Nonlinear model reduction based on smooth orthogonal decomposition
Document Type
Conference Proceeding
Date of Original Version
12-1-2007
Abstract
Large scale nonlinear model reduction based on smooth orthogonal decomposition (SOD) is presented. SOD is a multivariate time series analysis tool that provides optimal, low-dimensional representation of time series that are as smooth in time as possible. SOD is used to identify linear subspaces containing linear and nonlinear normal modes and span by smooth orthogonal modes (SOMs). Large finite element model (FEM) of a vibrating cantilever beam in a two-well potential is used to illustrate the model reduction. The SOMs of the simulated unforced, undamped FEM are used for model reduction. The performance of damped, forced FEM is then compared with three and five SOM based reduced-order models for various forcing parameters and close agreement is observed even for three SOM based reduced order model.
Publication Title, e.g., Journal
Proceedings of the 9th IASTED International Conference on Control and Applications, CA 2007
Citation/Publisher Attribution
Chelidze, David, and Gregory Chelidze. "Nonlinear model reduction based on smooth orthogonal decomposition." Proceedings of the 9th IASTED International Conference on Control and Applications, CA 2007 (2007): 325-330. https://digitalcommons.uri.edu/mcise_facpubs/100