Title: NARMAX model identification using a randomised approach

Authors: Pedro Felipe Leite Retes; Luis Antonio Aguirre

Addresses: Programa de Pós Graduação em Engenharia Elétrica, Universidade Federal de Minas Gerais, Av. Antonio Carlos 6627, 31270-901, Belo Horizonte, MG, Brazil ' Departmento de Engenharia Eletrônica, Universidade Federal de Minas Gerais, Av. Antonio Carlos 6627, 31270-901, Belo Horizonte, MG, Brazil

Abstract: Structure selection is one of the most critical steps in nonlinear system identification. A large family of methods, based on model prediction error, use concepts and tools from linear algebra. Other methods, based on model simulation error, have to deal with non-convex optimisation problems. More recently a family of methods have been put forward that has probabilistic setting. The randomised algorithm for model structure selection (RaMSS) belongs to this family and it has been shown to be effective to select regressors for NARX models. In the present paper, such a method is extended to cope with NARMAX models. The performance of the proposed method is illustrated using simulated and experimental data. It is shown that the proposed method is capable of correctly selecting model structures from simulation data. The method was also applied to experimental data with successful results.

Keywords: NARMAX; nonlinear models; ELS; OLS; randomised algorithm for model structure selection; RaMSS; NARX; system identification; parameter estimation.

DOI: 10.1504/IJMIC.2019.098779

International Journal of Modelling, Identification and Control, 2019 Vol.31 No.3, pp.205 - 216

Received: 16 Mar 2018
Accepted: 16 Mar 2018

Published online: 02 Apr 2019 *

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