Title: Numerical analysis of H filter for system parameter identification

Authors: Vassilios A. Tsachouridis

Addresses: United Technologies Research Center, Ltd., 4th Floor Penrose Business Centre, Penrose Warf, Cork, T23XN53, Ireland

Abstract: The numerical analysis of the H filter algorithm for system parameter identification is addressed in this paper. After presenting the filter under study and the convergence characteristics of the algorithm, a first order perturbation analysis is conducted. More specifically, norm-wise bounds are derived for the sensitivities, condition numbers, forward errors, backward errors and rounding errors of the iterative computational operations of the algorithm. Consequently, numerical stability and accuracy rules can be potentially composed as measures of algorithmic reliability and dynamical numerical performance. For exploring the impact of incorporating the specific algorithm dynamics in the numerical analysis, a similar framework is presented for the computed solutions irrespective of the used algorithm. The paper further exploits the application of the synthesised numerical analysis framework to parameter identification of linear state space systems. Finally some numerical examples are given for demonstration purposes.

Keywords: H filter; numerical algorithms; error analysis; perturbation analysis.

DOI: 10.1504/IJMIC.2018.095340

International Journal of Modelling, Identification and Control, 2018 Vol.30 No.3, pp.163 - 183

Received: 29 Sep 2017
Accepted: 16 Oct 2017

Published online: 17 Sep 2018 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article