Control suggestion for a fan.

[itsmoked]

I have a refrigeration compressor that I'm controlling.
Part of the control process is running a fan that carries away the extracted heat after the refrigerant is compressed.
The ideal pressure at the compressor outlet is 225psig. As long as the process is not overloaded the compressor outlet pressure is controllable via how much heat is being removed by the fan.

I have absolute control of the fan speed.
The pressure really should not fall below 200psig.

I'm not very successfully controlling the pressure with the fan at the moment using a simple linear equation.

I know axial fan performance is NOT linear with speed though how that relates once the air has to also be blown thru a tight radiator is not very clear to me.

The pressure can go considerably higher than 230psig without harm.

I'm using a simple point slope equation:
Fan Speed = (20 * Pressure) - 4,000
This has the fan at full speed (numerically '600') when the pressure reaches 230psig.

I really don't have any idea of the correct style of control or equation I should be applying here.
Please give me your thoughts.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[PNachtwey]

It looks like you have the start of a feedforward.
You need a closed loop term.
What is the range of fan speeds?
From your equation the fan speed should be 500 when the pressure is 225.
Why not have a proportional band? between 200 and 250 psi?
The speed 0 at 200 psi 500 at 225 and 1000 at 250.
A non -linear approach may make the fan speed 1200 at 250 but I don't know the speed range.
This should be simple to get going.
A refinement would be to make the band non linear.
Another refinement would be to move the band up or down.
Keep it simple to start.



Peter Nachtwey
Delta Computer Systems

http://www.deltamotion.com

http://forum.deltamotion.com/

 

[itsmoked]

Hi Peter!

What is the range of fan speeds?

Anything between (reasonably) 200RPM and 1750RPM.

From your equation the fan speed should be 500 when the pressure is 225.

Yes. However, the above equation result is directly fed to a VFD that accepts it as 50.0Hz so the fan speed is NOT directly the output number. The motor is fed the power frequency (50.0Hz), in this example of; 500==>50.0==> 1750 x (50.0Hz/60.0Hz) = 1458RPM = Actual_Fan_Speed.

The 1750 is the speed that the fan would spin given the full 60Hz.
It's just another linear conversion.

Why not have a proportional band? between 200 and 250 psi?
The speed 0 at 200 psi 500 at 225 and 1000 at 250.

I like that. Given the thing above:
00.0 @ 200 psi (0 RPM)
30.0 @ 225 psi (875 RPM)
60.0 @ 250 psi (1750 RPM)
That would be a linear equation rising to the right passing down thru the X-Axis at 200, and up thru 60 at 250 psi. I think that's sort of what I've already done except I think I used only 230 psi as the 60.0Hz point and I'm seeing the pressure virtually never getting above about 210 psi. Decreasing the slope as you've suggested would probably much better center the operating point.

A non -linear approach may make the fan speed 1200 at 250 but I don't know the speed range.
This should be simple to get going.

200 to 1750RPM

A refinement would be to make the band non linear.
Another refinement would be to move the band up or down.

I just found the standard fan law is that the generated pressure of an axial fan goes up with the square of the velocity(rotational). The pressure is what drives the air thru the heat exchanger so the we can roughly assume the cooling goes up with the square of the RPM or frequency.

Does this mean the curve should be... a x² function? But we'd need less frequency increase per PSI so a natural log perhaps? ln(pressure)?

Keep it simple to start.

Sort of did with the mis-centered linear so maybe centering it would work but with the square function I bet not real well.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[MikeHalloran]

I'd scale it so the fan speed range starts at 100 rpm or so, not 0, because the fan has to cool itself, and below 100 (WAG), that just won't happen.

Also, if the compressor is running but the fan isn't, you will get phone calls, forever.

Or worse, self motivated folks may decide to 'fix' it.




Mike Halloran
Pembroke Pines, FL, USA

 

[djs]

Why not use a PID controller to close the loop? Does your VFD have a PID controller in the firmware? If so you can use that. More than likely the speed of the fan will depend not only on high side pressure but also ambient temperature, humidity and also the low side temperature. Too many variables for an open loop system.

 

[roydm]

I'm with djs simple proportional control with some reset.

It seems like you have a good idea of what the proportional band would be if it starts to oscillate at all increase the band until it stops then use Segler & Nicols equation.

If it's not a steady load I.e sudden changes perhaps some Derivative or feed forward.

 

[itsmoked]

Hi Mike; Yes, I agree a fan with no compressor will stir up the natives. Luckily in the application the fan cannot be heard below about 800rpm nor seen.


djs; If only these drives had PID! I could probably run it via the PLC.. Maybe I'll try that.

Hi Roy; Problem with this is the system runs the entire range of environments and loads. Sometimes really lightly loaded in cold ambients and other times staggeringly heavy loads in desert extremes and in all case varying in altitude and humidity. It will be hard to tune it and know that it didn't start oscillating horribly the minute it was out of my sight. The system will occasionally run up to 300psi with the fans unable to even keep up. Eventually I even have to run misters to prevent damaging head pressures. I suspect this would invite PID wind-up issues.




Keith Cress
kcress -

http://www.flaminsystems.com

 

[roydm]

Keith,
I can't imagine the conditions changing that quickly, more like minutes rather than seconds, the tuning should be very conservative, from your linear equation you already have an idea of the gain, perhaps use a setting of half that and just add a little reset action say 10 minutes per repeat or even less.

I assume the PLC sets the fan speed set-point for the drive and it has access to all the conditions could it have a simple lookup table based on the conditions?

Does the compressor run continuously when it's required or does it turn On/Off, if so you would need anti reset windup.

 

[itsmoked]

Roy, Certainly the conditions only change over many minutes perhaps 2 hours as it is a mobile application moving at about 90MPH. The PLC controls all refrigeration aspects including occasional fan reversals and knows about both the low-side and high-side pressures. The compressor(s) typically always run when sunlight is present. Multiple zone liquid line solenoids cycle continuously depending on the zone demands. One zone is a major consumer, at least as much as the other two zones combined if they're calling. The PLC is sending the speed out to the fan VFD.

Coming up with a table would be a daunting task. I wouldn't know where to start.

So you're suggesting a PI controller without a lot of I?

Keith Cress
kcress -

http://www.flaminsystems.com

 

[roydm]

Yes, the PID setpoint would be 225 psi, the integral action will wind the output up or down to achieve that.
I don't think you need to worry about the shape of the fan cooling slope, simple PI will take care of that, if the pressure goes above setpoint the I term will eventually wind the output up to 100% if the head pressure calls for it.

Your equation Fan speed = 20 * Pressure - 4000(X) takes care of the Proportional action all you need to do is add Integral, perhaps just change the value of X slowly depending on how far the Pressure is above or below 225#. On startup the value of X would be 4000, every few seconds the PLC would compare Pressure to setpoint and make a change to X proportional to the error.
The Zone solenoids cutting in and out will obviously upset the controller, perhaps they could cause a step change in the fan speed by adding or subtracting to X.

What is the range of your pressure transmitter, that effects the Proportional setting. I wouldn't limit it at 230 if it can go higher, it should take twice as much Integral action at 235 as it does at 230.

 

[amazing azza]

Does this mean the curve should be... a x2 function?

Would it be reasonable to manually run the fan at different frequencies and record the resulting psi's, followed by a session of curve fitting? Just to get some experimental data to corroborate (or not) the theoretical equation...

 

[itsmoked]

Hi azza.
That's still pretty hard since I don't have enough information to know what flow the PSI would provide thru the heat exchanger. If I could come up with a flow hood so I could actually determine the speed/flow numbers it would at least tell me what's going thru the exchanger. Of course then I'd need to understand its efficiency too.

The opportunity to do things like that has passed since the apparatus has left the area. Now I am relegated to trying things and then watching the telemetry to observe how well any particular attempt is working. Part of the problem is that California has been so cold for the past 2 months that there hasn't even been a chance do any real cooling. We had to artificially load the compressors just to check the refrigerant levels. We do that by forcing all the zone liquid line solenoids open and then we get a a couple of minutes of large refrigerant flow until the expansion valves shut us down.

And right now it's moving at 71mph thru Cincinnati in 46F. Sigh.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[amazing azza]

hi, itsmoked. How's your fan been doing? Did you get it figured out?

 

[itsmoked]

HI AA,

I've just stuck with:

Condensor Fan Speed calculation:
Full speed 60Hz@ 230PSI head pressure.
Zero speed 0Hz @ 190PSI head pressure.
Using the point-slope formula Y=20X - 4000 give the VFD frequency to be used as the speed reference.
VFD Speed REF(in Hertz) = 20 x (head pressure) - 4000
Any results that come out negative are to be rounded to Zero (0)!
NOTE: The VFD takes an integer with the least significant figure being 00.X.

and have not had to make any changes. (Crossing fingers)

Keith Cress
kcress -

http://www.flaminsystems.com

 

[amazing azza]

Simple is best