This expression refers to a virtual model of a process, product or service and is sometimes described as a bridge between the physical and digital world. Data from sensors are streamed from physical equipment to the virtual representation to create a continuous loop.
"In the past, you would have built a prototype of a car body, and crashed it and deformed it, put it under stress, in order to find out whether your design met specifications," explains Jan Mrosik, chief executive of German engineering group Siemens' digital factory division.
"With the digital twin we take this CAD [computer-aided design] data and simulate crash properties and deformation — everything around the mechanical part of this product."
This can make the design process quicker and less expensive, by reducing the number of physical prototypes needed and rooting out flaws or design inefficiencies.
The concept of a digital twin extends to factory floors, to enable simulation of machinery before anything is built. In the past, production lines had to be switched off to prepare for a new product being made, so downtime can be avoided.
Data from real-world finished products and manufacturing plants can be fed into the digital twins in order to tweak designs and improve performance.
There are, however, limitations to widespread adoption, say industry analysts. While digital twins are attractive for the design of mass-manufactured and expensive products, it can be harder to scale the technology for manufacturing facilities, because of their different configurations and use of equipment from different suppliers.