Industrial Digital Twins Scale Up in Scope, Complexity, and Applications

Simulation tools have long helped to speed product development by streamlining the prototyping process and reducing both cost and waste. It was only a matter of time before OEMs thought of applying such a useful concept to their manufacturing and logistics operations in the form of digital twins.

Like simulations, digital twins are virtual, computer-generated versions of physical systems, such as an industrial robot, a manufacturing process, the complex layout of an automated storage and retrieval system, or even the inputs and nodes of a global supply chain. What distinguishes twins from simulations, however, is that they incorporate real-time data from the physical systems and processes they emulate and often allow users to provide active feedback to those systems.

This active connection to real-world processes makes digital twins more of a virtual environment than a virtual thing. They do not simply provide a static model of some current state of affairs. Digital twins have the unique ability to model a system’s current design and future behavior to support a wide range of tasks, such as virtually commissioning systems, optimizing production, or enabling predictive maintenance.

“I was at a conference recently where somebody said, ‘Well, digital twins are just a new name for simulation;’ and I thought, ‘Nothing could be further from the truth,’” said Dale Tutt, vice president of Industry Strategy at Siemens Digital Industries Software. “Twins are more than the 3D geometry, the simulation, or the analytics. They tie all of those data points and functions together to provide a high-fidelity accurate model of a situation.”

The potential operational improvements that digital twins offer has put them on a launchpad for rapid adoption and growth. Research group MarketsandMarkets projects global demand for the technology will grow at a compound annual growth rate of 61.3% through 2028, expanding last year’s $10.1 billion market to $110.1 billion within that time frame.

Uneven adoption

The use of digital twins is expanding at different rates in different industrial sectors. One common pattern suggests that sectors where the Internet of Industrial Things (IIoT) took early root are fertile ground for digital twins. The automotive industry, for example, digitized its production floors years ago and is now expected to be the prime adopter of digital twin technology through 2028, according to MarketsandMarkets. This is not surprising given that the IIoT provided the sector with the abundant, rich, and real-time data on which digital twins rely.

But the link between the IIoT and digital twins might be more correlation than causation. “I think a lot of forward-thinking companies that see the benefit to digital technologies are more apt to adopt digital twins,” said Jeremy Wisdom, a robotic application engineer at ABB, which makes RobotStudio, an offline programming and simulation tool for robotic applications.

Tutt echoed the correlation. Aerospace and automotive industries both leveraged digitalization early on, but digital twins are also finding traction in entirely different sectors that are less familiar with the IIoT. “We’re starting to see naval shipyards digitize,” Tutt said. “And as energy and utilities undertake big infrastructure projects, they’re asking what they can learn from the manufacturing industry.”

Greenfield installations, such as battery gigafactories, are also implementing digital twins as part of a larger digitization strategy. Twins can provide data analytics to help optimize battery production and accelerate some of the aging processes. But they can also begin reaping benefits even before the factory floor is laid by virtually planning a new gigafactory’s infrastructure and commissioning equipment.

“In the past, companies built their manufacturing, tooling and assembly processes before digitalizing everything,” said Tutt. “Now, they are automating more of those processes. That has been a big change in manufacturing. Although engineering and manufacturing digitalization began in separate silos, today customers wanting to digitalize their processes will leverage digital twin technology to solve the most significant challenges they are facing.”

Twins help avoid common floorplan layout errors that necessitate costly factory redesigns. Avoiding such errors not only helps industrial engineers compress development cycles, it also reduces building cost and schedule overruns that plague almost every industry.

Companies in Siemens’ circle that are implementing twins to help make engineering and manufacturing decisions — such as planning or virtual commissioning — are seeing development times reduced anywhere between 25 and 30%, according to Tutt.

“Some are claiming up to 50% improvement,” he added. “But keep in mind that’s the first time they’re implementing the technology. Once you have that organic learning cycle, you might see a 20% reduction in development time for each program over time.”

Even a fraction of that improvement would be a dramatic result in terms of upfront development times and reduced overrun costs. But whether supporting a retrofit or new installation, it is important for companies to involve their process people earlier in the implementation of a digital twin, according to Doug Hixon, a robotics application specialist at ABB.

Right after a robot cell has been designed, for example, simulation tools such as RobotStudio allow engineers to program and simulate the robotic process virtually to work out any kinks before it goes live on the plant floor. “Otherwise, you won’t have your skilled people do that until you show up at the job site and you miss the opportunity,” Hixon said. “So, I think industry needs to look at things differently and say, ‘Hey, if we're going to invest in this, we need to bring in our skilled people at the beginning of the process.”

Models for modeling

Given the influence digital twins can have on strategic and operational decisions, it is natural to wonder how they are built to accurately and repeatably model the complexity of an active physical system incorporating components, subassemblies, and systems from dozens of different vendors.

Digital twins were developed, in part, to make complex systems more accessible, intuitive, and manageable. The technology would not have made it this far if it did not facilitate integration of digital data from disparate vendors. But it is also in the interest of industrial equipment vendors to make their product CAD-data compatible with digital twins if they wish to compete in an increasingly digitized marketplace, according to Hixon.

“From the perspective of ABB and the general industrial segment, we provide a CAD model of the complete cell with the robot, with the position or, if need be, any additional equipment as part of our engineering procedures,” Hixon said. “So that data is there and available to our customers.”

Software such as the Siemens Xcelerator platform and its Teamcenter product lifecycle management software, provide the sort of digital environment where multi-CAD systems can be virtually assembled. Such software is designed to pull in CAD and other data for different robot cells, SCADA systems, and other components to create a virtual production floor, for example, or a complex product. “I once saw a picture of an aircraft that had maybe five different CAD systems comprising the model, and the engineers were able to bring it all into Teamcenter and visualize a digital mock-up of everything,” Tutt said.

The fidelity with which that mock-up models its physical counterpart depends, in part, on the business need and what digital twin users aim to accomplish. When modeling a new automotive system, for example, the early models might have lower fidelity because they are in a more conceptual phase. As more data is gathered about the system over time, the models mature and get more detailed.

Alternatively, as twins themselves get more sophisticated, users can more easily optimize the virtual model first and then calibrate the physical system against it.

“When we go about twinning a [robotic] system, what tends to happen is people make adjustments to the digital side,” Hixon said. “The rule of thumb is that your digital model should be the ideal because you can make lateral adjustments to it and then the physical model should have any calibration changes you make for the end-of-arm tool and other elements. Once you absolutely calibrate something like our arm, that calibration holds true unless you replace a physical component of the arm.”

Present and future

The future capabilities of digital twins are constrained only by our ability to parse data and, of course, by the physical twins that define them. But it is clear that industry has only begun to explore the full potential of digital twins. There are many vectors along which twins can further evolve.

Being digital, twins are by definition modular and scalable. The twin of a robot cell at the front end of a production line can be virtually synched with other robots on the line — as well as the line itself and so on until the twin might conceivably encompass all production operations across a global enterprise.

In addition to integrating more components and systems into the stack, twins are also incorporating more technological domains into the virtual model. During this January’s Consumer Electronics Show in Las Vegas, Siemens and Sony announced a collaboration to develop a virtual reality (VR) headset for use in manufacturing applications, which pose very different user and environmental demands than consumer VR gear.

Much of that device’s design will advance virtually based on a digital twin combining the headset’s mechanicals, computational hardware, electronic subassemblies, displays, power supply, and other systems.

“The state of the art is getting to where we’re able to model this multidomain environment and really optimize the headset for the customer,” said Tutt, who added that the benefits of that capability could translate to other multidomain products such as cars or airplanes.

Coincidentally, the adoption of virtual and augmented reality (AR) on the manufacturing floor would open another path on which twins could evolve, according to ABB’s Hixon. Currently, most users view and interact with digital twins using a computer display or tablet. But AR/VR headsets provide a more immersive interface with twin data. Proposed plant-floor layouts, the motion path of a robot arm, or visualizations of processes buried within a physical machine can all be superimposed on a physical thing or environment via an AR display to allow users to view and interact with the virtual and physical models together in real time.

Such abilities would revolutionize planning, operational maintenance, remote troubleshooting, and training. “We can bring in robot teach pendants and other pieces that are part of peoples’ day-to-day manufacturing or maintenance operation so they can interact with digital twins when performing tasks on their specific equipment,” Hixon said.

Conclusion

Any OEM, integrator, or line builder can recognize the benefit of reducing the time and cost to develop, integrate, and commission industrial systems. Digital twins achieve these goals and also reduce risk by providing a digital environment in which different technologies and systems can be combined, tested, and validated before they are even installed.

With broader deployments and more investment, digital twins will see increasing innovation and sophistication. Additionally, new and emerging technologies such as AR/VR as well as artificial intelligence and edge computing will further expand the capabilities and accessibility of digital twins in manufacturing environments.

“The models are getting strong enough now that people are able to use them out of the box more and more without needing to make as many adjustments,” Tutt said. “And I think that's also helping to improve the fidelity and the quality of the modeling that they're doing.”