February 4, 2025
In my reading, I’ve come across various terms and definitions for different types of current and future digital twins. These have expanded my understanding of the breadth of use cases for digital twins in enterprise now and in the future of work. Here’s what I found:
Introduction
Digital twins can actually be traced back to the 1960s when NASA used them to study and simulate real spacecraft. Though the concept isn’t new, companies are finally starting to realize the promise of digital twins. Advancements in AI, IoT, cloud computing, and more are helping to make digital twins more useful, powerful, and accessible to nearly every industry.
According to McKinsey Digital, 70% of C-suite technology executives at large enterprises are exploring and investing in digital twins. For manufacturers in particular, some of the earliest adopters, digital twins can be seen as an evolution of traditional product lifecycle management tools. No matter the industry or asset - there are digital twins of simple machine components up to the level of entire organizations and cities - the goal of digital twins has always been to better understand the physical world.
What is it?
You often hear about the convergence of IT (information technology) and OT (operational technology) in the form of digital twins. Definitions vary but most agree that a digital twin is a virtual replica or digital model of a physical asset, product, process, or system.
What distinguishes a digital twin from a mere 3D representation/visualization or standalone simulation is the constant flow of information between the twin and its real-world counterpart. A digital twin is regularly updated with real-time data from the physical asset or system it represents, enabling new levels of simulation, monitoring, prediction, and experimentation.
In other words, a digital twin evolves or matures in tandem with its physical counterpart. It accurately reflects the condition and mimics the behavior of the real-world asset or system at any given moment and can be subjected to multiple “what if” scenarios. The results of these simulations are used to inform decisions about the real world.
Recipe for a digital twin
Many sources break down digital twins into layers of data and technologies, beginning with data. To simplify the concept even further, here’s a sort of recipe for a digital twin:
Data: Digital twins integrate data from various sources, including historical data (e.g. maintenance records) and real-time (IoT) sensor data (e.g. temperature), as well as operational data (e.g. information from ERP systems), design and engineering data (e.g. CAD files), behavioral data (e.g. traffic flow), contextual data sources (e.g. market trends), unstructured data (e.g. operator notes), and more.
Intelligence: Simulation software, modeling (physics-based and statistical), advanced analytics, and AI/machine learning algorithms which make sense of the deluge of data and allow us to apply the digital twin for prediction, decision making, and other practical purposes.
Visualization: 3D modeling and related technologies like augmented and virtual reality which allow us to view and interact with the digital twin.
Why?
Until recently, digital twins couldn’t do much more than represent an asset. Digital twins were bespoke (i.e. expensive), informed mainly by data from sensors on the physical asset, and used for basic monitoring.
As stated above, advances in computing power, AI/ML, etc. have unlocked greater possibilities. Today’s digital twins can synthesize data from IT, OT, and other disparate systems to replicate more complex objects and systems; and, in addition to monitoring real-world assets, can accurately simulate potential outcomes in order to predict, plan, and optimize.
The value lies in the ability to monitor and simulate changes to the real-world asset or process without physical intervention. Virtual output then becomes real input, and the insights gleaned from the digital twin can be applied to its physical counterpart.
Types of digital twins
Like definitions, classifications for digital twins vary and often overlap. Labels may refer to what the digital twin is replicating (e.g. product digital twin) or indicate a use case (e.g. operational digital twin).
Asset twin: Digital replica of a single physical asset like a machine or piece of equipment (e.g. engine, wind turbine). Sometimes referred to as a “status twin” or “product twin” The focus is on the condition and performance of the asset (as opposed to how it interacts with processes, people, etc.). Use cases include performance optimization and predictive maintenance.
You may also see component or parts twins: Digital replicas of individual components or parts within a larger asset (e.g. switch or valve). Of course, a piece of equipment that’s part of a larger system might also be considered a component. (As I mentioned, the term used often comes down to the use case.)
Some distinguish between asset twins and product twins:
Product twin: Digital replica of a physical product (e.g. vehicle or building), providing a holistic view of the product’s performance across its lifecycle. Use cases include design optimization and lifecycle management. (If you think of an asset as something that’s operational, then a product twin might evolve into an asset twin once it becomes operational.)
Process twin: Digital model of a specific process or workflow (e.g. assembly line, supply chain) encompassing all the steps involved, how they’re executed, and how materials, information, etc. flow through the steps. Use cases include identifying bottlenecks and testing process changes without disrupting physical operations.
System twin: Digital model of an entire (complex) system (e.g. factory, power grid) providing a holistic view of the interactions and dependencies between the various parts and processes that make up the whole. Specific use cases include optimizing energy distribution and predicting system-wide disruptions.
Most sources consider system twins to be the most comprehensive.
Additional terms I’ve come across
Performance, operational, and environmental twins are designed for specific use cases.
Performance twin: Models the performance and health of an asset or system over time (i.e. uses historical and real-time data) to ensure efficiency, predict maintenance, and optimize throughput.
Operational twin: Models the current state of a process or system to provide actionable insights for day-to-day operations (can help identify bottlenecks, make real-time adjustments in response to changing conditions, etc.)
Environmental twin: Models external environmental factors to understand and predict their impact on an asset or system in order to, for instance, ensure regulatory compliance or support sustainable design.
Comprehensive twin: Integrates multiple data sources and types of digital twins (performance, operational, and environmental twins) to create a holistic view of an asset or system across its lifecycle including all interdependencies (e.g. smart grid, aircraft fleet). Useful for monitoring, optimization, and decision making.
Hybrid twin: A digital twin created using both data-driven and physics-based models (Ansys). Alternatively, a digital twin that combines different modeling techniques and data types to achieve greater accuracy and insights.
Digital twin of an organization: According to Gartner, a DTO is a digital representation of an entire organization’s structure, including people, processes, systems, and data. It provides insights into complex operations to help “build resilience, manage disruptions, and sustain performance even in volatile market conditions.” Use cases include scenario planning and project prioritization.
Human digital twin: Digital twins of people? Imagine a doctor using a digital working model of your body, one that mirrors your current vitals, for more accurate diagnoses or personalized medical procedures. Marketers might use digital representations of specific customer types to conduct research. One article even floated a vision of the future in which every publicly listed company has its own digital twin to interact with investors.
Digital thread
That brings us to the concept of the digital thread. In most definitions, digital thread refers to a framework for integrating data from every stage of a product’s lifecycle - from design and manufacturing to maintenance and retirement - to provide a single source of truth for decision making and collaboration across departments and functions.
Oftentimes discussed in terms of product development, the purpose of a digital thread is to create consistency and transparency in the name of communication and traceability and to uncover insights from traditionally siloed or underutilized data to improve existing and future products.
A digital thread might also focus on the lifecycle of a system to, for instance, track how upgrades in one subsystem affect overall system performance. Whereas a digital thread of a product aims to improve traceability and quality, a digital thread of a system aids in managing and optimizing complex systems. Both maintain data continuity, ensure traceability, and enable communication across the value chain.
As for how a digital thread differs from a comprehensive digital twin, it would seem digital threads aren’t used for simulations and may or may not include live data. You might think of a digital thread as an advanced lifecycle management tool that serves as a single source of truth for all stakeholders. A comprehensive digital twin, on the other hand, focuses more on the current state and future projections of/for an asset or system.
Future-looking
It’s important to note that today’s digital twins are still largely applied in silos. The proliferation of digital twins, the accuracy of digital twins, and the dream of self-configuring, autonomous digital twins are limited by challenges like data quality and the reliability of AI/ML models.
Digital twins are still maturing. At the end of the day, it doesn’t really matter if you know the difference between a system digital twin and a comprehensive one. However you classify it, a digital twin comes down to the use case. What’s the purpose of creating a digital twin? That will dictate the data sources and algorithms underlying the twin and the kinds of simulations you’ll conduct with it.
