A second-moment approach for direct probabilistic model updating in structural dynamics

Discrepancies between experimentally measured data and computational predictions are unavoidable for structural dynamic systems. Model updating methods have been developed over the past three decades to reduce this gap. Well established model updating methods exist when both the model and experimental measurements are deterministic in nature. However in reality, experimental results may contain uncertainty, for example arising due to unknown experimental errors, or variability in nominally identical structures. Over the past two decades probabilistic approaches have been developed to incorporate uncertainties in computational models. In this paper, the natural frequencies and the eigenvectors of the system are measured and assumed to be uncertain. A random matrix approach is proposed and closed-form expressions are derived for the mean matrix and the covariance matrix of the updated stiffness matrix. A perturbation technique is used to obtain a usable expression for the covariance matrix.The new method is illustrated by three numerical examples highlighting the influence of the eigenfrequency uncertainties on the mean matrix and the influence of the eigenvector uncertainties on the covariance matrix.

► A direct stochastic updating is performed to account for experimental uncertainties.
► The mean and the covariance matrices of the updated stiffness matrix are derived.
► A perturbation technique has been used to obtain the covariance matrix.
► The mean stiffness matrix is found sensitive to the eigenfrequency distribution.
► The covariance stiffness matrix depends mainly on the eigenvector distribution.