Want Your Process to be Healthy? Don’t Allow Dead-Time Kill it!

4 Practical Ideas for Halting the ‘Killer of Control’

After being away from an Oil and Gas platform for some time a series of control issues prompted my return. Apparently the production team had struggled to regain control of a few key processes. Having worked on those very same systems several months prior I arrived with an idea of what I’d find. 

The newer production operators were routinely impatient in their control of different processes. Unable to get loops back to Set Point quickly they often pushed the Controller Output farther than necessary – whether opening or closing the valve too much or changing Set Points to chase wobbles.

As I arrived in the control room there was one Temperature loop in particular that I found swinging. A trend of the process showed the sinusoidal pattern indicative of instability. Such are the joys of assisting newcomers and dealing with Dead-Time.

Most processes that involve significant lag are difficult to control. In a previous post I shared that this particular platform was located offshore West Africa. Its equipment was lightheartedly referred to as ‘seasoned’ and it lacked the latest instrumentation. Indeed, most of the controls were pneumatic.

The supervisors were also seasoned and knew what was required when it came to meeting production goals. As with most engagements, first-hand knowledge of the platform’s original configuration and day-to-day operational experience were keys to success.

In my last post about PID Tuning I mentioned that Dead-Time is relatively easy to calculate. What I didn’t share is that controlling processes characterised by large Dead-Time isn’t exactly straightforward.

Dead-Time is often referred to as the ‘killer of control’ for good reason. Essentially it’s the delay that occurs between a change to the Controller Output and the response of the corresponding Process Variable.

For unsuspecting production and engineering staff the effects of Dead-Time will sneak up on you. Here are a few ideas worth considering when you’re faced with a large Dead-Time process:

1. Correct for a large Dead-Time value by lessening it

The Dead-Time value is dictated by the time that the Process Variable needs to respond to a given change in Controller Output. In addition to transportation lag the associated delay can be the result of a poorly placed sensor.

So, make sure that the process’ instrumentation is installed in a fashion that will minimise sample lag and speed up signal processing. If Dead-Time is the enemy of good control, then minimise it any way you can.

2. Tune the controller for a conservative response

Fast control is not always good control. Indeed, it rarely is. Consider where the PID sits in the overall process and how other PIDs – both upstream and downstream – might be impacted by the controller’s speed.

Also, keep in mind the controller’s need to adapt to changes. A fast, aggressively tuned controller will not have as broad a range of stability as one that’s tuned conservatively. Think before you tune! 

Tuning a PID controller aggressively can be counter-productive with processes characterized by large Dead-Time relative to the Process Time Constant. In its effort to correct for accumulating Error the Integral Term of an aggressively tuned controller will result in overshoot and can drive the process toward instability. In contrast conservative tuning parameters allow the controller to avoid overshoot while steadily reducing Error.

3. The IMC controller is more patient than most

In a way the IMC controller is more tolerant than traditional controllers as it accounts for Dead-Time. Essentially the Integral term has an initial and minimal effect on the IMC controller’s performance.

As a result it doesn’t instantly increase Controller Output in response to a growing Error value. That ‘patience’ allows the IMC controller to wait for any initial change to the Controller Output to take effect before making other adjustments.

4. Accelerate disturbance rejection with Cascade Control

If reducing lag can’t be achieved by simply moving the sensor closer to the disturbance, then adding a secondary control loop as part of a Cascade Control strategy can speed the process’ response time.

The secondary loop puts eyes on the primary source of disturbances and allows the process to counter more quickly, thereby limiting the disturbance’s impact. While applicable to many processes, Cascade Control has particular merit when applied to those characterised by long Dead-Time.

It’s the nature of most production staff to be impatient and to push for corrections to process upsets. As such, long Dead-Time processes present a challenge. They demand that we minimise the Dead-Time value whether through adjustments in instrumentation or by implementation of alternative control solutions.

First and foremost, awareness of Dead-Time is a key to devising an appropriate control schema and to keeping processes healthy.

If you have experience with controlling large Dead-Time processes and have uncovered other keys to success, please share them with me and the community here on PII.