Remote control nightmare.

This week I was called in to "fix" some control issues with a set of 3 pumps in a remote northern village.

The setup was one master control PLC with a serial modbus radio transmitter and 3 remote pumps (10 miles away).

The remote controlers had only water levels and flowmeters analog inputs and digital I/O, the pressure sensor was at the main site.

So they had a lot of spikes and lows. I figure the PID was not stable. Once I looked at the control loop, they had a PID but no analog loop to control any pumps. They had a set of points where some pumps would start alone or in combination witht another up to all pumps ON.

No water tower!

Wow, quite a chalenge. I manage to get the system not to cycle to fast and be within ±10 PSIG of the setpoint. (SP=65 PSIG)

To do this I had to add some pretty precise conditions that delt with there fire-pump and holding tank filling but it worked. I will of course submit some recommendation about using VFDs but here is something that fascinate me during the days I observed the water supply.

Pump # 1 alone would give us 180 GPM at 65 PSIG, if I open a BIG leak it could go up to 240 GPM at only 35 PSIG

Pump # 2 alone would give us 270 GPM at 65 PSIG, if I open a BIG leak it could go up to 390 GPM at only 35 PSIG

Pump # 3 alone would give us 170 GPM at 65 PSIG, if I open a BIG leak it could go up to 275 GPM at only 35 PSIG

Now, ALL Pumps ON would give us 480 GPM at 65 PSIG, if I open a BIG leak it could go up to 510 GPM at only 35 PSIG

If we added the "alone" readings we got 620 GPM at 65 PSIG and 905 GPM at only 35 PSIG

I understand that the pressure variation augment to the square of the speed variation but the difference between 2 pumps runnig at max Flow was only 10% from 3 pumps. We got about 450 GPM with 2 pumps and 480 with 3 of them.

A soon as we would start the third pump we could see the flow drop sharply on the 2 other pumps.

Can you expand on this?

Thanks in advance.

Well, i'm not an expert on any type of fluid system, so I may be talking totally out of my a** here... If i'm too far off base, hopefully someone else will jump in and slap me down.

How are the pump suctions set up? It sounds like they don't have sufficient pump head to drive all three pumps at the same time. In effect, the pumps are "cavitating" or losing efficiency due to a pressure loss at the pump intakes with all three pumps running. With only two pumps running, it sounds like the inlet pressure is being maintained high enough to keep the pumps in the "higher efficiency" region.

It may not have anything to do with the pumps but with the dynamics of the piping system. IOW, without your "big leak" the max water that can flow down the system at 65 PSI is 480 GPM. Period. You can add pumps #4, #5, #6, #7, #8, and #9, and as long as the pressure is still 65, you won't get any more flow. Most single impeller pump curves are fairly flat at some pressure out to a given flow rate, at which point the curve falls off rapidly.

Thinking of it electrically, no matter how many 9 volt batteries you hook in parallel, you are not going to get more than 9 milliamps to flow through a 1K ohm resistor - but the current delivered by each battery will fall off as more batteries are connected in.

I think both Tom and Stalcup may be right.

If the pumps dont have NPSH (net positive suction pressure) they will cavitate. I dont think this is your problem.

I go more with Stalcup. As flow increases friction loss increases, increasing head. To get more flow you have to have more head thus you have to increase speed.

There is a limit of course ie a critical orifice (ie pipe) size) -- when this takes effective you cannot get more flow regardless of how much pressure you get.

I doubt if you have critical orifice. I believe your pumps are just fine and are just operating in accordance with their performance curves.

Of course if you increase pipe size you will reduce friction thus head and get more flow.
Dan Bentler

These are submersible pumps, all in there onw wells at close to 1 mile distance from each other.

In these places they need more water, they install another pump

In this case it does no good.

The man in charge at city all told me they where using close to 700 GPM ion hot summer days. This was based on his interpretation of pump speed (he trabnslated it to GPM) but now he has 3 flowmeters

This is not an uncommon phenomenon. It happens when the system curve (flow resistance vs. flow) is mostly friction without much static lift combined with a pump curve that is flat (the slope of the pump flow vs. pressure curve is small).

I've run into this many times, on pumps and blowers. The problem is that the backpressure rises as the square of the flow rate, so doubling the flow causes 4 times the pressure requirement. The pump output flow will be at the point where the pump curve intersects the system curve. If you have flat curves on the pumps, the increasing pressure pushes the pump back to the left on it's performance curve. Consequently, as you add pumps you get less flow from each pump. If the 65 psi was near the rating or max pressure capacity of the pumps what you saw is not surprising.

I try to avoid ever using or showing a pump curve without superimposing the system curve on it.

Doesnt that mean that it would be better to install a "booster" pump somewhere in the middle, rather than adding more pumps in the beginning of the line.

JesperMP said:

Doesnt that mean that it would be better to install a "booster" pump somewhere in the middle, rather than adding more pumps in the beginning of the line.

Click to expand...

It depends!

Better is a relative term, first of all.

You have to pay for the power, and flow times pressure is power, so from that standpoint it doesn't matter. If you have to build a new pump station for just the booster pump, the capital expense probably isn't worth it. If all you look at is control precision, them probably a booster pump is a good idea.

It is often possible to get pumps fitted with different impellers in the same housing to get significantly different flow vs. pressure performance. That would be the first thing I checked into.

None of the above is probably as cost effective as a stand pipe or even a water tower, which would even out the pumping load and demand.

I was also thinking about 4 times the pressure to achieve 2 times the flow. Isnt there a limit to how much pressure the pipes can take ?

Forgive me. I am obviously talking about stuff I know nothing about.

Its difficult to guess more accurately without knowing more about the pipe size and the dynamics of the system, but just looking at some tables I would estimate that a 6" pipe system of 1/2 mile in straight length is good for somewhere between 410 to 540 GPM at 65 PSI provided that the end of the pipe is open and net head is zero. 480 GPM in a 6" SCH40 pipe puts fluid velocity at 5-1/2 feet per second (you don't want to be over 7fps without other engineering considerations as a general rule of thumb).

JesperMP said:

Isnt there a limit to how much pressure the pipes can take ?

Click to expand...

Certainly, but unless the engineer totally blew the design process the working pressure of the pipe is greater than the maximum pressure that the pumps can produce. The allowable pressure for 6" Sch. 40, for example, is around 600 psi - burst pressure is several times higher.

Tom Jenkins said:

...This is not an uncommon phenomenon. It happens when the system curve (flow resistance vs. flow) is mostly friction without much static lift combined with a pump curve that is flat (the slope of the pump flow vs. pressure curve is small)...

Click to expand...

Again Tom you have guessed exactly. The Pumps are about 10 miles away but the at the same level as the city piping structures.

This is in a very nice touristic place and no tower could ever be installed. It would ruin the view.

They build the system many years ago but kept on adding pumps to where we are standing now.

How do you establish a system curve? Remember that we only have 3 pumps in on/off operation mode. One of them has an inverter which ic used has a very nice soft-starter (the other 3 have soft-start on them).

Sometimes some end-user will add equipments to a system to solve some irritants and they get a certain budget.

While they are at design stage they say:
- "Hey why not have this function also?"

That is where the **** its the fan. There budget was not for anything other than the fixing original irritant. Now they are not satisfyed of what they have BECAUSE they forgot what the first goal was.

In this case a man was driving many times a day to start or stop some pumps. He would even let run more than he needed. Hey more is better than less isn't it?

Thats before pipes begone to burst in houses in the middle of winter. 90 PSIG was a bit too much

So they fix this and with the same budget they which they can control it all.

I got called in as there last resort. Now its stable and I have to send them my recommendation. Not about the piping, but the controlers.

Thanks for your good inputs.

Thanks for all of the good information folks. I knew I kept paying you people for a reason .

I found this link to a page that talks about how to establish a system curve. I've never had to do anything like it myself, so I have ABSOLUTELY no idea how accurate it is. Any comments?

Tom Stalcup, that is a good description. For Pierre's purposes, though, he isn't likely to have the info (or the inclination) to calculate the head loss. For one thing there is a lot of fouling in old lines that causes friction to increase and the pressure to exceed the theoretical value. Besides, even the best pressure drop calculations are ±10%.

If you want to construct the system curve in an acctual application, you can do it with flow readings (if you have them) and pressure readings. Don't rely on the pump curve for the flow if you can help it.

If you know the static head (pressure) you can do it with one reading. Subtract the static head from the measured pressure, and use the formula Pdrop = k x Q². Back calculate for k, and then you can get the pressure at any flow by solving for the pressure drop and adding it to the static pressure. If you don't know the static head take three or four sets of flow and pressure reading and then throw the values into Excell to get the regression equation.

Tom Jenkins
I sorta knew this but you filled in the blank spots.
Thanks for the education

Dan Bentler