Use Derivative Action Responsibly

Derivative action can give you the fastest response in loop tuning, but only if you use it responsibly. Too much derivative will make the loop unstable. Too little deviative will not benefit response, and could result in, for instance, a jittery control valve, creating greater wear and shortening time to replacement.

By John Gerry, P.E., ExperTune February 1, 2002

KEY WORDS

Process and advanced control

Software and information integration

Loop-tuning software

PID (proportional-integral-derivative)

Process control valves

Sidebars: ONLINE

Derivative action can give you the fastest response in loop tuning, but only if you use it responsibly.

Too much derivative will make the loop unstable. Too little deviative will not benefit response, and could result in, for instance, a jittery control valve, creating greater wear and shortening time to replacement.

Stabilize with derivative

In PID (proportional-integral-derivative) loop tuning, adjustments can be made to one or more variables to create changes.

Using the appropriate amount of derivative (or D) allows using more proportional and especially more integral actions, resulting in a much faster-reacting control loop. Properly applied derivative action creates the most pronounced changes on second-order processes like temperature loops. But D action also can help the response of most loops.

For example, in an easy comparison, two controllers (see screen capture) running a temperature loop have the same proportional and integral terms. As an upset is applied, the PI-only curve swings visibly more widely at least four times, compared to the second, which adds derivative action.

A plot of dead time against process gain for each controller shows a 53% increase in robustness (see second screen capture). Clearly, D action can smooth the response, creating a more robust loop.

Controllers, processes differ

If setpoint response is important to a loop, then the effects of derivative on the setpoint should be examined. Some controllers allow removing derivative action from setpoint changes. In a simulated comparison of Honeywell Plantscape A and B algorithms, each with derivative, algorithm A reacts slightly more quickly with a large initial spike in the controller output. Algorithm B doesn’t react as quickly when it does respond (see third screen capture).

Most processes can be helped by derivative action, except those with almost pure dead time, which should not use D. These are somewhat rare. An old rule of thumb is to not use D on noisy loops.

Process control always has tradeoffs. If the loop is noisy, D action will make the control valve move more, causing more wear on the valve, decreasing its life. This is why using just a little bit of D can harm the loop: it does little to improve performance and wears out the valve. When using D, use the full and proper amount. Additional filtering can help counteract control valve wear, if the filter is the right size.

Other cautions; bad rap

Apply derivative only on controllers that limit the derivative gain. Manufacturers limit derivative gain by applying a first- or second-order filter to the process variable or to the error signal when the user enters a setting for D.

Without such filters, using D action in the presence of any amount of noise would continually the smash controller output into the upper or lower limits. In the lower plot on the fourth screen capture, the blue line shows such an event. The example demonstrates, in part, why D often receives a bad rap. The red plot shows the same controller with the proper D gain limit.

There is another caution with parallel-type controllers. (“Comparison of PID Control Algorithms,” Control Engineering , March ’87, explains PID controller types.) On many of these controllers, the D gain limit changes with the dialed-in value of controller gain.

Dial a gain of 1, and everything works as expected. Dial in progressively smaller controller gains, and the D gain limit slowly vanishes. Put in a large controller gain, and the D gain limit is so close to the actual D that it ends up canceling the D action. With parallel controllers, do not use D unless the controller gain is close to 1.

Filtering helps

A temperature-loop simulation applied to two identical PID controllers shows how additional filtering can make a difference. With a filter, performance is unchanged, robustness hurt slightly, but valve travel and reversals are dramatically reduced. In this case, the filter is probably worth it. Make sure that any filter added isn’t “big” enough to hurt loop performance.

Properly set, derivative action improves the response of most loops. Even with the proper setting, be careful to examine tradeoffs in valve wear. Analysis software can help in comparing options, and provide decision-making tools for loop tuning.

For more suppliers, go to www.controleng.com/buyersguide ; visit www.controleng.com/freeinfo .

Author Information

John Gerry, P.E., is president of ExperTune (Hubertus, Wis.;

ONLINE

Visit

Two other screen captures show that extra filtering can help reduce the control valve jitter associated with derivative, with just a slight reduction in robustness.

“Basics of Proportional-Integral-Derivative Control,” March ’98, p.135, provides a tutorial on PID control.

“How to Control Processes with Large Dead Times,” Control Engineering, March ’98, p.145, which includes more on robustness plots.

ONLINE

Visit

Two other screen captures show that extra filtering can help reduce the control valve jitter associated with derivative, with just a slight reduction in robustness.

“Basics of Proportional-Integral-Derivative Control,” March ’98, p.135, provides a tutorial on PID control.

“How to Control Processes with Large Dead Times,” Control Engineering, March ’98, p.145, which includes more on robustness plots.

ONLINE

Visit

Two other screen captures show that extra filtering can help reduce the control valve jitter associated with derivative, with just a slight reduction in robustness.

“Basics of Proportional-Integral-Derivative Control,” March ’98, p.135, provides a tutorial on PID control.

“How to Control Processes with Large Dead Times,” Control Engineering, March ’98, p.145, which includes more on robustness plots.

ONLINE

Visit

Two other screen captures show that extra filtering can help reduce the control valve jitter associated with derivative, with just a slight reduction in robustness.

“Basics of Proportional-Integral-Derivative Control,” March ’98, p.135, provides a tutorial on PID control.

“How to Control Processes with Large Dead Times,” Control Engineering, March ’98, p.145, which includes more on robustness plots.

ONLINE

Visit

Two other screen captures show that extra filtering can help reduce the control valve jitter associated with derivative, with just a slight reduction in robustness.

“Basics of Proportional-Integral-Derivative Control,” March ’98, p.135, provides a tutorial on PID control.

“How to Control Processes with Large Dead Times,” Control Engineering, March ’98, p.145, which includes more on robustness plots.

ONLINE

Visit

Two other screen captures show that extra filtering can help reduce the control valve jitter associated with derivative, with just a slight reduction in robustness.

“Basics of Proportional-Integral-Derivative Control,” March ’98, p.135, provides a tutorial on PID control.

“How to Control Processes with Large Dead Times,” Control Engineering, March ’98, p.145, which includes more on robustness plots.

ONLINE

Visit

Two other screen captures show that extra filtering can help reduce the control valve jitter associated with derivative, with just a slight reduction in robustness.

“Basics of Proportional-Integral-Derivative Control,” March ’98, p.135, provides a tutorial on PID control.

“How to Control Processes with Large Dead Times,” Control Engineering, March ’98, p.145, which includes more on robustness plots.