Closed loop heat transfer pump control problem

[StoneCold]

Greetings All
I have a closed loop heat transfer pump control problem that I am just not sure the best way to solve. Let me explain. We are a batch chemical plant. I have a chilled fluid loop that is recirculated by two centrifugal pumps. One pump is running and the other is in standby. You have to switch them manually if you want to run the other pump. Both pumps are on VFD's that feed information into a PLC. The current primary control method for the pump is to try to keep the discharge pressure at a set value(50 psig). The discharge pressure is measured with a pressure transmitter in the common discharge header downstream of the pump. The flowrate is variable depending on how many users are attached to the system it is measured with a flowmeter. Minimum flow is around 40 gpm and maximum flow is about 400 gpm. Flow beyond 400 gpm causes cavitation in the suction of the pump. So to try to minimize the cavitation of the pump we decrease the discharge pressure set point in 5 psig steps until the flowrate is less than 400gpm. The problem with this is there is no control logic to increase the discharge pressure again to try to optimize the amount of flow to the users.

The question is, with discharge pressure, flowrate, Motor Amps, Kilowatts available how should we be controlling these pumps so that I can optimize the service pressure and flow without needing two independent variables to use in the control logic?

 

[Latexman]

Solve the cavitation problem out to end-of-curve flow conditions, then only control on the discharge pressure.

Good luck,
Latexman

To a ChE, the glass is always full - 1/2 air and 1/2 water.

 

[ashtree]

The solution above is the only one available with one variable, but it may be easier to implement a solution if you accept a second variable which is flow. In that case you would control pressure but would have flow limit of 400gpm at which point you start backing off the pressure, by reducing pump speed.

Regards
Ashtree
"Any water can be made potable if you filter it through enough money"

 

[georgeverghese]

Agree with Ashtree - translating this into process controls language, we would have

a) A reverse acting discharge pressure controller that is set at 50psig which sets the pump speed
b) A reverse acting high set flow controller (set at 400gpm) cuts into the output from the PIC (at a low signal selector) to override the signal sent to the pump VFD.

 

[don1980]

Stonecold - possibly your question is being misunderstood. Based on my understanding, which is same as that for Latexman, the answer is an obvious and simple one. Fix the cavitation problem so you have a pump that can operate effectively over the needed range (range of flow and pressure), and then control the pressure so the process performs as it's intended to do. It seems apparent that the pump doesn't have enough NPSH-A to operate over the range of conditions that you need. NPSH-R increases as the pump runs out on the curve. You may be able to solve this problem simply by increasing the level in the pump suction drum, if there is one. Otherwise, I think you need a different pump - one that has a lower NPSH-R.

 

[georgeverghese]

The problem with the suggested solution would be when one or more users begin to demand more flow such that total flow exceeds 400gpm. In that case, the VFD would continue to slow down, header pressure would collapse, and all users will be starved of supply.

Another solution would be to find some user (or users) that could selected to be partially starved of supply in the event that total flow exceeds 400gpm. The high set FIC suggested previously would then be used to override the TIC output signal each at these selected users to keep total flow at no more than 400gpm (rather than to reduce pump speed as suggested previously).

Agree that if your total normal demand does exceed 400gpm, then these solutions wouldnt be the right approach. In this case, you'd have to either
(a) somehow increase NPSHa /decrease suction line frictional losses or
(b) add a NPSH booster pump upstream or
(c) changeout the pump altogether for one with lower NPSHr.

 

[ashtree]

To add to the list above put a Maric valve in the discharge line that limits flow to 400gpm.

All that aside i suspect it may be cheaper to change the control logic to accept a second variable.

Regards
Ashtree
"Any water can be made potable if you filter it through enough money"

 

[LittleInch]

What does"closed loop" mean? If this is a pressurised loop why can't you just increase the static pressure?

What does the pump curve look like?

Two inputs into a plc is simple. Just take the lower output through a low selector block and you can control on any number of inputs very easily.

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[StoneCold]

Thank you for taking the time to reply. Sorry I have been out for a few days.

The original design conditions were fine for the pump. But like everything things change, and now cavitation at the suction is a problem because the fluid has a much higher viscosity.
Changing out the pump or the piping is not going to happen.
So I guess we are really talking about a clever way to control on two variables and try to optimize the system.
I would like to go back to a PID loop for the system discharge pressure with the set point at 50 psig.
The quandary I have is how to override the pressure set point in such a way that I can keep my flow rate at 400 gpm or less and then automatically ramp the pressure back up as my flow rate falls off.
It seems like there should be some cascade to the loop.
IF GPM > 400
Then (Pressure Setpoint)*K*400/GPM= controlled variable
K being some arbitrary factor to correct the flow
Else
Controlled variable= Pressure setpoint(50)

I don't see how a low block selector would solve this problem on its own.

Regards
StoneCold

 

[georgeverghese]

With the limited information posted, I would prefer to go for the second process control option, since
a) Header pressure is still maintained during the override
b) The override control action is limited to one or two selected users which may be the cause for this high flow and can be temporarily starved of coolant supply. None of the other users will be affected in this option.

This override action is enabled by the high set FIC which would cut out the output signal from the TIC located on each of these high demand users through a low signal selector block. Ask a process controls engineer to help to explain how this works if you are not currently familiar with override control blocks - certainly not rocket science, so shouldnt take you long to get the picture.

On the other hand, what do you think of Little Inch's suggestion to increase coolant expansion drum pressure to get you more NPSHa ? If the corresponding pump shutoff head, in some rare low flow case, would be higher than the process design pressure of the coolant supply system, you could install a high pressure instrumented trip loop and a blocked flow PSV on the pump discharge??

 

[LittleInch]

"I don't see how a low block selector would solve this problem on its own."

Let me try and explain.

In your case you have two variables being controlled in a PID loop which outputs a speed signal -
One tries to maintain 50psig
one tries to maintain 400m3/hr

In low demand, the lowest speed signal will be set by the 50psi controller as the loop trying to maintain 400 gpm would output a higher speed leading to a higher pressure

At high demand the lowest speed signal will be set by the flow controller if >400gpm as the pressure controller will try and increase speed to try and maintain pressure.

As flow reduces< 400gpm then once the pressure rises to 50 psi or higher, the pressure control loop will send out the lowest signal and seamlessly "take over" the speed control

The low selector block just takes the lowest speed signal from the two loops and outputs that to the pump controller.

You could add in any other controlling elements such as temperature, motor amps, manual control you name it, the one outputting the lowest speed takes precedence.

Thus it's not a matter of "override", its a simple, reliable well proven method of seamlessly changing control of an item from one PID loop to another and back again automatically.

If you can't fix it mechanically (only you know) then this is the next best thing. We don't know your system so all this talk of "starving" systems isn't clear. If they work at 45 psig, but just get a bit hotter then so be it - your system has limits and you work within them.

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[StoneCold]

georgeverghese
There are no control valves on the individual users. They simply connect a hose to the system and turn it on. Unfortunately there are about 30 connection points available so they can connect more users than the chiller can practically serve.

LittleInch.
Thanks for walking me through the logic that you mentioned with the low block selector.
I totally get what you are saying now and I think that is a great idea.
I was thinking the selector block was on the inlet of the PID loop but in reality it is on the outlet side of two loops, just determining the value of the 4 to 20 ma signal.

Thank you very much.

Best Regards

StoneCold

 

[LittleInch]

I love them once I found out about them as it got over all this "override" idea.

It can work with any number of inputs, even a manual one if you really want to fiddle with it but then it limits the speed to what one of the others won't let you go faster than it wants. If you find motor amps are causing an issue you can add that in as well...

Remember - More details = better answers
Also: If you get a response it's polite to respond to it.

 

[georgeverghese]

Well, if there are no control valves or any other automated valves on these users, can see why you want to somehow reduce the main supply pump speed.

But before we drop this idea of choking the supply to one or more of these users, can you not retrofit an ON-OFF valve at least to these optional users ? Then, at flows greater than say 410gpm, you'd close these valves, and at some lower flow, you'd re open the supply to these users. This concept may be prone to severe cycling though. A continously modulating control valve on each of these optional users would be much better.

The last resort in my opinion would be the cheaper scheme to drop main pump speed through the high set override FIC as described earlier through the low select. In this case, supply flow to all existing users will drop as the more recent users are hooked into the supply grid.

 

[Rspain]

I essentially agree with Little Inch, one observation however.

Your system has only 1 'external variable' (as stated): flow demand. For a given flow demand, there is only one frequency that can match a given (discharge pressure, amperage, etc.)

No problem to run independent PID loops and select the most restrictive with a smooth change.

In the case of discharge pressure and flow rate, situation is clear.
For a given flow demand, each PID loop will suggest a different output (actually I think it is the P factor that shall decide the switching with a reset in I), you will choose the one that results in a point more to the left of the pump's characteristic curve.

Just be carefull in case you add many parallel conditions, to make sure that you don't get into an infinite unstable back and forth loop because one control loop pushes you into another loop, and viceversa (since both are actually dependant to each other and you have set incompatible set points).

I don't know if I have been able to make my point clear.

Sorry if not,
BR,
Roberto