Time domain effects of model order reduction

It is suggested that a preliminary and essential step in the design of flight control systems for highly augmented aircraft is obtaining an accurate open-loop dynamics model. Very large-order open-loop state-space models are constructed from analytical and empirical data obtained from knowledge of the vehicle's aerodynamics, propulsion, and structure dynamics. A balancing methodology for reducing a very large-order state-space representation to a more practical size is discussed. The balancing algorithm has a frequency domain error bound that guarantees the the magnitude of the reduced-order model's frequency response will be bounded. It is suggested that implementing techniques to preserve the number of nonminimum-phase zeros while performing model order reduction, and using residualizaton to match the steady-state magnitude, should improve the time history responses. A 140th-order aeroservoelastic model and a fourth-order critically damped system with an oscillator are considered as examples.<<ETX>>