DI Water Centrifugal Pump Flow PID control

Camcanpro

Member
Join Date
Jul 2011
Location
Toronto
Posts
15
I need to control DI water flow rate at about 45 kg/min. The water come out from a day tank and water will pass a centrifugal pump, a E+H Mass flow meter and a SPX flow control valve. We use PID to control the openness of the valve to maintain the flow rate. The valve is opened at around 15% by PID loop to keep flow rate at 45Kg/min. Centrifugal pump is controlled by a starter. Please see attached file for the PV and SP trend. PV is blue color and SP is red color, Green is the CV. PV is oscillating around the SP, Our current PID is: P: 18, I: 10.5 1/s and D is 0.5 s, the loop update time is 100ms.

(1) What cause the oscillation?
(2) My target is to maintain PV to be within 1% of SP.
(3) I have spent a lot of time to tune this loop, including using AB PIDE auto tune and Rstune, but none of them worked. I just don't understand where the oscillation coming from?

Thanks for help.

DI water PID cotnrol.jpg
 
It would seem you have some disturbance affecting the system that will require some detective work from you.

You might start with a copy of the P&ID diagram. With the system running as depicted in the trend (pump on, level steadyish), put the loop in manual and see if the oscillations persist. (you may have to control the valve manually to ensure the tank doesn't empty or drain). If the oscillations persist while the loop is in manual, that means you have some disturbance that is affecting the level. It could be one of many things that may require you to go out by the process and snoop around (put loop back in auto first):
1) the valve positioner could be hunting (hopefully there is a positioner - if not, that could very well be your problem). Is the valve trim properly specified or could the valve be oversized.
2) there could be a pressure makeup or vent on the tank that is cycling open and closed.
3) there could be something downstream that is inducing pressure fluctuations (which affects flow from a centrifugal pump).
4) Look at the P&ID and process to figure out / rule out sources of the disturbance.
5) Is there a vortex breaker at the discharge of the tank?
6) Et cetera (sometimes you may feel or hear the offending disturbance while you're out there)

Also, most often, tanks are there to allow for slight variations in inventory. Are you sure the spec you're trying to control to is based on a real need?

Good luck and let us know what you find.
 
Last edited:
I agree with the disturbance statement. Does the flow change if the valve position is held constant?

This problem is similar to another recent problem where someone wanted to control flow. I do a lot of flow control. One thing that makes flow control easy is to keep the pressures constant. If the pressures aren't constant then pressure sensors are needed so one can adjust the gains as the pressure drops changes.
 
P&ID of the flow control loop

Thanks for all the reply. Please see attachment for P&ID. There are two such kinds of flow control loop, Using centrifugal pump and flow cotnrol valve to maintain flow rate. I did the test, even put the flow control valve in manual, the flow still ocsilate. Totally we have 4 such kinds of loop in our project, each one has such disturbance problem. For other loop which have PD VFD pump control, almost no disturbance. So I guess the problem coming from the centrifugal pump and pressure not constant.

Other color represent another bulk ingredient PV and SP which Setpoint will follow master water flow rate at certain ration.

 

P&ID flow control.jpg
 
Yes, you've got it. If you put the flow control valve in manual mode then the pump curve should be constant. Either the speed control on the pump is very sloppy or there is something possibly down stream that is causing a changing back pressure. That would change the system curve.

You know about pump and system curves don't you?
http://www.engineeringtoolbox.com/pump-system-curves-d_635.html
 
This seems an unusual arrangement, particularly the tank/piping/instrumentation on the right. If I had to bet on one thing, I'd say it's the whole backpressure arrangement on the right; what controls the backpressure valve in the vertical line to the far right of the picture and to what value is it set? Can you put this in manual or change its setting to see what effect it might be having?

This whole arrangement begs further questions:
1) What's the relative volumes of the tanks?
2) Is the right-most tank one of many that is fed by the left-most tank via a header arrangement?
3) What's downstream of the right-most tank?
4) What maintains headspace pressure on each of the tanks?
5) Just what are the overall process objectives/constraints and how is this arrangement intended to satisfy it?
6) Which flow are you having a problem with?

I suspect this is less of a tuning problem and more of a process design / control objectives specification problem...
 
...or there is something possibly down stream that is causing a changing back pressure.
A boiler perhaps? If it is using De-Ionized water from a Day Tank, a good guess is this pump is a Boiler Feedwater Pump. Boiler water level control is difficult, because the boiler rocks along fat, dumb, and happy at a set pressure, until it suddenly needs a large amount of water, the pressure drops off, but not too much water can be added, or the boiler will be overfilled and lose its air pocket and cause a steam explosion.

I am not up-to-date on boiler controls, but in the US, the final boiler water level could not be controlled with a PLC. It must have a certified and tested control device. At least that is the way it used to be. The old McDonnell-Miller Boiler Level Pump Controller was almost a standard. It worked at any boiler pressure (up to the maximum rated controller pressure).

http://unitedstates.xylemappliedwat.../series-93-low-water-cut-offpump-controllers/
 
Last edited:
In my opinion this PID loop just needs better tuning.

The most obvious pointer to this is the reaction to the left of the value bar. The PV starts to rise, and the CV starts to close to correct the PV, but it is not reacting quickly enough. Eventually CV is taken low enough to bring the PV back down to SP. But it all looks too slow and sloppy.

The PID control loop ought to be able to deal with "disturbances", and I see no reason why you have to start with a stable flow with the loop in manual.
 
In my opinion this PID loop just needs better tuning.
Camcanpro can try increasing the gains but the distrubances will always be noticed. The question is can the gains be high enough to stay within 1% of the SP.


and I see no reason why you have to start with a stable flow with the loop in manual.
You don't but we found out that either pump velocity control is not steady or the down stream back pressure is changing the system curve and therefore the operating point.
 
Camcanpro can try increasing the gains but the distrubances will always be noticed. The question is can the gains be high enough to stay within 1% of the SP.



You don't but we found out that either pump velocity control is not steady or the down stream back pressure is changing the system curve and therefore the operating point.

Peter, I respect the length and breadth of your knowledge immensely, but in this instance I have to say that controlling flowrate to within 1% tolerance with a PID is achievable in most situations, even when the gods are against us with destination back-pressures, source gas breakthrough, pump cavitation, crude vessel top-pressure controls etc. etc.

Of course I understand that sudden and unpredictable disturbances caused by one or more of many events can really "challenge" the tried and tested PID algorithms, and in these situations we do our best with lots of D gain - although sometimes we cannot rely on the PID on its own and add some sort of additional control (aka - fudges) when we see, or can predict, these disturbances.

There are many cases where accurate, controllable, flow control is essential - blending, dosing, pasteurisation, etc.

Blending and Dosing are very similar processes, but accurate ratio control is required, so we may be modulating the SPs on the "slaved" flow loops.

Pasteurisation is a real challenge, as the flowrate through the pasteuriser must be maintained accurately.

When the OBV (Outlet Buffer Tank) becomes full, the product must be purged out of the pasteuriser to the OBV, replacing the product with water. So we get a source pressure change when the inlet changes over from the Inlet Buffer Tank (IBV), to the water main. When the purge volume is achieved, the water in the pasteuriser goes into recirculation, and we suddenly lose the OBV back-pressure (head).

Similarly, it all happens in reverse when we want to go "forward-flow" again, the water in the pasteuriser is diverted to drain when the product is re-introduced (a step-change in the source pressure from the IBV), and after the "prime" volume, the outlet route is changed over to the OBV (a step-change in the outlet pressure).

All the time these pressure changes are taking place, the flowrates and temperatures in the heat-exchangers must be maintained accurately, to maintain the number of "Pasteurisation Units" (a calculation which has exponential terms). Often, cascade PID controllers are used to speed up the response times, especially in the temperature control PIDs.

To say that we must have stable flow with a fixed control valve, or pump speed, to be able to achieve accurate flow control with a PID is wrong - we put PID control in place to couteract the effects of source/destination pressures.

All that said, our hardware must be up to the job as well. Control Valves must react quickly enough to the control system demands, drives must react quickly enough to the control system demands (most are shipped with acceleration/deceleration ramps enabled by default - i take them out). I don't want my fast-acting PIDs being throttled by sluggish response from drives and control valves.

The trend that the OP posted clearly shows a typical "unresponsive" PID reaction to a change in the PV, it is slow to respond to changes in flow, and when it does respond, it overshoots, and then has to sluggishly respond in the opposite direction, and because it is slow to respond, it overshoots again.

I'll say it again - this loop just needs tuning....
 
In my opinion this PID loop just needs better tuning.

The most obvious pointer to this is the reaction to the left of the value bar. The PV starts to rise, and the CV starts to close to correct the PV, but it is not reacting quickly enough. Eventually CV is taken low enough to bring the PV back down to SP. But it all looks too slow and sloppy.

The PID control loop ought to be able to deal with "disturbances", and I see no reason why you have to start with a stable flow with the loop in manual.

Peter, I respect the length and breadth of your knowledge immensely, but in this instance I have to say that controlling flowrate to within 1% tolerance with a PID is achievable in most situations,
I agree that the gains can be increased as the response does look sluggish. You may be right but we don't know for sure. I didn't expect to get any feed back on increasing the gains until the work week started. My point is that we don't know. The key is to increase the rate of response of the PID by decreasing the integrator time constant and see how that goes. If that doesn't work then knowing the down stream back pressure will be key.
 
The PID loop is fast enough to respond to these upsets.. IF:

A. Your flowmeter is fast enough to detect the disturbances immediately.

B. Your control valve is fast enough to respond immediately to the upset.

C. You don't mind wearing out control valves chasing the disturbances.

If, on the other hand, you can determine the physical cause of the disturbances and remedy the source of the problem, rather than merely attempting to respond to it, then you will have arrived at a superior solution.
 
The PID loop is fast enough to respond
I wasn't referring to how fast the PID updates. I was referring to how short the integrator time constant can be and not oscillate. If the integrator time constant can be 3 to 5 times shorter or faster than the time constant for the disturbances then the error should never exceed 1%.

The update time of 100 millisecond seems to be plenty fast enough. I am assuming there is 12 second between divisions on the trend. The current integrator time constant is 10.5 seconds. The disturbance seems to occur over about 6 seconds before the integrator winds up to correct. What if the integrator time constant is set to 2 seconds? Now the integrator may be able to wind up faster than the error can deviate 1%. The valve should be able to change that fast.
 
Working with a control valve opening of 15% suggests that your valve is way over sized at process conditions given. It may be that your controller has been tuned very loose because of your oversize control valve hence the apparent drifting of your PV?

Flow control loops are typically low gain (less than 1), low reset time (or high reset repeat rate) and No D action required.
 

Similar Topics

Hi, I am working on automating an industrial fabric shrinkage tester to replace its outdated electronics with a PLC. To get the tank's water level...
Replies
14
Views
502
Hi all, i have an application in which i need to read the water level in a 1 1/2" diameter tube that is 24 in. tall. the sensor will be mounted...
Replies
27
Views
1,824
Is it good idea to use ultrasonic clamp type of flowmeter for water dosing? I already using the flowmeter pulse output by setting to 10L per...
Replies
14
Views
1,257
I have an issue where I have a DP transmitter on a horz tank reading liquid level. Due to odd and unforeseen circumstances the tank filled to the...
Replies
17
Views
2,576
The Scadametrics people have a module that interfaces water meters to PLC protocols. It's called an Ethermeter. It can count pulses and has some...
Replies
6
Views
2,141
Back
Top Bottom