You may have formed an impression of digital twins based on The Matrix movies or The Sims video game series. But advances in cloud infrastructure, edge computing, IoT, distributed data management platforms, and machine learning capabilities have transformed digital twins from science fiction to a more mainstream business capability.
Bringing digital twin capabilities to companies requires technologists to cross the chasm and better understand their business’s operations and operational technologies (OT). “CIOs and IT leaders need to understand that OT is a different world than IT and a perfect digital twin is the merger of both,” says Jens Beck, partner of data management and innovation at Syntax.
For a long time, businesses could afford a separation between OT and IT, but that’s no longer the case for manufacturers, construction, retail, and other businesses that must connect the physical and digital worlds. A digital twin is one conduit to enable this connection, which has operational benefits for optimizing production and improving quality. Sometimes, more importantly, it can drive strategic benefits when machine learning on real-world data is used to improve products, services, and business processes.
I spoke to experts from a variety of fields to identify seven preliminary steps for technologists new to digital twins who are contemplating developing one.
1. Research successful deployments
Before brainstorming opportunities and diving into any new technology area, I always advise people to research the companies, use cases, and benefits delivered by early adopters. For digital twins, there are many examples in manufacturing, construction, healthcare, and other areas, including the human brain itself.
Leaders in any emerging technology area look for stories to inspire adoption. Some should be inspirational and help illustrate the art of the possible, while others must be pragmatic and demonstrate business outcomes to entice supporters. If your business’s direct competitors have successfully deployed digital twins, highlighting their use cases often creates a sense of urgency.
2. Identify game-changing opportunities
Building a digital twin is expensive; for example, one group estimates the cost of developing a digital twin for a commercial office building at between $1.2 million and $1.7 million. So, before developing a digital twin, the team should document a product vision, consider the business rationale, and estimate the financial benefits.
Sometimes, a game-changing objective drives investment, and Abhijit Mazumder, CIO of TCS, shared an example. “In 2020, TCS collaborated with a local non-government organization to address the problem of emerging COVID-19 hotspots,” he shares. “An enterprise digital twin simulated processes and situations to model factors—virus characteristics, demographic heterogeneity, and mobility patterns—that influenced spread. The digital twin of the city served as an ‘in-silico’ experimentation to explore effective interventions without compromising public health and safety.”
3. Consider life-cycle management
There’s a time and expense involved in developing a digital twin, but there are also ongoing support costs to ensure models deliver accurate results. David Talby, CTO of John Snow Labs, shares three disciplines to embrace before experimenting with digital twins:
- Have a clear business use case—don’t just experiment with technology for its own sake.
- Make sure that the population of digital twins that you use to create your model, service, or simulation is representative of real-world people.
- Have an MLOps toolset in place to quickly and reliably move from developing to deploying a digital twin.
Talby's key recommendation is to consider elements of the full life cycle up-front, especially the functions to support machine learning models and instrument automated deployments.
4. Leverage system design tools
With a business case and life cycle designed, what tools should teams consider to begin their planning and experiments? Arjun Chandar, CEO at IndustrialML, suggests using CAD software or simulation tools as a "way to experiment with digital twins on the design engineering side [and] estimate the effects of physical environments on newly designed products.”
Here are some examples of system design tools used in specialized fields:
- Autodesk digital twins, used in construction, engineering, and architecture.
- Bentley infrastructure digital twins, used in areas such as cell towers and water systems.
- General Electric digital twins, used for equipment, networks, and manufacturing processes.
- Siemens digital twins, used for designing, developing, and manufacturing consumer goods.
- Bosch digital twins, used for smart buildings, including space management and predictive maintenance.
These are only a handful of examples, but the key lesson for technologists working on digital twins is to become familiar with the industrial platforms used by operational teams.
5. Define usage personas and opportunities
Whenever technologists embark on a technology program, it’s critical to identify the end-users and usage personas for the resulting platforms. IT leaders should define who benefits most from the digital twin, and very often, it’s people working in operations that are the primary benefactors.
“The digital twin’s main capability is to merge OT/IT data and to put those data sets into context via data analytics or AI/ML if required,” says Beck. “But its real power lies in enabling OT, such as the engineers, maintenance, and other technicians, to retrieve data points—since they fully understand them.”
Understanding the user personas is one step, and the next one is to identify what parts of their workflow and operations stand to benefit from a digital twin’s data collection, machine learning predictions, and scenario planning capabilities.
“On the manufacturing and operations sides, IT leaders can opt to model their physical production area to simulate product flow, or they can model the assembly or logistics steps for putting a new product together,” says Chandar. “All of these use cases can be scaled, and generative AI can supplement traditional finite element analysis to test new products virtually. Production setups can be digitalized and simulated for any new products before physically setting up factory lines, and digital representations of work processes can be developed for all products in a factory.”
6. Architect a scalable data platform
Digital twins generate petabytes of data or more that must be secured, analyzed, and used to maintain machine learning models. One critical architecture consideration is designing the data model and flows for collecting IoT real-time data streams and the data management architecture for the digital twin.
Harry Powell, head of industry solutions at TigerGraph, says, “When creating a digital twin of a moderately sized organization, you will need millions of data points and relationships. To query that data, it will require traversing or hopping across dozens of links to understand the relationships between thousands of objects.”
Many data management platforms support real-time analytics and large-scale machine learning models. But digital twins used to simulate the behavior across thousands or more entities, such as manufacturing components or smart buildings, will need a data model that enables querying on entities and their relationships
Powell continues, “Today, companies are creating digital twins using graph databases to support various operational analyses and glean actionable and timely business intelligence. The construction of a detailed digital model could be high-level, containing only the largest components of the business such as entire factories, warehouses, and supply lines, or it can be more granular, modeling individual machines in the factory, warehouse racks, and delivery trucks.”
7. Establish cloud and emerging tech competencies
Installing digital twin platforms, integrating data from thousands of IoT sensors, and establishing scalable data platforms all require IT to have a core competency in deploying technology infrastructure at scale. While IT teams consider use cases and experiment with digital twin platform capabilities, IT leaders must consider the cloud, infrastructure, integration, and devices required to support a production-ready digital twin.
Beck provides this recommendation on infrastructure: “To scale digital twins, IT chiefs will find themselves leaning towards the cloud while still having some technology at the edge, such as hyperscalers, IoT device management, and data science.”
Beyond infrastructure, Mazumder recommends developing competencies to support emerging devices and leveraging analytics. “Digital twin success starts with a strong digital core, enabled by cloud-native applications like AI/ML and AR/VR, and helping orgs process data and applications regardless of infrastructure,” he says.
Conclusion
Digital twins have enormous potential, but until recently, the scale and complexity were out of reach to many businesses without advanced technology capabilities. That’s no longer the case, and IT leaders who learn and partner with operations have an opportunity to bring digital twin capabilities to their organizations.