). Above 15 fps can be a problem (water hammer and even pipe wall erosion).* Your staff engineers and/or operations people should be able to tell you what your city’s targets are. It may even be covered in your public works standards or a previous water system master plan**. In fact, I should have suggested looking at these documents at the start of my comments, but it slipped my mind. That’s the problem with turning 55 this weekend.  If you find results that are outside the norm they could indicate remaining problems with the model or they could indicate real problems in the system. The former provides clues for additional debugging. The later provides clues for future system enhancements.

You should also compare model results to real world data for the system. Most modern distribution systems have lots of relevant telemetry data to work with for this comparison. You may or may not get the go-ahead for hydraulic and water quality calibrations, but at the very least your model should produce results that reasonably reflect reality. You are not trying to get exact correlations (it’s actually impossible), but you are trying to get close enough that you can have confidence that the model actually depicts how the system functions now and predicts how it will function in the future. The books by Walski, et al, cover calibration and acceptable correlation in detail.

For example, are the pumps in the model matching the pumps in the system for flows and suction and discharge pressures (there should be pressure gauges at the pumping stations and maybe flow meters, too)? Are the pumps in the system turning on and off in the same way that the RULES in the model specify? Are the tank levels in the model fluctuating in a way that mimics how they fluctuate in the field? Are node pressures in the model indicative of the range of node pressures seen in the field? Are known problems in the system (e.g. low pressures, low chlorine residual, etc) showing up in the model? Your engineering staff and operations people should be able to come up with additional questions like these.

In addition to modeling current conditions, this model can be used as a foundation for looking at future conditions. EPANET cannot handle future growth in one model like WaterCAD, H2ONet, etc. can.*** But, hey, it’s free. Before you can even think about modeling future conditions, you need to get your base model (this one) as close to perfect as you can get it: fixed, modified, debugged, and running reliably. Then, make a copy of this model and add projected future growth from your General Plan, any Specific Plans, etc. This would include new demand areas, new pipes and nodes, maybe even new pumping stations, tanks, etc. It may also be appropriate to modify unit demands and demand patterns, change the RULES, etc., etc., etc. to reflect likely future operations. You could also take this second model and go through the same process to create a third model that looks farther out. If you create models for the future, be sure to include the year in the file name. For example, you might name the current model BW_2013.net (for Bloomington Water 2013), the first copy BW_2018.net, and the second copy BW_2023.net. I wouldn’t try to go beyond three models with EPANET. If you later find an error in the base model, it has to be corrected in the two future models. Any more models and the chance for error continues to rise.

Conclusions

I think I have beat this horse to death several times over. Again, I hope this has been helpful to you. The model has a good foundation, but there is still quite a bit to do to get it ready to use. However, it is all correctable and if you follow my suggestions, I think you will end up with a model you can be proud of.

Footnotes

* Depending on how you model fire flows, you may exceed 15 fps in the model (though usually not in real life). Good fire flow modeling is deliberately conservative, including sometimes modeling fewer fire hydrants than would actually be used by the fire department. Also, for normal fires, the fire department will likely use about one-half to two-thirds the flow required for that type of development and that you have modeled. (The difference between these two flows is a safety factor owned by the fire department.) Here is some info I previously posted on Eng-Tips. However, your fire department’s criteria will govern.

http://www.eng-tips.com/viewthread.cfm?qid=290322

http://www.eng-tips.com/viewthread.cfm?qid=283996

** I tried Googling for a water system master plan for your city, but did not find one. If your city doesn’t have one, I suggest they engage a good consultant to prepare one. A water master plan for your city will not be cheap (probably in the $100k-$300k range), but it is money well spent. The consultants who are capable of preparing a water master are more likely to use one of the commercial water modeling programs than EPANET because these program have many more bells and whistles and a more powerful interface. Fortunately, many of the most used programs on the market are actually based on EPANET and can read EPANET files directly (

http://en.wikipedia.org/wiki/EPANET).

*** Well, you could add in all the projected future growth and the associated water system improvements in the current model, then change the open/close status all the new pipes, and the demand status for all new nodes, etc, but that approach is messy and prone to error.


==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[BigInch]

fel, Send the city a bill for your services. You at least deserve an acknowledgement for your work, even if it is a letter stating that they won't pay.

Independent events are seldomly independent.

 

[fel3]

BigInch…

I was mostly trying to help a poor intern whose boss gave him a huge task for which he is ill-suited. With the information and suggestions I provided boedekek, he has a much better chance of making this thing work and gaining confidence in his own abilities. Frankly, given his circumstances, I think he did a creditable job. I also provided him with some ammo in case he wants to tread in that minefield that is explaining the real world to a supervisor. I just hope he treads lightly but firmly.

Besides, it gave me something interesting to do at after work while I was waiting for the temperature to go down (I'm in Fresno, CA). Also, my wife works until at least 6:00 p.m., so getting home right after 5:00 just means I have to cook.

I must admit, though, that this was not entirely an altruistic exercise. I have done lots of water modeling over the years and lots of debugging other peoples' models (several man-years worth between the two), but I haven't done any in the last three years or so. Thus, this was an excellent opportunity to resharpen my own skills.

Fred

==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[fel3]

One more thing that I just thought of: the best way to compare different pump curves in the same pumping station (e.g. PS SOUTH CENTRAL) is to put the data in Excel and plot all of them together. Unfortunately, EPANET can't do this.

==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[BigInch]

fel, I understand your motivation. That's why most of us regulars are ... well ..regulars. We need to help people and enjoy honing our skills at the same time we look for new things to learn ourselves. We know we get there when we can effectively teach the subject or provide comments of value and quality, such as you have made here.

My comment in this case was meant to recognize your dedication and professionalism, which I must say (more clearly now) are very impressive and IMO well above and beyond what anybody should expect, even if they paid for it, never mind if given in the helping spirit in which you offered it, totally free. I could only give you a star, so I did that, but truly I think you should get a medal of honor (and $2500) for this one. I just wanted that city to recognize the contribution you have made to them and to their employee ... in one way or another. Nothing else seems right under the circumstances.

I also understand why you're doing this while trying to keep cool inside the house. I'm in Dubai where it is often over 100F ... at midnight.



Independent events are seldomly independent.

 

[fel3]

BigInch…

Thank you for the kind words. I enjoy the opportunity to teach/mentor, often more than the work itself.

My younger son-in-law is an avionics technician in the U.S. Air Force and he spent part of last summer/fall in nearby Qatar. Even Fresno's heat did not adequately prepare him for the Persian Gulf.

Fred

==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill

 

[77JQX]

fel3 - I agree with BigInch. Quite an impressive set of responses to the OP's question - way above and beyond the normal here, and done in a respectful and professional manner. I had downloaded the model and was about to comment on it, but you posted first, and in more depth than I was willing to go into. It's been a pleasure watching you disect the model and provide guidance.