Performance Evaluation of Iterative Signal Design for System Identification

Abstract

In this paper, iterative system identification is investigated where the amplitude spectra of the perturbation signals are optimised according to the present estimate of the model parameters. The iterative signal design is compared for single-input single-output (SISO) open loop systems grouped into six different categories based on the system order and dynamic characteristic. Three different signal-to-noise ratios (SNRs) are tested. The performance of the iterative signal design is assessed by taking the ratios of the performance measures of the estimated and initial models to the actual model for a single iteration. These performance measures are defined based on errors in the frequency response. It was found that the iterative signal design is very effective in reducing the model error, except for high order lowpass systems.