Probabilistic Digital Twin for Damage-Aware Rotorcraft Path-Tracking Optimization Under Uncertainty

This paper develops a probabilistic digital twin-based methodology for optimal control of a rotorcraft that has incurred damage during a flight. A sudden change in the health state of the rotorcraft can significantly impact flight control. In the proposed digital twin approach, a probabilistic rotorcraft dynamic surrogate model is trained for fast time series prediction of future rotorcraft states for different damage levels for the damage mode of interest. The current health state is diagnosed to provide a probabilistic damage estimate, and future rotorcraft states are predicted given the current position, heading, the updated health condition, and candidate control settings. The diagnosis uncertainty as well as the prognosis uncertainty are quantified and propagated through time to obtain stochastic future rotorcraft states. The updated health estimate and the probabilistic, damage-sensitive prognosis surrogate are used for optimization under uncertainty to obtain the pilot control inputs that minimize the error between the planned and predicted flight path while satisfying probabilistic stability constraints for the rotorcraft. The methodology is demonstrated by performing simulation experiments on a helicopter that incurs damage on its rear horizontal stabilizer during flight.