In short
A digital twin is a digital model of a physical asset, line or process, fed with real-time data, that lets you monitor, simulate and predict its behaviour.
A digital twin differs from a static 3D model: it is continuously updated with data from the field. This makes it possible to monitor the current state and test what-if scenarios without putting the real system at risk.
In the process industries digital twins are used to optimise process parameters, analyse energy consumption and train operators. The value comes from the accuracy of the model and the quality of the data that feeds it.