Dynamic Analysis and Identification of Octopole Active Magnetic Bearing System

碩士 === 國立中正大學 === 機械系 === 93 === With the development of technology, high rotational speed machines become the mainstream to mechanical design recently. The magnetic bearing is regarded as an ideal bearing which has many superior characteristics than the conventional bearing. The characteristics inc...

Full description

Bibliographic Details
Main Authors: Kung Yung-Hsiang, 龔詠翔
Other Authors: Shiau Ting-Nung
Format: Others
Language:en_US
Published: 2005
Online Access:http://ndltd.ncl.edu.tw/handle/48646520818500088762
Description
Summary:碩士 === 國立中正大學 === 機械系 === 93 === With the development of technology, high rotational speed machines become the mainstream to mechanical design recently. The magnetic bearing is regarded as an ideal bearing which has many superior characteristics than the conventional bearing. The characteristics include non-contact, non-lubrication, widely range of working temperature and long life working etc.. In this study, the analysis of system characteristics and parameter identification are focused on the octoploe electromagnetic bearing system. In the analysis of system characteristics, finite element method (FEM) and Lagrange approach are employed to derive the system equations of motion. And the decentralized PD controller is adopted to control the rotor system and system stability is also analyzed. The numerical integration method is used to investigate the system characteristics. The hybrid identification method is combination of the method proposed by Yasuda et al. and Gauss-Newton algorithm, and it uses the simulation data which is employed by Runge-Kutta method. After discussing analysis of the electromagnetic force model, a simple model of the non-linear magnetic force is proposed by this study and its identified results have a good accidence with the real system. By simplifying the items of Fourier series expansion and collocating with convergence of Gauss-Newton algorithm, accurate parameters can be solved faster and the error is lower than 5%. However, Kp and Kd of the decentralized PD controller influence the identified results and the system characteristics. According to the numerical results of this study, identified results of the hybrid method are better than Yasuda’s method in the bad state of the controller.