ePrints@IIScePrints@IISc Home | About | Browse | Latest Additions | Advanced Search | Contact | Help

Improved Well-Conditioned Model Order Reduction Method Based on Multilevel Krylov Subspaces

Kumar, Neeraj and Vinoy, Kalarickaparambil J and Gopalakrishnan, Srinivasan (2018) Improved Well-Conditioned Model Order Reduction Method Based on Multilevel Krylov Subspaces. In: IEEE MICROWAVE AND WIRELESS COMPONENTS LETTERS, 28 (12). pp. 1065-1067.

[img] PDF
Iee_Mic_Wir_Com_Let_28-12_1065_2018.pdf - Published Version
Restricted to Registered users only

Download (357kB)
Official URL: http://dx.doi.org/10.1109/LMWC.2018.2878957

Abstract

Reduced order models (ROMs) based on the asymptotic waveform evaluation enable fast and efficient parametric analysis of large-scale matrix systems exhibiting nonlinear dependence on certain desired parameter(s). However, they are known to be narrowband due to the inherently ill-conditioned moment generation process. While well-conditioned approaches exist that enforce the moment-matching criteria by introducing some correction terms, these are not optimal and are difficult to parallelize. This letter introduces a well-conditioned multilevel Krylov model order reduction (WMKMOR) technique which is accurate over a larger band and faster to set up than existing approaches. Also, the multiple levels of Krylov subspaces in WMKMOR are generated independently. The improvement in ROM bandwidth using this technique is demonstrated for a finite-element model of an electromagnetic scattering example.

Item Type: Journal Article
Additional Information: This article belongs to IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
Keywords: Asymptotic waveform evaluation (AWE); finite-element method (FEM); model order reduction (MOR); scattering
Department/Centre: Division of Mechanical Sciences > Aerospace Engineering(Formerly Aeronautical Engineering)
Depositing User: Francis Jayakanth
Date Deposited: 22 Jan 2019 13:03
Last Modified: 30 Jan 2019 05:59
URI: http://eprints.iisc.ac.in/id/eprint/61375

Actions (login required)

View Item View Item