Search

We propose a probabilistic methodology for data-driven updating of non-Gaussian high-dimensional symmetric positive-definite matrices involved in computational models. We cast the data-driven updating as a Bayesian identification of the symmetric positive-definite matrices. The posterior thus obtained exhibits several hyperparameters that control the dispersion of the prior and the weight of the weighted distance that represents the model-data misfit in the likelihood function. Using an identification criterion that quantifies the agreement between the predictions and the data, we identify these hyperparameters so as to obtain not only improved predictions but also a probabilistic representation of model uncertainties. The numerical implementation of the Bayesian inversion by using a Markov chain Monte Carlo (MCMC) method is computationally challenging because the support of the posterior is restricted to a set of symmetric positive-definite matrices and the dimensionality of the problem grows with the square of the matrix dimension and hence can be high. We thus use a transformation of measure to set up the Markov chain in terms of real-valued state variables whose distribution is Gaussian under the prior but non-Gaussian under the posterior. This transformation of measure allows us to sample the posterior using an Itô-SDE-based MCMC method that inherits computational efficiency in high dimension from leveraging the gradient of the posterior. We apply this methodology to a problem of a data-driven updating of a reduced-order model in computational structural dynamics. • Bayesian inversion of reduced matrices of reduced-order models. • MaxEnt prior from random matrix theory for UQ in computational mechanics. • Transformation of measure to enable the application of gradient-informed MCMC. • Use of gradient for computational efficiency in high dimension. • Determination of hyperparameters by means of an identification criterion. [ABSTRACT FROM AUTHOR]

Abstract: This paper presents a probabilistic upscaling of mechanics models. A reduced-order probabilistic model is constructed as a coarse-scale representation of a specified fine-scale model whose probabilistic structure can be accurately determined. Equivalence of the fine- and coarse-scale representations is identified such that a reduction in the requisite degrees of freedom can be achieved while accuracy in certain quantities of interest is maintained. A significant stochastic model reduction can a priori be expected if a separation of spatial and temporal scales exists between the fine- and coarse-scale representations. The upscaling of probabilistic models is subsequently formulated as an optimization problem suitable for practical computations. An illustration in stochastic structural dynamics is provided to demonstrate the proposed framework. [Copyright &y& Elsevier]

Abstract: This paper addresses the inversion of probabilistic models for the dynamical behaviour of structures using experimental data sets of measured frequency-domain transfer functions. The inversion is formulated as the minimization, with respect to the unknown parameters to be identified, of an objective function that measures a distance between the data and the model. Two such distances are proposed, based on either the loglikelihood function, or the relative entropy. As a comprehensive example, a probabilistic model for the dynamical behaviour of a slender beam is inverted using simulated data. The methodology is then applied to a civil and environmental engineering case history involving the identification of a probabilistic model for ground-borne vibrations from real experimental data. [Copyright &y& Elsevier]