Key points
– 4 Most Common Data Collection and Controller Tuning Pitfalls
Ask technicians and engineers about manually tuning PIDs. Their responses usually fall into one of three different groups:
1) Some who will say it’s a “dark art” and won’t stray from an OEM’s default parameters
2) The majority who will describe approaches that are based on seemingly random trial-and-error
3) Very few who will have truly figured it out
My
career has thankfully allowed me to rub shoulders with enough
practitioners from that last group and with their guidance I’ve
cultivated a list of tuning tips that have served me well.
The topic of manual loop tuning reminded me of consultancy services I provided to a Food & Beverage supplier located in the UK. It was a chocolate products manufacturer and as with most confectioners this one relied on PID controllers to maintain temperature of the plant’s tempering and enrobing systems.
It’s essential to establish the cause-and-effect relationship between your test and the loop’s response.
Over several months leading up to my contract control across the plant had steadily deteriorated, resulting in a noticeable loss in the “snap” that high quality chocolate should have. Something as seemingly trivial as poor control was causing the plant to waste costly inputs and even discard product that failed to meet production spec.
The plant
manager understood that production could stray beyond designated
tolerances for only so long before the business would suffer. During my
initial discussion with him I learned that the slip in performance
coincided with the retirement of a senior engineer.
He’d been
the one that checked the PIDs every 6-9 months and who singled out those
that needed occasional adjustments. It quickly became obvious to me
that the retiree was one of those few who had truly figured it out. That
assumption was later confirmed by others on staff.
I spent a number of days learning about the processes, investigating root-causes and tuning individual PIDs. Ultimately I found that the plant’s degraded performance resulted from a series of pumps that had been upgraded across the plant.
Their dynamics directly impacted each loop’s ability to regulate control. What surprised the young technician who’d assisted me was how systematic the processes of testing, analysing and tuning PIDs could be – even when manually tuning loops.
Fortunately for practitioners most plants are now equipped with a DCS or other control system that supports the use of tuning software. That’s key to my current endeavours in the process optimisation sphere. Even so, there are four basic tips for collecting good test data that I always share with those who still tune their PIDs manually:
1. Start at a Steady-State
This is a must if tuning loops manually, and most tuning products include this requirement as well. By steadying the process it’s easier to calculate values for Process Gain, Process Time Constant, and Process Dead-Time which can then be converted into PID tuning parameters.
2. Bump Beyond the Noise
It’s important to drive the Process Variable beyond any noise that’s apparent in the process signal. I recommend tests that are 3-5 times the size of the noise band. That should reveal a process’ dynamic behaviour which can then be modelled and controlled.
3. Set a Speedy Sample Rate
The data collection speed matters when tuning PIDs, and faster is always better. While 1-minute data may be suitable for very slow Temperature loops, data sampled at 1-second is needed for fast Temperature, Pressure, and Flow loops. If the data is too slow, it may miss important details.
4. Don’t Use Disturbance Data
It’s essential to establish the cause-and-effect relationship between your test and the loop’s response. That means it’s necessary to keep all other loops “quiet” during testing. If an upstream process influences your test data, then it’d be best to start again and get it right.
As with most things there are best-practices for tuning PID controllers that can make the work less daunting. And while the four tips shared in this post may seem straight forward, few practitioners apply them when collecting test data for analysis, modelling and optimisation.
Previously I offered thoughts on root-cause analysis and optimisation – a Systems Thinker approach which also applies to manual loop tuning. In my next post I’ll share simple, repeatable steps for calculating both process and controller tuning parameters.
