A digital twin is a digital copy of a real structure or machine, kept up to date with data from it, so it can be checked and simulated without touching the real one. The copy can be a 3D model of a building, a model of a jet engine, or a model of a whole factory. The word "twin" is used because it follows the real object as it changes.
Here is how it works in general. First, a model of the object is built from its drawings and specifications. Sensors on the real object then send readings, such as temperature, vibration or position, and these update the model so that it shows the object's present condition. Engineers can look at the model to see what is happening inside, or run tests on it. For example, they can ask what would happen if the load doubled. The answer comes from the model, so the real object is not put at risk. The results guide decisions on repair and operation.
Engineers describe three levels. A digital model only holds the design. A digital shadow is also updated with live readings. A full digital twin can also send commands back to the real object. The model must be calibrated, which means its answers are compared with the real readings and adjusted until they agree. A twin can mislead when a sensor gives wrong numbers, or when the object is repaired and the twin is not updated. More detail in the model needs more computing power, so engineers include only the details that matter for the question being asked.
Take, for example, a jet engine with sensors that report its temperatures on each flight. The twin shows how much life each blade has left and predicts when it needs repair; see jet engines. Consider a building. The twin shows the position of every pipe and wire, and the live temperature of each room; see building information modelling. Imagine a chemical plant. Operators test a new setting on its twin before changing the real plant; see plant digital twins.
A twin is built on the digital thread and sensor data from industrial IoT. It enables predictive maintenance and uses engineering simulation. Related pages are autonomy simulation and operator training simulators.
How is it different from a plain simulation? A simulation is a model run with assumed numbers. A digital twin is fed live data from one real object, so it shows that object's state now.
Appears in
- Network digital twin↗ Observability and monitoringA bank searches a copy of its whole network to find which firewall rule stops its branches reaching the loan system.
- Network change verification↗ Observability and monitoringA network team tries a router change on a model of its network first, to see what breaks before changing the real one.
- 3D medical imaging↗ Medical imagingA surgeon rotates a 3D model of a patient's heart, built from scan slices, before planning the operation.
- Plant digital twin↗ Predictive maintenance softwareEngineers test a new setting on the copy of a plant before changing the real plant.
- Civil infrastructure design software↗ Construction technologyA bridge is modelled in full, and the same model is later kept up to date with sensors on the real bridge.
- Construction progress monitoring↗ Construction technologyThe scan of a building site is compared with the design model to show the work that is behind.
- Virtual lab↗ UpskillingA student runs a chemistry experiment on a simulated lab bench that behaves like a real one.