Abstract

Model updating techniques are widely used by analysts in order toverify and eventually improve the correlation between analytical and experimental results.The application of such techniques in large scale models is made difficult by the inverse nature of the problem which allows multiple solutions. Also the increasing computational power of computers and theimprovement of analytic algorithms to manage large problems generates an accelerated tendency to over-discretize the FE models, which reduces even more the ratio between measured and analytical degrees of freedom, and increase substantially the numbers of potential adjusting parameters. Thus, the situation points towards the use of reduction techniques to solve the matching incompatibility without loosing accuracy of the analytical results. This paper deals with the tuning of FE models using experimental measures. It considers the use of appropriate cost functions that express the discrepancies between the analytical and experimental models in the modal space. The technique is successfully implemented in a general purpose optimization package used for industrial applications. This environment allows an open choice of the design parameters, and to perform easily parametric studies, statistical analyses, multi-objective optimizations.