Experimental identification of stochastic processes using an uncertain computational non-linear dynamical model

Abstract

The problem presented deals with tubes bundles in Pressurized Water Reactors. The final objective is to identify a model of the external loads applied to theses tubes bundles through the knowledge of dynamical responses. In complex dynamical systems, such an identification is difficult due to the size of the computational model and due to the high number of parameters to be identified. As a consequence, a simplified computational model is constructed. The introduction of such a simplified model introduces model uncertainties. We are first interested in the implementation (modelling and identification) of a probabilistic approach of uncertainties in the mean computational model using the non-parametric probabilistic approach for parameter uncertainties and model uncertainties. In addition, a probabilistic model for the stochastic loads is constructed to take into account model uncertainties in the probabilistic model of the stochastic loads. Finally, the non-linear stochastic dynamical system submited to the uncertain stochastic loads is used to identify the probability model of its uncertainties. In a first part, the theory is presented. The second part is devoted to the validation of the theory in presenting an application. © 2008 IOP Publishing Ltd.