It’s not every day that oil and gas IT gets into the New York Times. This felt rather curious for one who has been writing about the subject for over 20 years, almost an infringement of one’s personal space. On April 19th, Steve Lohr covered GE’s ‘pivot’ away from its grandiose plans to be a ‘top 10 software company by 2020’ as the then CEO Jeff Immelt stated back in 2015. Lohr reported that Predix GE has been pared back from being an ‘operating system for the industrial internet’ to a ‘set of software tools to help write applications rather than being connected to layers of code for automating data analysis.’ From platform to app. Quite a come down!
Immelt was both a visionary and cheerleader for the big data/artificial intelligence/analytics movement. GE began its digital push for real in 2011, with the opening of its digital business unit in San Ramon, California. As Immelt loved to explain, the idea was to harden the open source Hadoop-esque software environment that has spun out of the Google/Facebook brigade and develop a data and analytics platform for ‘big’ industrial data. Immelt managed to convince many that using the GAFA’s technology would bring the GAFA’s success. It took me a while to unpick this and I admit to being somewhat wise after the event, but in my 2017 editorial, I described this as the ‘great misunderstanding.’ I highlighted how poorly the business models of the GAFAs mapped across to the industrial domain. This was GE’s first problem.
But there was an even bigger problem with the whole notion of ‘big’ data. When I take a snap with my digital camera, a couple of megabytes eventually find their way into the iCloud (and the Google cloud and, I think, chez Microsoft too). Multiply that by a couple of billion photographers snapping away daily and you do indeed have big data. Most of the peta-exabytes of data that we are told are being produced every day, year or whatever, are of this nature. We could call this ‘capricious’ big data. This is not to be confused with the ‘nasty’ big data that Facebook and the now defunct Cambridge Analytica purvey, but I digress. Immelt’s plan to recast the GAFA business model into industry relied on the ‘fact’ of big data. Back in 2013, GE reported that ‘a modern aircraft collects around a terabyte of information per day.’
At the time, I found this hard to believe, although I have heard it repeated often subsequently. In a short email exchange with GE I learned that (in 2017) ‘GE engines produce data snapshots at various points in flight, such as take-off, climb and cruise. The snapshots include up to 1,000 different measurement parameters, and each engine can pump out between 50 to 200 megabytes of data per flight depending on the flight time.’ So we are at most in gigabyte not terabyte country.
My skepticism in the context of oil and gas big data comes from my 20 plus years of tracking the industry. One outfit that has been doing analytics for a very long time is Intelligent Solutions whose ‘Top down reservoir model’ has been using AI/analytics for a couple of decades. A while back, I asked CEO Shahab Mohaghegh how they managed with the big data issue. He looked surprised. The software ran on a PC and seemingly there were no problems with data overload. When you think about it, this is not all that surprising since reservoir models are made on historical data that predates any data explosion. But does the data explosion even exist?
There are good reasons to think not. Oilfield data (and, I bet, airline data) is extremely un-capricious. Sensors are not ubiquitous, low-cost and ‘Nest like.’ They are esoteric devices that are very expensive to deploy, especially in Atex/explosive environments. They also require expensive maintenance and regular calibration. Their data may be rather hard to get at, as it can be more or less concealed from the outside world in proprietary formats. That this is an issue today is evident from our report in this issue from the ABC Wellsite Automation conference.
I don’t want to give the impression that I am just beating-up on Jeff Immelt and GE. There are many others companies which have promised the moon in this context. The whole world is hooked on the big data/AI meme. But when you hear an oil CEO citing a geophysical processing center as an example of AI/big data, you know that something has gone wrong in the Chinese whispers that emanate from the consultants, through the IT department and into the boardroom.
Another checkpoint in my evaluation of the bigness of big data came from a remark I caught at the 2017 EU Spire conference on the future of plant operations made by BASF’s head of R&D, Achim Stammer who opined that, ‘in the process industry, I would say that we do not have big data.’
Of course, none of the above will stop ‘the media’ (Oil IT Journal included) from reporting on the excitement and hype that the big data/AI movement has brought. In this very issue we bring you an on-the-spot report from the Data science in energy event co-hosted earlier this year by France’s premier R&D establishments Inria and IFPen. There is indeed a lot going on in this space, as many researchers are enthusiastically developing away with the open source, big data tools. Elsewhere in this issue we report on multiple AI-related developments, in natural language processing, in AI-driven multiphase flow metering and on developments in DNV GL’s Veracity, a competing platform to GE’s Predix and on applications of Google’s TensorFlow. We also dutifully report on what I consider to be very improbable developments in the tulipomania that is blockchain. Just remember, as the Romans had it, caveat emptor!
© Oil IT Journal - all rights reserved.