Key highlights
- The article shows how modern AI tools like Grok can instantly and accurately diagnose process issues, such as magnetic flowmeter errors or steam turbine oscillations.
- Drawing on historical disasters like Three Mile Island and Chernobyl, it makes the case that smarter automation and better human-machine interfaces, now powered by AI, can help prevent operator errors and improve overall plant safety.
If you recall a time with only three broadcast TV channels and a rotary dial, which required you to get out of your recliner to select one of the other two networks, the phrase “Danger, Will Robinson!” might be familiar. Many of us who grew up in the 1960s may have imagined a future like the one depicted in the then-contemporaneous, broadcast television series Lost in Space. Surely by Y2K or soon thereafter, intrepid scientists would be loading up their families on a craft less reliable than the family station wagon.
After crash-landing on uncharted, and often hostile, planets, Robinson is frequently accompanied and advised by “Robot,” a robotic friend/babysitter. His pal is dependable and useful until it’s hacked by the nefarious Dr. Smith, whose bad intentions somehow remained invisible to the other characters.
A few years after the TV show was cancelled, the Three Mile Island nuclear accident happened. Its modest impact was exacerbated by the film, The China Syndrome, which was followed a few years later by the 1986 nuclear disaster at Chernobyl. Both accidents were attributed, at least in part, to human operators confounded by their human-machine interfaces (HMI), misinterpreting them, and disabling automation systems that may have prevented those accidents. Sometime in the interim, the infamous “man and dog” automation joke became widespread. It describes a control room with two employees: a man to feed the dog, and the dog to bite the man if he tries to touch anything.
Present day artificial intelligence (AI) capabilities and anticipated developments might compel some to think we’ll have a robot to be our friend and babysitter or replace us entirely. After downloadingGrok 3 Beta to my phone, I decided to try it.
I asked Grok if it knew what a magnetic flowmeter is. Without delay, it listed an accurate and comprehensive answer, even explaining its principle of operation (based on Faraday’s Law of electromagnetic induction), key features (suitable for conductive fluids, high turndown, etc), and limitations. It was presented instantly.
I gave Grok a troubleshooting question—flow indication was incorrect/low. What could be wrong? Grok suggested plausible causes and remedies—again instantly. It answered with 10 probable causes, as well as what to check for and next steps. Among its suggestions were improper installation, fluid conductivity issues, entrainment issues, fouling or coating, electromagnetic interference (EMI) and calibration.
Unlike Robinson’s Robot, Grok sounded more like the computer on Star Trek. I tried some process questions. Why is my steam turbine speed oscillating? Again instantly, the AI engine produced four categories of potential issues (control system issues, mechanical issues, process conditions and system interactions). Each category had two or three more specific possibilities, which included what I already knew to be the problem (deadband and hysteresis).
Get your subscription to Control's tri-weekly newsletter.
In each case, Grok offers more in-depth analysis, next steps to troubleshoot the issue, and asked if I had more details to share. I typed specifics on the turbine inlet valve, actuator manufacturer and oscillation frequency. I received five pages of advice, such as eliminating high-frequency mechanical vibrations (bearing issues) and focusing on the control system.
What isn’t obvious is how many Three-Mile Island power plants had to fire up to produce these results, or how many concurrent end users create cartoons and manipulate images for their amusement. Will performance diminish when they become legion? It isn’t a huge stretch to imagine a mobile platform—perhaps one of the anthropomorphic robots produced by the same company—becoming certified for hazardous atmospheres, and following operators around the plant on rounds with sophisticated sensors that hear frequencies amidst the cacophony for which we wear hearing protection, and “eyes” that see UV and IR. They could work hours in extreme cold or heat. They could prioritize safety issues, warning their human companions of dangerous noise, heat, voltage or a restricted tube in a process heater that could burst into flame.
As Robot would say, corrugated arms flailing, “Warning! Warning!” The hurdles of persuading our human operators to accept such assistants will be the biggest challenge.
