Coevolutionary Algorithm for Multivariable Discrete Linear Time-variant System Identification

System identification is the basis for control system design. For linear time-invariant systems have a variety of identification methods, identification methods for nonlinear dynamic system is still in the exploratory stage. Nonlinear identification method based on neural network is a simple and effective general method that does not require too much priori experience about the system to be identified. Through training and learning, the network weights are corrected to achieve the purpose of system identification. The paper is about the identification of multivariable nonlinear dynamic system based on PID neural network. The structure and algorithm of PID neural network are introduced and the properties and characteristics are analyzed. The system identification is completed and the results are fast convergence.

Download Full-text

Continuous-time linear time-varying system identification with a frequency-domain kernel-based estimator

IET Control Theory and Applications ◽

10.1049/iet-cta.2016.0385 ◽

2017 ◽

Vol 11 (4) ◽

pp. 457-465 ◽

Cited By ~ 7

Author(s):

John Lataire ◽

Rik Pintelon ◽

Dario Piga ◽

Roland Tóth

Keyword(s):

System Identification ◽

Frequency Domain ◽

Continuous Time ◽

Linear Time ◽

Time Varying ◽

Time Linear

Download Full-text

Floquet Experimental Modal Analysis for System Identification of Linear Time-Periodic Systems

Volume 5: 6th International Conference on Multibody Systems, Nonlinear Dynamics, and Control, Parts A, B, and C ◽

10.1115/detc2007-34790 ◽

2007 ◽

Cited By ~ 8

Author(s):

Matthew S. Allen

Keyword(s):

System Identification ◽

Linear Time ◽

Time Varying ◽

Response Model ◽

State Matrix ◽

Linear Time Invariant ◽

Response Data ◽

Time Varying Systems ◽

Time Invariant ◽

Time Periodic

A variety of systems can be faithfully modeled as linear with coefficients that vary periodically with time or Linear Time-Periodic (LTP). Examples include anisotropic rotorbearing systems, wind turbines, satellite systems, etc… A number of powerful techniques have been presented in the past few decades, so that one might expect to model or control an LTP system with relative ease compared to time varying systems in general. However, few, if any, methods exist for experimentally characterizing LTP systems. This work seeks to produce a set of tools that can be used to characterize LTP systems completely through experiment. While such an approach is commonplace for LTI systems, all current methods for time varying systems require either that the system parameters vary slowly with time or else simply identify a few parameters of a pre-defined model to response data. A previous work presented two methods by which system identification techniques for linear time invariant (LTI) systems could be used to identify a response model for an LTP system from free response data. One of these allows the system’s model order to be determined exactly as if the system were linear time-invariant. This work presents a means whereby the response model identified in the previous work can be used to generate the full state transition matrix and the underlying time varying state matrix from an identified LTP response model and illustrates the entire system-identification process using simulated response data for a Jeffcott rotor in anisotropic bearings.

Download Full-text