The executable digital twin leverages advanced tools like simulation for faster, better results.
This episode of Designing the Future is brought to you by Siemens Digital Indutries Software.
In engineering, moving a concept from idea to eventual hardware has always been a challenge. And when those engineering projects are large and complex, the need to establish a single source of truth to keep multiple engineers and development teams working together is essential. Configuration control of conventional CAD processes help, but today’s computer-aided engineering includes development tools that go far beyond rendering, such as computational flow dynamics and simulation. The ability to iterate quickly and virtually puts a premium on project management.
The solution is the digital twin, which promises to maintain order within a rapidly accelerating design iteration process. And the executable digital twin is the key to real-world, cost-effective application of the digital twin concept.
Joining engineering.com’s Jim Anderton to describe the executable digital twin are three experts from Siemens Digital Industries Software: Ian McGann, Director of Innovation for Smart Technologies, Doctor Leoluca Scurria, Product Manager, Executable Digital Twin and Doctor Durrell Rittenberg, Director, Simcenter Experience Product Management.
The transcript below has been edited for clarity:
Jim Anderton: In engineering, moving a concept from idea to eventual hardware has always been a challenge. And when those engineering projects are large and complex, the need to establish a single source of truth to keep multiple engineers and development teams working together is essential. Configuration control of conventional CAD based processes help. But today’s computer rated engineering includes development tools that go far beyond rendering, such as computational fluid dynamics and simulation. Now the ability to iterate quickly and virtually puts a premium on project management.
The solution is the digital twin, which promises to maintain order within a rapidly accelerating design iteration process. And the executable digital twin is the key to the real world use of the digital twin concept. Joining me to describe the executable digital twin are three experts from Siemens Digital Industry Software. Ian McGann, director of Innovation for Smart Technologies, Dr. Leoluca Scurria, product manager for Executable Digital Twin, and Dr. Durrell Rittenberg, director SIM Center Experience Product Management.
Can we kick off just a sort of level set and establish something basic? What is a digital twin?
Ian McGann: Well, for us it’s a virtual representation of a physical asset.
Jim Anderton: It’s a very simple way of describing that, but it’s in a sense, isn’t everything a digital twin in a sense, the concept or the idea an engineer has for design is in a sense a digital twin and a rendering is in a way, a digitized twin. But we’re talking about something that’s a little bit different from just a CAD file, aren’t we?
Ian McGann: We are, yes. Like you mentioned, documentation can be a form of a digital twin as well. What we’re talking about is when you can capture the dynamics, the movements, the forces, the stresses. That’s for us in engineering, let’s say a more functional digital twin and capturing all of that, let’s say complexity of the structure. So we would say we have a complex digital twin or a comprehensive digital twin. So we’re capturing the fluid structure interactions, the mechanical, the electrical, everything in that digital twin model that represents and is linked and connected to the physical asset.
Jim Anderton: Leoluca, we’re talking about the executable digital twin. What is executable? What makes digital twin executable?
Leoluca Scurria: Yeah, so this links actually to the definition of digital twin because usually these digital twins are used for specific purposes. So when you have multiple representations of different, let’s say physics and behaviors of the assets and the way we make it executable, basically we package it and we package it in a way that can be used also for other purposes. And this is key because our customers put a lot of efforts and invest a lot in the creation of digital twins. And basically our goal is then to enable them to leverage these descriptions, these behavioral representations also outside of the usual, let’s say, environment. So you can imagine a digital twin that gets created during the design process then can be reused also on the machine to optimize the control or the maintenance, and basically can extend the value creation over the entire product life cycle.
Jim Anderton: With a rendering, a rendering is a digitized version of what used to be an ink on paper hard asset, something that was a physical touchpoint you could hold in your hand and you could refer to. So if there was a question, it would be, let’s check the print. Now it’s let’s check the rendering, let’s check the CAD file. Does digital twin have that same relationship to a physical asset? Is this something where you’d think on the shop floor, or if there’s a question of a manufacturing engineer talking to design engineer, you’d say, let’s check the digital twin?
Leoluca Scurria: Well, that’s one possibility, and that links to the one source of truth that we always worked over. So, what you have with rendering is really an image or like a visual representation we instead look at representing the physics of the system. So when we go into how is actually this asset meant to work, then we can look at the digital twin and say, Hey, this is how we see the physics working and interacting with external world.
Ian McGann: Well, like Leoluca says, it’s brilliant in that and in this same reference that you had related to the CAD file, like you’d look at the CAD file, but that’s always a point in time. Once the machine is made, things change and the CAD file is still representing the perfect scenario the way we want it to build it, but it doesn’t represent that moment in time. And it’s the same if you just looked at a physics model. You’d see that’s the way we intended it to work.
But what we’ve done with the digital twins is we’ve gone that step further, is we’ve made a connection between the physical asset, the machine, and that physics based model or complex model, let’s say, and the model updates into match the virtual in real time. So if you take a look at it, you’re on the shop floor, you say, okay, let’s take a look at the model. You’re going to see exactly the state of the machine, the physics behind the machine, at that point in time. But the cool thing is you can go back in time as well. You can say, well, what was it last week? Did it change?
Durrell Rittenberg: Well, one of the things also is to think about the types of challenges that an executable digital twin can really address. And Ian, one of my favorite examples is we’ve done a fair amount of work in the automotive industry, but in other industries as well. And we work with one of the larger, actually work with all of them, but this is one of the larger German auto manufacturers. And they came to us with one of the challenges they have, which is, Hey, you’re doing such a great job of simulation that you’ve actually reduced the need for physical prototypes, except we still need physical prototypes. We need to be able to test them. We need to be able to validate them. What can you do within Siemens and your digital strategies that might help us improve what we get from a physical model? And so we started by going back to the engineering groups and saying, okay, well what kind of digital twins do we have that we might be able to leverage?
And we actually chose vehicle dynamics, it’s not important. What is important is that working with that team, we were able to show them how a digital twin could actually help them on the physical side. And they went from setup times that were in the three day, four day timeframe down to about four hours, which was fantastic. But more importantly, within those four hours, once they got it out on the track, they were getting 10 times more information because they were using the executable digital twin as part of the process.
As a result, their models, they didn’t have to go through five days of testing, they could get it done in one day. Which means that the amount of time, calendar time that they have, wall clock time, they can use this physical prototype for other types of analysis, went way up. So it’s the kind of thing that extension of the engineering that’s done typically in design and extending it into the physical prototype. And ultimately those models with the executable digital twins can be deployed either in the AV testing mode, so they can put it inside of a model that runs for AV testing, or it can actually be deployed on the vehicle itself, which means that your car has an executable digital twin running in the background that helps the car improve tuning of the suspension system, the breaking system, the safety systems. All of that now is leveraged based on the work that was done by the engineering teams back in the design process.
And it’s that kind of connectivity that the executable digital twin is bringing to modern engineering context, that this connectivity, this representation is really changing how people think about what they do on the engineering side and how that engineering work can be transformed and used in different contexts. And I think that that kind of helps phrase exactly the types of problems we’re trying to get after. Yes, the technology is based on lots of different types of engineering and types of model reductions and these processes are real time, but it’s important to take that step back and think about how this is helping modern engineering corporations and companies around the world tackling difficult types of problems and do things in a very different way. So it’s really quite exciting.
Jim Anderton: Durrell, you brought up a couple of things which are worth unpacking that feel like a paradigm shift. Look, traditionally in product development process from engineering perspective, it doesn’t have to be automotive, it could be almost anywhere. There’s a design iterate, test, break, redesign, iterate. And there used to be a saying my old engineering boss used to say, there is no design, there is only redesign. And so the process is about how fast can we iterate our way to an acceptable solution and then we go.
And we iterate our way, historically in many cases by building prototypes and then real world testing them and then breaking them. Now we’re looking at a world in which not only can we virtually test things and break them, but Durrell is talking about a world in which we can deploy product into the field and then get real time feedback about the performance of that product and cycle that back into the redesign process. So are we blurring the lines between where design stops and production begins? Are we going to see a future where redesign is constant for everything from the shoes in our feet to the cars we drive?
Ian McGann: So that’s a good point. So we have to combine the shift left strategy that we see in our customers. So less prototypes. We need to achieve the same output in shorter time reducing the prototype cost. And so we have two ways of doing it. So one way is to basically accelerate the design process by combining the digital twin with the physical testing to get more information during the design phase. But also we need to connect the products that are out there in the field creating very useful and meaningful data connected back to the design phases. And this creates, let’s say an ecosystem, a digital thread throughout the entire life cycle that allows a quicker innovation, a quicker improvement of the product themselves. So this is the ultimate goal to really create an ecosystem where the information can flow across the enterprise and the product lifecycle.
One of the things I love about our customers is they always push us. They challenge us a little bit and the latest challenges, well yeah, I want to know what’s going on in the vehicle at this point in time, but I want to know, well, the example they have is a battery degradation. They want to be able to take the battery that’s in the vehicle and resell that at the end of its life, but they don’t want the battery obviously to be damaged when it gets resold. So there’s an optimal point where you say, “Okay, now the battery is the perfect point to resell it. Let’s take it out as a vehicle put another one in and then give that battery a second life.” So the analysis and the work that you have to do, you don’t want to put sensors all over the battery.
So we use digital twins to give us these virtual sensors to do battery degradation analysis in real time and then report back to either the tier one, if that’s the owner, let’s say of the battery if you’re leasing the vehicle or the OEM themselves or even the owner of the vehicle. So he could get that information and say, “Ah, you know what? Time to change my battery. I’m going to install it in my house as a backup generator and go get a new one.” But it’s that type of thinking that I love. That’s what our customers are bringing to it that we didn’t have before.
Jim Anderton: Durrell, that’s an interesting point is that we live in an IOT age where we’re talking about a future in which we have sensors embedded in everything and many sensors embedded in a product. The feedback I hear from individuals working in this sector, there’s a worry that we’re going to overwhelm engineers with data and that processing that data is going to be a factor. The simulation community turn around and say, “We’re not going to need 10,000 data points for a tennis shoe. We’re going to simulate the product up front basically, and we’re going to optimize it to the point where we can do three sensors and we can get the actionable information we need. Is simulation going to basically sort of wrestle that data overload problem to the ground, do you think?
Durrell Rittenberg: I think, there’s two pieces to that. One is obviously simulations are getting faster, more accurate, and you can get more information in that design phase and that actually can increase confidence for sure. But there’s another piece to that which is, there are strategies with machine learning and AI that we’re using that help kind of take that information and actually provide insight, not just data. I mean, the problem we have right now is it’s easy to create so much data that you really find yourself in the, my favorite is Delta Airlines. The CEO actually talked at a conference I was at which was an engineering conference. He said for every flight they’re bringing back about a, I think it was four terabytes of test data for every engine and every system in an aircraft.
And he said, “We have all this information and someone said, “Well that’s great. What do you do with it?” He said, “Well, we send it back to the engine manufacturers, let them make sure that they know we have this data.” But ultimately what they want to get to is how do you get information back to the people who have to maintain that aircraft that you know what? Engine number two is probably getting close to end of on wing life and needs to be refreshed and how do you predict that. And that’s where the executable digital twin, which can be powered by IOT, which can be powered by machine learning algorithms and other strategies can help predict when you might need to do that. So that idea of predictive maintenance which is kind of where the goal of a lot of the work that’s being done in engineering today is really around that because there’s such a business reason to do it.
It also comes back to that performance optimization. Ian brought up a great point, which is how do you make those batteries last longer? Well, you do that by understanding how they’re performing. And you do that by understanding the information that’s coming out of those using this an encapsulated, executable digital twin strategy within that battery pack, and it’s giving you more information, more insight about how the battery’s performing that you wouldn’t get otherwise. And it’s that insight that allows you to say, you know what? It is time to slap that on the wall and make it a Tesla battery or whatever, the one you put in your garage that can power the house.
So that’s the kind of thing we need to be really thinking about, it’s a shift in the way we think about how information could be used and the kinds of things that these executable digital twins can provide in terms of information and insight into a complex engineering problem. It takes it out of the engineering domain and takes it back to the guy who’s got his new Lucid or whatever the most recent, EVS and it’s telling them, “Hey, you know what? Look, the way you drive is impacting the battery, you might want to consider changing it or maybe it makes a shift inside the actual drive mechanism to save battery life.” Anyway, these are the kind of things we need to think about. It’s a lot more than just the technology. It’s what you get out of it.
Ian McGann: My little addition, so the machine learning aspect of this I think is quite interesting. We have one customer who is using the digital twins combined with physical assets. So they have a physical asset, they have a digital twin, physics based digital twin combined with it. And what they want to do is generate massive, massive amounts of data for the purpose of machine learning. They don’t have a history, that’s the problem with it. So in order to get that data they’re using the digital twin models with defects programmed into them that allows them to annotate or label the information far more accurately. So that when they’re creating that machine learning algorithms, they now have labeled information of thousands and thousands of data points that they wouldn’t have gotten until the systems were deployed. So that’s the other use case where you have the digital twins are complimenting the machine learning and actually generating more data for you. But it’s smart data, it’s insightful data.
Jim Anderton: Product design has always been constrained by manufacturability. You can design anything, but can you make it? I’d say we’ve got some technologies like additive manufacturing, which have removed those constraints to a great extent. If you can imagine a shape and test that shape virtually you can make it now. Is the executable digital twin, will that work with technologies like additive or process automation do you think, to alter the way engineers go about the design process? Are they more free now with this technology?
Leoluca Scurria: So when we look at innovative production systems like additive manufacturing, the knowledge that the customer have on the production processes is limited. And often when we look at manufacturing, most of the decision making are based on prior knowledge about the production process. With executable digital twin, basically, we can maximize and give smart data to our customer based on few prototypes of the innovative production system. And that basically allows to come up with a, let’s say, effective production process much quicker.
We actually had a talk with an aircraft manufacturer that was trying to optimize some manufacturing application for composites. They were saying like, okay, we have these new production processes that we want to speed up, but we don’t know what is the right starting point to optimize our parameters. So, that’s where we started engaging and talking about how we can really reduce that lead time from initial prototypes to, let’s say, actionable production processes through virtual sensors, performance optimizations and performance predictions. And that you can only do it if you can combine the few information that you have from initial prototypes with a digital twin to maximize not only the information, but really the insights about your production process. That’s where the real value of executing by digital twin is, it’s really to transform data or big data that you have in two insights and actionable information.
Jim Anderton: Durrell, we see in business software world, the movement towards software as a service rather than selling a CD-ROM with a package in it and then mailing updates. An individual in the electric motor industry was in conversation with recently mentioned that they felt that the future for that ubiquitous product, which is used in engineering manufacturing everywhere, was to perhaps go away from selling an electric motor to a customer, but actually selling power by the hour and a model used by the jet engine industry before. Because this ability to feed real realtime information back means that the electric motor manufacturer could schedule preventative maintenance or even swap the motor out without the customer even being aware of the performance of that motor.
So, in a perfect world that imagine that you’re leasing the motor and the motor manufacturer sends a technician who services it or even replaces it, perhaps without the customer even knowing that individual’s coming. So it sort of worry free, trouble free. If you extrapolate that world, you’re talking about a potential world in which everything from the clothes on our back to the shoes on our feet to the robots that build our products down there no longer exist as an owned asset on the factory floor. The executable digital twin, does that play into that future?
Durrell Rittenberg: It absolutely does. And actually I think the example that you just gave with the electric motor or even with the gas turbine or wind turbine, really doesn’t matter if you start to think about it as, as an organization, we basically get paid for how much power we generate with our wind turbine. And if we can start to figure out ways through the executable digital twin to improve outcomes by looking at not just one wind turbine for example, but maybe looking at all the wind turbines that we have in a wind farm. Within those wind farm wind turbines, we have executable digital twins running within the motor that gives us information about the mechanics of the motor and whether it needs to be serviced. We can also look at wind performance. We can start to bring together all this information and use that as a decision making platform to optimize how much power we’re generating. That’s going to help someone who’s in the business of selling that power maximize their return.
It also has an environmental impact that we don’t often think about that. But the idea that you can improve not only the efficiency, which has a direct impact to how green a particular type of energy might be, you start to think about how an executable digital twin strategy can be part of a sustainability effort within an organization. There’s another example which is kind of slightly tangential to the wind power, but it has to do with food manufacturing. We don’t think about food manufacturing the same way that we think about building a jet aircraft. But food manufacturers generate a tremendous amount of product and one of the main people working with right now, they make cheese puffs. We all like cheese puffs. They’re delicious. But you’ve probably been in a situation where you took a cheese puff out and you bit down on that guy and it was super hard, you almost broke your teeth and you’re wondering what the heck is going on? Well, it turns out that food science, the process by which they make those cheese puffs, that whole process is heavy engineering.
So if you can come up with this executable digital twin that can provide insight to the manufacturer that, hey, my extrusion mixture is off just a little bit, and it has that outcome where you get things that are just not sellable, that executable digital twin is going to save them a tremendous amount of energy because they can actually optimize their output, it creates a better product for everyone, and it does it at a more sustainable way in the context of how much power he uses. So when we think about these digital twins, we really need to be thinking about how executable digital twins and this overall digital thread from requirements through the design, through the actual manufacturing, and ultimately into the bag of cheese puffs, all of that, there’s digital twins across that entire workflow. The more that we can start to leverage that, the better outcomes in any one piece of that.
But this is the kind of thing we should be thinking about as an industry because we need to figure out how engineering is going to help us address some of the major challenges we have as a population. These are not just making novel devices. This is about really looking at how we can impact the world in a positive way.
Jim Anderton: Ian, Durrell has just touched on sustainability and there’s no way to have any talk about engineering at all in this day and age without talking about sustainability. And how many times have we talked to … I talked to manufacturers, engineering firms to say, “Great, I’d love to reduce my carbon footprint, but I make valves. I’m not in business to reduce my carbon footprint.” Can you talk about how the executable digital twin can help them square that circle, which is address sustainability issues, but not lose focus on the core business?
Ian McGann: If you look at where the valves, pumps, et cetera, will be used in the end, it’s going to be in … Well, we have one example. Customer is a water reservoir. And it’s covering a country. Right? So this is quite a …. 16,000 pumping stations at extremely complex set up. And the problem is that you’re pumping water from one location to another, and by the time you get the real data back, it’s too late. You’ve already pumped too much or you’ve already distributed too much water. Right? So that the information doesn’t come in from the real sensors in time.
So we create a digital twins of the entire setup, all the pumps, all the stations, everything. And we then start optimizing that. So it’s a information leads to insight, leads to optimization. And I think that’s the point that we’re getting to this, is that if your valves, your pumps, if they have these digital twins connected to them already, you’re selling that information as well as the value of the pump. Right?
And that’s connecting to a bigger system. And that bigger system is basically conserving energy by saying, “Well, you don’t need to pump so much because we’ve predicted that it will be full. So stop now before the real data gets there.” So it’s like an extremely advanced model predictive controller. But the model is of the entire system, and to the levels of complexity that you normally wouldn’t see in its standard MPC.
So yeah, we have examples of customers and they don’t ask us for a digital twin. That’s the nice thing, is they don’t say, “Oh, we need digital twin technology.” No. They come in and say, “We want to be more energy efficient. Can you help us? We have this target. We want to …” You know, “How can we get there?” And in pretty much all of those cases, the combination of the executable digital twin and the physical sensors that we have and can distribute throughout the systems is what’s allowing them to get there.
Leoluca Scurria: And another extremely important point is that, as Ian was mentioning, that we have customers which have system in place since 30 years. Right? And you don’t want to go there and say, “Look, I got an amazing solution for you, but you have to go back to your system and put, I don’t know, 1000 sensors of it so we can optimize it.” With executable digital twins, we can be what is called brown field compatible. Right? So we can transform the information that you have from your system into meaningful insights.
And that’s what drives in the end your, let’s say, the business value of it. Because we can really reuse what is already in place. And then you can imagine that there are also companies that then use it also to improve their business models. So they can say, “Okay, are you a power user? Then you need to have this type of capabilities on your machine, this type of performances.” And as it is software based, they can easily switch it on and off depending on the customer need, to optimize also the usage and their business model in the end.
Jim Anderton: As a question, sort of a wrap up question, can I just go around the horn quickly and ask about the executable digital twin as a starting place? How can an engineering organization now which uses conventional software-driven design processes make that transition? What is the first step that they need to do? Durrell.
Durrell Rittenberg: It’s a great question. Of course, it’s going to depend largely on the type of engineering they’re doing, but it’s kind of going back to what I talked about in the automotive sense. It’s really starting to look at where are those challenges that we want to address, and looking at the kind of information we have to start. And that’s really where we would come in and work with an organization is to try to understand, “Okay, what is your ultimate outcome? What are you looking to do? And let’s take a look at the assets you have in the context of the water system.” And using that as a start point, really trying to connect the dots.
Siemens, as an organization, as you know is somewhat broad in terms of the kind of things we do. We have motion control, we got the factory automation bit, we’ve got the software, the part that we fit into. But all of those pieces are part of this broader ecosystem of digitalization that organizations are going through. And it’s really taking a look at, “What are the outcomes?” And starting to look at the engineering information that’s available as a start point.
Because what we find is oftentimes there’s enough information to get started within an existing design process, that we can start there as long as we know where we’re going. And that’s kind of the key thing. And we can help with that as well.
Jim Anderton: Leoluca, your recommendation, how do they start?
Leoluca Scurria: Yes, indeed. So I want to link a bit to what Durrell was saying because what I always say when I talk to customers is like, “Okay, we have this amazing technology that really solve business problems. We are expert in digital twin, executable digital twin, but you customer are the expert in your application, and you only know your problems.” So we always start with having a conversation, and understanding how we can find a compelling problem that we can solve with executable digital twin.
And then from there we can move in different directions. So that depends on the stage of digitalization of the company. So you have some enterprises that already have a broader adoption of digital twins. So in that case is more shifting the paradigm to also use in service and for maintenance. And we have companies which are, let’s say, that use less digital twins. And in that case, we can also, let’s say, go on a journey together starting from digital twin creation and then leverage this digital twin also outside of the, let’s say, more economical usage.
Jim Anderton: Ian, what’s the first step?
Ian McGann: Well, the first step is creating your digital twin. Right? So that’s the first step. And if you go back five years ago, it was an extremely difficult process. And we worked really hard to make that one button click and it’s done. And it’s an executable digital twin that you can now deploy, manage across your machines. And I think that’s the point is people might have looked at the digital twin technology three years ago, four years ago, and they said, ‘Ah, it’s too difficult. It’s not working.” Or, “We’re not getting the connections between the physical assets and the virtual like we want.” I’d say, “Take a look at it again. And what you’ll find is that that creation process, the deployment process, that’s in place now, the management process is a lot, lot easier than it was five years ago.”
Jim Anderton: Durrell Rittenberg. Leoluca Scuria, Ian McGann, thanks for joining me on the show today. And thank you for watching. See you next time on Designing the Future.
