Improved solution for system identification equations by Epsilon-DecompositionMatrix eigenvalue theory is used to examine the source of ill-conditioning in linear algebraic equations. This approach highlights the crucial role played by the zero and near-zero eigenvalues and corresponding eigenvectors of poorly conditioned systems. Insight gained from this approach is used to significantly improve a recently developed solution procedure called Epsilon-Decomposition (E-D). E-D is an efficient alternative to Singular Value Decomposition (SVD) for ill-conditioned systems arising in parameter estimation and system identification studies. The efficiency of the improved E-D over SVD resides in the need to only obtain the zero and near-zero eigenvalues of the coefficient matrix as opposed to all of its eigenvalues and vectors (as required by SVD). Thus, the efficiency of E-D is significant for large matrices with small rank deficiency.
Document ID
19900042235
Acquisition Source
Legacy CDMS
Document Type
Conference Paper
Authors
Ojalvo, Irving U. (Bridgeport, University CT, United States)
Date Acquired
August 14, 2013
Publication Date
January 1, 1990
Subject Category
Numerical Analysis
Report/Patent Number
AIAA PAPER 90-1146
Meeting Information
Meeting: AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference