Reduced-Order Modeling
Reduced-order models compress a high-fidelity simulation into a much smaller one that keeps the behavior that matters.
Smaller but faithful
A full simulation may track millions of unknowns, far too many to evaluate repeatedly or in real time. A reduced-order model (ROM) captures the dominant behavior of that system with a handful of variables, running orders of magnitude faster while preserving the responses of interest.
How reduction works
Many high-dimensional systems actually evolve on a low-dimensional structure. Techniques such as projecting the dynamics onto a small set of dominant modes — found from simulation or experimental data — exploit this to build a compact model that reproduces the important dynamics.
Where ROMs fit
- Real-time estimation and control, where the full model is too slow.
- Rapid design sweeps and optimization loops.
- Digital twins that must keep pace with the physical machine.
Knowing the limits
A ROM is built to be accurate for a specific range of conditions and a specific set of outputs. Pushed outside that range, or asked about behavior its modes do not capture, it can be quietly inaccurate. Responsible use tracks the ROM’s validity envelope and falls back to full fidelity when needed.
ROM versus surrogate
A ROM reduces the governing equations while keeping physical structure; a data-driven surrogate simply fits inputs to outputs. ROMs tend to extrapolate more gracefully because they retain physics, which makes them attractive for the control and twin roles in a fusion plant where trust in the edges matters.