ABSTRACT

This chapter consists of seven sections. Section 5.2 shows how to obtain an estimate of the process step response from the FSF model parameters. Section 5.3 discusses the topic of smoothing the step response estimate using a reduced order FSF model. Section 5.4 analyzes the errors introduced by using a reduced order model. Section 5.5 derives confidence bounds for the process frequency response and step response estimates obtained from an FSF model. Section 5.6 presents a generalized least squares algorithm for identification of multi-input , single-output systems in which the P RESS criterion is used to determine both the FSF process model order for each subsystem and the number of terms to be included in the noise model. Section 5.7 presents an industrial case study using data collected from the Sunoco refinery in Sarnia, Canada.