G5 Doctoral dissertation (article)

Online Time and Frequency Domain Identification of a Resonating Mechanical System in Electric Drives


Publication Details
Authors: Nevaranta Niko
Publisher: Lappeenranta University of Technology
Publication year: 2016
Language: English
Related Journal or Series Information: Acta Universitatis Lappeenrantaensis
ISBN: 978-952-265-995-8
JUFO-Level of this publication:

Abstract
In modern machinery, the dynamic performance of electric drives is often limited by the mechanical characteristics of the system such as flexibilities. With the growing demand for high-performance machinery, there is an increasing need for techniques to estimate mathematical models in real time that describe these mechanical systems and the possible changes in order to obtain a high-performance control. At the same time, requirements for high reliability are continuously increasing, which significantly motivates to improve system identification methods for the diagnostics and condition monitoring of mechanical parts in electric drives. A proper real-time system identification method is of great importance in order to obtain an analytical model that sufficiently represents the most important characteristics of the identified system. Even though many identification methods have been proposed in the system identification literature, there is a strong motivation to develop computationally efficient algorithms for online frequency response estimation. Especially, online nonparametric identification could provide several opportunities for fault diagnostics and robust controller design. In this doctoral dissertation, the online system identification of a resonating mechanical system in an electrical drive is studied. The discussion covers closed-loop identification approaches, which are based on both time and frequency domain observations. The time domain identification approach employs a closed-loop output error-based identification routine. In addition, two different types of frequency domain identification approaches are proposed that are based on a time-frequency representation of signals by applying sliding-DFT and Kalman filters. It is shown that the proposed online frequency domain methods provide a good alternative to the conventional time domain online identification solutions. Theoretical approaches are tested with experimental mechanical test setups that can be regarded as resonating two-mass systems. The experimental results confirm the feasibility of the identification methods by verifying the obtained models according to a given validation criterion, thereby showing that the system dynamics can be identified with an accuracy that makes it possible to apply the proposed approaches for online frequency response analysis.

Last updated on 2018-19-10 at 08:49