Are your control loops more restless than they need to be?
Most PID controllers are tuned once and then left alone for years, while the process does change. The twelve signs below are each small, but they often point to loops that operators keep correcting, that keep raising alarms or that come closer to the trip limit than necessary. Check them for your plant. One minute, and your result right away.
Twelve signs
0 of 12 answered
- 01 / 12
Are there control loops that operators regularly switch to manual, or where they nudge the setpoint to keep the process under control?
- 02 / 12
Are there control loops that keep giving the same alarm, even when nothing is really wrong?
- 03 / 12
During disturbances, do process values come close to a trip limit, or has a control loop ever contributed to a trip or near-trip?
- 04 / 12
Do control loops regularly show overshoot or recurring oscillations around the SP?
- 05 / 12
After a disturbance or setpoint change, does it often take longer than you would expect before the PV is stable around the SP again?
- 06 / 12
Does the control behaviour clearly change with a different load, production capacity, recipe or operating conditions?
- 07 / 12
Are there control loops whose output keeps moving up and down, while the SP and process load are fairly constant?
- 08 / 12
Are there control loops that make many or large corrections, while that has hardly any visible effect on the PV?
- 09 / 12
Do outputs regularly run into their minimum or maximum, so the controller temporarily has no control room left?
- 10 / 12
Are there control loops where the PV stays above or below the SP for a long time, without the controller removing that offset?
- 11 / 12
Are there control loops that react strongly to small fluctuations or measurement noise in the PV, so the output keeps correcting unnecessarily?
- 12 / 12
Have the P, I and any D settings of your control loops mainly grown over time or been adjusted by feel, rather than substantiated from process behaviour or trend data?
- About your plant · 1 / 3
What is your role?
You may skip this. It helps us put your answers in context.
- About your plant · 2 / 3
Roughly how many control loops does your plant have?
You may skip this. It helps us put your answers in context.
- About your plant · 3 / 3
When were the control loops last reviewed systematically?
You may skip this. It helps us put your answers in context.
0 of 12 signs recognised
No signs.
Your control loops seem to do what they should. If you have doubts about one loop in particular, check the signs for that loop.
One or two signs.
That happens in almost every plant and is usually no cause for concern. Keep a particular eye on overshoot and outputs that run into their limit: that is where trips start.
Enough signs to take a closer look.
This kind of unrest rarely stands out: no fault goes off, but operators correct and the process runs with extra margin. Because nobody tracks performance per control loop, it never makes it onto a list. Whether it is down to the tuning or to a valve or measurement, you can largely check yourself.
Strong indications that your control loops run more restlessly than necessary.
This many signs cost your operators attention every day, make the process run with a safety margin and bring it closer to the trip limit than you would like. The cause is not always the tuning: sometimes it is a valve or a measurement. Which of the three, you can largely check yourself.
For {aantal} of the {totaal} questions you did not know. That is a sign in itself: without insight into your control loops, nobody knows whether they run well.
What your signs mean
- Close to the trip limit
A restless loop leaves little margin. When a disturbance hits, the trip limit is reached sooner, and then one unfortunate coincidence is enough for an unplanned shutdown.
How to check
Look up the largest disturbance of the past few months and see how far the PV overshot relative to the trip limit. If it gets close while the disturbance itself was not exceptional, the loop damps too slowly or overshoots. Also look at the last trips: if one of the loops was already oscillating just before, the control played a part.
This check is in your breakdown. Request your breakdown
- Overshoot or oscillation
The loop periodically heads towards the alarm or trip value. This is the silent cause behind many unplanned trips that nobody recognises as a control problem.
How to check
Measure the time between two peaks of the oscillation in the trend. A regular swing with a fixed period usually points to the tuning. If the loop oscillates in the same rhythm as another loop or a pump, they are keeping each other in motion and must be tuned together, not each on its own.
This check is in your breakdown. Request your breakdown
- Output at minimum or maximum
No control room means no grip on the process. This is where trips start, and it often points to a valve or range that no longer suits the current load.
How to check
Count how often and how long the output sat at 0 or 100% in the past week, and at which load. If it happens mainly at high or low load, the valve or the range no longer suits the operating point. If the PV overshoots as soon as the output leaves its limit, the I action has kept winding up against that limit (windup).
This check is in your breakdown. Request your breakdown
- Recurring alarms
A hunting process value keeps touching an alarm limit. After a while nobody pays attention any more, and then the alarm that does matter no longer stands out.
How to check
Count in the alarm log how often this alarm went off in the past month and how long it stayed active each time. Many short alarms that clear by themselves point to a process value swinging around the limit: then the problem is usually in the control, not in the limit. Put the alarm times next to the trend of PV and output to see whether they coincide with the oscillation.
This check is in your breakdown. Request your breakdown
- Manual intervention
Operators do not trust the loop. In practice that means correcting all day long and a way of running the process that differs per operator.
How to check
Look up in the history how often the loop went to manual or the setpoint was moved in the past month, and in which situation: during a load change, at start-up or with a particular shift. If it always happens in the same situation, you know where the control falls short. Also ask the operators what they do with the output then: that is often exactly what the controller itself should be doing.
This check is in your breakdown. Request your breakdown
- Slow recovery after a disturbance
After every disturbance the process is out of spec for a long time and operators have to correct. That is why it is often run with extra margin.
How to check
Find a recent disturbance or setpoint change in the trend and measure how long it takes until the PV stays within a narrow band around the SP again. Compare that with the time the process itself needs to respond to a valve movement. If recovery takes many times longer, the controller is leaving time on the table. How much faster it can go without overshooting follows from the process dynamics.
This check is in your breakdown. Request your breakdown
- Persistent offset from the SP
The controller does not remove the error. Usually too weak an I action, sometimes a valve at its limit or a measurement that is wrong.
How to check
Look at what the output does while the offset persists. If the output slowly keeps moving, the I action is too weak. If the output sits at 0 or 100%, the controller has no room left and the cause is the valve or the range. If the output does not move and the PV looks remarkably flat, check the measurement first.
This check is in your breakdown. Request your breakdown
- Behaviour depends on load
One setting for all conditions only works at one operating point. Outside that point operators intervene, or the process runs with extra margin.
How to check
Compare two periods in the trend: one at low and one at high load. If the loop oscillates at one operating point and is sluggish at the other, the process itself changes with the load. Then one set of settings never fits everywhere, and tuning is needed that takes the whole range into account.
This check is in your breakdown. Request your breakdown
- Output moves continuously
Unrest that wears the valve and never lets the process really settle. Often too sharp a P or I action, sometimes a valve that no longer follows cleanly.
How to check
Put PV and output side by side in the trend for an hour, in a period with a fixed setpoint and a constant load. If the output moves along with the PV in a fixed rhythm, the problem is usually in the tuning. If the output moves in steps or a sawtooth while the PV sticks and then suddenly jumps, look at the valve first.
This check is in your breakdown. Request your breakdown
- Much correction, little effect
Suspicious for the valve: stiction or oversizing. Then part of the problem is not in the tuning, but in the valve itself.
How to check
Test whether the valve sticks, in consultation with the operator and with a calm process: put the controller in manual and give the output a few small steps of 0.5 to 1% in the same direction. If the PV only responds after two or three steps, and then suddenly, the valve is sticking (stiction). The question then is how much of the unrest comes from the valve and how much from the tuning; that can be extracted from the trend data.
This check is in your breakdown. Request your breakdown
- Reacts to measurement noise
Unnecessary valve traffic and wear. Usually a missing filter or a D action that does not suit the signal.
How to check
Zoom in on a few minutes of trend at the highest resolution you have. If the PV jumps up and down in small, fast steps and the output follows those steps, the controller is amplifying noise. How much damping the loop tolerates without becoming slower on real disturbances follows from the process dynamics.
This check is in your breakdown. Request your breakdown
- Settings grown over time
The knowledge is gone. Nobody can explain why the controller is set the way it is, so nobody dares to touch it.
How to check
Look up when the settings were last changed and why, for example in the settings list, the DCS audit log or the change register. Write down the current Kp, Ti and Td. If there is no justification anywhere, measuring the process dynamics (dead time, time constant and gain) is the first step: that yields settings you can explain.
This check is in your breakdown. Request your breakdown
Not yet possible to assess
- Close to the trip limit
A restless loop leaves little margin. When a disturbance hits, the trip limit is reached sooner, and then one unfortunate coincidence is enough for an unplanned shutdown.
How to check
Look up the largest disturbance of the past few months and see how far the PV overshot relative to the trip limit. If it gets close while the disturbance itself was not exceptional, the loop damps too slowly or overshoots. Also look at the last trips: if one of the loops was already oscillating just before, the control played a part.
This check is in your breakdown. Request your breakdown
- Overshoot or oscillation
The loop periodically heads towards the alarm or trip value. This is the silent cause behind many unplanned trips that nobody recognises as a control problem.
How to check
Measure the time between two peaks of the oscillation in the trend. A regular swing with a fixed period usually points to the tuning. If the loop oscillates in the same rhythm as another loop or a pump, they are keeping each other in motion and must be tuned together, not each on its own.
This check is in your breakdown. Request your breakdown
- Output at minimum or maximum
No control room means no grip on the process. This is where trips start, and it often points to a valve or range that no longer suits the current load.
How to check
Count how often and how long the output sat at 0 or 100% in the past week, and at which load. If it happens mainly at high or low load, the valve or the range no longer suits the operating point. If the PV overshoots as soon as the output leaves its limit, the I action has kept winding up against that limit (windup).
This check is in your breakdown. Request your breakdown
- Recurring alarms
A hunting process value keeps touching an alarm limit. After a while nobody pays attention any more, and then the alarm that does matter no longer stands out.
How to check
Count in the alarm log how often this alarm went off in the past month and how long it stayed active each time. Many short alarms that clear by themselves point to a process value swinging around the limit: then the problem is usually in the control, not in the limit. Put the alarm times next to the trend of PV and output to see whether they coincide with the oscillation.
This check is in your breakdown. Request your breakdown
- Manual intervention
Operators do not trust the loop. In practice that means correcting all day long and a way of running the process that differs per operator.
How to check
Look up in the history how often the loop went to manual or the setpoint was moved in the past month, and in which situation: during a load change, at start-up or with a particular shift. If it always happens in the same situation, you know where the control falls short. Also ask the operators what they do with the output then: that is often exactly what the controller itself should be doing.
This check is in your breakdown. Request your breakdown
- Slow recovery after a disturbance
After every disturbance the process is out of spec for a long time and operators have to correct. That is why it is often run with extra margin.
How to check
Find a recent disturbance or setpoint change in the trend and measure how long it takes until the PV stays within a narrow band around the SP again. Compare that with the time the process itself needs to respond to a valve movement. If recovery takes many times longer, the controller is leaving time on the table. How much faster it can go without overshooting follows from the process dynamics.
This check is in your breakdown. Request your breakdown
- Persistent offset from the SP
The controller does not remove the error. Usually too weak an I action, sometimes a valve at its limit or a measurement that is wrong.
How to check
Look at what the output does while the offset persists. If the output slowly keeps moving, the I action is too weak. If the output sits at 0 or 100%, the controller has no room left and the cause is the valve or the range. If the output does not move and the PV looks remarkably flat, check the measurement first.
This check is in your breakdown. Request your breakdown
- Behaviour depends on load
One setting for all conditions only works at one operating point. Outside that point operators intervene, or the process runs with extra margin.
How to check
Compare two periods in the trend: one at low and one at high load. If the loop oscillates at one operating point and is sluggish at the other, the process itself changes with the load. Then one set of settings never fits everywhere, and tuning is needed that takes the whole range into account.
This check is in your breakdown. Request your breakdown
- Output moves continuously
Unrest that wears the valve and never lets the process really settle. Often too sharp a P or I action, sometimes a valve that no longer follows cleanly.
How to check
Put PV and output side by side in the trend for an hour, in a period with a fixed setpoint and a constant load. If the output moves along with the PV in a fixed rhythm, the problem is usually in the tuning. If the output moves in steps or a sawtooth while the PV sticks and then suddenly jumps, look at the valve first.
This check is in your breakdown. Request your breakdown
- Much correction, little effect
Suspicious for the valve: stiction or oversizing. Then part of the problem is not in the tuning, but in the valve itself.
How to check
Test whether the valve sticks, in consultation with the operator and with a calm process: put the controller in manual and give the output a few small steps of 0.5 to 1% in the same direction. If the PV only responds after two or three steps, and then suddenly, the valve is sticking (stiction). The question then is how much of the unrest comes from the valve and how much from the tuning; that can be extracted from the trend data.
This check is in your breakdown. Request your breakdown
- Reacts to measurement noise
Unnecessary valve traffic and wear. Usually a missing filter or a D action that does not suit the signal.
How to check
Zoom in on a few minutes of trend at the highest resolution you have. If the PV jumps up and down in small, fast steps and the output follows those steps, the controller is amplifying noise. How much damping the loop tolerates without becoming slower on real disturbances follows from the process dynamics.
This check is in your breakdown. Request your breakdown
- Settings grown over time
The knowledge is gone. Nobody can explain why the controller is set the way it is, so nobody dares to touch it.
How to check
Look up when the settings were last changed and why, for example in the settings list, the DCS audit log or the change register. Write down the current Kp, Ti and Td. If there is no justification anywhere, measuring the process dynamics (dead time, time constant and gain) is the first step: that yields settings you can explain.
This check is in your breakdown. Request your breakdown
Prepared by Brent Zwart, Process Control Engineer.
Your breakdown is on its way.
Check your inbox, and your spam folder to be sure. Our personal response to your answers will follow within one to two working days.
Would you rather talk straight away?
Tell us which control loop is running restlessly. In a first conversation we often already know whether there is something to gain.