When I sat down in my office this morning, I was greeted by the operations manager, who pointed out an entry from last night’s operations log: “Did you see? The boiler steam vent is in manual because the pressure reading whacked out and made the vent open.” I always appreciate (not) such descriptions of instruments gone awry, like “whacked out,” “went crazy,” “died,” etc.—most too colorful for me to repeat in this dignified, professional journal. It means I have to go to the control house and stare at trends, diagnostics and related measurements to determine what really went wrong.
But after all these years I “get” the perspective of operators. We give them measurements and controls to do their job: operate and optimize a complex process, while looking after their crew’s safety and many millions of dollars in fixed assets, precious metal catalysts and inventory. They don’t really have time to analyze or speculate about the root cause of measurement maladies.
The plant in question has been employing digitally integrated transmitters for more than 16 years. Such transmitters support relentless measurement status updates. If the device or the system detects a lapse in communications or any of the sensors or silicon necessary for a valid, timely and truthful measurement, it flags it instantly to all the users of the data. So I find it disheartening when a fatal device failure isn’t flagged as “bad” before the loop reacts. It’s this sort of thing that can lead to a lot of loops being left in manual.
Closing a loop is a fundamental goal of automation and control. Let an algorithm look after mundane tasks like “relieve the steam header before it lifts a safety valve” and free the operator to look after more critical and creative tasks. Control professionals have been doing this since the days of pneumatics. And today, we can make an algorithm “smart” by employing mode shedding—degrading to manual and holding last output when a smart device flags its data as bad or uncertain. Along with the unparalleled precision of digitally characterized measurement devices communicating over a purely digital medium (e.g. fieldbus or wireless), we have unprecedented tools at our disposal for robust, simple, closed-loop control.
Loops in manual might be a KPI of your controls group, as having too many loops in manual means your controls are unreliable. But many “loops” are always in “manual,” as they rely on a thinking human to take action to mitigate a consequence. This is the definition of an alarm. We know we have dozens if not thousands of them, and many are claimed as layers of protection in the design of our processes. There are metrics for an acceptable alarm load, and our standards say if we exceed 10 in 10 minutes, we exceed “manageable” and risk operator error. What happens when we confront the operator with more information from more “things”?
The purveyors of systems and software to exploit the coming Industrial Internet of Things (IIoT) are aiming to create some loops, or as XMPro’s Steve Wedler says, “Close the loop by orchestrating complete event flows all the way from diverse sensors and data sources to specific business actions.” I like Steve’s concept, but we also must be mindful of the law of unintended consequences, and the age-old caveat of all computing machines: garbage in, garbage out. If the bassoon player in the orchestra is blasting B-flats during my Mozart concerto, no one is going to enjoy the music. Actionable information is essential for the effectiveness of the “loop,” as it is for closed-loop control and open-loop alarms, and we need to silence the whacked-out, derelict players.
When process control professionals had input into standards for digital communications between devices and control systems, they were, thankfully, very mindful of how bad data impacted the robustness of loops. That’s why the standards they created incorporate measurement and signal status. If we’re going to create value from diverse sensors and data sources, process control professionals have to voice and incorporate the same aesthetic; validate every signal and configure “loops” (preferably by default) to utilize the status.
Especially if you’re aiming to use the old stuff—like our plant’s 16-year-old pressure transmitter that “whacked out” in the middle of the night, diligent data validation, perhaps including device upgrade or replacement, may be warranted.