The importance of energy criteria for selecting modes in reduced order modeling

Suparno Bhattacharyya, Joseph P. Cusumano

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Scopus citations

Abstract

We study the performance of the proper orthogonal decomposition when used for model reduction of an Euler-Bernoulli beam subjected to periodic impulses. We assess the accuracy of reduced order models (ROMs) obtained using steady-state displacement time series. The spatiotemporal localization of the applied impulses is tuned to control the number of excited modes in, and hence the effective dimensionality of, the system’s response. We find that when the impacts are significantly localized (i.e., are more impulsive), the conventional variance-based mode selection criterion can lead to inaccurate ROMs. We show that this arises when the reduced subspace capturing a fixed amount (say, 99.9%) of the total data variance underestimates the energy input and/or dissipated in the ROM, leading to energy imbalance. We thus propose a new energy closure criterion that provides an improved method for generating ROMs. The energetics of the resulting ROMs properly reflect those of the full system, and yield simulations that accurately represent the system’s true behavior.

Original languageEnglish (US)
Title of host publication31st Conference on Mechanical Vibration and Noise
PublisherAmerican Society of Mechanical Engineers (ASME)
ISBN (Electronic)9780791859285
DOIs
StatePublished - Jan 1 2019
EventASME 2019 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference, IDETC-CIE 2019 - Anaheim, United States
Duration: Aug 18 2019Aug 21 2019

Publication series

NameProceedings of the ASME Design Engineering Technical Conference
Volume8

Conference

ConferenceASME 2019 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference, IDETC-CIE 2019
Country/TerritoryUnited States
CityAnaheim
Period8/18/198/21/19

All Science Journal Classification (ASJC) codes

  • Mechanical Engineering
  • Computer Graphics and Computer-Aided Design
  • Computer Science Applications
  • Modeling and Simulation

Fingerprint

Dive into the research topics of 'The importance of energy criteria for selecting modes in reduced order modeling'. Together they form a unique fingerprint.

Cite this