Physics vs Data-Driven Twins

A twin can be first-principles physics, machine-learned surrogates, or a hybrid - and Kronos uses fast learned models inside a physics-consistent frame.

The concept

Physics models are trustworthy and extrapolate but can be too slow for control; data-driven surrogates are fast but reliable only where trained. The design question is where to place each.

How Kronos uses it

Kronos runs learned surrogates for speed inside a physics envelope that enforces conservation laws and safe limits, with uncertainty carried through so the twin flags when it is extrapolating rather than guessing silently.

Honest note

Surrogates accelerate; the physics envelope and safety limits always have final say.