Stochastic modeling of uncertainties in computational dynamics and applications

Christian Soize (1)

1 LaM - Laboratoire de Mécanique

Ageneric microscale stochastic model is presented for a large class of randomanisotropic elastic microstructures allowing a parametric analysis of the RepresentativeVolume Element (RVE) size to be performed. This new approach can be useful for a direct experimental identification of random anisotropic elastic microstructures when the standard method cannot easily be applied. Such a RVE is used to construct the macroscopic properties in the context of stochastic homogenization . The probability analysis is not performed as usual for a given particular random microstructure defined in terms of its constituents, but is performed for a large class of random anisotropic elastic microstructures. For this class, the probability distribution of the random effective stiffness tensor is explicitly constructed and allows a full probability analysis of the RVE size to be carried out with respect to the spatial correlation length of the random microstructure.

Mots clés: uncertaintiy quantification; computational dynamics; nonparametric probabilistic method; modeling errors; model uncertainties; random matrix; statistical inverse problem; random unertainties

Domaines: Probabilités [math.PR]; Vibrations [physics.class-ph]; Vibrations [physics.class-ph]; Acoustique [physics.class-ph]; Acoustique [physics.class-ph]

Christian Soize. Stochastic modeling of uncertainties in computational dynamics and applications. 2nd LNCC Meeting on Computational Modelling, Aug 2006, Petropolis, RJ, Brazil. pp.1-19 (Keynote Lecture). ⟨hal-00689703⟩