Model-reduction techniques for Bayesian finite element model updating using dynamic response data

Reliability sensitivity analysis of stochastic finite element models

“…To cope with this difficulty the proposed reliability sensitivity analysis is combined with a model reduction technique in the present formulation. In particular, a method based on component-mode synthesis is implemented in order to define a reduced-order model for the structural system [14,15,17,16]. For completeness, some of the basic equations involved in the model reduction technique are reviewed in the following two subsections.…”

“…Note that this transformation matrix maps the mode shapes of the reduced-order model to the independent physical coordinates of the original unreduced model. The solution for the modal responses together with the non-linear equation for the evolution of the set of variables y(t, w, θ) can be obtained by an appropriate step-by-step integration scheme [32,17]. Finally, it is observed that the previous representation of the solution can be extended for the case of non-classically damped systems by recasting the equation of motion of the reduced-order system into a first-order state-space form or by solving a quadratic eigenvalue problem [33][34][35].…”

“…More specifically, after the division of the structure into components the model reduction technique involves two basic steps: definition of sets of component modes; and coupling of the component-mode models to form a reduced-order system model. If the division of the structure into components is guided by a certain parametrization scheme in terms of the system parameters, substantial computational savings can be achieved in estimating the reliability sensitivity measures avoiding system reanalyses for different values of the parameters [17,16]. Moreover, the drastic reduction in computational efforts is obtained without compromising the accuracy of the sensitivity estimates.…”

Finite element model updating of a space vehicle first stage motor based on experimental test results

“…The experiment results and the FEA results from the software will be considered for updating. The selection of the parameters for updation is crucial because the FE model of the real structure is affected by updating the selected parameters [10]. The important issues are the number of preferred and selected parameters from the set.…”

Methodology for model updating of mechanical components with local nonlinearities

“…Another model updating strategy proposed by [11] for nonlinear vibrations of structures, is based on proper orthogonal decomposition and its nonlinear generalizations as well as auto-associative neural networks [9] used data from relatively large-scale experimental soil-foundationsuperstructure interaction (SFSI) systems to develop reduced-order computational models for response prediction, employing trained neural networks [10] applied a particle filtering algorithm on experimentally measured tip accelerations using Bayesian principles to estimate the changes in damping and flexural rigidity of a beam. Another Bayesian approach was proposed by Jensen et al [12], wherein a Bayesian FE model updating strategy using dynamic response data is employed for structural response prediction.…”