Water distribution controls for a small plant 6

[amazing azza]

Hello friends, I am planning a water distribution system for a small plant. There are a number of machines that use water (ie. continuously if machine is on), and there are a number of sinks and hose stations which may be opened at any moment for an arbitrary length of time.

What are some of the typical ways to control supplying water to this mix of users? Here is what I can think of:

1. Have the supply pump be on 100% of the time and install an overpressure valve at the pump outlet that sends pump output back into the water tank. This way the pressure in the supply line is always equal to (or less than) the rating of the overpressure valve.

2. Have the supply pump be on 100% of the time and pump water in a loop around the plant that leads back into the water tank. This way the supply line is always pressurized at the pressure that the pump can deliver. (Naturally, this value would be chosen wisely).

3. Have the supply pump be on 100% of the time and have a control valve at pump outlet that sends the output of the pump back into the water tank depending on the pressure in the line. Using PID control, the pressure in the supply line would be kept constant.

4. Install a pressure switch on the pump outlet such that if someone opens a valve (or machine turns on), the pressure in the line will fall and the pump will turn on (maybe with inverter control). Pump would turn off when the pressure in the line spikes as the valve is eventually closed.

I experimented with #4 a bit, but found that in real life the line does not stay pressurized, but rather falls to low values (~0.5 barg) fairly quickly once the pump is turned off.

 

[jwill23]

Amazing azza I am attempting to this same project and have read the whole thread. I have some confusion on what sensors/drive/PLC you have determined works best. I first tried the whole floweter setup but found the meter to not be responsive enough and sometimes just doesn't work. My setup now is that the the pumps run 100% of the time with a recirculation loop. I am wanting to only run the pump when needed by machine or operator.

 

[itsmoked]

jwill;

You need a PLC. Do you have any experience with PLCs? If so what make?

It will need to support an analog input.

It will either need to have an analog output to pass the needed speed to the VFD or it could use a protocol like MODBUS to command the VFD's speed.

You will need a VFD capable of running the pump motor. (The pump motor must be threee phase.)

You will need a pressure transducer that spans from 0.0 to the pressure you want to operate at plus a 25 or 50% so any accidental water hammer doesn't wreck the pressure transducer. The transducer analog out needs to match the PLC's analog in offering. Preferably both should be 4-20mA as that's typically easier to run as a sheilded twisted pair.

You need a check valve sized for the piping.

You need a small pressure tank.
~1/2 gallon for 3/4" water line.
~1 gallon for a 1" water line.
~2 gallon for a 1-1/2" water line.
~5 gallon for a 2" water line.

You'll also want things like RUN/STOP switches to disable it all.**

If I was going to do this I'd use inexpensive but quality Automation Direct stuff and I could get everything there except the pressure tank. I'd choose the CLICK! series PLC and get the CPU version that does both the Analog in and the Analog out. I'd use the newer Ethernet CPUs as they're really sweet and allow modifying the logic without even stopping the ladder execution.

I would also use the 4" color touch screen HMI (enormous bang-for-the-buck) so I didn't need switches** and so I could display lots of things like the pressure and the pressure setting and a page of good stuff like the PID settings while developing the PLC code of the application.

You may also want to use the standard flow meter with a pulse-out to show the actual flow occurring on the HMI since the above system does not know or actually care about flow. The pulse-out flow signal will be fine for showing water consumption even though it's completely useless for controlling the pump. You could totalize the flow too.

Operational note: I've recently learned not to run a VFD on a standard motor at less than 1/4 of full speed, or better, 1/3 of full speed as it can overheat the motor while pumping absolutely nothing anyway. So, keep that in mind in whatever algorithm you come up with.


Keith Cress
kcress -

http://www.flaminsystems.com

 

[waross]

Hi Keith. Is there any reason why you can't use the P function of a PID controller directly to the PLC? Set the PLC to either run at a minimum of 35% or stop the motor when the input falls below about 35%.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[itsmoked]

Hey Bill.

Possibly, certainly worth a try, however a lot of very popular lower-end drives don't have PID control. Often, too, they have to be in vector mode before they allow the PID to run. Conversely there are a couple of drives out there that have PLCs built into them that would probably handle this job. Because of the fairly non-linear aspect of centrifugal pumping I'm not clear on how a typical drive's P would manage. I'd like to try it sometime and almost did last month. But, after a week of turmoil on a very low dollar job I just went with a Grundfos SCALA2. I was seriously impressed with everything about it. Worked like a charm. ~$600+ bucks though, but that was less than me gathering everything, putting it together and writing the PLC stuff. The SCALA2 has a tank but it's so small and optimized that you can't find it in its stylized package. You can tell it's doing exactly what I described above. It's so quiet you can't hear it standing in the redwoods about twenty feet away.

PIDs in PLCs turn out to be pretty simple too. You can grass-roots one in about 5 ladder rungs. I've done many that way as the PLCs I use usually don't offer native PID functionality and upgrading to one that costs 50% more just so they do kinda galls me.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[waross]

The unit I suggested will output a signal to control a VFD.
We used Ramp and Soak controllers to dry softwood lumber years ago. CRT (Constant Rising Temperature) drying schedules were popular. Before that we used large case pneumatic controllers with a cam turning with the circular recording chart to drive the set point up scale. Changing a schedule for a different species meant removing and replacing the set-point cam.
Part of our design and build task was detailing and having a cam about 10" in diameter cut from a thin aluminum sheet.


Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[jwill23]

Keith I’m not too keen on PLC and have never programmed one myself. I was thinking that all this could have been done using pressure switch/transducer with p&id loop on a drive. I definitely didn’t think I would have to purchase and additional tank. I’m just not sure this process will work for mine. I may need to drop back and punt so to speak.

 

[itsmoked]

Hi jwill; I'd be interested to know if that worked. I think boundary conditions will cause a problem with your scheme because it can leave the pump idling with little flow actually occurring.

You could probably do that with a large pressure tank in the, I don't know, maybe 50 gallon rsnge and then use the VFD'd pump exactly like a normal ON/OFF pressure switch operated water system but instead using a spooling UP/DOWN pump instead of a bang ON/OFF. It would take some tinkering.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[jwill23]

Keith I was thinking that as the pressure drops too low.(determined by the flow of the pump) Than the drive could kick in the motor and run until an “over-pressurized” state and then shutoff. My thinking is if the machines or operators are using it that the pump would stay running. My problem I think I would have is pressure drop through the check valve too quickly than the pump would run just as much and I would get no benefit from it.

 

[jwill23]

After some further investigating online I did find a flow switch with a controller. Do you think this could be viable option either with or without the pressure switch? Or would I experience some other irregular running issues. I am worried that maybe the flow wouldn't be enough to engage the flow switch. Which is why I was leaning to the pressure switch.

 

[itsmoked]

Could you tell us some more about your plant?

Flow?

Lowest service flow?

Highest expected flow?

Does the flow stop entirely?

Fed from a well, or a tank, or municipal water supply?

Is this a booster system or does it provide most or all the pressure?

What kind of facility is it, an apartment complex, a single dwelling, a cannery?




Keith Cress
kcress -

http://www.flaminsystems.com

 

[amazing azza]

Hi guys, I suppose a progress update is overdue, as I have gone through a few iterations on the design of this thing.

The first way I had it running was with an external temp/process controller acting as a PID controller. It has a nice display for SP/PV and an easy to use interface. The start signal came from a low alarm and the stop signal was done via the "sleep mode" of the VFD. (Sleep mode is when the drive turns off if the command frequncy falls below a certain threshold. Very similar to what waross was mentioning). This worked, but the temp controller was not a very sensitive controller, and if you closed off the consumer slowly, the PID output did not fall as expected of it (this is undesired behaviour). Perhaps if I used the VFD's internal PID option it would have worked better, but I did not investigate it further.

The second way was with PLC assistance. I added a PLC, obviously If the pump was running, the program would periodically force-drop the drive frequency. The start signal came from a low alarm as before, but now the stop signal was driven by another low alarm that resulted (or not) from the momentary frequency drop. This worked surprisingly well (thanks itsmoked!), and I ran it this way for close to a year.

But I just can't stop messing with it. Recently, I was changing tank level sensing from using internal relays on a flowline ultrasonic level sensor (those relays are crap btw, don't use them. The sensor itself is ok, just don't use the relays) to a plain old analog pressure reading, seeing how I had a PLC now. I decided to also try using the built-in VFD PID functions instead of relying on an external controller.

Thus, the third way fed the process value signal directly to the VFD, which did its own PID. The VFD also forwarded this process value to the temperature controller, which was now only functioning as a display and providing alarms. The VFDs PID capabilities turned out to be *far* better than the temperature controller's, albeit a tad more difficult to puzzle through. This is how I am running it now.

 

[itsmoked]

Great feedback azza!

My thought was that a temp controller PID is not going to understand the cubic X[sup]3[/sup] response of the system (centrifugal pump) it was controlling very well.

Do you have a pressure tank involved?

What period do you use to slow the pump for checking for continued flow?

Did your VFD PID have a "Pump Mode" you're using?

In regards to the VFD's PID are you running P.I.D. or just P.I. or just P? Did you figure out how to tune it?

Keith Cress
kcress -

http://www.flaminsystems.com

 

[waross]

Did you set up the proportional band on the original PID or use the default setting?
The proportional band setting has a great influence on the response characteristics.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[amazing azza]

The main problem with the temp controller turned out to be its integration (and differentiation) time slice. On paper it claims 150 ms update cycle, but after using it I think that applies only to the proportional term. Integration appears to be done over 1 second slices. This leads to overcompensation. The only thing you can really do in this situation is turn down Ki. But even so, it was a totally usable system if you didn't mind an occasional oscillation around the setpoint.

I do not have a pressure tank. The flow checks are done once every minute. The VFD PID itself does not have a "pump" setting. It only concerns itself with things immediately related to PID, such as: PV<->frequency mapping, Kp/Ki/Kd settings, and such. There is one weird setting - response time. One can adjust the response time to be slower or faster (with K's being fixed). But this is claimed to be used for things like conveying, whereas for pumps, the manual specifically instructs to use a slower response. Being myself, of course I tried the turbo mode It tripped the drive as it tried to spin up too fast, heh. Didn't sound very pleasant either. The "pump" mode is present as a macro and affects the V/F curve (linear vs quadratic vs vector), overload threshold and minimum frequency limit. I did not investigate how, if any, this changes the PID response. But, yes, I do have it set to "pump".

With regard to PID settings, I played with 3 combinations P, PI, PID and found that the D term was almost never useful in this application. In the end I settled on PI. My tuning method is Ziegler–Nichols-ish. First, the controller is configured for P-only and the Kp (gain) is increased until oscillations appear. From there I turn it back down to where it is stable, but still has a large offset. Then I start adding the I term. Again, keep adding until it shows signs of instability, then go back to an intermediate setting.

 

[itsmoked]

Thanks for those fascinating details!

What VFD did you use?

Keith Cress
kcress -

http://www.flaminsystems.com

 

[amazing azza]

Glad to be able to share it with you all

For the water treatment plant I use Toshiba VF-S15's. Very happy with them. The waste water plant will use Siemens's, it will be an interesting head-to-head.

 

[waross]

I'm a believer in KISS, Keep It SSimple.
I start with P only. It may not be perfect, and may show deviations on the recording chart but often it produces satisfactory results.
eg; 10% proportional band with 5% offset will always be within 5% under normal operation.
Don't use Integral unless you must.
Derivative is mostly for very slow acting processes. eg; Temperature control with large thermal masses.
By contrast there are probably millions of water supply systems running on on-off control with deadbands or differentials of 30% to 50% of full scale.
How much effect will a 5% deviation have on wash-down hose performance for example?
Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[itsmoked]

More excellent points Bill.

Thanks Azza. I've also found Mitsubishi drives to be nice too.

Keith Cress
kcress -

http://www.flaminsystems.com