BAYESIAN CALIBRATION OF A FINITE ELEMENT C-SPAR MODEL USING DIGITAL IMAGE CORRELATION

Bayesian calibration is a statistical framework for combining data from experimental tests and numerical models, while formally accounting for uncertainty due to: i) unknown model inputs, ii) experimental observation errors, and iii) inaccuracies in the assumed model physics. Accurately quantifying such uncertainty can help ensure consistency when comparing Finite Element models with structural test data, and provide statistical confidence metrics in the predictions made by these models. This capability will be invaluable in enabling improved aircraft certification processes informed by virtual testing using combined data from mathematical models and component-level structural tests. In this paper, Digital Image Correlation (DIC) data from preliminary compression tests of a composite C-spar are used to calibrate a Finite Element model of the spar, implemented in ABAQUS. Torsional springs are used to model the imperfectly clamped boundary conditions of the test specimen, which are considered uncertain. The aim of this study is to learn about the stiffness of these springs, as well as an uncertain longitudinal modulus and ply thickness. The primary contribution of this paper is to address challenges associated with calibrating the full-field nodal displacement output using the high-dimensional data produced by data-rich structural tests, in a efficient approach using Gaussian process emulators. Fitting the model to such a large volume of data is the key challenge addressed.