Covariance Function Realization Algorithms for AeroElastic Dynamic Modeling

Poorly damped structural resonance modes induced by aero(servo)elastic interaction can be monitored and modeled via data-based estimation techniques. Such estimation techniques can use time-domain measurements of input/output behavior to formulate a dynamical model suitable for control system design. With the availability of a model, a model-based feedback control design technique can be used to dampen the resonance modes. This paper demonstrates the use of a class of realization algorithms that is able to deliver a consistent estimate of (poorly damped) structural resonance modes without user intervention. The new algorithm can provide consistent estimation of structural parameters, such as damping and location of resonance modes, based on random but quasi-stationary excitation and output measurement signals. This is done by first estimating standard auto- and cross-covariance functions. Subsequently, the covariance function estimates are used in a realization algorithm that only requires a standard Singular Value Decomposition to extract a multivariable state space realization that models the dynamics of the aero(servo)elastic interaction. The ability to handle quasi-stationary signals allows in-situ parameter estimation in the presence of gust disturbances. The estimation method uses a modified version of the well-known Subspace algorithm to formulate a discrete-time model directly on the basis of covariance function estimates. The procedure is demonstrated on a high-fidelity F/A18AAW aircraft model for which noise perturbed data is simulated at different dynamic pressures for in-situ identification purposes.