Sanity CHeck - Process Control

ASF

Lifetime Supporting Member
Join Date
Jun 2012
Location
Australia
Posts
3,907
Hi all,


I have a process with multiple required outcomes, with different devices controlling different process variables, but with each of them affecting the other. Just looking for a sanity check to make sure that (a) it's possible to achieve what I'm trying to achieve, and (b) that I've got it worked out correctly.


Attached is a sketch of the process (please excuse the roughness, I do have proper CAD drawings but they include a lot more ancillary equipment and are a bit convoluted to work through, so in the interests of simplicity I've just drawn in the critical equipment). We have a chamber through which we circulate air using a supply and return air fan. We can draw the supply air from outside and exhaust the return air, or we can recirculate the return air directly to the supply air, or any proportion of the two, using the motorised dampers MD5, 6 and 10. Any exhausted air is drawn through a heat exchanger to warm the supply air coming in. We monitor the fresh air flow using FT1, and the recirculated air flow using FT2. We monitor the pressure in the chamber using PT1.

Dampers MD1 and 4 are not relevant to the application (I don't think so, anyway. I've drawn them in in case someone has an idea for a way they could be used to enhance the process control). They exist because the actual application has duty/standby fans and so we need to be able to open/close as the duty changes to prevent air short circuiting via the standby fan.

All dampers are analog modulating dampers, and all fans are on VSD.

We have three objectives:
1. Maintain a set negative pressure setpoint in the chamber (e.g. -15Pa)
2. Maintain a set airflow through the chamber (e.g. 500L/s)
3. Maintain a set fresh/recirculated ratio of air (e.g. 30%/70%)

To break that down into individually controllable items, I am thinking as follows:
1. Total (actual) airflow through chamber = FT1 + FT2
2. Airflow Setpoint through FT1 = Total airflow setpoint * ratio of fresh air
3. Airflow Setpoint through FT2 = Total airflow setpoint * ratio of recirculated air

And then to control all of these items, I am thinking as follows:

1. Supply Air Fan varies speed to control total air flow (FT1 + FT2)
2. Return Air Fan varies speed to control chamber pressure
3. Supply Air Damper MD10 varies position to control the airflow through FT1
4. Return Air Dampers MD5 and MD6 vary position inversely in tandem to control airflow through FT2

I am envisioning that I will have to have a startup sequence of sorts because each of the process variables depend on the others to an extent. Here's what I'm currently thinking:
1. Open MD10 and MD6 fully, close MD5 (100% fresh air, 0% recirculation)
2. Start return air fan until chamber pressure setpoint reached
3. Start supply air fan until total airflow reached. Return air fan will ramp up to maintain negative pressure setpoint as supply air fan speed increases
4. Start modulating MD4, MD5 and MD10 to achieve required airflows through FT1 and FT2. Supply Air Fan may ramp up or down as airflow is disturbed but as long as I don't have it tuned too aggressively, it should ultimately settle out.

Anyone got any thoughts/expertise/advice to share?
 
Last edited:
Image doesn't look to be attaching. Try this:
33xirls.jpg
 
Holy interdependencies Batman!!!!

Since since application is moving air I suppose nothing is linear too.
Are there are formulas for the flow through the fan as a function of RPM and pressure difference?
Are the pressure sensors so provide feedback that the pressure in the chamber is -15 pa?
What about pressure sensor in other places?

I wouldn't call MD1 irrelevant. If MD1 closes it is easier to reduce the pressure in the chamber. Of course slowing down the supply relative to the return fan will do that too but that may reduce the flow through the chamber by too much.

I know how I would approach this. I would make a cost function using the 3 criteria you gave. I would have a weight for each criteria and square the errors between the desired and actual values for each of the 3 criteria.

If there were formulas this would be easy but if there are no formulas I would start tweaking on coefficients/gains for each fan and dampers. After tweaking a gain I would check if the cost function dropped. I would keep tweaking a gain until the cost function was minimized for that gain and move on to the next gain and tweak that until the cost function dropped to its minimum. I would go through all the gains and probably go through them again a few times since the gains will interact. All the while I would be minimizing the cost function.
 
Since since application is moving air I suppose nothing is linear too.
Are there are formulas for the flow through the fan as a function of RPM and pressure difference?
I have the fan documentation which has a graph on it, I'll attach it separately below.



Are the pressure sensors so provide feedback that the pressure in the chamber is -15 pa?
What about pressure sensor in other places?
There is an analog pressure sensor in the chamber, but no others (except for DP switches on fans to prove airflow)


I wouldn't call MD1 irrelevant. If MD1 closes it is easier to reduce the pressure in the chamber. Of course slowing down the supply relative to the return fan will do that too but that may reduce the flow through the chamber by too much.
That's the sort of thing I suspected it might be useful for, which is why I left it as a modulating damper just in case :) My current plan is to keep that option in my back pocket in case I need it, but to avoid it if possible because we already have enough interdependencies!

I know how I would approach this. I would make a cost function using the 3 criteria you gave. I would have a weight for each criteria and square the errors between the desired and actual values for each of the 3 criteria.
Can you explain a little more what you mean by a "cost function"?

If there were formulas this would be easy but if there are no formulas I would start tweaking on coefficients/gains for each fan and dampers. After tweaking a gain I would check if the cost function dropped. I would keep tweaking a gain until the cost function was minimized for that gain and move on to the next gain and tweak that until the cost function dropped to its minimum. I would go through all the gains and probably go through them again a few times since the gains will interact. All the while I would be minimizing the cost function.
Hopefully once I understand the concept of a cost function a little more I'll understand what you mean here.
 
Interesting. I’m just approaching this at first impression just to maybe give you some outside looking in suggestions.

I know you said you don’t think MD4 or MD1 are relevant, but to me they look like the two main valves you need to control.

I would use MD5 for recirc and MD6 to release to exhaust.
The system does seem very redundant however. If these aren’t very big fans then I don’t see any reason to be modulating fan speed. Most fans or centrifugal blowers like to run on the higher side of their curve. Modulating the fan speeds would require you to have a minimum speed that is high enough that the fan is actually doing useful work. You may find that the fan doesn’t produce a useful amount of flow until it is over half speed. By that time, you may find that the last half of the motor speed doesn’t have very much affect. In which case, it would be pointless to modulate the motor speed.

UNLESS

In order to keep a negative pressure, the return fan will need to be moving more air than the supply. Slowing the supply down would help.


In my mind, in order to keep a negative pressure in the chamber, you will need MD4, MD10, MD1, and MD6 open. The rest should be closed. It’s not going to have a negative pressure if the air is not moving.

Recirculating the air at 100% should equalize the pressure to zero..I think. So it seems that the recirc is just there to recoup some product or good air. Say like an idle state or something.

All in all, I think your approach is correct with what you got. Sometimes a fresh look from someone else helps and I hope I didn’t steer you off track. Thoughts popped in my brain and I just wrote them down.
 
Thanks Seth. It's a closed loop so at yes, the return fans will need to move more air than the supply fans to achieve/maintain a negative pressure.

Good point too about the fans working best near the top of their curve - the more I think about it, the more I think the fans speed might end up having to be static, and the flow rates controlled by the dampers. I can probably permit the fan's "fixed speed" to be adjusted on the SCADA within reasonable limits - as long as it stays above such a speed as to achieve the required airflow with MD1 fully open, the damper should be able to trim it as necessary. But I'm quite (read: totally) unexperienced in this area, so definitely not sure!
 
ASF, hi. I have been working on several air handling units over the last year. I was new to this, and there are some things we (our company) have learned.

Our application originally started with 4 rooftop units, one of which is a "make-up" air unit, meaning it draws in outside air. The other 3 are climate control units and do not take in outside air. They have burners and cooling coils for heat and cool. Of course supply and return fan arrays. The 3 climate control units use a damper configuration similar to yours, and when the outside temp, plus the dewpoint or "enthalpy" allows using outside air to cool or heat, we use the dampers in "economizer" mode, I. E. "Free" heating or cooling. These are for a large office building at Ford.
I am also working on a test cell with 3 other units, and the controls for that are vastly different.
A few cautions--most of the damper wiring, freeze stat, smoke detector etc are hard wired, because failure to have dampers open when starting up can cause damage, I. E. duct implosion, and other duct damage.

You are right in using very slow gain controls in your PID's, I have 6 PID's for each rooftop unit and all totaled 13 for the test cell units.
Modulating the dampers is good, I modulate a minimum outside air damper and the economizer dampers, plus the fan speeds for static pressure in the spaces.
I probably didn't give you too much help, just be careful with the damper action and the fans ramping up too fast. I could send you some flow charts of the units if it will be of interest to you. Good luck, and I've always wanted to see Australia.
 
ASF,
If you can, can you tell us what this process is actually doing? Just curious.
At first glance, it looked like a quench box. Then saultgeorge mentioned HVAC and that makes more sense.
 
We have three objectives:
1. Maintain a set negative pressure setpoint in the chamber (e.g. -15Pa)
2. Maintain a set airflow through the chamber (e.g. 500L/s)
3. Maintain a set fresh/recirculated ratio of air (e.g. 30%/70%)

To break that down into individually controllable items, I am thinking as follows:
1. Total (actual) airflow through chamber = FT1 + FT2
2. Airflow Setpoint through FT1 = Total airflow setpoint * ratio of fresh air
3. Airflow Setpoint through FT2 = Total airflow setpoint * ratio of recirculated air

And then to control all of these items, I am thinking as follows:

1. Supply Air Fan varies speed to control total air flow (FT1 + FT2)
2. Return Air Fan varies speed to control chamber pressure
3. Supply Air Damper MD10 varies position to control the airflow through FT1
4. Return Air Dampers MD5 and MD6 vary position inversely in tandem to control airflow through FT2
Anyone got any thoughts/expertise/advice to share?

Three thoughts:

1. It is clear from the description there will be multiple interacting control loops. In order to achieve stable operation that meets process objectives, these loops will need to be decoupled implicitly (i.e., different closed-loop relative response rates) or explicitly through complexity of the control system design (e.g., classical control with decouplers, or model-based multi-input, multi-output). Without this decoupling, it is likely the loops with "fight" each other, with the control system magnifying process disturbances.

The implicit approach is generally more intuitive and less costly to implement. Sometimes it happens naturally, such as between a flow/temperature regulation where flow is almost always much faster to control than temperature. In response to a disturbance, flow will quickly settle to its target (maybe a moving target) while temperature is still working to get back to its set point.

In the above process description you have competing flow and pressure objectives, and these could have very similar relative responses, leading to undesirable control loop interaction.

Your options include: (1) take actions to impact open-loop time constants, or (2) tune the closed-loop responses in priority of objectives, with slowest response being lowest priority. An example of option (1) would be accel/decel settings of the fan VFD's. Another might be set-up of the damper controls. It is probably best to do some experimentation with the system once it is built to see if there are any natural, inherent differences in open-loop response. One way to do this is by bumping each final control element ("CV") individually while holding the others constant, and examining a trend chart of the process measurements ("PVs").

2. The described strategy sounds like a good starting point, though I would consider initially simplifying it as follows: (a) for strategy step #3, do not actively control MD10; start with leaving it at 100% or mostly open, and (b) choose either MD5 or MD6 to achieve the desired recirculation ratio, not both in tandem (one of them stays at a fixed position). I think the MD5 or MD6 choice would depend on whether total flow or recirculation ratio is more important to regulate. My thought here is to let fan speed dominate control with minimum (or constant) pressure drops in the system.

3. It is a very good idea to have a well planned start-up strategy, as you have already considered.

In general, my inclination is to start with the least complexity necessary to meet objectives, and add manual/automatic control elements as issues are discovered. (The exception would be process/personnel safety-related features, which would be included and tested at process start-up.) Usually, the less moving parts, the easier to troubleshoot in the long run.
 
ASF,

Nice job defining the problem and providing that simplified process diagram.

Maybe the process designer isn't available to consult on how to control it. I suspect they took a shortcut by specifying all the dampers and VFDs so it could be adapted to work in the field or with uncertain requirements (so that ductwork and equipment sizing would not have to be rigorously designed/balanced/optimized).

I would consider this system a 'balanced draft' system and you'll likely have to experiment a bit to figure out suitable ways to control it. Balanced draft systems will typically have their inlet/outlet dampers (or inlet/outlet VFDs) linked via some ratio, bias, or feedforward control arrangement.

I’d like to think the VFD solution might give better turndown and less likely to produce unpleasant noise/harmonics under some operating conditions.

A shot in the dark (based on some of my experiences with forced draft boilers with exhaust gas recirculation)….

An optimistically simple system utilizing the VFDs would be:

MD1 and MD4 fixed open

supply air fan speed (S1) determined by PID output to control inlet flow (F1) setpoint (you did say total flow, but I prefer inlet and was hoping you could transform your objective)

return air fan speed (S2) determined by S1 + PID output (bias) to control pressure setpoint (the PID output limits would need to be established, eg [-20, 20]. (You’ll also want to field-specify process-meaningful VFD limits)

MD5 position determined by PID output to control flow ratio (F2/(F1+F2)) setpoint (this ratio will likely behave better than F1/F2 and can be transformed to nominally achieve your stated objective)

MD6 and MD10 (and maybe MD1) position may follow a schedule based on F1 setpoint (imagine effectively resizing ductwork based on rate).

As suggested by Mispeld, you should likely strive to detune some loop’s performance relative to others to achieve overall graceful and robust behavior.

Other considerations that were not discussed are what turndown requirements are, what are expected disturbances, and any expected seasonal variations in behavior.

Good luck and I’d be interested in hearing how you make out.
 
use a fan to control the flow.
use a valve to control the pressure in the chamber, it should be in the inlet obvious.
and one valve in the line for old air, to make the ratio.
the rest is extra and not needed for the process.
have a flow sensor in the inlet and one in the old airline. now you can control the ratio and the flow.
 
saultgeorge said:
You are right in using very slow gain controls in your PID's, I have 6 PID's for each rooftop unit and all totaled 13 for the test cell units.
Modulating the dampers is good, I modulate a minimum outside air damper and the economizer dampers, plus the fan speeds for static pressure in the spaces.
I probably didn't give you too much help, just be careful with the damper action and the fans ramping up too fast. I could send you some flow charts of the units if it will be of interest to you. Good luck, and I've always wanted to see Australia.
Thanks George. All my dampers have analog position feedback, and are spring return to either open or closed as the application demands. So if a damper fails, (a) I'll know about it, and (b) it'll fail into a position that the system can hobble along until it's fixed.

seth350 said:
If you can, can you tell us what this process is actually doing? Just curious.
At first glance, it looked like a quench box. Then saultgeorge mentioned HVAC and that makes more sense.
Yes, it's a HVAC system of sorts, for a specific type of process. There are other process parameters and equipment in the picture - but those a re straightforward so I eliminated them from the diagram for clarity.


Mispeld said:
Three thoughts:

1. It is clear from the description there will be multiple interacting control loops. In order to achieve stable operation that meets process objectives, these loops will need to be decoupled implicitly (i.e., different closed-loop relative response rates) or explicitly through complexity of the control system design (e.g., classical control with decouplers, or model-based multi-input, multi-output). Without this decoupling, it is likely the loops with "fight" each other, with the control system magnifying process disturbances.

The implicit approach is generally more intuitive and less costly to implement. Sometimes it happens naturally, such as between a flow/temperature regulation where flow is almost always much faster to control than temperature. In response to a disturbance, flow will quickly settle to its target (maybe a moving target) while temperature is still working to get back to its set point.

In the above process description you have competing flow and pressure objectives, and these could have very similar relative responses, leading to undesirable control loop interaction.

Your options include: (1) take actions to impact open-loop time constants, or (2) tune the closed-loop responses in priority of objectives, with slowest response being lowest priority. An example of option (1) would be accel/decel settings of the fan VFD's. Another might be set-up of the damper controls. It is probably best to do some experimentation with the system once it is built to see if there are any natural, inherent differences in open-loop response. One way to do this is by bumping each final control element ("CV") individually while holding the others constant, and examining a trend chart of the process measurements ("PVs").
So, if I understand correctly, I should make sure that the most important loop responds the fastest, so that any other loops follow behind once the most important loop is more or less stabilised. Am I following you correctly? Probably in this case, the most important thing is to maintain a negative pressure setpoint in the room. It's not a disaster if it rises to -10 or -5Pa briefly during a disturbance, as long as it stays negative and returns to the setpoint shortly afterward. So, if I give the return air fan a much faster accel/decel rate than the supply air fan, the pressure will stabilise quickly and the flow will slowly ramp up/down behind it. The dampers controlling the outside/recirc air ratio will then need to be tuned lower again, so that they just very slowly make adjustments as the flow settles out. This seems to be a combination of your two options if I understand correctly - the accel/decel discrepancy is an example of option 1, and the deliberate "de-tuning" of the outside/recirc damper PID would be an example of option 2. Am I right?
Mispeld said:
2. The described strategy sounds like a good starting point, though I would consider initially simplifying it as follows: (a) for strategy step #3, do not actively control MD10; start with leaving it at 100% or mostly open, and (b) choose either MD5 or MD6 to achieve the desired recirculation ratio, not both in tandem (one of them stays at a fixed position). I think the MD5 or MD6 choice would depend on whether total flow or recirculation ratio is more important to regulate. My thought here is to let fan speed dominate control with minimum (or constant) pressure drops in the system.
That's a good idea. Total flow is important to the process outcome, whereas the inside/outside air ratio is mostly just a case of efficiency. I'll have a play with a few options and see which gives me the best results.
jamesau said:
I’d like to think the VFD solution might give better turndown and less likely to produce unpleasant noise/harmonics under some operating conditions.
This is probably true. While the point has been made that fans operate best at the "top of their curve", perhaps as long as the fans have been well sized to the application (which I believe they have), I will be able to use fan speed control rather than adjusting dampers.

jamesau said:
An optimistically simple system utilizing the VFDs would be:

MD1 and MD4 fixed open

supply air fan speed (S1) determined by PID output to control inlet flow (F1) setpoint (you did say total flow, but I prefer inlet and was hoping you could transform your objective)
Air flow through the chamber is a reasonably important process condition for the application, so I think I need to stick with total flow in this case.
jamesau said:
return air fan speed (S2) determined by S1 + PID output (bias) to control pressure setpoint (the PID output limits would need to be established, eg [-20, 20]. (You’ll also want to field-specify process-meaningful VFD limits)
I don't think I'm fully catching your meaning here. I'm a bit of a hack when it comes to PID control at the moment - just enough knowledge to be dangerous, as they say - so you may have to dumb it down for me a little :) (or maybe I'm just misunderstanding what you mean and it's actually quite simple?)

Thanks all, please keep throwing around any ideas/advice/thoughts, I'm grateful for all of it. I'll make sure to report on my findings in due course!
 
So, if I understand correctly, I should make sure that the most important loop responds the fastest, so that any other loops follow behind once the most important loop is more or less stabilised. Am I following you correctly? Probably in this case, the most important thing is to maintain a negative pressure setpoint in the room. It's not a disaster if it rises to -10 or -5Pa briefly during a disturbance, as long as it stays negative and returns to the setpoint shortly afterward. So, if I give the return air fan a much faster accel/decel rate than the supply air fan, the pressure will stabilise quickly and the flow will slowly ramp up/down behind it. The dampers controlling the outside/recirc air ratio will then need to be tuned lower again, so that they just very slowly make adjustments as the flow settles out. This seems to be a combination of your two options if I understand correctly - the accel/decel discrepancy is an example of option 1, and the deliberate "de-tuning" of the outside/recirc damper PID would be an example of option 2. Am I right?

This essentially captures what I was trying to describe. The one point I would emphasize is to first understand the open loop responses of proposed loops in the basic design. (This basic design suggested three independent loops for chamber pressure, total flow, and recirculation ratio.) The relative gains and time constants of the individual loops may allow for stable operation without intentionally de-tuning individual loops. It can be difficult and costly to determine these responses and interactions without actually running tests on the equipment.

One example of process impact by equipment design for this application is the extent of air leakage into the chamber. If it is tightly sealed, the pressure response and instrument performance will be much different than if the chamber has openings to allow product or people to move through. With high leakage, the pressure loop may be naturally slow to respond to exhaust fan speed, and it turns out that total flow can be more aggressively tuned even though it is not highest priority.

In this application you have the flexibility (or curse) of more control elements --fans and dampers -- than measurements. It is difficult to specify the optimal arrangement of the loops without knowing the extent of interactions, which can be effectively evaluated by equipment testing during start-up.

My experience base is not heavily in HVAC-type control systems, though my instinct was to start with basically the same arrangement originally proposed. It is advisable to reserve time in the schedule for testing, tuning, and possible reconfiguration of the overall strategy. And/or bring in a consultant you trust that "does these things all the time" if the additional cost is justified.
 
I don't think I'm fully catching your meaning here. I'm a bit of a hack when it comes to PID control at the moment - just enough knowledge to be dangerous, as they say - so you may have to dumb it down for me a little :) (or maybe I'm just misunderstanding what you mean and it's actually quite simple?)

I was suggesting that the same PID output value used to set the speed of the supply air fan be added to (bias) the output of the pressure controller; this sum would then set the speed of the return air fan. (Basically, as more flow is called for, both fans will increase their respective outputs by the same amount, assuming the pressure setpoint is satisfied). This would require more programming overhead.

Regardless of how you go about it, you'll have to experiment a bit to determine how best to control this process.

Good luck and have fun with it.
 

Similar Topics

Hello all, I have a Controllogix 1756-L61 with some RIO. There are a couple of 1734-OB4E's that have gone bad. (no output voltage) My boss found...
Replies
10
Views
1,011
I have 2 Absolute Encoders 8192 steps per Rev. (Model TR Electronic CEV65m-11003) These Encoders communicate to the PLC5-60 via Parallel push/pull...
Replies
3
Views
1,486
Hi, I'm just trying to clear data for any type of String. The following results in only the first character being zeroed: tSINT := 0...
Replies
25
Views
8,101
Hi all, I'm generating AOIs for lots of our shipped components and, for the purpose of eliminating superfluous functionality/bloat, I got stuck...
Replies
6
Views
2,236
Hey guys, I've recently been asked to review some code at a water treatment plant that isn't working right; seems like sometimes it'll work, but...
Replies
7
Views
5,212
Back
Top Bottom