Reduced-Order physics-based digital twins: From computational models to flight-tested reality

At Cnam, Paris, September 10th 2026, 3 p.m.

Charbel Farhat, Vivian Church Hoff Professor of Aircraft Structures,
Department of Aeronautics and Astronautics & Institute for Computational and Mathematical Engineering,
Stanford University, Palo Alto, California, USA

Well-resolved viscous CFD models coupled with turbulence modeling remain computationally prohibitive for real-time digital twinning, while purely data-driven surrogates lack both robustness and generality, typically restricting predictions to a small number of scalar quantities of interest. This talk presents an end-to-end, physics-based framework bridging this gap via hierarchical projection-based model order reduction and nonparametric stochastic calibration.

The framework establishes a systematic progression from high-dimensional models to an operational digital twin instance via hyperreduction, probabilistic representation of uncertainty, and subsequent data-driven model updating. To capture unsteady nonlinear fluid dynamics in real time, it incorporates a novel trajectory-based perturbation formulation using localized tangent operators to build a database of reduced-order counterparts.

Validated using real flight data from an F-16D aircraft executing aggressive maneuvers under the United States Air Force HAVE MIRAGE project, the RANS-based hyperreduced stochastic surrogate of the full-order aerodynamic system delivers physics-consistent online predictions of aerodynamic loads at 20 Hz -- achieving speedups up to 57× ahead of real-time on a single CPU.