Premium
Back‐Transformation into Physical Configuration Space after Model Order Reduction onto a General Subspace
Author(s) -
Lein Claudius,
Beitelschmidt Michael
Publication year - 2015
Publication title -
pamm
Language(s) - English
Resource type - Journals
ISSN - 1617-7061
DOI - 10.1002/pamm.201510021
Subject(s) - subspace topology , linear subspace , model order reduction , robustness (evolution) , computer science , algorithm , projection (relational algebra) , transformation (genetics) , transformation matrix , mathematical optimization , theoretical computer science , mathematics , artificial intelligence , geometry , kinematics , biochemistry , chemistry , physics , classical mechanics , gene
Model order reduction (MOR) techniques that project onto a general subspace are common practice. In Elastic Multi‐Body‐Dynamics (EMBD) distinct interface coordinates are required for the interconnection of the elastic structure and the multi‐body system. For that reason, a physical interpretation of the coordinates is mandatory, which is not the case for reduced order models in a general subspace. To make alternative MOR techniques accessible for EMBD, a back‐projection approach was introduced by [1]. Therein, the system is projected back onto the physical configuration space, which requires the inversion of the master partition of the projection matrix. But the procedure lacks of robustness and generality. A novel approach is introduced by generating additional master coordinates using sensor placement methods, e.g. the Effective‐Independence‐algorithm (EfI) [2]. By using a rank criterion for the automatic selection of additional coordinates, the improved back‐transformation performs properly and without damaging the reduced order model at a fairly small computational overhead, which is demonstrated at the example of a gear box housing. (© 2015 Wiley‐VCH Verlag GmbH & Co. KGaA, Weinheim)