Use of the singular value decomposition method to detect ill-conditioning of structural identification problems

✍ Scribed by W.M. Hasan; E. Viola

• Publisher: Elsevier Science
• Year: 1997
• Tongue: English
• Weight: 676 KB
• Volume: 63
• Category: Article
• ISSN: 0045-7949
• DOI: 10.1016/s0045-7949(96)00361-6

✦ Synopsis

AWract-Parametric identification problems, often, require solving systems of algebraic equations via mversion of matrices of noisy data. Generally, these problems are badly conditioned and, therefore, small perturbations in the experimental data result in magnified errors in the ident&ed parameters. In this paper the singular value decomposition method is used to investigate ill-conditioning of two time domain, physical and modal, identification methods. It is seen that ill-conditioning is influenced by many factors: type and pomt of application of the excitation, choice of the paramters to be identified, choice of the experimental data and sampling interval of the response. A 3 d.o.f system is used to demonstrate the convergence of the physical-parameters identification method using quantities which enhance the conditton of the problem. It is not our aim to establish rigorous rules between condition number and parameters of the two methods examined, since ill-conditioning may vary from one problem to another. Our aim is to address the attention to the important issue of selecting the optimal quantities which make the tdcntifratton problem well-conditioned.