TABLE OF CONTENTS In most practical applications of system identification to aircraft problems, parameter estimation methods are applied to measured data from smallamplitude maneuvers executed about a trimmed flight condition. For these maneuvers, a linear aerodynamic model can be assumed. However, results from this analysis are only valid locally, that is, near the flight condition where the maneuver was performed. Information about the aerodynamic characteristics over a wider range of conditions can be obtained by analyzing maneuvers involving large variations in angle of attack, sideslip angle, and control deflections.
Analysis of large-amplitude maneuvers requires postulation of a model that might involve a relatively large number of parameters, e.g., including higherorder terms in a multivariate Taylor series expansion. Very often the increased model complexity cannot be supported by the information content in the data. This can result in parameter estimates with low accuracy, or the parameter estimation might fail.
To overcome these problems, a procedure known as data partitioning can be used (cf. Ref. 26). The idea is to divide the data points from a maneuver or set of maneuvers covering a large range of some important independent variables into partitions, where each partition contains the data points with values of important independent variables that lie within small ranges. This converts a modeling problem that might require a complicated nonlinear model structure and many model parameters into a series of simpler problems that require only linear model structures and just a few model parameters for each of the simpler problems.
Data...
