Etude probabiliste du comportement non-linéaire des brasures utilisées dans les modules de puissance

Abstract

The objective of this paper is to develop a general framework or updating the predictions of models of structures using observations gathered from the monitoring of these structures. A general Bayesian updating scheme is developed, combining prior information on model parameters and monitoring data (including measurement uncertainties). A Markov chain Monte-Carlo (MCMC) sampling method is used for computing the posterior probability density functions (PDF) of input random variables. Then the updated PDF of the response quantities of interest is computed from the posterior PDFs. The approach is illustrated on the example of fatigue crack growth under homogeneous cyclic loading.