Overview

The previous chapter presented the basic concepts of state-space model identification using the frequency-response method. The example applications were based on canonical model structures, which are state-space equivalents of SISO transfer functions. In this chapter, the concepts and examples focus on the identification of physical model structures for flight-vehicle dynamics. This is the most challenging aspect of system identification, and good results depend heavily on a high-quality MIMO frequency-response database. All of the previous steps of the overall methodology (see Fig. 2.1) must be carefully completed to ensure that the extracted model has reasonable values for the identified parameters and provides good predictive accuracy. Proper collection of flight-test data, kinematic consistency and state reconstruction, advanced FFTs, multi-input conditioning, and multiwindow optimization are all important to achieving the necessary frequency-response database quality. The initial investigation of model structures and key parameter values using SISO transfer-function and canonical identification methods is also very important to establishing the correct physical model structure and initial parameter guesses. Errors in any of these steps will degrade the accuracy and utility of the MIMO physical model an ultimate product of many system identification studies.

The identification of physical models of flight-vehicle dynamics is discussed extensively in the literature. Iliff170 is an excellent and succinct reference on the stability and control derivative model structure and identification for fixed-wing applications using time-domain methods. Examples of some recent fixed-wing identification studies include the X-3135 ^, 82, and the F-18.179 An excellent overview of fixed-wing and rotorcraft applications at the...