Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations TugboatEng on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!
Negative pressures and disconnected nodes 2
- Thread starter makoy888
- Start date
That's not a lot of information, but I can give you a few things to start with:
1. For disconnected nodes, find all nodes that are truly disconnected (if any) and connect them to the network. I have seen this in other people's models quite a few times, both here and at work. I never make this mistake myself, of course. Never.![[smile]](/data/assets/smilies/smile.gif "[smile] [smile]")
Sometimes, however, there are other problems with the model that can generate a "disconnected nodes" error when there aren't actually any disconnected nodes. For this case, the solution is to find and fix the other problems. For example, a model that can't complete its iterations can generate this error.
2. Negative pressures can have several causes, including, but not limited to the following:
a. Tank(s) and/or reservoir(s) that are too low relative to the elevations of the other nodes. Check all elevations to make sure they are correct. Increasing the diameter of a tank won't change pressures, although this would cause pressures to increase and decrease more slowly during a time period simulation. Regardless, the proper thing to do is to accurately model the physical elements in the system, rather than make things up.
b. Pipes that are closed and control valves and check valves that hinder or block flow through their pipes. Check for all these. I once inherited a model from a city of 50,000 people that had a check valve incorrectly inserted into a 12" main. When I tried to rerun the last consultant's scenarios, I didn't get his results. As it turned out, one of the city's young engineers had tried to teach himself the program (not EPANET), but instead of working on a copy of the model, he used the master version of the model. As I dug deeper into the model, I found some pipes with changed diameters, a tank that I had designed a few years before had been deleted, an entire section of the city had been disconnected (it had it's own tank, so the model would run, even if incorrectly), etc. Rather than try to untangle this mess, we asked the city to get a clean copy of the model from the previous consultant.
c. Pumps that are off or that have bad curve data. Check this.
d. Fixed inflows that are entered as demands (positive numbers), rather than as inflows (negative numbers). This is an easy typo to make and it often results from not following the ridiculous sign convention that most water modeling software uses. Considering inflows and demands from a mass balance perspective, inflows should be positive and demands should be negative, not the other way around. I have only used one water modeling program that got this correct.
e. Pipes that are too small for the flows involved. Check for typos, too. I have seen models with 10" pipes entered as 1", 24" pipes entered as 2", etc.
f. Pipes with unreasonably low C-factors. I have seen quite a few models where a default value of 100 was used, when the actual pipes were probably in the 120-140 range.
Adding controls to engage more flow only makes sense if there is more flow to engage in the actual system. Adding flow that doesn't exist turns your model into a fantasy.
If you upload a copy of the model, I MAY have time to look at it over the next few days.
There is probably more, but I'm late for dinner, so I will stop now.
Fred
==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill
1. For disconnected nodes, find all nodes that are truly disconnected (if any) and connect them to the network. I have seen this in other people's models quite a few times, both here and at work. I never make this mistake myself, of course. Never.
Sometimes, however, there are other problems with the model that can generate a "disconnected nodes" error when there aren't actually any disconnected nodes. For this case, the solution is to find and fix the other problems. For example, a model that can't complete its iterations can generate this error.2. Negative pressures can have several causes, including, but not limited to the following:
a. Tank(s) and/or reservoir(s) that are too low relative to the elevations of the other nodes. Check all elevations to make sure they are correct. Increasing the diameter of a tank won't change pressures, although this would cause pressures to increase and decrease more slowly during a time period simulation. Regardless, the proper thing to do is to accurately model the physical elements in the system, rather than make things up.
b. Pipes that are closed and control valves and check valves that hinder or block flow through their pipes. Check for all these. I once inherited a model from a city of 50,000 people that had a check valve incorrectly inserted into a 12" main. When I tried to rerun the last consultant's scenarios, I didn't get his results. As it turned out, one of the city's young engineers had tried to teach himself the program (not EPANET), but instead of working on a copy of the model, he used the master version of the model. As I dug deeper into the model, I found some pipes with changed diameters, a tank that I had designed a few years before had been deleted, an entire section of the city had been disconnected (it had it's own tank, so the model would run, even if incorrectly), etc. Rather than try to untangle this mess, we asked the city to get a clean copy of the model from the previous consultant.
c. Pumps that are off or that have bad curve data. Check this.
d. Fixed inflows that are entered as demands (positive numbers), rather than as inflows (negative numbers). This is an easy typo to make and it often results from not following the ridiculous sign convention that most water modeling software uses. Considering inflows and demands from a mass balance perspective, inflows should be positive and demands should be negative, not the other way around. I have only used one water modeling program that got this correct.
e. Pipes that are too small for the flows involved. Check for typos, too. I have seen models with 10" pipes entered as 1", 24" pipes entered as 2", etc.
f. Pipes with unreasonably low C-factors. I have seen quite a few models where a default value of 100 was used, when the actual pipes were probably in the 120-140 range.
Adding controls to engage more flow only makes sense if there is more flow to engage in the actual system. Adding flow that doesn't exist turns your model into a fantasy.
If you upload a copy of the model, I MAY have time to look at it over the next few days.
There is probably more, but I'm late for dinner, so I will stop now.
Fred
==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill
- Thread starter
- #3
- Thread starter
- #4
- Thread starter
- #5
makoy888....
I took a look at your model, ran it, played with it, and exported the .net file to review. Here are my observations. This is not a complete critique of the model, but I think there is more than enough here to get you started and maybe even enough to get you finished. Please ask again if you have additional questions.
GAPS
[ol 1]
[li]The pipes and nodes N/NW/W of and served by SiladReservoir are not connected to the rest of the system. I suspect there should be something* connecting Node i-107 and Node n-106.[/li]
[li]The pipes and nodes SW of and served by MangaReservoir are not connected to the rest of the system. I suspect there should be something* connecting the tank or Node 107 to Node 62A and/or Node 62. Downstream of MangaReservoir is another gap at AlibaughReservoir. Here, I suspect there should be something* connecting the tank or Node i-124 to Node n-123.[/li]
[li]Immediately S of PoblancionReservoirs are two gaps, one between Node 103 and Node 53 and one between Node 97 and Node 88. I suspect there should be something* closing these gaps.[/li]
[/ol]
As you can see, there is a pattern here. When I look at each tank, I see one or two gaps in the system and I am pretty sure they should not be there. Also, your Water Treatment Plant is not connected to the system. What this means is that each disconnected part of the system is operating independently and no water is moving between them. I suspect that is probably not the actual case.
BTW, I inserted fictitious pipes in all 16 gaps (1m long, 25mm diameter, C=140) and got the model to run without generating any error messages. Most of the pressure problems are still there, but at least the model runs.
* That "something" might be:
[ul]
[li]A pipe, if the isolated tank is supposed to float on the main part of the system.[/li]
[li]A pump, if the isolated tank is higher than the main part of the system.[/li]
[li]A pressure reducing valve or pressure sustaining valve, if the isolated tank is lower than the main part of the system.[/li]
[li]Duct Tape, if all else fails.
[/li]
[/ul]
TANKS
At first glance, the tank elevations seem to be OK with respect to their service areas. The tanks are pretty small from my experience. However, if they really are small and are modeled correctly, then this is just an observation, not a critique of the model. BTW, if the Water Treatment Plant puts out a fixed or nearly fixed HGL (e.g. from variable speed pumps or gravity feed from the clearwell), then I would model it as a reservoir.
PIPES
[ol 1]
[li]I only looked at the first time step (T = 0:00 hrs). I noticed that some pipes are closed (e.g. Pipe 150, Pipe 152, Pipe 159, etc.). Should these pipes be closed? For example, Pipe 150 has an initial status of OPEN, but the first time slice has it CLOSED. I didn't find any controls or check valve that caused this, so it must be something else that I didn't find yet.[/li]
[li]The pipes are relatively small (50mm to 200mm). But, like the tanks, if they really are that small, then this is just an observation. At T = 0:00 most of the flow velocities are very low. I didn't check at the other timesteps.[/li]
[/ol]
NODES
[ol 1]
[li]I only looked at the first time step (T = 0:00 hrs). Most of the pressures looked normal to slightly high. I would expect the need for pressure regulators at some buildings to protect the plumbing and fixtures inside. The plumbing code I use here in California requires building pressure regulators for pressures exceeding 80 psi, or 56 meters.[/li]
[li]However, check the pressures at the following nodes: Well, i-104, i-124, 97, 103, 104, i-107, and 109. These nodes have pressure heads that are many times the Earth-Moon distance. Seriously. The biggest number I found was at Node 109: 2.18x10^7 m, which is nearly 57x the Earth-Moon distance. This is either magic or an error. This appears to be related to the gaps and closed pipe in the system, since each of these nodes is next to a gap and/or has a closed pipe attached. Fix the gaps, open the closed pipes, and I suspect these high pressure heads go away.[/li]
[/ol]
CONTROLS
[ol 1]
[li]I don't see any controls. Should there be any? If there are any pumps that fill higher tanks or PRVs or PSVs that prevent lower tanks from overfilling, then there should be something.[/li]
[/ol]
PATTERNS
[ol 1]
[li]Pattern 1 looks like a reasonable diurnal curve.[/li]
[li]Pattern 2 is just straight 1's, which would be correct for fixed inflows and/or demands.[/li]
[li]Most nodes have Pattern 1, which makes sense.[/li]
[li]I don't see anything in the model that resembles a fire flow. I typically do static and extended period models, and haven't done a water quality model in years. The EPS is useful for modeling how tank levels fluctuate, and that is a legitimate purpose here. However, for a system this size, the fire flow is likely to be the biggest stress on the system. For fire flow, I suggest creating a second EPS, but this time with fixed HGLs in the tanks and use the timesteps to move the fire flow around the system (here, time is irrelevant). When I do a static model with fire flows, I typically make the first time step the Average Daily Demand (this is mostly for debugging the model), make the second time step the Peak Hour Demand, and make all additional time step Maximum Day Demand Plus Fire Flow. I set up fire flow demands at the critical nodes throughout the system then turn them on and off one at a time at each time step. If you search for my recent posts, I covered this a couple times for other Eng-Tips members.[/li]
[/ol]
REACTIONS
It doesn't appear to me that you are modeling water quality, so why (under TIMES) do you have the Quality Timestep set at 0:05, but the rest of the model is based on a Hydraulic Timestep of 1:00? If you are only modeling hydraulics, and a timestep of 1:00 is sufficient for that purpose, then I suggest changing the Quality Timestep to 1:00 as well. This will reduce the total number of iterations EPANET has to calculate and will save you time because this model actually a bit of time to completely run.
I hope this helps.
Fred
==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill
I took a look at your model, ran it, played with it, and exported the .net file to review. Here are my observations. This is not a complete critique of the model, but I think there is more than enough here to get you started and maybe even enough to get you finished. Please ask again if you have additional questions.
GAPS
[ol 1]
[li]The pipes and nodes N/NW/W of and served by SiladReservoir are not connected to the rest of the system. I suspect there should be something* connecting Node i-107 and Node n-106.[/li]
[li]The pipes and nodes SW of and served by MangaReservoir are not connected to the rest of the system. I suspect there should be something* connecting the tank or Node 107 to Node 62A and/or Node 62. Downstream of MangaReservoir is another gap at AlibaughReservoir. Here, I suspect there should be something* connecting the tank or Node i-124 to Node n-123.[/li]
[li]Immediately S of PoblancionReservoirs are two gaps, one between Node 103 and Node 53 and one between Node 97 and Node 88. I suspect there should be something* closing these gaps.[/li]
[/ol]
As you can see, there is a pattern here. When I look at each tank, I see one or two gaps in the system and I am pretty sure they should not be there. Also, your Water Treatment Plant is not connected to the system. What this means is that each disconnected part of the system is operating independently and no water is moving between them. I suspect that is probably not the actual case.
BTW, I inserted fictitious pipes in all 16 gaps (1m long, 25mm diameter, C=140) and got the model to run without generating any error messages. Most of the pressure problems are still there, but at least the model runs.
* That "something" might be:
[ul]
[li]A pipe, if the isolated tank is supposed to float on the main part of the system.[/li]
[li]A pump, if the isolated tank is higher than the main part of the system.[/li]
[li]A pressure reducing valve or pressure sustaining valve, if the isolated tank is lower than the main part of the system.[/li]
[li]Duct Tape, if all else fails.
[/li][/ul]
TANKS
At first glance, the tank elevations seem to be OK with respect to their service areas. The tanks are pretty small from my experience. However, if they really are small and are modeled correctly, then this is just an observation, not a critique of the model. BTW, if the Water Treatment Plant puts out a fixed or nearly fixed HGL (e.g. from variable speed pumps or gravity feed from the clearwell), then I would model it as a reservoir.
PIPES
[ol 1]
[li]I only looked at the first time step (T = 0:00 hrs). I noticed that some pipes are closed (e.g. Pipe 150, Pipe 152, Pipe 159, etc.). Should these pipes be closed? For example, Pipe 150 has an initial status of OPEN, but the first time slice has it CLOSED. I didn't find any controls or check valve that caused this, so it must be something else that I didn't find yet.[/li]
[li]The pipes are relatively small (50mm to 200mm). But, like the tanks, if they really are that small, then this is just an observation. At T = 0:00 most of the flow velocities are very low. I didn't check at the other timesteps.[/li]
[/ol]
NODES
[ol 1]
[li]I only looked at the first time step (T = 0:00 hrs). Most of the pressures looked normal to slightly high. I would expect the need for pressure regulators at some buildings to protect the plumbing and fixtures inside. The plumbing code I use here in California requires building pressure regulators for pressures exceeding 80 psi, or 56 meters.[/li]
[li]However, check the pressures at the following nodes: Well, i-104, i-124, 97, 103, 104, i-107, and 109. These nodes have pressure heads that are many times the Earth-Moon distance. Seriously. The biggest number I found was at Node 109: 2.18x10^7 m, which is nearly 57x the Earth-Moon distance. This is either magic or an error.
This appears to be related to the gaps and closed pipe in the system, since each of these nodes is next to a gap and/or has a closed pipe attached. Fix the gaps, open the closed pipes, and I suspect these high pressure heads go away.[/li][/ol]
CONTROLS
[ol 1]
[li]I don't see any controls. Should there be any? If there are any pumps that fill higher tanks or PRVs or PSVs that prevent lower tanks from overfilling, then there should be something.[/li]
[/ol]
PATTERNS
[ol 1]
[li]Pattern 1 looks like a reasonable diurnal curve.[/li]
[li]Pattern 2 is just straight 1's, which would be correct for fixed inflows and/or demands.[/li]
[li]Most nodes have Pattern 1, which makes sense.[/li]
[li]I don't see anything in the model that resembles a fire flow. I typically do static and extended period models, and haven't done a water quality model in years. The EPS is useful for modeling how tank levels fluctuate, and that is a legitimate purpose here. However, for a system this size, the fire flow is likely to be the biggest stress on the system. For fire flow, I suggest creating a second EPS, but this time with fixed HGLs in the tanks and use the timesteps to move the fire flow around the system (here, time is irrelevant). When I do a static model with fire flows, I typically make the first time step the Average Daily Demand (this is mostly for debugging the model), make the second time step the Peak Hour Demand, and make all additional time step Maximum Day Demand Plus Fire Flow. I set up fire flow demands at the critical nodes throughout the system then turn them on and off one at a time at each time step. If you search for my recent posts, I covered this a couple times for other Eng-Tips members.[/li]
[/ol]
REACTIONS
It doesn't appear to me that you are modeling water quality, so why (under TIMES) do you have the Quality Timestep set at 0:05, but the rest of the model is based on a Hydraulic Timestep of 1:00? If you are only modeling hydraulics, and a timestep of 1:00 is sufficient for that purpose, then I suggest changing the Quality Timestep to 1:00 as well. This will reduce the total number of iterations EPANET has to calculate and will save you time because this model actually a bit of time to completely run.
I hope this helps.
Fred
==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill
- Thread starter
- #7
I tried your suggestions and it run successfully on single period setting. I tried running it again using 72 hours time period but failed.
GAPS:
This is the way I was told to model this WDS. They said that the gaps would give pipes with unit headlosses within the allowable limit(10m/km)
Tanks:
1. I modelled an existing system so the tank sizes are as is.
2. For WTP, when you say fixed Hydraulic Grade line (HGL), you mean fixed inflows coming from both surface and spring water?
PIPES:
I slso encountered closed pipe status in some version of the model though they were initially open. I dont know why.
NODES:
![[bigsmile]](/data/assets/smilies/bigsmile.gif "[bigsmile] [bigsmile]")
Fixed the pressures after I closed the gaps just like you said.
PATTERNS:
this is interesting as i have never modelled this before. i am working on the fire flow EPS right now and I am searching the the threads you mentioned about it. But I may have questions as i go along. I hope its ok.
Thanks again this is very helpful.
GAPS:
This is the way I was told to model this WDS. They said that the gaps would give pipes with unit headlosses within the allowable limit(10m/km)
Tanks:
1. I modelled an existing system so the tank sizes are as is.
2. For WTP, when you say fixed Hydraulic Grade line (HGL), you mean fixed inflows coming from both surface and spring water?
PIPES:
I slso encountered closed pipe status in some version of the model though they were initially open. I dont know why.
NODES:
Fixed the pressures after I closed the gaps just like you said.PATTERNS:
this is interesting as i have never modelled this before. i am working on the fire flow EPS right now and I am searching the the threads you mentioned about it. But I may have questions as i go along. I hope its ok.
Thanks again this is very helpful.
-
1
- #8
makoy888....
If single period works and extended period doesn't work, I suggest looking at changes in tank volumes over time. You may have drained one or more tanks down to zero volume. (The EPS ran for me, but my gap-filling pipes and other edits are certainly different from yours.)
I don't understand the rationale you were given for creating gaps. It makes no sense at all. If the actual distribution system doesn't have these gaps, then the model shouldn't have them either. Instead, the model needs to include whatever elements exist there, such as pipes, pumps, control valves, etc. Also, how does one compute unit head losses for pipes that aren't in the model?
In addition, the gaps prevent water from moving from one zone to the next, which means the extended period simulation will not come even remotely close to reflecting actual system operation. The goal is to create a model that accurately describes the actual distribution system.
BTW, I rarely care about unit head losses unless it's for long-distance transmission mains. What matters more to me are total head losses (and the resulting downstream pressures) and pipeline flow velocities. I'm OK with a short pipe with a high unit head loss as long as the velocity is within reason.
Your model already uses a reservoir for the WTP. An EPANET reservoir has a fixed HGL (unlike an EPANET tank), which means that no matter how inflows and outflows change during your extended period simulation, the HGL of the WTP never changes. In most systems I have looked at, this is a valid assumption because the outlet of the WTP usually has a fixed or nearly fixed HGL (either clearwell water surface or outlet of variable speed pumps pumping out of the clearwell).
The closed pipe situation is going to take more study. If you can't solve it, post your updated model and I can take another look.
Fred
==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill
If single period works and extended period doesn't work, I suggest looking at changes in tank volumes over time. You may have drained one or more tanks down to zero volume. (The EPS ran for me, but my gap-filling pipes and other edits are certainly different from yours.)
I don't understand the rationale you were given for creating gaps. It makes no sense at all. If the actual distribution system doesn't have these gaps, then the model shouldn't have them either. Instead, the model needs to include whatever elements exist there, such as pipes, pumps, control valves, etc. Also, how does one compute unit head losses for pipes that aren't in the model?
In addition, the gaps prevent water from moving from one zone to the next, which means the extended period simulation will not come even remotely close to reflecting actual system operation. The goal is to create a model that accurately describes the actual distribution system.
BTW, I rarely care about unit head losses unless it's for long-distance transmission mains. What matters more to me are total head losses (and the resulting downstream pressures) and pipeline flow velocities. I'm OK with a short pipe with a high unit head loss as long as the velocity is within reason.
Your model already uses a reservoir for the WTP. An EPANET reservoir has a fixed HGL (unlike an EPANET tank), which means that no matter how inflows and outflows change during your extended period simulation, the HGL of the WTP never changes. In most systems I have looked at, this is a valid assumption because the outlet of the WTP usually has a fixed or nearly fixed HGL (either clearwell water surface or outlet of variable speed pumps pumping out of the clearwell).
The closed pipe situation is going to take more study. If you can't solve it, post your updated model and I can take another look.
Fred
==========
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill
- Thread starter
- #9
I have so many versions of the model. Attached here is the model of the system without gaps (following your advice)and with the original (existing) pipe sizes. the first model uploaded contains a much larger pipe sizes. Those were the result of my attempts at trying to make the system run. this version wont even run using the single period.
- Thread starter
- #10
Makov888…
I haven't forgotten about this. Unfortunately, my free time the past couple weeks has been very limited. I hope to dig into the latest model this weekend.
Fred
============
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill
I haven't forgotten about this. Unfortunately, my free time the past couple weeks has been very limited. I hope to dig into the latest model this weekend.
Fred
============
"Is it the only lesson of history that mankind is unteachable?"
--Winston S. Churchill
Lima Charlie
Mechanical
Dear all,
I see to be having the same challenge with negative pressure. I have checked all elevations and they seem fine. All nodes are at less elevation to reservoir and demand at the 6 points = 1.2 litres/second. I have attached the file. Someone help!
This is just a simple system I did for learning purposes.
Regards,
Lima
I see to be having the same challenge with negative pressure. I have checked all elevations and they seem fine. All nodes are at less elevation to reservoir and demand at the 6 points = 1.2 litres/second. I have attached the file. Someone help!
This is just a simple system I did for learning purposes.
Regards,
Lima
-
1
- #13
Based on your flow and diameters for pipes 3 & 4 you have a flow rate of 294.4 l/s in each which is giving you a total headloss in each of these pipes of 1,058 m and 3,609 m, respectively. You need to verify that your flows and or diameters are correct.
- Status
Similar threads
- Locked
- Question
