One of the earliest use cases for AI was in operations – specifically in maintenance. This is not my field and I admit to having been surprised by the notion that there were significant benefits to come from ‘optimizing’ maintenance. As I see it there are two kinds of potential benefits here. An ‘optimal’ maintenance program ought to result in less breakdowns, less production loss and consequently more revenue. That’s good but it does not necessarily mean spending less. This ‘optimal’ may cost more. The other potential benefit involves using super smart analytics to achieve the same (or perhaps better) results with a reduced spend on maintenance. The first kind of benefits are like apple pie. The second kind sounds a bit dodgy.
Imagine if an operator is using AI to achieve the same with less spend. And imagine that things go wrong. There is a major accident. An offshore field blows up. There are deaths and a trial. The maintenance program is called into question. Figures are produced showing a year on year reduction in inspection and maintenance which meant that some bad kit was not spotted in time. The judge asks, ‘How did you establish this reduced maintenance program?’ The operator is forced to admit that the AI system came up with the new program. ‘And how did it achieve such a result?’ ‘Well we can’t actually answer that question because the reasoning is kind of hidden in the neural net – we don’t actually understand how AI comes to its conclusions’. Maybe in a couple of years time defendants will be answering the same question with ‘Because ChatGPT told us it would be OK’, but I digress.
You may think that this is preposterous. I think it is preposterous myself. But that was before I listened to one of the talks given at the 2023 CERA week on ‘Transforming workflows and operations with AI and data’. This was presented by SLB’s (formerly Schlumberger) Sujit Kumar who explained how ‘Digital transformation is […] helping companies do more and achieve better performance, with less. Using AI, data analytics, and machine-learning techniques, it is now possible to realize tangible cost and time savings across data-intensive processes, reducing tasks that took months and years into days and hours’.
The first parts of Kumar’s talk correspond to the apple pie style of benefits, one applied AI to prolific data streams to optimize well performance, the other was a ‘smart gas plant’ developed for DCP midstream with AI boutique Geminus. But it was Kumar’s concluding remarks that made my mouth drop. In his summary he stated, ‘Cost reduction was substantial in both examples. Once you have activities like a smart maintenance program [ in place ], you are not doing maintenance as often as before. You are not doing inspection as often as before. You are reducing a huge amount of human footprint and machinery footprint’. Wow! I would not like to be a defendant with a witness like that! Imagine the CSB video interview of the inspectors who were ‘let go’.
~
On an unrelated topic, while, as one does, futzing around on my smartphone on the subject of Steven Spielberg’s excellent autobiographical film The Fabelmans, I read up a bit on his dad, Arnold Spielberg . Arnold was quite a star in the IT world, migrating from the field of electronical engineering and setting up GE’s Industrial Computer Department in 1957, before working on what might be termed the digital transformation of point of sale systems. Why am I telling you this? Well, next time you hear a presenter dissing the old folks who ‘don’t get digital’, think of Arnold. Next time you hear a speaker trotting out the old trope that digital transformation is something shiny and new, think of GE in 1957. Arnold died in 2020, aged 103. More of his interesting life on Wikipedia.
© Oil IT Journal - all rights reserved.