FOM vs ROM calibration

Simulated systems inherently differ from real-world systems due to non-linearities and uncertainties. As a result, every simulation requires calibration against test bench experiments, including reduced-order models (ROMs) for virtual thermal sensing.

While it is possible to calibrate the original full-order model (FOM) and then reduce it, this approach is highly inefficient. Running thousands of transient simulations to fit the parameters of a full-order model with millions of variables could take weeks, whereas calibrating a reduced-order model can be achieved in just a few minutes.