Overview

The theory and procedures discussed through Chapter 10 address the identification of a high-quality MIMO frequency-response database from the initial frequency-sweep time-history records. This MIMO database of conditioned frequency responses and partial coherences, shown in Fig. 2.1, constitutes the non-parametric model of Sec. 1.4. As the flowchart in Fig. 2.1 shows, there are two classes of parametric models that can be obtained at this point: transfer-function and state-space representations. The detailed topics of parametric model identification will be the focus of the next three chapters.

This chapter discusses the determination of transfer-function representations, which are the parametric model forms that are the simplest to extract from the numerical frequency-response databases. Transfer-function models are (linear) input-to-output descriptions of the dynamic system; they can be represented by pole-zero descriptions. A transfer-function fit that best matches the frequency-response data (magnitude and phase) on a Bode plot over the frequency range of interest is obtained.

For many applications these models are found to be quite sufficient, including handling-qualities analysis, actuator and other subsystem models, aeroelastic mode determination, and models for root-locus-based control system design. Even if the ultimate goal is the determination of a fully coupled state-space representation, as discussed in Chapters 12 and 13, obtaining transfer-function models is a useful intermediate step that provides information on the fundamental dynamic characteristics and a good estimate of key parameter values.

The transfer-function identification approach presented in this chapter is based on the lower-order equivalent system (LOES) concepts initially put forth by...