Recursive parameter estimation of a mechanical system in frequency domain

Abstract

Frequency-domain identification and parameter estimation methods are well established and commonly applied for commissioning and diagnostics purposes in electric drives. In this paper, the feasibility of a recursive least squares parameter estimation algorithm from frequency-domain observations is studied. The identification problem is treated from two different perspectives: first, by estimating a discrete autoregressive model with exogenous terms (ARX) from the discrete Fourier transforms (DFTs) of the input-output signals obtained from the identification experiment and second, a nonparametric model that is fitted in terms of least squares regression. Both proposed identification approaches are studied by simulations and experimentally validated by a closed-loop-controlled servomechanism.