Non-linear identification using a genetic algorithm approach for model selection

Michael F. Platten, Jan R. Wright, Keith Worden, Jonathan E. Cooper, Grigorios Dimitriadis

Research output: Chapter in Book/Conference proceedingConference contribution

Abstract

The Non-Linear Resonant Decay Method is an approach for the identification of non-linear systems with large numbers of degrees of freedom. The identified non-linear model is expressed in linear modal space and comprises the modal model of the underlying linear system with additional terms representing the non-linear behaviour. Potentially, a large number of these non-linear terms will exist but not all of them will be significant. The problem of deciding which and how many terms are required for an accurate identification has previously been addressed using the Forward Selection and Backward Elimination techniques. In this paper, a Genetic Algorithm optimisation is proposed as an alternative to those methods. A simulated 5-DOF lumped parameter non-linear system is used to demonstrate the proposed optimisation. The use of separate data sets for the identification and validation of the modal model is also investigated. It is found that the Genetic Algorithm approach yields significantly better results than the Backward Elimination and Forward Selection algorithms in many cases.
Original languageEnglish
Title of host publicationConference Proceedings of the Society for Experimental Mechanics Series|Conf. Proc. Soc. Exp. Mech. Ser.
Publication statusPublished - 2005
Event23rd Conference and Exposition on Structural Dynamics 2005, IMAC-XXIII - Orlando, FL
Duration: 1 Jul 2005 → …

Conference

Conference23rd Conference and Exposition on Structural Dynamics 2005, IMAC-XXIII
CityOrlando, FL
Period1/07/05 → …

Fingerprint

Dive into the research topics of 'Non-linear identification using a genetic algorithm approach for model selection'. Together they form a unique fingerprint.

Cite this