Digital Twins: Where Are We Now?

An overview of digital twin technology and a summary of where the technology stands today.

The concept of the digital twin isn’t exactly new, but like many things in Industry 4.0, digital twins are emerging more frequently as a result of the convergence of technologies that enable their existence.

Thanks to the cloud, the Internet of Things (IoT), advanced simulation and vast amounts of compute, digital twins are definitely an emerging technology to watch out for, and industry is spending a lot of money to keep abreast of its developments.

In this article, we will be taking a look at the basics of the digital twin, as well as speaking to representatives of various companies involved in digital twin development to find out where this technology stands right now.

Turbofan digital twins. (Image courtesy of ANSYS.)

Turbofan digital twins. (Image courtesy of ANSYS.)

Digital Twin 101

A digital twin is a virtual (and often 3D) representation of a real-life system.

There is a dataflow between the twin and its real-life counterpart, and, via sensors on the real-life system, the twin is updated so that the system’s status can be monitored, virtually, in real time. And this process is not just for status updates of the real-life system. The twin can be programmed to run what-if scenarios and optimal parameters can be determined for functionality.

A digital twin can often comprise different layers, with each virtual layer representing an equivalent physical layer. These layers can include physics, materials, structures, electronics, fluids and so on.

As long as the system is being measured, it can be modeled in the digital twin, and the more data you capture from the real-life system, the more robust and accurate the digital copy can be. And our ability to capture, process and simulate in the era of big data and the cloud has made all of this possible. The digital twin is truly a child of Industry 4.0.

A lot of information can be gained from the digital twin—be it diagnostic or prognostic. Such information can be used to reduce downtime and increase return on investment. It can even be fed from actual service life back into the product development phase, enabling a closed loop product lifecycle in which future products are continuously optimized. This last point is a common theme within IoT-driven product development. In the past, product development generally ended at the point where the product was handed over to the customer. 

Digital twin overview (Image courtesy of Deloitte.)

Digital twin overview (Image courtesy of Deloitte.)

Those days are now long gone, with product feedback extending far beyond a product’s handover into service and even beyond that point into product disposal. As a product grows in efficiency, so does its twin—and vice versa. The twin can be used as a sandbox for testing new ideas that can inform its real-life counterpart. And the real-life product can inform the twin. The dataflow is truly a two-way process, and the digital version and hardware can evolve together.

“A digital twin offers a new way of interfacing with the real world by creating an immersive, accurate digital representation of the real world.” said Heather Kerrick, senior research engineer at Autodesk.

“Our research is exploring a digital twin as a programming environment for hardware such as 6-axis robot arms, in addition to monitoring a facility [where] a user can actively act on what they see and make changes to their equipment and processes from within the digital twin and see those changes reflected in the twin’s physical counterpart.”

A Twin for All Occasions

Digital twins can come in many forms depending on the industry and use case.

For example, a production line may have a digital twin that shows the status of the line, and which may be connected to the feeder and transfer lines that show any stoppages. The same digital twin may even be connected to a finite element analysis platform, which can show how changes to a product’s design can affect the facility’s real-life operations and efficiency.

At the extreme end, the production could be the assembly of a fighter jet, with every single tangible component and system being mirrored in a 3D model. Want to know why the landing gear is stuck? Just click the landing gear on the model and it will zoom into a subsystem. Keep zooming in until you find the problem. Want to test some new control surfaces on the jet? Just add them to the digital twin and simulate it.

Or take the example of a 3D-printed bridge in the Netherlands, which was designed by MX3D with the aid of Autodesk. Both the prototype and the actual final version of the bridge have digital twins that were used for different purposes.

“Our ‘smart bridge’ prototype at Autodesk’s Pier 9 Technology Center has an operative digital twin that is based on a Revit model—it’s a simple structure, so no need to construct the digital twin on a captured as-built model,” said Alec Shuldiner, IoT researcher at Autodesk and project manager for the collaboration with MX3D.

“That’s used to show the location of sensors and, via Autodesk’s Project Dasher 360, the (near) real-time data generated by those sensors.We use machine learning algorithms to generate intelligence on top of that sensor data but don’t currently show that intelligence (e.g., occupancy counts) in the digital twin. In future, we will upgrade the digital twin to show not only the data streams but the ML outputs as well.

The [final version] bridge is largely complete, but the sensor net has not yet been assembled or applied,” continued Shuldiner. “However, we anticipate later this year rolling out its digital twin that will be identical to the Pier 9 bridge twin—location of sensors, real-time data streams generated by those sensors, ML-based intelligence derived from those data streams, including real-time status of the structure as a whole—with one important exception:  the underlying model for the digital twin will be based on a scan of the as-built bridge (necessary because the as-built and design diverge significantly).”

As you can see, digital twins are all things to all people, and their applications vary according to industry and use case. What data can be represented in the twin really comes down to the ingenuity of the twin’s designer, and what it is that needs to be measured. And if it can be measured, it can be managed. It is this increased insight through efficient product lifecycle management that makes the digital twin an attractive proposal.

“The convergence of a physical asset with its digital definition through a digital twin is going to have a multiplier effect throughout the organization: from new products and business models to better engineering decisions, improved efficiency and serviceability.” said Francois Lamy, Vice President, PLM Solutions, IoT Solutions Group, PTC.

“By analyzing assets against their real-world usage, manufacturers are going to be able to make better decisions to improve the product and to better market and sell it. The digital record of the asset from design, manufacturing, and into the field will offer a great opportunity to improve profitability with future sales, recalls, or upgrades.”

“However,” continued Lamy, “manufacturers need to start managing and making sense of data with a smart, connected product development process before they can achieve digital twin. By organizing product data in a digital product definition, proactively designing products for connectivity capabilities, and ensuring a digital thread throughout the entire product lifecycle, manufacturers can take the next step to a complete digital twin.”

Where Are We Now?

The digital twin was identified as a Top 10 Strategic Technology Trend for 2018 by market analyst giant Gartner. Consequently, it’s no surprise that the digital twin made its way onto the 2017 Gartner Hype Cycle for Emerging Technologies.

Gartner's Hype Cycle for Emerging Technologies 2017. (Image courtesy of Gartner,Inc.)

Gartner’s Hype Cycle for Emerging Technologies 2017. (Image courtesy of Gartner,Inc.)

As you can see from the Gartner Hype Cycle chart, digital twins are currently in the innovation trigger phase, meaning there is low adoption and their usage is currently primarily limited to research institutes and R&D departments of companies. The chart shows that the technology’s plateau of productivity (and widespread adoption) will be reached in 5 to 10 years. Gartner itself estimates that 1-5 percent of assets currently have digital twins.

According to a recent survey by Gartner, 48 percent of businesses currently using IoT technology are looking to implement a digital twin in 2018. Also according to Gartner, 50 percent of manufacturers with assets exceeding $5 billion will have at least one digital twin initiative launched for products or assets by the year 2020.

What does this all tell us? Well, it tells us that the digital twin is an emerging technology that has captured the imagination of manufacturers, but that it is not yet ready for the main stage. When it is ready for prime time, however, we can expect the charge to be led by high-end manufacturing companies that actually have the funds to invest in its development. The big guys will be taking the risks of early adoption, because they stand to be the ones to benefit from it the most.

Mirroring Complex Systems

So now that we know what a digital twin actually is, what stage of technical complexity are we currently achieving in terms of digital twin technology?

“Smart products are complex systems of systems, and require the simulation of multi-domain models and interpretation of sensors and edge devices to accurately predict performance,” said Tom Maurer, senior director of Strategy at Siemens PLM Software.  “Verifying the performance of Level 5 autonomous drive systems is—if not the most complex—it is one of the most critical applications of the digital twin today. Without digital simulation, it is estimated that it would take over 14 billion kilometers of physical testing to validate performance.”

That’s a lot of test driving, for sure. To give you some idea of perspective, that’s the equivalent of roughly driving to Pluto…and back.

For clarification, a Level 5 autonomous car is achieved when everyone in the vehicle becomes a passenger, and there is no need for human intervention or decision-making. It is, effectively, the holy grail of autonomous vehicle research and, naturally, due to the complexity of all the systems involved, can benefit greatly from digital twin technology.

As mentioned previously, the charge toward mass adoption of the digital twin will be led by the big manufacturing companies.

“While nearly all industries are adopting some form of digital twin technology, the biggest interest appears to be from industrial equipment (such as heating and cooling systems, fluid pumping systems, etc.) and the energy industry (such as wind turbines, oil and gas, etc.),” said Sameer Kher, director of Product Development at ANSYS.

“We see initial applications of digital twins in high value areas in order to justify deployment costs,” continued Kher. “High value areas could either represent expensive equipment such as wind turbines or high cost of downtime applications such as pumping systems in oil and gas. Over time, we expect digital twins to permeate to all products—even in the consumer space.”

So, maybe we can expect to see digital twins of electric vehicles, smartphones…even digital twins of our homes that incorporate security, plumbing, electrics and heating into one easily accessible system, constructed from various building information modeling (BIM) and product data. One of the key selling points of digital twin technology is that manufacturers can recycle product and manufacturing data into useful assets for use in the twin. This is where model-based design (MBD) comes into play.

“Our customers are reusing digital models and other development information throughout the product development, launch and support process,” said Maurer of Siemens PLM Software.“Reusable product information evolves from past program experience to 1D systems models that lead to optimized 3D models used in manufacturing planning and life-cycle support.”

Technological Challenges

As with any emerging technology, there will be hurdles to overcome before widespread adoption of digital twin technology is achieved.

“There remain several technological challenges to solve—these include challenges with sharing model IP, the need for faster and more accurate system-level representations of the detailed physics models, and connectivity/deployability (at scale) with existing IIoT platforms,” said Kher. “ANSYS’ solution is intended to be IIoT platform neutral. We are about to announce a new product that helps address most of these challenges by allowing customers to reuse a lot of existing IP to build accurate models. With prebuilt connectors to popular IIoT platforms and runtime model generation capabilities, we will make the connectivity and deployment of digital twins much easier for customers.”

These sentiments are in keeping with what was described in last month’s article on cloud-based simulation. IP security is a concern, especially when dealing with sensitive industries such as aerospace.   

Siemens has observed a different set of challenges.

“[There is a]need for a comprehensive digital twin supporting product, production and performance of smart products and manufacturing operations,” said Maurer. “Complex product development requires cross-domain solutions to recognize and understand the impact of the other domains—a true systems model. This information needs to be connected through a comprehensive digital thread connecting the virtual world of development with the insight real operational performance comparing the predictive analytics with the performance analytics to make informed decisions in both environments.”

Being an emerging technology that is constantly evolving, it’s quite likely that issues may arise that we haven’t even predicted.

“Most companies underestimate the complexity of managing digital twins over time. The digital twin configuration will be vastly different and constantly changing depending on the business model, target scenario and position in the supply chain.” said Marc Lind, Senior Vice President of  Strategy at Aras.

“In many cases businesses will end up having a wide variety of different Digital Twin configurations for different product lines and various ways of delivering customer value. The diversity of Digital Twin configurations within a single company means that data model flexibility, openness, and upgradability will be critical factors to success.”


The digital twin represents the perfect fusion of software, hardware and IoT-enabled feedback, and importantly allows users to apply 3D model data across the entire product lifecycle. Gone are the days of designing a part in CAD, only to throw that 3D model into an archive. Digital twins can transform previously dumb CAD models into dynamic and living system components at the heart of the twin.

The companies we spoke to have very different outlooks on the current challenges of digital twin development, as well as different views on how the future of the digital twin will evolve.

Some companies such as Siemens see the digital twin as being a purely cloud-based asset due to the elasticity and scalability of the available compute required to run a digital twin, while companies such as ANSYS see the future of the digital twin as being a hybrid of cloud and edge computing.

Aras are also seeing the future of cloud computing as a combination of the two.

“Digital Twin strategies include a number of different aspects from collecting time series data streams, conducting simulations, performing analytics, managing the digital twin configuration, and others.” said Lind. “At Aras we believe that digital twin initiatives will take advantage of both cloud-based resources and existing data center environments in hybrid scenarios.”

There are so many different opinions from companies on this topic that one thing is for sure: digital twins are going to evolve dramatically in the next few years, based on the divergence of solutions to the different challenges observed by each company.

And once the big guns have mastered the technology associated with building digital twins—and once a state of IoT agnosticism has been achieved—we can expect digital twin development to trickle down into consumer goods and reach a higher state of adoption.

If you’d like to read some more use cases or know more about the digital twin in the general context of Industry 4.0, there is an interesting report from Deloitte University Press at this link right here.