Time-domain formulation in computational dynamics for linear viscoelastic media with model uncertainties and stochastic excitation

The paper is devoted to the computational time-domain formulation of linear viscoelastic systems submitted to a nonstationary stochastic excitation and in presence of model uncertainties which are modeled in the framework of the probability theory. The objective is to introduce and to develop an adapted and complete formulation of such a problem in the context of computational mechanics. A reduced-order model in the time domain with stochastic excitation is derived from the computational model. For the reduced-order model, the stochastic modeling of both computational model-parameters uncertainties and modeling errors is carried out using the nonparametric probabilistic approach and the random matrix theory. We present a new formulation of model uncertainties to construct the random operators for viscoelastic media. We then obtained a linear Stochastic Integro-Differential Equation (SIDE) with random operators and with a stochastic nonhomogeneous part (stochastic excitation). A time discretization of this SIDE is proposed. In a first step, the SIDE is transformed to a linear Itô Stochastic Differential Equation (ISDE) with random operators. Then the ISDE is discretized using an extension of the Störmer-Verlet scheme which is a particularly well adapted algorithm for long-time good behavior of the numerical solution. Finally, for the stochastic solver and statistical estimations of the random responses, we propose to use the Monte Carlo simulation for Gaussian and non-Gaussian excitations. ∗Corresponding author. Email addresses: [email protected] (C. Soize), [email protected] (I. E. Poloskov) Preprint submitted to Computers and Mathematics with Applications October 11, 2012 ha l-0 07 46 28 0, v er si on 1 28 O ct 2 01 2 Author manuscript, published in "Computers and Mathematics with Applications 64, 11 (2012) Pages: 3594-3612" DOI : 10.1016/j.camwa.2012.09.010