What is a digital twin?

Other

29 July 2025

What is a digital twin?

A digital twin is a digital representation of a physical object or system. It might represent a bridge, a vehicle, a production process or a building. Data from the physical system keeps the model up to date.

A digital twin goes beyond a static drawing or a visual model. It combines information about how a system is built with data about how it behaves. This can help people monitor its condition, test scenarios and decide what to do next.

• Monitoring and synchronisation. Data from the physical system updates the digital model, so it can reflect the state of the real object.
• Intervention and control. In some applications, insights from the model can lead to adjustments to the physical system. Those decisions may be made by people or by an automated control process.

How does it differ from a simulation?

A simulation can test what might happen under a set of conditions without being connected to the real system. A digital twin maintains a connection with its physical counterpart through data.

Simulations can be part of a digital twin. The wider system may also combine data analysis, models, sensors and control mechanisms. The combination depends on the purpose of the project.

The parts of a digital twin

1. The physical system

This is the real object or process that the model represents. Examples include a bridge, an engine or a production line. Its relevant characteristics determine which data needs to be collected.

2. Sensors and data collection

Sensors can measure temperature, pressure, movement, vibration or load. Other records may provide information about the design, maintenance history or operation of the system.

3. Cleaning and preparing the data

Raw data can contain gaps, noise or inconsistent measurements. It needs to be checked, filtered and prepared before it is used in a model. Otherwise, unreliable input can lead to unreliable conclusions.

4. Analysis and interpretation

The data is interpreted using knowledge of the system. This might include physical relationships or operating rules. Combining that knowledge with measurements can help identify trends and deviations.

5. Modelling and representation

The digital model represents the aspects of the physical system that matter for the task. It may use statistical methods, physical simulations or other analytical models. New data and insights can be used to update it.

6. Planning and scenario analysis

A model can be used to ask what might happen if conditions change. For example, how could a change in load affect a structure? What might happen if maintenance is postponed? These scenarios support decisions without first making the change to the real system.

7. Decisions and actions

People can use the results to plan maintenance or change a process. Where a system has suitable control mechanisms, some actions may be automated. The required checks and human oversight depend on the application.

What can digital twins be used for?

The value of a digital twin lies in the questions it helps answer. Typical aims include:

• Understanding the current condition. Monitoring can help reveal changes or unusual behaviour.
• Planning maintenance. Trends in the data may help identify components that need attention.
• Improving a process. Scenario analysis can compare possible changes before they are put into practice.
• Supporting decisions. A model can turn complex measurements into information that people can interpret.
• Following a system over time. A model can be maintained as the physical system and its use change.

Possible applications

• Manufacturing. Models of machines and production processes can support monitoring, maintenance and planning.
• Infrastructure. Models of bridges or buildings can support assessments of their condition.
• Healthcare. Research and development also explore models that combine information relevant to a patient or care process.
• Transport. Models of vehicles and aircraft can help analyse their operation and performance.
• Cities. Models can bring together information about traffic, energy use and infrastructure.

The challenges of building a digital twin

A digital twin brings together several technical disciplines. Building a useful model requires more than collecting as much data as possible.

Complexity and scale

The amount of data and the calculations required can vary greatly. The infrastructure needs to suit the task, from a small device to a larger computing environment.

Connecting different technologies

Sensors, data processing, simulations and control systems need to work together. A project may require several kinds of specialist knowledge and clear agreements about the interfaces between them.

Long term support

A bridge or an aircraft may remain in use for many years. The model also needs maintenance as software, equipment and the physical system change. A useful design allows those changes to be made without rebuilding everything.

Choosing the right level of detail

Collecting too little information can limit the model. Collecting unnecessary detail can make it more expensive and harder to maintain. The right balance follows from the questions the model needs to answer.

Reliability and safety

Decisions based on a model can affect real equipment and people. Errors and uncertainty need to be recognised, especially where a model is connected to a control process. A simulation result is not a guarantee of what will happen in practice.

How a digital twin is developed

Step 1: Define the purpose and system

Start with the physical system and the decision you want to support. Are you investigating maintenance, performance or planning? That purpose determines the requirements for the data and model.

Step 2: Collect relevant data

Identify the measurements and records needed for the task. Consider how they will be collected and whether they are reliable enough for the intended use.

Step 3: Clean and connect the data

Prepare the data and bring the relevant sources together. Check how missing values, inconsistent measurements and errors will be handled.

Step 4: Build the digital model

Choose a representation that suits the purpose. It may combine physical, statistical or other modelling methods. Allow for updates and extensions.

Step 5: Analyse and test scenarios

Use the model to investigate specific questions. Compare its results with observations from the real system and examine where the model is uncertain.

Step 6: Connect the results to decisions

Decide how people will use the findings and whether any actions should be automated. Define the limits and checks before connecting the model to a physical process.

Step 7: Maintain the model

Update the model as the physical system changes. Its data sources, assumptions and outputs need ongoing attention.

Examples of digital twins

A bridge

Imagine a model intended to support the maintenance of a bridge. Sensors could measure load, vibration and temperature. The model could combine those measurements with information about the structure and its materials.

Engineers could investigate changes over time and compare scenarios involving traffic or weather. The model would need to be kept up to date as the structure or its use changes.

Healthcare

A model can bring together different kinds of information relevant to a patient or care process. Such applications require particular attention to the quality of the data, the model's limitations and the role of professional judgement.

A production process

A manufacturer might connect measurements from machines to a model of a production process. The aim could be to understand interruptions, compare operating conditions or support maintenance planning. The usefulness of the model depends on its data and how well it represents the process.

Frequently asked questions about digital twins

What is the difference between a digital twin and a digital shadow?

A digital shadow usually describes a representation that receives data from the physical system. The term digital twin is often used for a wider connection that can also support decisions or actions affecting the physical system. Terminology varies between applications, so it helps to describe what a particular system actually does.

Are digital twins new?

Many of their components, such as simulations, data analysis and control systems, have existed for decades. The concept brings those components together around a physical counterpart and an ongoing data connection.

Does a digital twin have to be an exact copy?

No. A model focuses on the aspects relevant to its purpose. Omitting unnecessary detail can make it easier to use and maintain.

How often does a digital twin need updating?

That depends on what the model is used for and how quickly the physical system changes. Some applications need frequent updates; others can work with less frequent measurements.

Can a digital twin make decisions automatically?

Some systems can use model outputs in an automated control process. In others, people interpret the results and decide what action to take. Automatic control is a design choice that requires its own checks.

Connecting models with the physical world

A digital twin connects a model with information from a real object or system. It can support monitoring, scenario analysis and decisions, provided that the model and its data suit the purpose.

Building and maintaining that connection takes specialist work. Clear aims, suitable data and an understanding of the model's limits matter more than adding detail for its own sake.