Premium
A model‐free method for identification of mass modifications
Author(s) -
Suwała Grzegorz,
Jankowski Łukasz
Publication year - 2012
Publication title -
structural control and health monitoring
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.587
H-Index - 62
eISSN - 1545-2263
pISSN - 1545-2255
DOI - 10.1002/stc.417
Subject(s) - truss , impulse response , hessian matrix , parametric statistics , parametric model , impulse (physics) , algorithm , sensitivity (control systems) , convergence (economics) , minification , convolution (computer science) , control theory (sociology) , computer science , mathematics , mathematical optimization , engineering , artificial neural network , structural engineering , mathematical analysis , artificial intelligence , statistics , physics , control (management) , quantum mechanics , electronic engineering , economics , economic growth
SUMMARY In this paper, a model‐free methodology for off‐line identification of modifications of structural mass is proposed and verified experimentally. The methodology of the virtual distortion method is used: the modifications are modeled by the equivalent pseudo‐loads that act in the related degrees of freedom of the unmodified structure; their influence on the response is computed using a convolution of the pseudo‐loads with the experimentally obtained local impulse responses. As a result, experimentally measured data are directly used to model the response of the modified structure in a non‐parametric way. The approach obviates the need for a parametric numerical model of the structure and for laborious initial updating of its parameters. Moreover, no topological information about the structure is required, besides potential locations of the modifications. The identification is stated as a problem of minimization of the discrepancy between the measured and the modeled responses of the modified structure. The formulation allows the adjoint variable method to be used for a quick first‐ and second‐order sensitivity analysis, so that Hessian‐based optimization algorithms can be used for fast convergence. The proposed methodology was experimentally verified using a 3D truss structure with 70 elements. Mass modifications in a single node and in two nodes were considered. Given the initially measured local impulse responses, a single sensor and single excitation were sufficient for the identification. Copyright © 2010 John Wiley & Sons, Ltd.