Identification of Hammerstein Systems with Multiple-Type Discontinuous Nonlinearities
| Autor: | Po-Yen Tseng, 曾柏巖 |
|---|---|
| Rok vydání: | 2012 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 100 The identification of Hammerstein systems with multiple-type nonlinearity is studied. A new algorithm is proposed to circumvent the problems associated with traditional Maximum Likelihood (ML) type methods: 1. the nonlinearity is not continuously differentiable; 2. improper initial values may lead to erroneous results. The new method is composed of two phases. In the first phase, the nonlinearity is approximated by Fourier sine series. The approximated nonlinearity and dynamic linear element are estimated along with noise dynamics by a least square type method designed to ensure global optimum. In the second phase, the output of the nonlinearity is estimated by filtering the system output reversely through the linear and noise dynamics acquired in the previous phase. The nonlinear characteristics can then be identified by the system input and the estimated nonlinear output. Numerical examples for several types of multiple-type nonlinearity are given to demonstrate the effectiveness of the proposed method. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|