Introduction
Companies can begin the journey by starting with just one digital twin that has a data product at its core, evolving it over time to provide increasingly powerful predictive capabilities and the foundation of an enterprise metaverse. McKinsey & Co.
Relevance
Digital twin technology provides capabilities to enable new business models and decision support systems by incorporating data collected from multiple sources and analytic capabilities.
The concept has seen increasing interest in recent years in both academia and industry as indicated by the growth of publications, articles, and commercial marketing (VanDerHorn and Mahadevan 2021). Thus, digital twins are among the most promising digital technologies being developed at present.
Origins
The concept originates is nearly 15 years old and stems from the field of product lifecylce management Grieves (2014).
The development of the concept was motivated shift from the predominantly paper-based and manual product data to a digital model of the product.
Focusing on connecting physical with digital systems, the digital twin has similarities to concepts like Cyber Physical Systems (CPS, control of production systems) and Internet of Things (IoT, networking physical devices).
Definition
The concept is constantly evolving as it expands to new industries and use cases. This has already led to a variety of definitions that threaten to blur the concept and may lead to ineffective implementations of the technology.
VanDerHorn and Mahadevan (2021) propose a consolidated and generalized definition, with clearly established characteristics to distinguish what constitutes a digital twin and what does not.
A digital twin is “a virtual representation of a physical system (and its associated environment and processes) that is updated through the exchange of information between the physical and virtual systems.” VanDerHorn and Mahadevan (2021, 2)
The digital twin can be characterized by three primary components:
- A physical reality,
- a virtual representation, and
- interconnections that exchange information between the physical reality and virtual representation
Components
The physical reality that is to be represented by the digital twin can be decomposed into the physical system (the system of interest), the physical environment (where the physical system resides), and physical processes (the mechanism by which the entities of the system under interest undergo changes in state) (VanDerHorn and Mahadevan 2021, 3).
The virtual representation reflects the physical reality in an abstracted form in the virtual space. It can also be decomposed into the virtual system, the virtual environment, and the virtual processes. The abstractions of the physical reality often take the form of data models (data structures) and/or behavior models (mathematical or computational models). The virtual system, the virtual environment, and the virtual processes may contain multiple models of the physical counterparts at different levels of abstraction, which may or may not interact with each other (VanDerHorn and Mahadevan 2021, 4).
Virtual processes play an essential role in the digital twin because they allow future states to be simulated This includes diagnosis and prognosis, design review, simulation and optimization, as well as “what-if” scenarios and sensitivity analyses. Applying virtual processes to the virtual system changes the virtual system states in a similar way that physical processes change the states in a physical system. This allows the user of the digital twin to hypothesize how the physical system would behave under similar physical processes. Thus, decisions can be made about what actions to take based on whether the simulated results match the desired results (e.g., predictive maintenance) (VanDerHorn and Mahadevan 2021, 4).
The physical-to-virtual connection via a data interconnection enables the process by which data collected from the physical reality is used to update the states maintained in the virtual representation. In general. It encompasses three steps (VanDerHorn and Mahadevan 2021, 5):
- The process of collecting the relevant information including the direct measuring of the physical reality,
- interpreting the collected data to a form that is consistent with the level of abstraction and
- the updating process that uses the data to update the states of the virtual representation.
Systems of digital twins
Business value can be increased by interconnecting digital twins to complex ecosystems.
Such interconnected digital twins allow, e.g., to
- simulate the complex relationships among different entities and generate richer behavioral insights and
- incorporate dependencies and correlations into the models (by analyzing the interactions)
Example
Combining the digital twins of customers with the twins of retail and online shops would enable the creation of outstanding omnichannel experiences that seamlessly support the customer journey across all channels.
Implementation types
According to VanDerHorn and Mahadevan (2021), current industry implementations of digital twins can be generally grouped into three categories:
- Digital twin component solutions (e.g., provided by Microsoft),
- commercial off-the-shelf solutions (e.g., provided by OEMs like GE or Siemens), or
- custom hybrid combinations.
Examples
McKinsey and Co. outline following examples for digital twins:
- SoFi stadium (in California)
- To help optimize stadium management and operations, a digital twin aggregates multiple data sources, including information about the stadium’s structure and real-time footfall data.
- SpaceX
- A digital twin of the Dragon capsule enables operators to monitor and adjust trajectories, loads, and propulsion systems. The goal is to maximize reliability and safety during transport.
- Anheuser-Busch InBev
- A brewing and supply chain digital twin enables brewers to adjust inputs (for instance, “add more hops to mixer #3”) based on active conditions and can automatically compensate for production bottlenecks when, for example, vats are full. It also gives the company’s production engineers remote assistance and AR capabilities for live troubleshooting on how to fix pump leaks and other common issues.
Future application fields
The growing consensus is that digital twin implementation will expand in many areas including healthcare as promising field (Saracco 2019).
Digital twin applications in healthcare and education imply that the concept will be extended from artificial systems to people.
The availability of millions of digital twins of people would make inference possible (e.g., by applying AI) and help practitioners gain knowledge that could be used for precision medicine and proactive health care.
Technology challenges
Saracco (2019, 62) outlines following challenges related to the design and implementation of digital twins:
Interoperability of digital models, data interconnection, extended digital twins, and data privacy, availability, and ownership.
Interoperability of digital models: there are different ways to create digital models, which hampers portability of models within and across different areas.
Data interconnection: synchronization between the physical twin and its virtual representation requires pervasive and affordable communication channels (e.g., 5G or LoraWan).
Extended digital twins: when the concept evolves to go beyond mirroring and shadowing a physical counterpart (e.g., by taking action on behalf of its physical counterparts), twins need to be open to data exchanges beyond the twin relationship, while preserving their identities and the association with their physical doubles.
Data privacy, availability, and ownership: data architecture must ensure that data is secure, portable, and can be controlled by the owner.