Skip to main content

Digital twin definition

Digital twins, models, shadows, simulations, and control systems are frequently used interchangeably, leading to confusion in procurement, implementation, and expectation setting. The purpose of this document is to clearly define these distinct concepts to ensure that buyers and suppliers have a shared understanding of what is being requested, delivered, and expected.

By establishing precise definitions, organisations can make more informed decisions when specifying requirements, assessing capabilities, and ensuring that delivered solutions align with their intended use cases. This clarity reduces ambiguity, improves procurement efficiency, and helps both technical and non-technical stakeholders understand the fundamental differences between these digital representations.

The National Digital Twin Programme is developing the tools, technologies, and operating frameworks necessary to enable interoperability, scalability, and adoption across these interconnected concepts. While Digital Twins remain a primary focus, the programme also addresses the requirements of digital models, digital shadows, simulations, and control systems to ensure a comprehensive and integrated approach to digital transformation.

Full Definition

A digital twin:

  • is a digital representation of a physical-world entity or environment or a process that includes a two-way data[1] flow into and out of the physical world in a timeframe that is appropriate for the required decisions and assumptions;
  • replicates its physical-world counterpart to a known tolerance, performs without statistical bias[2], and is able to predict what the associated physical-world entity would do given a stimulus; and
  • provides a level of prediction confidence within specified validation envelope and an associated assumption set, based on physical parameters and specifications, including after integration with other digital twins.

Understanding key Terms in the Digital Twin definition

The bullet points above define the essential characteristics a system must meet to qualify as a Digital Twin. To provide clarity, each key term has been broken down below. Click on any term to expand and see its detailed explanation.

1

Be associated with a physical-world entity or environment or a process that includes a two-way data.

In order to generate digital twins of physical entities, environments or a process, there will likely be a “template” for each type of physical-world counterpart to be twinned. This template must allow for a two-way data flow into and out of the physical world. When such a two-way flow of data is established, a template becomes a digital twin of the specific instance of the physical entity or environment or the process from which it receives, and to which it sends, data. For example, a specific car type may have a template used to create a digital twin, but every car manufactured will have a unique digital twin instance associated with it.
2

A timeframe that is appropriate for the required decisions and assumptions.

A digital twin must run in a timeframe that allows it to be run synchronously with its physical counterpart. This can be achieved if the two-way communication flow into, and out of, the physical world is appropriate for the required decisions and assumptions – in other words in a timeframe that allows the digital twin to be updated so it remains sufficiently aligned with the physical entity, environment or process. For example, if a manufacturing plant runs on a twenty-four-hour schedule with eight non-production hours, the twin must be capable of updating to current status within that twenty-four-hour window. Equally, an aircraft engine digital twin must be capable of running to the current status of the physical entity after landing and before the status is next updated. A digital twin may have the capability of incorporating the outputs of a simulation that may be run in faster-than-real-time, however, this capability is not a requirement of a digital twin.
3

Replicates its physical-world counterpart to a known tolerance, performs without statistical bias.

A digital twin must reliably replicate its physical-world counterpart in every way known to be relevant to the use case it has been designed for. A statistical, data or simulation-based model can be a digital twin only if it has been trained specifically on a physics-based (or theory-based) representation of the physical entity with a known and unbiased level of uncertainty, or if it has been trained using the physical-world data provided by the specific physical counterpart. A physics-based model can only be a digital twin if it is associated with a known tolerance and fidelity.
4

Have a specified validation envelope.

A digital twin must only replicate its physical-world counterpart within known validation windows – the operational limits within which it accurately represents and influences reality. For instance, if the twin is a physics-based model trained on data within a known temperature range there can be confidence within that temperature range. Outside of that range there will be a point at which the physics and data changes, and the digital twin will no longer be valid. To meet the criteria of a digital twin the validation envelope over all relevant parameters must be specified and associated with the digital twin and taken into consideration in “what-if” questions. If the physical world counterpart unexpectedly operates outside of the validation range, an immediate re-assessment of the digital twin must be undertaken.
5

Have an associated assumption set.

Even within a validation envelope, the digital twin will only reflect specific characteristics of the physical entity. It may reflect size, shape, weight, motion and temperature for instance, but not model the effect of an external force. The assumption set specifies explicitly what the digital twin accounts for and what it deliberately excludes. Consider a digital twin of a car brake system: the digital twin could be classed as such without the ability to model the change to the physical system resulting from a collision with another car, providing that this is captured in the linked assumption set.
6

Be based on physical parameters and specifications.

A digital twin must represent physical world limitations and specifications within its validation envelope. Where a digital twin has been developed for a new design of a physical entity, it cannot contain components or properties that cannot be replicated in the physical world. For example, a digital twin must not contain models of metals with properties that cannot be produced in reality. In order to analyse the impact of idealised properties, a digital twin would need to be used with a separate model or simulation.
These constraints prevent misapplication of a digital twins in scenarios they were not designed to handle, and ensures they can be used for test and evaluation and safety case development. Further, these requirements, when combined with common set of data exchange protocols, ensures digital twins are fully composable with other digital twins.

Digital Twin States

A digital twin can be in one of the two following states at a point in time:

  • Connected – if it is currently exchanging data into and out of the physical world in a timeframe that is appropriate for the required decisions and assumptions.
  • Semi-connected – if it utilises synthetic data but retains at least one physical world data feed. Consider a digital twin that is dependent on both ambient temperature and external vibration. If the current ambient temperature was still drawn from the physical world, but the vibration level was simulated using a model to understand the impact of stress on the physical world object, this would be deemed “semi-connected”.

A digital twin may move between these states and remain a digital twin, provided its current state is recorded. If a digital twin disconnects it can only return to being a digital twin if the reconnection occurs within a timeframe in which it will still accurately replicate its physical world counterpart.

Common Application Areas

Digital twins of assets and capabilities

This area encapsulates component, platform,  system and manufacturing (factory) digital twins. The existence of digital twins of individual components within different industry partners which need to be integrated at the platform level poses a significant challenge, but is proven technology in industry. However, even in this most well-known sector there are complications driven by the multi-fidelity requirement of digital twins and the need for assured outputs in an appropriate timeframe.

Digital twins of environments

This area encapsulates infrastructure, transport, manufacturing, environment and mega-structure digital twins. A mega-structure is defined as a complex amalgamation (or ecosystem) of both component, platform, infrastructure and environment twins such as would be found in an urban environment.   This requires complex multi-fidelity twins with real time inputs to ensure accurate representation of an environment and the ability to output to infrastructure in that environment.  This is currently unproven technology with substantial open research and technological challenges.

Digital twins in biology and medicine

In this area, efforts are focussed on digital twins of aspects of a human, such as a human heart. Digital twins of human organs or body structure give huge benefits for medicine and equipment design, however, even at this level, linking such a physical digital twin to a modelling framework that could output the forces exerted by a body on surroundings or clothing is not yet possible.

Clarifying Digital Twins, Digital Shadows, Models, Simulations and Control Systems

Not all digital representations qualify as digital twins. To ensure clarity, it is important to distinguish between digital twins, digital shadows, and standalone models or simulations.

  • Digital Model – a digital representation that is not dynamically linked to a physical entity, environment or a process.
  • Digital Shadow – a representation which is updated to reflect changes in the physical world by a one-way flow of data —from the physical to the digital—but which is unable to directly control or influence its physical world counterpart.
  • Simulation – a representation of a process or system that exists or could exist in the physical world. Simulations are based on digital models and execute those models over time to study behaviour under various conditions. They are used for predictive analysis, testing, and exploring hypothetical scenarios, but unlike a digital shadow, they are not required to maintain a live connection with a physical counterpart.
  • Control System – a system that manages and regulates the behaviour of other devices or processes by applying control actions based on predefined rules. Control systems operate in two main forms:
    • Open-loop control – where the control action is applied without feedback from the system’s output.
    • Closed-loop control – where the control action adjusts based on feedback from the system’s output.

A Control System differs from a Digital Twin in that a Digital Twin is capable of undertaking predictions, whereas a Control System cannot.

Example:
A Car Suspension in
Development and Operation

Concept Example Purpose
Digital Model A 3D computer-aided design (CAD) model of the suspension system, including material properties and structural design. Development – provides a structured representation for design and engineering.
Simulation Running finite element analysis (FEA) on the digital model to predict how the suspension reacts under different forces and stresses. Testing – analyses performance before building a physical prototype.
Digital Shadow Real-time sensor data from the suspension system (e.g., shock absorber pressure, wheel displacement) is sent to a dashboard but does not influence the car. Monitoring – reflects live performance data without sending feedback to the physical system.
Control System The car’s active suspension adjusts stiffness based on driving conditions. In open-loop mode, it follows preset settings; in closed-loop mode, it adapts based on sensor feedback. Regulation and management – ensures real-time control of the suspension system.
Digital Twin The suspension’s digital twin analyses upcoming road conditions using real-time and historical data. If the car is approaching a pothole or rough terrain, the digital twin predicts the impact and adjusts the suspension stiffness in advance to improve handling and comfort. Optimisation and decision-making – enhances performance by anticipating conditions and adjusting proactively.

References

[1] In some fields, the term ‘data’ may have different interpretations. Here, ‘data’ should be understood in its broadest sense, referring to representations of facts, observations, or concepts. In a broader conceptual framework, ‘information’ refers to data that has been processed or contextualised to provide meaning, while ‘knowledge’ represents structured understanding derived from information through experience, reasoning, or analysis. For simplicity, when we use ‘data’ in this document, we mean all three—data, information, and knowledge—unless otherwise specified.

[2] Here, bias is a statistical term. Without bias means that an estimator, when used repeatedly, does not systematically overestimate or underestimate the true value of what it is measuring. In statistical terms, an estimator is unbiased if its expected value equals the actual value of the parameter it is estimating. This ensures that any errors are due to random variation rather than a systematic skew in one direction.

Full definition
Understanding key Terms in the Digital Twin definition
Digital Twin States
Common Application Areas
Clarifying Digital Twins, Digital Shadows, Models, Simulations and Control Systems
Reference

Digital twins, models, shadows, simulations, and control systems are frequently used interchangeably, leading to confusion in procurement, implementation, and expectation setting. The purpose of this document is to clearly define these distinct concepts to ensure that buyers and suppliers have a shared understanding of what is being requested, delivered, and expected.

By establishing precise definitions, organisations can make more informed decisions when specifying requirements, assessing capabilities, and ensuring that delivered solutions align with their intended use cases. This clarity reduces ambiguity, improves procurement efficiency, and helps both technical and non-technical stakeholders understand the fundamental differences between these digital representations.

The National Digital Twin Programme is developing the tools, technologies, and operating frameworks necessary to enable interoperability, scalability, and adoption across these interconnected concepts. While Digital Twins remain a primary focus, the programme also addresses the requirements of digital models, digital shadows, simulations, and control systems to ensure a comprehensive and integrated approach to digital transformation.

Full Definition

A digital twin:

  • is a digital representation of a physical-world entity or environment or a process that includes a two-way data[1] flow into and out of the physical world in a timeframe that is appropriate for the required decisions and assumptions;
  • replicates its physical-world counterpart to a known tolerance, performs without statistical bias[2], and is able to predict what the associated physical-world entity would do given a stimulus; and
  • provides a level of prediction confidence within specified validation envelope and an associated assumption set, based on physical parameters and specifications, including after integration with other digital twins.

Understanding key Terms in the Digital Twin definition

The bullet points above define the essential characteristics a system must meet to qualify as a Digital Twin. To provide clarity, each key term has been broken down below. Click on any term to expand and see its detailed explanation.

1

Be associated with a physical-world entity or environment or a process that includes a two-way data.

In order to generate digital twins of physical entities, environments or a process, there will likely be a “template” for each type of physical-world counterpart to be twinned. This template must allow for a two-way data flow into and out of the physical world. When such a two-way flow of data is established, a template becomes a digital twin of the specific instance of the physical entity or environment or the process from which it receives, and to which it sends, data. For example, a specific car type may have a template used to create a digital twin, but every car manufactured will have a unique digital twin instance associated with it.
2

A timeframe that is appropriate for the required decisions and assumptions.

A digital twin must run in a timeframe that allows it to be run synchronously with its physical counterpart. This can be achieved if the two-way communication flow into, and out of, the physical world is appropriate for the required decisions and assumptions – in other words in a timeframe that allows the digital twin to be updated so it remains sufficiently aligned with the physical entity, environment or process. For example, if a manufacturing plant runs on a twenty-four-hour schedule with eight non-production hours, the twin must be capable of updating to current status within that twenty-four-hour window. Equally, an aircraft engine digital twin must be capable of running to the current status of the physical entity after landing and before the status is next updated. A digital twin may have the capability of incorporating the outputs of a simulation that may be run in faster-than-real-time, however, this capability is not a requirement of a digital twin.
3

Replicates its physical-world counterpart to a known tolerance, performs without statistical bias.

A digital twin must reliably replicate its physical-world counterpart in every way known to be relevant to the use case it has been designed for. A statistical, data or simulation-based model can be a digital twin only if it has been trained specifically on a physics-based (or theory-based) representation of the physical entity with a known and unbiased level of uncertainty, or if it has been trained using the physical-world data provided by the specific physical counterpart. A physics-based model can only be a digital twin if it is associated with a known tolerance and fidelity.
4

Have a specified validation envelope.

A digital twin must only replicate its physical-world counterpart within known validation windows – the operational limits within which it accurately represents and influences reality. For instance, if the twin is a physics-based model trained on data within a known temperature range there can be confidence within that temperature range. Outside of that range there will be a point at which the physics and data changes, and the digital twin will no longer be valid. To meet the criteria of a digital twin the validation envelope over all relevant parameters must be specified and associated with the digital twin and taken into consideration in “what-if” questions. If the physical world counterpart unexpectedly operates outside of the validation range, an immediate re-assessment of the digital twin must be undertaken.
5

Have an associated assumption set.

Even within a validation envelope, the digital twin will only reflect specific characteristics of the physical entity. It may reflect size, shape, weight, motion and temperature for instance, but not model the effect of an external force. The assumption set specifies explicitly what the digital twin accounts for and what it deliberately excludes. Consider a digital twin of a car brake system: the digital twin could be classed as such without the ability to model the change to the physical system resulting from a collision with another car, providing that this is captured in the linked assumption set.
6

Be based on physical parameters and specifications.

A digital twin must represent physical world limitations and specifications within its validation envelope. Where a digital twin has been developed for a new design of a physical entity, it cannot contain components or properties that cannot be replicated in the physical world. For example, a digital twin must not contain models of metals with properties that cannot be produced in reality. In order to analyse the impact of idealised properties, a digital twin would need to be used with a separate model or simulation.
These constraints prevent misapplication of a digital twins in scenarios they were not designed to handle, and ensures they can be used for test and evaluation and safety case development. Further, these requirements, when combined with common set of data exchange protocols, ensures digital twins are fully composable with other digital twins.

Digital Twin States

A digital twin can be in one of the two following states at a point in time:

  • Connected – if it is currently exchanging data into and out of the physical world in a timeframe that is appropriate for the required decisions and assumptions.
  • Semi-connected – if it utilises synthetic data but retains at least one physical world data feed. Consider a digital twin that is dependent on both ambient temperature and external vibration. If the current ambient temperature was still drawn from the physical world, but the vibration level was simulated using a model to understand the impact of stress on the physical world object, this would be deemed “semi-connected”.

A digital twin may move between these states and remain a digital twin, provided its current state is recorded. If a digital twin disconnects it can only return to being a digital twin if the reconnection occurs within a timeframe in which it will still accurately replicate its physical world counterpart.

Common Application Areas

Digital twins of assets and capabilities

This area encapsulates component, platform,  system and manufacturing (factory) digital twins. The existence of digital twins of individual components within different industry partners which need to be integrated at the platform level poses a significant challenge, but is proven technology in industry. However, even in this most well-known sector there are complications driven by the multi-fidelity requirement of digital twins and the need for assured outputs in an appropriate timeframe.

Digital twins of environments

This area encapsulates infrastructure, transport, manufacturing, environment and mega-structure digital twins. A mega-structure is defined as a complex amalgamation (or ecosystem) of both component, platform, infrastructure and environment twins such as would be found in an urban environment.   This requires complex multi-fidelity twins with real time inputs to ensure accurate representation of an environment and the ability to output to infrastructure in that environment.  This is currently unproven technology with substantial open research and technological challenges.

Digital twins in biology and medicine

In this area, efforts are focussed on digital twins of aspects of a human, such as a human heart. Digital twins of human organs or body structure give huge benefits for medicine and equipment design, however, even at this level, linking such a physical digital twin to a modelling framework that could output the forces exerted by a body on surroundings or clothing is not yet possible.

Clarifying Digital Twins, Digital Shadows, Models, Simulations and Control Systems

Not all digital representations qualify as digital twins. To ensure clarity, it is important to distinguish between digital twins, digital shadows, and standalone models or simulations.

  • Digital Model – a digital representation that is not dynamically linked to a physical entity, environment or a process.
  • Digital Shadow – a representation which is updated to reflect changes in the physical world by a one-way flow of data —from the physical to the digital—but which is unable to directly control or influence its physical world counterpart.
  • Simulation – a representation of a process or system that exists or could exist in the physical world. Simulations are based on digital models and execute those models over time to study behaviour under various conditions. They are used for predictive analysis, testing, and exploring hypothetical scenarios, but unlike a digital shadow, they are not required to maintain a live connection with a physical counterpart.
  • Control System – a system that manages and regulates the behaviour of other devices or processes by applying control actions based on predefined rules. Control systems operate in two main forms:
    • Open-loop control – where the control action is applied without feedback from the system’s output.
    • Closed-loop control – where the control action adjusts based on feedback from the system’s output.

A Control System differs from a Digital Twin in that a Digital Twin is capable of undertaking predictions, whereas a Control System cannot.

Example:
A Car Suspension in
Development and Operation

Concept Example Purpose
Digital Model A 3D computer-aided design (CAD) model of the suspension system, including material properties and structural design. Development – provides a structured representation for design and engineering.
Simulation Running finite element analysis (FEA) on the digital model to predict how the suspension reacts under different forces and stresses. Testing – analyses performance before building a physical prototype.
Digital Shadow Real-time sensor data from the suspension system (e.g., shock absorber pressure, wheel displacement) is sent to a dashboard but does not influence the car. Monitoring – reflects live performance data without sending feedback to the physical system.
Control System The car’s active suspension adjusts stiffness based on driving conditions. In open-loop mode, it follows preset settings; in closed-loop mode, it adapts based on sensor feedback. Regulation and management – ensures real-time control of the suspension system.
Digital Twin The suspension’s digital twin analyses upcoming road conditions using real-time and historical data. If the car is approaching a pothole or rough terrain, the digital twin predicts the impact and adjusts the suspension stiffness in advance to improve handling and comfort. Optimisation and decision-making – enhances performance by anticipating conditions and adjusting proactively.

References

[1] In some fields, the term ‘data’ may have different interpretations. Here, ‘data’ should be understood in its broadest sense, referring to representations of facts, observations, or concepts. In a broader conceptual framework, ‘information’ refers to data that has been processed or contextualised to provide meaning, while ‘knowledge’ represents structured understanding derived from information through experience, reasoning, or analysis. For simplicity, when we use ‘data’ in this document, we mean all three—data, information, and knowledge—unless otherwise specified.

[2] Here, bias is a statistical term. Without bias means that an estimator, when used repeatedly, does not systematically overestimate or underestimate the true value of what it is measuring. In statistical terms, an estimator is unbiased if its expected value equals the actual value of the parameter it is estimating. This ensures that any errors are due to random variation rather than a systematic skew in one direction.