Parameter identification of structural dynamic models by inverse statistical analysis

Abstract

Computational model updating techniques are well established in the field of structural dynamics. These methods make use of experimental modal data to correct numerical models for further investigations. Nevertheless, the updated model only reflects the dynamical behaviour of the structure that was tested. There are many things to mention which will certainly lead to different test results: different environmental conditions during testing, a slightly modified test set-up, a test on a nominal identical but different structure (from serial production), etc. However, to perform stochastic simulations a number of assumptions must be made to establish the uncertain input. Consequently, an inverse approach is needed which is capable to identify a stochastic numerical model from uncertain experimental data. This work focuses on the development of a parameter-based approach for stochastic model updating in the field of structural dynamics using first order sensitivities. Mean value parameters and a full covariance matrix of the numerical model can be identified from uncertain experimental modal data. Therefore experimental eigenfrequencies and mode shapes with mean values and convariances are necessary.