In a panel of experts at the Analysis, Simulation & Systems Engineering Software Summit (ASSESS) Congress, engineers debated the definition of the Digital Twin as well as the role simulation and IoT will play in its inevitable expansion.
Vendors have also weighed in on the topic, so strap in for a bumpy ride.
What is the Digital Twin?
The Digital Twin is a tool that can potentially account for the whole system of a product or service. It keeps track of all the information about a system you need and from that information assists in the decision-making process. (Image courtesy of Wikipedia.)
Tom Maurer, Senior Director of Strategy at Siemens PLM Software, defines the Digital Twin as “a digital model that accurately represents a product, production process or the performance of a product or production system in operation.”
He added that by “using design, simulation, manufacturing and analytics software, users can create and validate model-based Digital Twins of their products and production operations.”
So, a few things to unravel here. First is the notion that the twin is a representation of a real thing. This precludes simulations from counting as twins all by themselves. “A Digital Twin isn’t a twin until, you know, it has a twin. A physical product must exist,” joked ASSESS delegate Anne-Marie Giroux, Researcher at Hydro-Québec.
The next piece of information we learned from Maurer is that the Digital Twin is often related to model-based thinking because it links real world data to a systems engineering model of the whole lifecycle of a physical product or service. Though a twin can focus on one aspect of a product, its full potential is unleashed only when its usefulness spans multiple silos in an organization.
“If we take the life cycle we have design, manufacturing, services and operations, and then end of life. The benefits of the Digital Twin for each step is different,” explained Fouad El Khaldi, General Manager of Industry Strategy & Innovation at ESI Group at ASSESS. “For design, the twin’s main purpose is to set the performance of the product for the lifecycle. For manufacturing, it’s to optimize the process and reduce costs. For services, it is to reduce the operational cost and to use predictive methods. The idea is to bring simulation into this lifecycle information: you have a physical asset where you link it to a parallel Digital Twin.”
Olivier Ribet, vice president of Industry at Dassault Systèmes, noted that a sufficiently robust Digital Twin could lead us to do away with the concept of job roles all together. He said, “To unify and understand the enormous and diverse information about the 3DEXPERIENCE Twin (Dassault Systèmes’ branded term for the Digital Twin -e.d.), innovators have to overcome traditional, siloed-expert thinking. Of course you need the capabilities to scientifically and physically simulate all the pieces working together as intended. But, engineers also need methods and tools to foster a social dimension to their structured, physical and procedural information. It is about people, which has a social component that often requires coaching and change management to enable breakthrough thinking and open innovation.”
So, what does a system-level twin look like? Well, to the individual user it will look any another other Digital Twin, since it will only given them the information that they find interesting. After all, why would sales or marketing need to know everything the engineering team would need to know?
“The Digital Twin will mean different things to different people,” agreed another ASSESS delegate, who wished to remain anonymous, “If you do operate a hydro electric plant, it doesn’t matter what the initial design is. But, if you have a certified family of aircraft then it really does matter what the original design records were and how to integrate them.”
Ultimately, a Digital Twin will unify all the data an organization needs.
“The digital model matures through the product lifecycle during design, manufacturing and operation,” said Robert Harwood, Global Industry Director at ANSYS. “This digital connectivity through the life cycle can be described as a Digital Thread, which is not linear but circular, with data from all stages being fed back into the product ideation and creation stages.”
Maurer agreed: “We connect this information with the Digital Thread. We see the Digital Twin as a level of intelligence to predict real world performance and the Digital Thread is the connectivity and context for business decisions. It connects the design, operation and simulation information together.”
So, now we have the Digital Twin and Digital Thread. What’s the difference?
Well, Doug Macdonald, director of Product Marketing at Aras describes Digital Thread as the connective tissue between the Digital Twin, IoT and any other data source. It’s the channel through which information is fed back into simulations, product design and other aspects of the Digital Twin.
“MBSE [model-based systems engineering] is the starting point for the Digital Thread and serves as the foundation for downstream cross-functional design,” suggested Macdonald. “Using the insights gained from IoT data, product simulations can be run to better understand failure modes, leading to design improvements over time. A manufacturer or asset owner can link the Digital Twin to its service, manufacturing, design history, real time IoT data and simulation models specific to its configuration and expected failure modes in the DFMEA (Design Failure Mode and Effect Analysis). Comparing these simulation outputs with actual results provides valuable insights into the condition of the asset.”
So, with MBSE we are back to systems engineering. Using MBSE tool, engineers will be able to make a system-wide simulation of a product or service for the Digital Twin. Similar to MBSE, the Digital Twin can incorporate 3-D data/simulations, characterizations of the 3-D data/simulations using response surface models, 1-D simulations and 0-D simulations. The 1-D and 0-D simulations, as well as the response surface models are used to speed up the system models so they are no longer waiting for slower 3-D simulation.
“High-Tech innovators know that only a model based systems engineering (MBSE) approach provides the speed and accuracy to master today’s and tomorrow’s complexity of cyber-physical systems,” agreed Ribet. “In order to understand the actual customer experience, and how innovations impact that experience, the Digital Twin needs to be able to simulate the plant, product or service at system level, where you can efficiently manage its behavioral aspects that are driven by hardware, software and content-enabled functions.”
So, after all that, what is a Digital Twin? Srivathsan Govindarajan, Vice President, SAP Digital Twin, summarized it best:
“A Digital Twin is a dynamic digital representation of a live physical object and needs to represent specific aspects of physical objects like shape, working state and structural behavior. Digital Twins will dynamically change in near real-time as the state of the physical object changes.”
Common Misconceptions About the Digital Twin
It is inconceivable to think a Digital Twin needs to flood a user with every conceivable piece of information about the physical twin. (Image created using imgflip.com)
However, no one needs a tool that can account for everything from the subatomic to the cosmic scale. That’s also not the point of a Digital Twin. The Digital Twin simply links all the physical data inputs and digital model processes useful to an organization.
How Elaborate is The Digital Twin?”
“Our vision of a Digital Twin is a potentially comprehensive digital equivalent of a unique physical asset of something in the real world,” said Mike Campbell executive vice president of PTC’s ThingWorx IoT and AR technology. “The content you capture of a Digital Twin is based on a use case. What you monitor is what you care about.”
“When talking to clients, all of that Digital Twin data can be overwhelming,” added Campbell. “We frame up the questions as ‘What is your problem? What do you want to solve?’ Then we come up with a Digital Twin situation for you in particular.”
Let’s use Occam’s Razor for a second to dispel the myth that a Digital Twin needs every conceivable piece of information about it’s physical twin.
- No one would be satisfied with the tool as it would overwhelm them
- There would be no way to filter the data dump.
- There is no computer system available to crunch all the numbers
“Ask yourself: ‘What are the decisions I want to make and what do I need to know to make those decisions?’ Then we get into the appropriateness,” suggested Steve Coy, Founder and CEO, TimeLike Systems. “One thing people are doing now is they are sticking a bazillion data recorders all over things. They come back with a device and say ‘What do we do now?’ Print it out and use it for insulation?”
Just like every other device in the engineering toolkit, the Digital Twin will necessarily involve simplifications and assumptions in order to be useful. If skipping lessons on string theory is okay for finite element analysis (FEA), then it’s okay for the Digital Twin.
“It’s not going to be every [physical] aspect of the digital aspect that you have,” said another anonymous ASSESS delegate. “It will be one aspect you are really interested in and if you are interested in other aspects then you need a different model. That is a reality. You don’t have a multi-modal, elasticity, system level and every other simulation in [the Digital Twin]. That isn’t the way it will be.”
In other words, a Digital Twin is simply a virtual representation of all the information users need to supplement their work—no more, no less. It’s a question of scope. Sure, an organization can gather more data than that one user might need. But that would simply mean there are more Digital Twins for each asset, user or relationship or one Digital Twin that filters data accessible by a user’s role.
“Different people interacting with the Digital Twin might want different lenses of it,” agreed Campbell. “Your boss has a different Digital Twin of you than your doctor does. One looks at salaries and work history, the other medical history.”
Is the Digital Twin Just a CAD File?
So now that we have limited the scope of the Digital Twin there is a tendency to swing the pendulum too far the other way and conclude that the Digital Twin just a CAD or CAE model. The fallacy here lies in confusing the twin with a model. As stated above, for a Digital Twin to count as such, it needs a physical counterpart with which it can interact.
“While the term Digital Twin is often confused with a 3D CAD model, in reality, the Digital Twin is significantly more complex,” explained Macdonald. “The Digital Twin refers to a specific real-world asset in-service in the field and represents the exact configuration of the product at a point in time. By combining insights from IoT with an exact product configuration, service and manufacturing processes can be optimized and design improvements identified.”
Giroux agreed: “Well, design models have existed for about 40 years now. To me, the twin is something that is a recent concept. The added value is inputting data into that model from operations and real life. The twin takes data in from the real environment with the aging of the assets. I do not see any benefit of the twin if it’s only at the design stage.”
Perhaps some of the confusion about Digital Twins spawns from the idea that they can be used before a product is live. “The Digital Twin is a means to design and optimize end-user experiences,” said Ribet. “It is used before a real product or service is produced, and during the lifetime to the real product, the Digital Twin is used to monitor and adapt the real twin’s functions and performance. For that purpose, the Digital Twin has to be able to behave like its real twin, being equipped with all its knowledge, capabilities and characteristics.”
Remember that the Digital Twin is a system-level tool, which encompasses design, and notice how Ribet still references a “real twin.” This means that once you hook some sensors up to a prototype and link the data back to a digital model you now have a Digital Twin that is live before the product is.
In the end what sets a CAD model apart from the Digital Twin? The CAD model is not automatically changing in response to changes in a physical asset.
Why Do You Need a Digital Twin?
So, what makes this Digital Twin so useful? Why would anyone want to have a virtual representation matching what they can already see in real life?
The experts suggest that there are currently five major uses for the Digital Twin:
1. Validate Your Model with Real-World Data
“These devices do not exist in a vacuum. They exist in some environment so to model them and understand them we also need to understand their operational environment and how they interact with it.” - Steve Coy, TimeLike Systems
2. Track Operations for Decision Support and to Alert Users about Anomalies
“To me, it can become a twin only when that device is in operation. Then you begin to have some feedback on how [the physical twin] is really aging and being operated,” - Anne-Marie Giroux, Researcher at Hydro-Québec and ASSESS delegate.
“Many uses of the Digital Twin will be in real-time or close to real-time. You want decision support? That is not a year or week later: it is now,” - ASSESS delegate.
3. Predict How Assets will Change Over Time
“The insights we gain for the manufacturing and use of a product are optimizing our operations. It’s a question of how to satisfy orders, predict the performance of the factory and identify root causes,” - Tom Maurer, Senior Director of Strategy at Siemens PLM Software
“Using the Digital Twin in simulation can both improve operational procedures, and in contingency planning, it can even be embedded in the control system loop,” - Srivathsan Govindarajan, Vice President, SAP Digital Twin.
4. Get Design Feedback
“In the future, Digital Twins in manufacturing will help detect potential quality issues earlier on, or even improve the quality of the product being manufactured through delivery of new insights,” - Srivathsan Govindarajan, Management specialist at SAP.
“Some companies try to leverage what exists in data already from understanding their design. So, it’s richer than a simple model for design. They are using data from previous experience,” - Fouad El Khaldi, General Manager of Industry Strategy & Innovation at ESI and ASSESS delegate.
5. Discover New Revenue Streams
“New revenue streams can be created from the data itself. The Digital Twin has initially focused on complex, high-cost assets, but today the cost models for sensors, communication, analytics and simulation are such that it is possible to develop a Digital Twin of almost any product,” - Robert Harwood, Global Industry Director at ANSYS
The Digital Twin is clearly a comprehensive tool, with the ability to help out countless verticals in an organization. If you are only looking at the data, or only using the data to accomplish a single task, then you probably aren’t making the most of your Digital Twin set up.
So, how do you build a Digital Twin?
Harwood explained that your typical Digital Twin comprises of three basic components:
- Connected products, typically utilizing the IoT
- A Digital Thread that collects all the data and merges it into one framework
- The ability to convert this data into something actionable using analytics or simulation
Let’s dig deeper into what both simulation and the IoT have to offer the Digital Twin.
What Role Will Simulation Play with the Digital Twin?
Simulation will be one of many resources in the Digital Twins, along with artificial intelligence. and analytics. (Image courtesy of ANSYS.)
However, Maurer noted that when designing the physical aspect of a Digital Twin, CAE has “the ability to use this simulated data as the baseline, to define sensor locations and to determine what we need to study in order to understand the products’ operations.”
By feeding real-world data collected from the twin back into designs, engineers can improve future models of a product or even its current operation in the field.
Maurer said that, “the actual performance data can be compared to the Digital Twin, enabling actions that create successful business outcomes for optimization and new product introduction.”
In other words, engineers will be able to use simulations linked to the Digital Twin to predict how the physical twin will perform in a real-world environment instead of the ideal and perceived worst-case conditions outlined in the design process.
“The Digital Twin by itself doesn’t do any good unless it has interactions with its environment. So, you also need to model its environment,” explained Coy. "Depending on the nature of the interactions with the environment, and how well it is instrumented, that may well be the bigger challenge. For the system itself, you have the option to put in whatever sensors you need, but for its environment you don’t get to do that.”
“You can think of simulation as the secret sauce in the Digital Twin,” joked Harwood. “The low hanging fruit of the Digital Twin value-add comes through optimizing product operations and maintenance. However, by bringing in simulation we can leverage the Digital Twin to get real world product behavior insights. These insights can change the design of the product and how that product is then manufactured. The concept of the Digital Thread helps bring this level of digital connectivity across the product lifecycle.”
Now, one might be tempted to call “foul play” here, since 3D simulations are quite slow to respond with results. That might make it hard to believe that simulation can benefit a production line or equipment in the field.
Nevertheless, this is why it will be important to link up various simulation technologies to Digital Twins to meet the application at hand. Sure, you might need some slower 3D simulations for a twin in the design or prototyping stage. But for operations, 1D simulations and 1D characterizations of 3D simulations are often sufficient.
“Traditional simulation performed during the design phase of a product’s lifecycle can be perceived as a slow process, because many different use cases must be investigated based upon best estimates of the conditions the product will be subjected to in the real world,” argued Macdonald. “But, with the benefit of IoT data, actual operating conditions can be simulated with confidence, quickly yielding actionable insights.”
One of the most tantalizing possibilities is using simulation to assess how long the product will remain operational. Imagine a Digital Twin that keeps track of its own mortality, its own wear and tear. Using simulation, the twin can then estimate its remaining working-life and report back to maintenance.
“You use predictive maintenance to see how long the physical twin will operate to proactively shut it down and to schedule the shut down,” said an ASSESS delegate. “This would be opposed to a machine breaking down, which is much more catastrophic.”
What Role Will the Internet of Things Play within the Digital Twin?
A lot of the benefits we hear boasted about the IoT we also hear repeated for the Digital Twin: preventative maintenance, analytics and AI-optimized systems. So, what sets them apart?
“The IoT informs the Digital Twin with real-world performance insights,” said Maurer. “Merging this real-world insight with the predictive digital model allows users to take action on the insight, creating better products, optimizing production operations and developing new operational business models.”
In other words, the IoT acts as the bridge between the physical and the digital. Everything will be fed into the Digital Twin, from the temperature and moisture of the environment to the production status of the current batch.
There is no doubt that the IoT will be a popular option to gain insights about a product, but it won’t be the only one. Traditional sensors can also contribute to the Digital Twin, so don’t toss out that SCADA data yet.
Of course, the IoT does offer more flexibility with respect to the product’s mobility, location and monetization options. Take Caterpillar’s product-as-a-service model of selling the capacity to move dirt, rather than selling equipment outright.
A sure sign that the Digital Twin is a new technology is the anxiety it engenders regarding privacy and ownership. The idea of sharing this much data with suppliers, and therefore potentially competitors, might sound a little scary, especially when you’re concerned about IP. It certainly caused a stir at ASSESS.
“There is also a legal aspect,” noted an anonymous ASSESS delegate, before rhetorically asking: “Will every operator want to feed their operational data back to the manufacturer? Who owns the data of the Digital Twin and who gets to reap the benefits? Does the operator want to give their manufacturing data to the manufacturer of the device? Does he want to give it to all competitive manufacturers of these devices? From the operator’s stand point he may not want to give it to everyone. So, who owns the data?”
It’s great to use data to improve manufacturing operations, but most manufacturers only have a small pool of equipment data on which to base their optimizations.
Their suppliers, however, might have hundreds or thousands of connected equipment feeding them data. This larger dataset translates into more optimization potential, which has led many vendors to believe that customers will be sharing their data with suppliers.
“Service is where IoT offers the greatest insight,” said Campbell. “Wouldn’t it be great if the service tool were informed about what’s going on with the product now? Use predictive analytics to get the parts and get them to the maintenance crew faster.”
Macdonald agreed: “Together, IoT data matched with the Digital Twin enables an organization to understand more about how a product is being used by the customer. The benefits to the customer are significant: new insights can enable customers to better schedule maintenance, optimize use of resources, proactively address potential product failures and avoid product downtime. Ultimately, the Digital Twin is key to improving the product over time based on understanding of actual usage.”
This is all made possible with the IoT and a slew of Digital Twins, each being monitored from a central location that recommends maintenance cycles. In this case, the ownership of the data becomes a little cloudy especially if the equipment is now rented.
Clearly, we are just at the start of the Digital Twin story, and the truth is that it’s still being written.
Khaldi summed up this nicely when he said, “We are at the beginning of the Digital Twin mega trend. Today, the main focus is on services like predictive maintenance. But in near future, we will need to integrate the manufacturing into the Digital Twin to reliably handle this predictive maintenance and design.”
The big question is: What will you be using the Digital Twin for?
For more on ASSESS and how to become a member, click this link.
For more on building an IoT system for your Digital Twin read this eBook: Comparing Platforms for Developing New Internet of Things Products.
For more on the Digital Twin, watch this webinar: Using Digital Twins for Fast, Low-Risk Virtual Commissioning.