Steel production generates intense heat and consumes huge amounts of energy; transporting hot metal from the blast furnace to torpedo car to the steel plant requires the use of industrial ladles that handle large volumes of metal at temperatures up to 2700 degrees F. But the process is subject to so many variables—the equipment, its lining materials, the steel recipe—it is ripe for thermo process optimization.

Tata Steel Nederland (TSN) created a digital twin of its steel ladle operation as part of a broader effort to decarbonize; a TSN researcher described it on a webcast as a “complex system of different demands and choices to make.” For example, running the operation at slightly lower temperatures reduces wear on the ladle’s parts—potentially reducing production losses—but actually increases CO2 emissions.

“You can’t make the right decision with only one model of one installation,” says Paul Van Beurden, principal researcher. TSN now runs dozens, if not hundreds, of what-if models on its digital twin, getting answers in seconds about what adjustments would reduce emissions while keeping efficiency high. The thermo management application also uses AI to make predictions about equipment failure, among other things.

Digital twins are virtual representations of a physical item or an entire system, based on data captured and shared through sensors and Internet of Things (IoT) devices. An accurate digital twin enables rich 3D modeling and what-if analysis. With the addition of machine learning (ML) and advanced data analytics capabilities, it’s possible to quickly derive powerful insights about the system being modeled. The implications are huge. As demonstrated at TSN, digital twins can save serious money, improve product designs, elevate efficiency and productivity, and boost sustainability.

Until now, heavy industries that work with large assets and that use product lifecycle management (PLM) systems—including oil and gas extraction, aerospace, and automotive—have been leaders in digital twin adoption. But that has changed in recent years because the components that make digital twins possible and useful are now much less expensive, easier to use, and easier to access. Digital twins are now being used in retail, utilities, healthcare, and even smart cities.

When implemented effectively, digital twins can serve as catalysts for strategic decisions. They can provide visibility into an organization’s processes and ways to improve them, in turn strengthening customer experience and relationships. Digital twins provide a sandbox where innovations can be tested and refined before they’re launched into the real world. They afford businesses a cost- and time-efficient way to design smarter products and assets while capturing more information about them. They let companies make products better, faster, and safer. They also create new revenue opportunities, including service offerings that power as-a-service business models, which remove the burden of large capital outlays and lifetime maintenance from customers and keep them connected with service providers.

The making of a digital twin: Access sensors, collect data

The digital twin should not be confused with the digital model or simulation alone, which has the digital part but does not represent an existing physical object. The digital twin mirrors its physical doppelgänger in every way. This includes changing over time as it captures a constant, as-close-to-real-time-as-possible data stream from that object. This concept traces back to a presentation given in 2002 by product lifecycle management expert Michael Grieves. For product manufacturers, the earliest adopters of digital twins, the concept meant more efficient production, less waste, and better predictive maintenance.

To create a digital twin, you need to record the base information about the object and capture how it’s performing and being used. Sensors do that, whether in a machine in a factory or a tire on the road. The evolution of the Industrial Internet of Things (IIoT) has pushed digital twins even further, as IIoT-enabled machines and components collect and feed data in real time—meaning they’re self-reporting their own conditions. Steadily declining sensor costs over the last two decades have played a role in the increase of IIoT—and by extension, digital twins. (That trend seems to have reversed modestly since 2020, according to Electronics Sourcing—but anything less than sustained, dramatic price hikes would be unlikely to slow the continued rise of smart factories and digital twins.)

What is a digital thread, and how is it different from a digital twin?

A digital twin mirrors the current state of a piece of equipment, a system, or a facility, and it’s often used for real-time status monitoring and predictive maintenance or troubleshooting.

Digital threads, another concept arising from the product lifecycle management world, are similar but have some key differences.

A digital thread is a continuous, connected stream of information about the full lifecycle of an intelligent asset, from design through decommission. It’s commonly used to understand and improve the way the product is made, whether through changes to design, constituent parts and subsystems, maintenance plans, or operating procedures.

A digital thread might incorporate data from one or more digital twins, along with other data sources such as the equipment used to manufacture the asset, the operating context, and customer service records.

By providing a full, centralized record of changes over time, a digital thread can increase visibility and collaboration across departments, particularly engineering design and manufacturing.

As the cost and complexity of digital twins has fallen, adoption has spread beyond manufacturing to many different types of businesses. “Manufacturing companies can tap increasing opportunities to gain insights. The interest is in getting all this data off the factory floor and analyzed, not only to help understand what’s going on today but also to forecast what plant operators can do in the future to improve productivity and quality,” says C. V. Ramachandran, digital transformation and operations improvement expert at PA Consulting.

Opening doors to operational improvements

The Digital Twin program, a collaboration among six Dutch universities, an industrial consortium, and the Dutch government, aims to find ways to use digital twins to respond to companies’ biggest challenges.

Bayu Jayawardhana, project leader and professor at the University of Groningen, says the key objective of the program is for a digital twin to be used as a tool for decision-making, not just on a machine but also at an operational level. For example, maintenance and repair of assets is traditionally expensive because of labor and downtime; predictive maintenance, with data provided by a digital twin, is more efficient. “To do that means you need to have something to allow you to predict the behavior of your assets in the future,” he says.

The program also works with consumer goods company Philips, which has made electric razors for decades. The company wanted to improve the time-to-market and design process by using data from a line of smart shavers that users can customize based on their skin sensitivity. Sensors embedded in the shavers collect information on how they’re being used, which is then fed into a digital twin to advise the consumer about choosing and using a specific shaver.

Digital twins for factory workers

Digital twins can even be used for people. At Texas State University’s Ingram School of Engineering, Jesus Jimenez, an industrial engineering professor, is directing research into making digital twins of workers in a factory.

Jimenez says the idea is to look at not only productivity but also research and ergonomics of, for example, a machine operator or forklift driver. The research team uses sensors and motion capture systems with cameras that record the way a study subject moves while performing different actions commonly found in manufacturing, like lifting and reaching. Data on the physiology of the operator—heart rate, calories burned, respiration rate—are collected and put into the twin to help predict things like fatigue.

“You can see how to train your operators better, and you can also look at things that are providing feedback on when the human will be ready for a transition to a different job or a different schedule,” says Jimenez.

According to an article from the journal Sensors, “Digital twins, in this context, provide a granular view of worker health, well-being, and interaction with industrial systems, which enables [health and safety] professionals to anticipate potential threats, simulate various safety scenarios, and innovate evidence-backed strategies.” For instance, if a human digital twin can predict fatigue or high stress levels in a worker operating heavy machinery, immediate interventions, such as breaks or rotations, can be recommended to circumvent potential hazards.

Jimenez’s team is expanding their research to include cognitive functions to help better understand how humans in these contexts make decisions and errors, also factoring in cognitive decline. Here, too, sensors provide the data for workers’ digital twins. “There are lots of opportunities, like creating training tools for the operator,” Jimenez says. “We could do something like capture the highly skilled operators and save a copy of that person.” So, for example, when someone retires, all their years of knowledge and experience wouldn’t be lost; rather, their successful approaches to solving problems could be passed on to other employees.

Other use cases for digital twins

In addition to use on the factory floor, experts point to other possibilities for digital twins:

Logistics

Digitally enabling supply chains has the potential for creating more efficient, resilient, and sustainable global trade. By twinning products in transit and the transit process itself, real-time visibility into where a shipment is would support better decision-making if, say, an enormous ship got stuck in the bank of the Suez Canal. (Not that this would ever happen, of course.)

A digital twin could identify and solve many what-if scenarios to support better decision-making for highly complex supply chains. But the vision of creating digital twins of supply chains has a major obstacle to overcome: sharing intelligence. Companies often hold their data close, which gets in the way of efficiency. “The best solution we have found is to simulate the supply chain really well to the warehouse level and then, depending on how close we are with our supplier, to make them part of a digital twin,” says Ramachandran of PA Consulting.

Pharmaceuticals

Digital twins, along with AI and machine learning, can accelerate drug discovery and testing by first modeling drugs before going to clinical trials.

Charles Cooney, professor emeritus of chemical engineering at the Massachusetts Institute of Technology, believes digital twins will be a very powerful tool for improving the quality and lowering the costs involved in developing biotherapeutics by speeding up regulatory approvals. Cooney says these drug therapy products are the pharmaceutical industry’s fastest-growing segment because there is a high degree of specificity for the product to treat its target disease. Digital twins with robust mechanistic models, supported by process analytical technology and analytical methods for important quality attributes, will improve assessment of safety and efficacy and speed up the approval process, he explains.

“You can begin to see how the technology is helping you improve the process for the patient. And at the end of the day, the most important thing that we have to do is ensure the safety and efficacy of the patient experience,” says Cooney.

Oncology

According to an article in the journal Frontiers in Artificial Intelligence, Memorial Sloan Kettering Cancer Center is using digital twin technology to create a personalized virtual model that replicates a patient's tumor characteristics, genetics, and response to treatment. This digital twin is created by integrating comprehensive patient data, including imaging scans, genetic profiles, and clinical records.

Smart cities

Singapore has implemented a digital twin of its city-state. This ambitious project involves the integration of vast amounts of data from multiple sources, including transportation systems, utilities, and infrastructure networks. By fusing these diverse data sets, Singapore's digital twin offers a dynamic real-time representation of the city's functioning to improve traffic congestion and increase sustainability.