Droning On

From Reuters, via Twitter:

General Electric Co has begun testing autonomous drones and robotic “crawlers” to inspect refineries, factories, railroads and other industrial equipment with an eye on capturing a bigger slice of the $40 billion (31.6 billion pounds) companies around the globe spend annually on inspections.

In trials with customers, aerial drones and robots are able to move around and inside remote or dangerous facilities while photographing corrosion or taking temperature, vibration or gas readings that can be analysed by computer algorithms and artificial intelligence, Alex Tepper, head of business development at Avitas Systems, a startup GE formed for this business, told Reuters.

Hmmmm. I think somebody might be overselling something to impress a journalist here. I have heard of drones being used to scope pipeline routes and to look for leaks, which makes perfect sense. Normally this is done by helicopter, so a drone is simply a cheaper and easier way of doing the same thing. And the insides of pipelines are inspected by a sort of robot called an “intelligent pig”, which detects corrosion among other things. This is an evolving technology, but it has been around a long time. I have also seen remote control helicopters used to inspect flares.

But carrying out inspections of refineries and factories? No such facility is that remote, they are all manned to some degree. Why not just send an inspector? And a dangerous facility? Okay, I get that drones and crawlers could be useful in assessing the damage done to a plant that has just blown itself to smithereens or leaked poisonous gas everywhere, but is this their target customer? One that can’t operate its facilities safely?

This looks to me like a solution in search of a problem. The throwaway line about artificial intelligence points in that direction. Photos from a drone might give an inspection team some useful idea on the condition of something that is hard to reach, as will temperature readings, but they’ll not be analysed using artificial intelligence or even an algorithm. If and when drones are used on refineries and in factories, they’ll not be autonomous.

Share

15 thoughts on “Droning On

  1. The points you make here are similar to my thinking on autonomous shipping- i.e. Tank/Bulk/Container vessel drones.

    It’s a great technology, but do we really want it? The American Bureau of Shipping is starting to suggest, like you, that it’s a solution going begging of a problem.

    There is one twist, though, in that the cost savings are marginal, compared with the cost of cleaning up after an accident.

    The truth is that whilst manning a vessel is a significant chunk of the cost of operating it, crew cost savings are dwarfed by the costs associated with an Exxon Valdez, a Höegh Osaka or similar. You could maybe build a business case to support an autonomous ship in isolation (i.e. savings on crew over 10 years may be greater than the difference between a robo ship and a regular one), but the opportunity cost of doing so when shipping is on it’s arse makes the idea a dumb one. You want one automated bulk carrier or a fleet of six sister ships, bought at a knockdown price from an operator who is really struggling?

    However, with Rolls Royce pushing out super-shiny renders of the “ship of the future” with tedious regularity, the concept isn’t not going to go away. But it still seems like an idea short of an application- a real ‘doing it because we can’ moment.

    I blame the engineers.

  2. My BEng(Hons) project, so many, many years (more than a couple of decades) ago, involved image analysis for high definition sonar for autonomous submersibles for the North Sea.

    Dunno what happened to that after I graduated.

    I blame the engineers.

    I blame the sales weasels.

  3. If you look at successful autonomous enterprises then most of them have a “one to many” application. Meaning that one person can now do the job that previously took many to do. The best example being autonomous mine trucks where one person in central control now performs the job of what used to entail say 20 employees. I think a “one to many” sniff test is quite indicative of the merits of a proposed scheme.

    Sightly off topic but relevant to new technologically superior innovations and refineries there is a famous product failure in the US. This company invented a new system that was 100% effective in detecting hydrocarbon leakage in soils adjacent to hydrocarbon tanks, the technology was sound, the application was very effective and it had regulator and EPA approval, this was tipped to be a billion dollar market, why wouldn’t it be with this type of perfect solution although it failed drastically. None of their prospective clients wanted a system that may result in their actual pollution levels being identified.

  4. A lot of autonomous drone/self-driving car stuff is just about PR and boosting stock prices. Idiot investors who think it all means something.

    I’m not working in AI, but I work in software, and AI is of interest, and it’s all hugely overcooked. There’s lots of stuff based around Bayesian logic (like spam filters, message routing systems, detecting credit card theft) that use prior data to predict how to process new data, but it’s all “roughly right”. It’s a very useful thing for guiding users.

    Credit card fraud prevention is a good example. The software produces lots of false positives, it misses lots of fraud too. But it gets it right enough that passing what is still a small number of cases to humans to call card owners saves money. The “misses lots of fraud” is significant. You can only apply AI to things where the odd failure doesn’t matter.

  5. “A lot of autonomous drone/self-driving car stuff is just about PR and boosting stock prices. Idiot investors who think it all means something.”

    Hear! Hear!

  6. I’m old enough to remember when the UK switched from coal gas to natural gas. That meant connecting the entire UK to a national gas pipeline, rather than each town/city having its independent local grid (coal gas aka town gas).

    The main pipeline ran through our school playing field from somewhere near Coventry to somewhere between Leamington and Warwick. Every week (same bat time, same bat channel), a helicopter with (I guess a gas sniffer) used to fly that route checking for leaks. This was about 1968.

  7. I am reminded of the chap who came to me to report a suspected leak; this was years ago on a petrochemical plant. You’ll have to imagine his Irish accent.

    “Sor, there’s a smell.”

    “What does it smell like?”

    “Fucking chemicals.”

  8. Drones need pilots and are still quite expensive. From the article it appears that GE is going for the safety market. Given that safety is to protect humans, it’s probably still more effective to have the humans carry kit, H2S detectors, dosimeters, etc.
    We’ll have unmanned petrochemical plants and oil rigs before we need drones.

  9. S E
    Autonomous submersibles?
    Never heard of them. Did you mean ROVs? (Remotely operated vehicles)

  10. @ James- there’s a difference ‘twixt ROV’s and AUV’s

    My employer does both, through their defense bit. AUV’s are fire and forget: they cruise around doing what they do with no intervention from surface folks. ROV’s are more like drones- piloted, but from afar.

  11. “In trials with customers, aerial drones and robots are able to move around and inside remote or dangerous facilities while photographing corrosion or taking temperature, vibration or gas readings that can be analysed by computer algorithms and artificial intelligence”

    This sounds like it will be as successful as those computer dating sites that promise to match you with your perfect companion.

  12. There is one twist, though, in that the cost savings are marginal, compared with the cost of cleaning up after an accident.

    Yeah, it doesn’t seem worth the risk.

  13. A lot of autonomous drone/self-driving car stuff is just about PR and boosting stock prices. Idiot investors who think it all means something.

    Indeed, that’s the impression I got.

  14. Given that safety is to protect humans, it’s probably still more effective to have the humans carry kit, H2S detectors, dosimeters, etc.

    Exactly.

    We’ll have unmanned petrochemical plants and oil rigs before we need drones.

    We already have unmanned oil platforms. As soon as the equipment gets slightly complicated (i.e. large rotating stuff like compressors, generators, etc.) you need it manned.

  15. This sounds like it will be as successful as those computer dating sites that promise to match you with your perfect companion.

    Heh!

Leave a Reply

Your email address will not be published. Required fields are marked *