Digital Twin Consortium is leading an effort to harmonize the understanding of a digital twin, capture best practices, and propel the innovation of digital twin technology. On December 3, 2020, the Consortium released our official definition of a digital twin. The process involved a cross-section of domain specialists, an extensive review of documentation on the subject, and a stress test of the definition based on digital twin use cases. The result is a definition that leverages established knowledge on the topic, is consensus-driven, and validated against practical considerations.
Digital Twin Consortium Definition
Below, please find the Consortium’s digital twin definition:
A digital twin is a virtual representation of real-world entities and processes, synchronized at a specified frequency and fidelity.
- Digital twin systems transform business by accelerating holistic understanding, optimal decision-making, and effective action.
- Digital twins use real-time and historical data to represent the past and present and simulate predicted futures.
- Digital twins are motivated by outcomes, tailored to use cases, powered by integration, built on data, guided by domain knowledge, and implemented in IT/OT systems.
The foundational elements of the definition are captured in the first sentence: the virtual representation, the real-world entities and processes it represents, and the mechanism by which the virtual and real-world entities are synchronized.
In the Consortium’s definition, virtual representation is a set of correlated digital models and supporting data (e.g. metadata, master data, reference data, etc.), which provide cohesive information about their subject matter.
To avoid confusion or any narrow connotations related to the term models, we looked to Miriam Webster for guidance. One of their definitions of the term model is “to produce a representation or simulation.” They define simulation as “the imitative representation of the functioning of one system or process by means of the functioning of another.”
We divided digital models into two categories:
- A representational model which consists of structured information which generally represents the states of entities or processes.
- A computational simulation model which is an executable model of a process and consists of data and algorithms that input and output representational models.
Digital models can be 3D meshes, 3D CAD Models, mathematical equations, databases, physics-based, AI-based, and even Excel spreadsheets. All these things can serve as digital representations or computational simulations related to the historical, present, and potential future states of real-world entities and processes – the subject matter of a digital twin.
The Real World
We chose entity over alternatives such as asset (not flexible enough), object (software connotations), thing (general aesthetics), and several other synonyms because our members needed a term that could work as effectively for an oil well as it could for a blood cell.
Miriam Webster’s definition includes the phrases independent, separate, self-contained existence, and “something that has separate and distinct existence and objective or conceptual reality.” We intentionally – and perhaps ironically - do not mean entity in the digital sense of the word, (i.e. a data record with a unique identifier).
The inclusion of process was a matter of some debate, but in the spirit of flexibility we decided it was important to include it. Entities connect and interact to form systems that implement processes, which happen in time. These processes can be productive (the delivery of perishable goods to a grocery store) as well as non-productive (the spoiling of those goods in transit without proper handling).
What is important to note is that the entities can refer not just to the assets of interest (e.g. perishable goods) but also the physical environment with which they interact (e.g. the temperature in the grocery store loading bay).
This leads into the third foundational element: the synchronization of the virtual and real-world that makes digital twin use cases possible.
This word choice was also the subject of intense debate, and we considered many alternatives from existing digital twin literature. In the end, candidates such as connected (too IoT-centric), mirrored (too abstract), and twinned (too circular), did not quite capture both the intent and the flexibility we needed from the formal definition. We will happily use them in supporting documentation in a narrower context.
In our definition to synchronize is to cause the virtual representation to match more-closely the real-world, and/or cause the real-world to match the virtual representation of a desired state more closely. All digital twins must have synchronization mechanisms that can mirror the real world in the virtualization through observation (i.e. the water is rising), and/or mirror the virtualization in the real world through intervention (i.e. open the floodgates).
Not only can these mechanisms be unidirectional or bi-directional, but they can also be multi-faceted. Observation mechanisms could include sensors, laser scans, satellite imaging, radar, videos, and intervention mechanisms include actuators and increasingly robots. But people have not been entirely automated out of this equation. The most common mechanisms still feature a human-in-the-loop.
Depending on how rapidly the real-world entities change and the means of real-to-virtual synchronization, the frequency of synchronization might vary (i.e. real time, daily, milestone, etc.). For digital twins that rely on multiple synchronization mechanisms, the frequency might vary by mechanism (i.e. sensors and laser scans). There may also be an observational synchronization frequency and an interventional synchronization frequency, which may vary per-entity as well.
Fidelity characterizes the degree of precision and accuracy applied to the virtual representation, as well as to its synchronization mechanisms. The fidelity only needs to be sufficient for the desired use-cases. For example, a patient using a mobile device for wayfinding in a hospital would not need reality modeled at a molecular level (although simulating the impact of a virus on their blood cells would). Regardless of the level of fidelity specified, it can only be trusted if it can be validated.
While there are many existing digital twin definitions – some written by our members – this one addresses the Consortium’s unique requirements: it is flexible enough to accommodate the practical needs of our members and can be applied as effectively to a smart city as a satellite. Most importantly, the Consortium’s digital twin definition is extensible and will serve as semantic scaffolding for future documentation we release related to taxonomies, technology platform stacks, security, case studies and more.
Please register for this IoT Solutions World Congress webinar featuring the authors of the Digital Twin Consortium definition in a panel discussion. For more information about the digital twin definition, click here.
About the Author
Sean Olcott, Technical Director, Gafcon (left) and Casey Mullen, Distinguished Architect, Information Architecture for Federated Digital Twins, Bentley Systems (right)