Inverse problems in stochastic computational dynamics

Abstract : This paper deals with robust updating of dynamical systems using stochastic computational models for which model and parameter uncertainties are taken into account by the nonparametric probabilistic approach. Such a problem is formulated as an inverse problem consisting in identifying the parameters of the mean computational model and the parameters of the probabilistic model of uncertainties. This inverse problem leads us to solve an optimization problem for which the objective function describes the capability of the uncertain computational model to best-fit the experimental data. Two objective functions are proposed. The methodology is applied in the context of the robust updating of a computational model of composite sandwich panels in the low- and medium- frequency ranges for which experimental results are available.