Digital twin

A computational representation of a physical asset, updated from its data.

The term covers at least three unrelated technologies. A geometric twin represents physical layout and is used for outage planning and training. A data-driven twin learns statistical relationships from historian data. A hybrid twin combines first-principles models of the physics with data-driven components for what the physics does not capture.

They answer different questions and fail differently. A data-driven twin is reliable inside the range of conditions it was trained on and returns an answer outside that range without signalling that it is extrapolating, which is a significant limitation on assets that operate transiently by design.

Proposals frequently do not state which kind is being offered, and a good deal of disappointment in this category comes from buying one and expecting the behaviour of another.