This paper addresses the problem of parametric correction of condensed finite element models using test data. A technique is proposed to correct individual elements or components within each condensed subdomain. It is shown how the original updating algorithm can be applied to models formed of condensed substructures with the additional cost of solving linear systems for obtaining contributions of the condensed subdomains to the derivatives of the objective function but the added capability of identifying modeling errors within the subdomains. The applicability of this technique to the updating of test-analysis correlation models is discussed when finite element matrices are condensed to a size intermediate between the experimental model and the full-order model. The procedure is illustrated using simulated data and it enables the identification of two disconnected modeling errors within the same subdomain without having to perform multiple condensations at each updating iteration.
展开▼
