Bentley’s Year in Infrastructure dives deep into digital twins
Depending on whom you’re speaking with, digital twins are either unfathomably complex and therefore out of reach, self-serving vendor-babble, or a really useful way of thinking about assets and technology. I’m hosting a series of panels on digital plants at Bentley Systems’ Year in Infrastructure this October and November and, honestly, it’s all about the latter — but if you’re stuck in the first two buckets, let’s talk.
First, what do I mean when I say “digital twin”? Easy: a digital version of a physical thing. The devil is, as they say, in the details: if the only digital thing you have is a PDF, that’s not really a twin because the PDF represents so little of the physical thing. A CAD model or one made via reality capture, like LIDAR or photogrammetry? Now we’re getting somewhere since we can use that for fly-throughs, mockups, measurements, and more. A 3D model plus a PDF of the operator instructions for a piece of equipment? Even better. We can associate the 3D representation of a pump, say, with the PDF, and have at our fingertips information we might need as we do our walkthrough. Then, a holy grail for many, if we can add actual operating data to the model, we can interrogate this intelligent model to see how the pump is performing and what an operator needs to do to keep our system in balance.
Do you need to get to that real-time place to see the benefits of a digital twin? No. Connecting various sources of static data (CAD, PDFs, maintenance schedules, spares lists, and so on) creates the potential for all sorts of what-if analyses.
Can a digital twin, even if not connected to live data, be complicated to create and maintain? Absolutely. You need to start with the latest CAD model and PDF, and then update both as physical reality changes — or the twin is no longer an accurate representation. Once lost, trust in that twin will be very hard to recapture so thinking through all that’s required for these updates is on everyone’s list of best practices when it comes to relying on digital twins in operations.
Does there need to be one mega-twin with every possible piece of information pertaining to the asset? Definitely not – that’s unwieldy, especially since you need to keep it all up-to-date. Focus on a twin that includes just the info that’s needed to answer the particular question you’re asking. In the case of this pump, if we’re concerned about operations, perhaps the installation instructions aren’t relevant – don’t add them into the twin. But perhaps DO add the PFD, simulation results, or other logic that was used to define why that pump was selected in the first place, in case it needs to be replaced. Also, consider having function-specific twins: an operating twin, for example, that’s used to monitor and tune daily production. A maintenance twin might help plan how the line could be shut down and restarted. And maybe a financial twin that’s used in fix-now/delay-repair cost/benefit trade-offs. These are all slices of data pertaining to the same physical object; if it’s easier to have the twin be one central data repository, then create different specific access points for different functions or user types.
Which, of course, brings us to the technology:: is this just vendor-babble? No. Vendors, of course, have their own spin on every aspect of this digital twin story, but it is a technology concept whose time has come — it’s the convergence of cloud storage and easy (secure, controlled) access to lots of different data types, served with front-ends that make it easy to consume. It’s the realization that we’re vastly under-utilizing the asset data we’ve been collecting for years; now we have the processing speed to actually make sense of it and use it for real-time decision making.
The advent of digital twins is also, IMHO, a reflection of the fact that business has never been tougher. Companies can’t afford to be complacent; they need to eke out every possible penny of profit, cut every cost, work towards greater sustainability — and can’t do that without data and without the scenario-testing that digital twins can enable. If you can take a twin offline and game out how to make massive changes, tuning that model until it’s optimized the way you need it to be –without ever affecting current operations– you can gain a significant advantage over competitors.
it’s that convergence of business case and technology that leads to the popularity of digital twins.
Are digital twins complex? Only as complicated as you want them to be. I’d suggest starting simple, with a specific business problem, and collect the data you need to answer just those questions. Don’t get overly ambitious – it’s better to start small, show success, and then grow.
Why would you want a digital twin? I’m hosting panels about digital plants –factories– so am biased, but there are many potential uses. In planning, twins can help make cost tradeoffs, simulate production, guide design and create a means for collaboration between different functions, many of whom may be more able to articulate what they need or want when they see it “in real life”, digitally. In construction or assembly, a plant that has millions of individual pieces becomes more accessible when it can be laid out and built up in stages, and again, stakeholders can more readily communicate about the process. Eventually, there’s a handover from construction to operations; that can go more smoothly if digital twins were used for virtual test and commissioning, ironing out issues well before the plant actually spins up. In operations, a digital twin can help with operator training, defining safety protocols, do what-if trade-offs, and a lot more. And when it comes time to decommission the plant, its deconstruction can be planned and safely accomplished by modeling it all first with the digital twin.
Everywhere here that I’ve written “plant”, think bridge, building, city — the principles are the same.
It all starts with a business need: serving the citizens of a city by better traffic planning. Making a building more economical by optimizing heating. Putting out that tiny bit more product without adding production lines. Digital twins improve visibility into the asset and enable analyses we couldn’t easily do before.
A hint: tune into the YII panels to hear what real people are doing with digital twins, across industries and disciplines.
Is there vendor-babble? Oh goodness yes. But remember that you’re in charge of the data, connections, implementations, and uses of the digital twin. There is hype, for sure, but I’d focus on three things: Making sure you can connect what you have to what you think you need –in other words, don’t convert your data; make the technology do the work. Make sure that you trust the vendor’s long-term vision –you don’t want to recreate this twin because of some sort of technology obsolescence. Yes, things change, but look at the very long view. Make sure you trust that this technology will be around for years. And, maybe most important, make sure that this technology, whatever it is, supports your business case — you’re not creating this digital twin because you have nothing better to do. You need to improve uptime, make more effective asset decisions, train people – whatever it is, make sure the technology supports that use.
We cover all of this and so much more in the three Digital Plant sessions — please join us! You do need to register but the event is free. Bentley CEO Greg Bentley and Microsoft’s CEO, Satya Nadella, kick off the Year in Infrastructure on October 20, 2020, at 11 AM Eastern US, and the Digital Plant sessions start right afterward, on October 20 at 1 PM Eastern. There are lots of other sessions spread across October and early November; check here for the agenda.
And a quick plug: I was a juror for the Reality Modeling category for this year’s Year in Infrastructure Awards, and they were again an impressive bunch. You can see all of the finalist presentations here and tune in for the Awards Ceremony on October 21 at 12:30 Eastern. there may be singing …