By Dr Jim Scotson, Digitalisation Lead
Digital twins have polarised the industry for years. Many have greeted them with huge enthusiasm as one of the most exciting tools of modern manufacturing. While for others, they’re an expensive way of making a factory look good on a large screen.
The idea of creating a digital twin promises a new, safe way to test ideas, train people, understand problems and make better decisions. But while many manufacturers like the idea and their impressive nature, many are left asking ‘but what can it actually do for me?’ Both interpretations are probably fair, but it really comes down to what you want to do with your digital twin that makes all the difference.
This thought struck me the other day when a manufacturing partner tactfully asked me the ‘what percentage of this is gimmick?’ and this is where I think the conversation about digital twins has gone wrong.
As long as it looks good
For a long time, we just focused on making twins looks more realistic. Better graphics, rendering, faster processors and 3D environments which all made it possible to have some genuinely spectacular virtual industrial environments. Platforms such as NVIDIA Omniverse, the Unreal Engine and Siemens Digital Twin Composer allow us to build visually rich, spatially accurate and physically realistic spaces capable of supporting sophisticated cumulation.
This is fantastic but this isn’t the thing that makes a digital twin useful. A beautifully rendered factory that doesn’t know what the real factory is doing is still, fundamentally, an ornament and an expensive one at that. Recreating every pipe, valve and cable tray complete with mysterious patch of oil make it look like a factory. But, if it doesn’t know that the batch 2 charger has slowed down or that the temperature is drifting it’s essentially just a waxwork offering no operational insight.
Connected technologies change everything
In my view, the biggest change in digital twins over the last few years, has been the move beyond what is looks like into the real beating heart of the twin. In other words, the technologies required to connect physical assets to digital environments. Sensors are cheaper, industrial networking is better, Internet of Things devices are everywhere, and cloud and database platforms have become easier to deploy. APIs make it easier for systems to talk to each other, computer vision allows us to turn a live camera feeds into usable information and industrial data that was once hidden inside isolated control systems can be collected, structured and used elsewhere. This changes everything.
A useful digital twin needs to understand what the physical plant is doing in real time. Not when somebody manually exports a CSV. That live connection takes a digital model from representing what a factory looks like, to representing what it’s actually doing. And that’s when things start to get interesting.
Add historical data so the twin can tell you what the plant has been doing, add analytics so it can begin identifying patterns and if you want to add AI into the mix, you can start to predict what might happen next. You can dip into optimisation, operating scenarios and into the world of predictive control all thanks to your digital twin. That’s when platforms like NVIDIA Omniverse become extremely powerful as they move from displaying an isolated virtual environment into an interactive digital representation connected to the real process underneath it.
Is the factory ready for a sibling?
Sometimes the issue with a digital twin wasn’t that it was useless, but that the factory wasn’t ready for a new digital sibling. You can’t build a live representation of a plant if you can’t reliably access plant data. If all you have is inconsistent, poorly structured information that’s spread across half a dozen systems that don’t speak to each other, you can’t expect a digital twin to work miracles.
This is why, at Glass Futures (GF), we’ve spent so much time discussing, not ‘do we need a digital twin?’ but ‘are we ready for one?’. Through PlantConnect, we’re helping manufacturers understand their level of digital readiness, how ready they are to actually accept digital technologies, before throwing more tech at the problems. Part of that is work built around what we describe as Technology Acceptance Level. The principle is simple. Before asking what shiny, new digital tool you should by, first consider and understand what your factory is currently capable of supporting. Ask yourself these questions:
- Do you have the right sensors and can you access their data?
- Is that data stored somewhere useful?
- Can different parts of the plant communicate?
- Do you have a historian?
- Can systems access information through APIs?
- Can cameras or other non-traditional data sources contribute information?
Perhaps most importantly, and the one we find comes up all the time is whether anybody is actually using the data that is already being generated? Manufacturing environments generate enormous amounts of useful information, but the reality is that much of it is simply trapped. It might sit inside a machine controller or an encrypted historian that only one person knows how to query, or it may be displayed on a screen and then disappear forever.
Often, it might be stored elsewhere but in such a difficult format that it’s too difficult to do anything with. Unlocking this information can be far more valuable than installing another layer of technology and can often be done cheaply.
Insignificant devices transform data gathering
One of the great opportunities created by the Internet of Things is that relatively simple devices can now collect useful information from equipment that was never designed to be part of a connected factory. For example, a vibration sensor, a temperature probe, an energy meter or a camera can suddenly provide another stream of operational information. Individually insignificant, once connected they can create the nervous system that your digital twin actually needs.
This is the part that interests us most at GF. We don’t start conversations asking, ‘what would you like your factory to look like in 3D?’ we begin with ‘what data do you already have? Where is it and can we access it? Can we trust it? And, what decisions would you like to make with it?’ Only then does it make sense to start building upwards. Once the infrastructure is there, the possibilities become much broader.
Endless possibilities as digital twins become easier and cheaper
The exciting thing is that building the digital foundations needed for an effective, transformative digital twin are becoming easier and cheaper. That’s why they’re finally becoming genuinely useful. This coupled with the fact we can feel like we’re not only ‘in’ the factory but that we’re connected to the physical factory in a useful way. Once this connection exists, the digital twin stops being something to show people and becomes something you use in a practical, affordable and valuable way.
And the possibilities for use are endless. You can train operators without disrupting production, test changes before making them, explore operating scenarios, compare predictions with reality and bring together simulation, live data and human experience all in one place.
Most importantly you can start reducing the risk associated with making decisions in complex, industrial environments. And thanks to our experience at GF and our powerful AI-driven twin, we can help manufacturers explore if they’re ready for one, help fill in any missing data gaps, then ultimately use their digital twin to transform their operations and make them more sustainable.