What is a Digital Twin?
A Digital Twin is a virtual representation of a physical object, process or system that is synchronised with data from that real system at an agreed frequency and fidelity, so that you can digitally monitor, analyse, predict and optimise its behaviour. A digital twin combines modelling, Sensor data, simulation and analytics. What sets it apart from an ordinary simulation is the link: the twin is continuously updated with real-time or near-real-time data from the OT environment. That makes the digital twin one of the building blocks of Industry 4.0 and an important meeting point between IT and OT.
🧠 Where does the term digital twin come from?
The idea is older than the name. From the 1960s onwards NASA used ground-based simulators that mirrored spacecraft; during Apollo 13 (1970) they helped engineers analyse the failure of the oxygen tank. The concept as we know it today was described by Michael Grieves in the early 2000s, in his course on Product Lifecycle Management at the University of Michigan. He identified three parts: the physical product, the virtual product and the data and information connection between them. Grieves credits the term “digital twin” to John Vickers of NASA. In 2010 NASA published a first working definition, aimed at better simulation of spacecraft.
There is still no single, universally accepted definition. The most important ones:
| Source | Definition in brief | Emphasis |
|---|---|---|
| Grieves (PLM) | Physical product, virtual product and the data connection between them | Product lifecycle |
| ISO 23247 | A “fit for purpose” digital representation of an observable manufacturing element, synchronised with that element | Manufacturing |
| Digital Twin Consortium | Virtual representation of real-world entities and processes, synchronised at a specified frequency and fidelity | Outcomes, IT/OT systems |
| NIST IR 8356 (2025) | Digital representation of a real-world entity, concept or notion, physical or perceived | Security and trust |
ISO 23247 counts personnel, equipment, material, processes, facilities, environment, products and documents among the “observable manufacturing elements” that can be given a twin.
🧱 What are the components of a digital twin?
At its core, every digital twin has three components:
- Physical system - for example a pump, a machine, a production line or an entire plant
- Digital counterpart - a 3D model, data structure, physics-based model or machine learning model
- Data exchange - via IIoT, SCADA, MES, sensors or fieldbuses
The reference architecture in ISO 23247-2 works this out into four domains:
| Domain | Role | Practical example |
|---|---|---|
| Observable manufacturing domain | The physical reality being tracked | Machines, material, operators |
| Data collection and device control | Collecting data and, where permitted, controlling devices | PLC, edge device, OPC UA server |
| Core domain | Model, synchronisation, simulation, analysis and optimisation | Simulation platform, historian, analytics |
| User domain | People and systems that use the twin | Dashboards, MES, ERP, other twins |
The standard deliberately leaves the synchronisation rate open: the application determines it.
🔄 What is the difference between a simulation, a digital shadow and a digital twin?
Many projects labelled “digital twin” are in fact a model or a shadow. The difference lies in the direction and automation of the data flow:
| Level | Physical → digital | Digital → physical | Typical use |
|---|---|---|---|
| Simulation / digital model | Manual or one-off | Manual | Design, virtual commissioning |
| Digital shadow | Automatic, continuous | Manual, a human decides | Monitoring, condition monitoring |
| Digital twin | Automatic, continuous | Automatic or via a controlled route | Optimisation, automated adjustment of setpoints |
A simulation is therefore a one-off analysis detached from the real system. A digital shadow follows the system live but sends nothing back. Strictly speaking, you only have a digital twin when results flow back automatically. For OT that write-back is precisely the sensitive part, because it touches the control system.
Grieves and Vickers also distinguish the digital twin prototype (before the asset is built), the digital twin instance (one per delivered unit, for its whole life) and the digital twin aggregate (the collection of instances, used to analyse and learn across the fleet).
🏭 What are digital twins used for in industry?
| Use case | What the twin does | Typical data |
|---|---|---|
| Predictive Maintenance | Predicting failures from anomalous behaviour and remaining useful life | Vibration, temperature, current |
| Production optimisation | Running scenarios before changing processes or setpoints | Process data from SCADA and the historian |
| Design and validation | Testing machines or lines digitally before the physical build | CAD, control code, process models |
| Virtual Commissioning | Testing PLC software against a simulation model | I/O signals, sequences |
| Planning and logistics | Optimising the sequence and routing of orders | Orders from MES and ERP |
| Remote monitoring | Live oversight via dashboards or 3D visualisation | Statuses, alarms |
| Energy management | Simulating consumption patterns and analysing efficiency | Energy meters, process data |
Beyond the factory there are large twins of infrastructure and even of the Earth. The European Destination Earth programme, launched by the European Commission in 2022 and implemented by ECMWF, ESA and EUMETSAT, builds digital twins for weather-induced extremes and climate change adaptation.
🔧 How do you connect a digital twin to OT and IT?
A digital twin draws data from every level of the Purdue Model. Common building blocks:
- PLC, SCADA and historian - supply live and historical process data
- OPC UA and MQTT - the usual protocols for exposing that data in a structured way, often in a Unified Namespace
- Edge Computing - pre-processing close to the machine, for example extracting features from vibration data sampled in the kilohertz range
- Time series databases - efficient storage of measurement series
- MES, ERP and CMMS - orders, planning and maintenance history, structured according to ISA-95
- Cloud and AI - computing power for simulation, analysis and model training
- Asset Administration Shell - the standardised digital twin of Industry 4.0 (IEC 63278-1), which records master data, documentation and submodels per asset
The synchronisation rate depends on the purpose. Indicative values:
| Use case | Typical update to the twin |
|---|---|
| Vibration analysis at the edge | Sampling in kHz, features per second to per minute |
| Process monitoring and optimisation | Seconds to minutes |
| Production planning | Minutes to hours, or per order |
| Master data and documentation (AAS) | On change |
🔐 What security risks does a digital twin introduce?
Security deserves early attention, because a digital twin gathers all knowledge about an installation in one place. NIST IR 8356 (February 2025) identifies five novel risk areas: massive instrumentation with low-cost IoT sensors, centralisation of all measurements, manipulation of what operators see, remote control through the twin, and standards that enable universal access. Whoever takes over the twin knows the process and can show an operator a false picture while the installation does something else.
Measures that work in practice:
- Zones and conduits - place the twin in its own zone following the zones and conduits model and IEC 62443, never inside the control network itself
- One-way traffic where possible - a digital shadow needs no write access; a read-only OPC UA server or a data diode is enough
- Controlled write-back - route setpoint changes through an approved path, with plausibility checks in the controller
- Defense in Depth - encrypt traffic with TLS, use a VPN for remote access, strong authentication and logging
- Manage cloud dependencies - record which data goes to the cloud and who can access it
For organisations covered by NIS2 and the Dutch Cybersecurity Act (Cyberbeveiligingswet), a twin with a connection to the control system simply falls within the risk analysis and duty of care.
🧭 How do you get started with a digital twin?
- Pick one concrete question - for example unplanned downtime of a pump set or the energy use of a furnace. A twin “of the whole plant” usually fails.
- Decide the level - model, shadow or twin. Almost always start as a shadow, without write access.
- Inventory the data - which sensors, tags and systems deliver what, at what quality and frequency? In a brownfield situation instrumentation is often missing.
- Choose open standards - OPC UA, MQTT and the Asset Administration Shell prevent vendor lock-in.
- Design security in - zone, conduit, access rights and monitoring belong in the first design, not added afterwards.
- Validate the model - compare predictions with actual measurements before you base decisions on them.
- Settle data ownership - since 12 September 2025 the European Data Act gives users of connected machines the right to access their machine data and to have it shared with a third party, such as the supplier of a twin.
❓ Frequently asked questions
What is the difference between a digital twin and a simulation?
A simulation is a model that is run once or independently of the real system. A digital twin is continuously updated with data from the physical system, so the model follows its current state. A simulation model is often the starting point for a digital twin, though.
What is a digital shadow?
A digital shadow is a digital representation that automatically receives data from the physical system but sends nothing back automatically. A human decides what happens with the insights. Many industrial “digital twins” are in fact digital shadows, which is actually an advantage for OT security.
Is there a standard for digital twins?
For manufacturing, the ISO 23247 series describes a digital twin framework, with a general part, a reference architecture, a part on digital representation and a part on information exchange. The Asset Administration Shell, specified in IEC 63278-1, standardises the digital twin of an asset. No single standard covers all digital twins.
What is the difference between a digital twin and an Asset Administration Shell?
A digital twin is the general concept of a digital counterpart of a physical object or process. The Asset Administration Shell is a concrete, standardised implementation from Industry 4.0, with fixed submodels and a fixed metamodel. A comprehensive digital twin can use an AAS as a standardised layer for asset data.
Is a digital twin a security risk for OT?
A digital twin centralises process data and sometimes control capabilities too, which makes it an attractive target. The risk is manageable by placing the twin in a separate zone, granting read-only access where possible and routing any write-back through a controlled path. That keeps a compromised digital twin separated from the control system.
Does every company need a digital twin?
A digital twin pays off mainly for expensive assets, complex processes or frequent unplanned downtime. For a simple installation, good monitoring and a historian often deliver most of the benefit. Start with one well-defined question and only expand the digital twin once it has paid for itself.
📌 In summary
A Digital Twin is a virtual representation of a physical system that is continuously synchronised with real data and used for monitoring, simulation, prediction and optimisation. The concept goes back to Grieves and NASA, is worked out for manufacturing in ISO 23247 and takes shape per asset in the Asset Administration Shell. Start small as a digital shadow, choose open standards and design the connection to OT securely from the outset, with Predictive Maintenance as a typical first use case.