International audienceThe context of this work is the setting up of decision aiding tools in robust design of mechanical structures in an uncertain environment. We propose to study the inverse problem of the parametric identification of mechanical structures starting from uncertain observations and to appreciate if the information gain obtained thanks to possible experimental measurements justifies the cost of these last. In this study, we use a probabilistic formulation of the inverse problems which combines, in the form of probability distribution, a prior information on the parameters and models of uncertainties on the real or simulated measurements. The a posteriori distribution permit to generate models thanks to a Monte Carlo method. ...