The Beginnings of a Rubric for Digital Twins
What you need to know to say "that is not a digital twin." Waterstons Innovation: The Dots
Today we are trying something a little different - below is a transcript of a conversation we’ve had as we begin to create a method to help us determine whether the thing you’ve made is actually a digital twin or is it just a fancy data platform. Enjoy, and let us know how you would decide if something is a Digital Twin or not!
What is this? This is The Dots, our newsletter about exciting things we find in the world of innovation. We imagine innovation as connecting the dots; putting together a jigsaw. Our puzzle pieces are the pieces of interesting information we absorb in the world, our partners and their products. This newsletter is about these dots we find and connect.
A Rubric for Digital Twins
Andrew: I thought it might be interesting to have a more conversational post about something we talk about a lot right now, which is Digital Twins. We've done a lot of other newsletters about it. So I thought it’s probably a good time to sort of come together and think about what we've learned.
And maybe begin thinking of the question that we started at, which is “what is a digital twin?” And beginning to try and craft a rubric, which is a word I really like right now, rubric, about how you would say something is a Digital Twin? Trying to get to a stage where you could say: “that thing is not a digital twin.”
Alex: It'd be a really great place to be. I think you're right there. I think it's such a broad subject. I think that the term digital twin, it's like the new Industry 4.0. It covers almost everything, depending on who you're talking to. And I think just very similarly to Industry 4.0, it started out as a very key core idea, and it's kind of drifted and grown into something bigger and more.
I think it is more marketing led now, actually, but there is still a core there of like what actually is it and what do you want to do with it that's important
Andrew: It's similar to AI. There is a core definition that has broadened out and out and out. And similar to industry 4.0 and AI is maybe what’s led to the drifting is that the core, right in the middle, is something useful and interesting. Something that's got people excited.
Alex: I think it's something useful, interesting, and extremely expensive and niche that's in the middle. And people have seen that and been like “wow, that looks good.” And then just naturally, because it's what we do as human beings, we've taken that core niche incredibly expensive thing and tried to make it less niche and less expensive, broadened it out.
Andrew: Well, yeah, when you hear Digital Twin, you think of this huge 3D platform that's making decisions for you. And yeah, of course, you want to ride that bandwagon, so you might end up calling your Power BI dashboard a Digital Twin, to try and get into that.
Alex: Yeah, exactly. I think that's a really, I think that's a good observation, right? It's a good place to start: is your Power BI dashboard a digital twin of your business? And there's a question behind all of that, which is, “is every digital twin essentially a model, a simulation of a physical object?”
What granularity or fidelity do you have to be at for it to warrant being an actual digital twin and not just a dashboard, or a report, or a model? Where do you tip over into that? How accurate or how much detail do you have to have in there for that to justify saying this is a digital twin?
Andrew: There's so many little bits in that statement you said. So many different ways that could break down the definition of a digital twin. The resolution, I think, is an interesting place to begin with. Because I think the resolution is a reason why you see a digital twin, in quotation marks, out in the wild, and go, “that's not a digital twin.” I think in a digital twin, the important word is twin. And it has to match something in reality.
So, I think a lot of the Digital twins you see don't perfectly match reality. They say “This is a digital twin of a city”, but it's not. It is a very specific part of a city. And I think that if I had to begin making a rubric, I might start there: you have to be a twin of something. You're not a “twin of a city”, you're the digital twin of the specific element of the energy grid of a city.
You know, it’s not a digital twin of a factory. You're a digital twin of the specific process inside of a factory, which is less exciting sounding, but I think more grounded in reality and realistic.
Alex: Yeah, I think there's that physical object, right? It has to be an actual physical object. Although there's a NASA definition, which is like, this is a twin of an object or process. But even then, when they're talking about process, what I think they're talking about is a physical process. Things are moving from place to place. It's not a digital twin of data flowing through a platform. That's not what they're interested in. What they're interested in is this thing that moves in the real world from here to there. And then someone picks up.
Andrew: Yeah, I imagine NASA has a lot of processes that move data about or whatever, but in some sense, they are a manufacturer of something. It's just the thing they are manufacturing is a spaceship.
I think that's another thing you see a lot of it - people making “digital twins” of an ephemeral idea, a business process or something that's a data collection pipeline. And I'm not sure if that counts.
Alex: No, I don't think so. I don't think so. I mean, a lot of this is really woolly, right? But I think there has to be a real-world tangible thing that you could reach out and touch or interrupt by standing in the way of it for it to qualify.
Andrew: Then that leads us to another interesting point, which is we've seen examples of people starting with quotation marks, a digital twin, rather than the other way around. “Oh, we made this digital twin about this thing that doesn't exist yet.” And then they build the real thing.
I'm guessing by our definition, that doesn't count. That's not a digital twin because a digital twin has to have a tangible asset it is attached to.
Alex: I don't know, I struggle with this. It's definitely not a digital twin before the physical object exists. Right? A definition of a digital twin is that you have a virtual model of a physical object and that virtual model is kept in line with the physical object in a timely manner, right? I don't know that it matters whether the model was built before or after the object.
Would it be that if I 3D printed my bridge and then I took a digital scanner to it and I scanned it and then I took that scan and turned that into a 3D model, would that then be a digital twin of the bridge?
Andrew: I think I see what you're getting at. And I think yes. I think before the physical version exists, it's not a digital twin, but when it does exist, it slots together, and becomes a digital twin.
I think the other assumption there is that the information is up to date. They match each other. A change in the physical one will affect the digital one and the change in the digital world affects the physical world as well.
Alex: Now that's an interesting addendum there, which is that it goes in both directions, that you can change the virtual one and it changes the physical one, because that's a different thing.
Andrew: Cause this is, well, this is why this is something I was getting hung up about. What is the difference between a simulation and a Digital Twin? And I think it's that information is bi-directional. It can come and go from the physical world. Whereas in the simulation the digital world is just the output. You are simulating something, but that doesn't feed back into the actual tangible object.
Alex: But, if you have a digital twin of your building that monitors airflow, air quality, temperature, window status, all that stuff, you might run a bunch of simulations on that model that you've built (that is a digital twin) to say: “if we open the windows in this way on this side of the building, it will cool the whole building down by four degrees centigrade.” Right? And when you run that simulation, you're not opening all those windows in real life too...
Andrew: Okay, well, I totally agree with that. I totally agree. I think maybe what I mean is that a digital twin encompasses the idea of a simulation. You have the ability to run your simulation in digital twin and go, “oh, I like that” and press a button and feed that back through to reality. Whereas in a straight-up simulation, you couldn't press that button and go make that reality.
Alex: I think that's true. Although again, if you have a digital twin of three robot arms on a factory line, and you are working out whether or not they're gonna hit each other as they move, and you're looking at if this one moves in and screws this screw in at the same time as this one moves in and pulls this chip, and at the same time as this one goes in and puts these tabs in, are they gonna clash? You're unlikely to take that, the output of that and press a button and immediately have that happen in the real world.
You're more likely to be like, “oh yeah, this is set up correctly. I kind of worked out that these things are not gonna hit each other.” That's still a simulation where there's no data flowing back in the other direction. I actually am not totally convinced that a digital twin has to flow data in both directions.
Andrew: I'm not sure. I'm not convinced the data can only flow one way. That's maybe something that maybe that's something we could go and do some research on and find someone to talk to about.
Alex: I think one of the interesting things about this is that I think you won't get that definition.
Andrew: Is the point of it all then to make a decision? If you can't make a decision off the back of it, is it not a digital twin?
Alex: I think that's the root of this. I mean, we went away, we talked to Professor Miles Elston at the Institute for Safe Autonomy. We grilled him about what he thought a digital twin was. He has a digital twin of the building that he works in, it's a beautiful building.
And he gave us this, his fundamental view was, well, yes, a digital twin needs to have data flowing in both directions. There needs to be a 3D model of the thing that you are twinning. There needs to be the ability to run a simulation in that model. That was that. That was his just baseline of this is a digital twin.
And then we kind of talked about it for a bit and he said, “or you could say that actually any kind of simulation or model that allows you to make a useful business decision could be described as a digital twin,” Which is like the opposite end of the scale to the place that he started with.
I mean, ultimately, I don't think it really matters whether you describe it as a digital twin or not. It's a tool that you are using to make business decisions of some description. And I think that's key.
Andrew: The idea of having the ability to make a decision in the definition is interesting because you actually mentioned earlier that the resolution of a digital twin should be granular enough to make a decision.
I think you only have a digital twin to the level of your decision-making ability. If you can't make a decision about every element of a city, I don't think you have a digital twin of the city. You have a digital twin of the element you are making a decision about.
Alex: I think that's right. I think actually the key to working out what your digital twin would look like is to start with what problem we're trying to solve and then to work backwards and then to be like, well, okay, if we want to solve that problem, how do we build a model? And actually, maybe this is fundamentally it. Can we build a model of the system or process or physical object that then allows us to make that decision appropriately?
What do you think?
We have lots of ideas about how to keep this Substack going, but would love your input. How do you want to receive our newsletter in the future?