The term “digital twin” pops up more and more as companies look for ways to drive innovation, and for good reason. Twins can help businesses monitor the state of equipment at remote sites, optimize product designs for specific needs, and even model the business itself. But many of these conversations skip over a key question. What is a Digital Twin?
The core concept of Digital Twin is a digital model of a physical object, system or process that exists (or could exist) in the physical world. Businesses can use Twins to monitor, analyze, and simulate their physical counterparts. At Accenture Labs, we put an emphasis on Twins with sophisticated representation and rich enough data to support both advanced analytics and intelligent reasoning.
A Digital Twin starts with data about the physical world. That can include data from design (such as CAD models), production (manufacturing data, parts manifests) and support of the item being Twinned (asset maintenance logs, IoT readings, and operations manuals). This data feeds into advanced analytics and AI models to derive insights. Paired with engineering simulations, it can drive “then-what” and “what-if” questions that would be expensive, slow, or impossible to pose with the physical product: Would this part hold up to repeated heavy use in a sub-zero environment? Would this device’s performance change if a certain part was made with a different metal?
Twins can also be used to look “backward,” answering questions like, "How often was the physical object used outside of the recommended conditions?” or “Under what conditions does it perform best?”
We then use the term Digital Thread to refer to the processes and technologies that maintain the connection between a Digital Twin and its physical counterpart. The Digital Thread brings together enterprise and IoT data, analytics and machine learning, and IT integrations to keep the physical and digital worlds in sync.
Our definition of a Digital Thread includes domain-specific knowledge related to processes, costs, and regulatory compliance. This helps to define how data is incorporated into the Twin, how that data stays current, and how data from the Twin can be applied. For an example, imagine a scenario from the oil and gas industry, where we have a Digital Twin of an oil well. A Digital Thread could include knowledge that dictates how that Twin may be treated differently depending on whether the Twin is used in design, operations, or decommissioning.
A well-designed and well-implemented Digital Thread can produce a rich, robust, and useful set of Digital Twins.
Different Twins for Different Purposes
These core definitions of Digital Twin and Digital Thread have broad applications. Ultimately, though, the structure of Digital Twins varies depending on the thing being twinned, and the operations and analyses that the twin has to support.
A twin of an automobile seat model may include the component list; that in turn points to the Digital Twin of each component, including its structure, features, assembly process, and CAD/CAM models. Such a Twin could be used to analyze the seat’s structural properties, and to help the manufacturer respond to (or prepare for) changes in their supply chain.
If a supplier changes the specs on a component part—such as reducing a load capacity to save cost—the digital twin can provide quick insight into which seat models are impacted and may need to be re-tested to make sure they meet their load requirements. A more sophisticated digital twin could also support simulations that predict which models are likely to fail the tests. Someday we may have digital twins with such high fidelity that testing can be done entirely in simulation.
Twinning an Instance
So far, we’ve described Twins that represent the design of a particular product. In the case of the car seat, the Twin from above represents the design that will be used for every physical “copy” made of that seat. It may be quite complex, but it’s also fairly static: It changes only when the underlying design changes.
But we can also create Twins for different instances of that seat design–a specific seat, used in a specific vehicle. Sensor data plays a bigger role in this Twin: The data looks less like specs, and more like streams. How much has the seat been sat in? Who has been sitting in it? What configurations does each driver use, and when do they change them? How much does the user squirm? How hot does the seat get?
A twin that models this usage can be used for many purposes, including knowing when this particular seat is wearing out, automatically determining when the seat configuration should be changed in response to real-time conditions, or determining the need for design changes in future releases of the product based on usage patterns.
With the wealth of real-world, real-time data available today via IoT, and the maturation of AI systems that can derive insight from that data, businesses can use Digital Twins to powerful advantage. Stay tuned to find out how Accenture Labs is building ever more sophisticated Twins to deliver value for our clients.