Tagungsband - DAGA 2024
Beitrag
A Framework for the Identification of Spatially Varying Elastic Material Properties via Modal Data
Zusammenfassung:
The steady increase in computational power and the advent of functionally graded materials give rise to characterization methods for heterogeneous materials. Confining the context to non-destructive testing sets the stage for methods like digital image correlation or digital volume correlation. Next to these overkill approaches, guided waves and data from modal analysis present fundamentally feasible approaches that are not mature yet. This work aims to develop a numerical method for identifying the spatially varying elastic material properties of a structure using modal data while considering uncertainty. We achieve this by using Bayesian inference, the finite element method, the generalized polynomial chaos expansion, the Karhunen-Loève expansion, and develop an acceleration modification and a generalization for several quantities of interest while proposing a stochastic material model for wood. The acceleration of the procedure implies that, once the polynomial chaos surrogate is trained, it could potentially be applied to an unlimited number of altered configurations. Stumping intuition, we show that providing the cross-correlation of multiple unknowns a priori is not necessarily beneficial. We present the first realistic covariance model used for describing wood by a random field. This facilitates non-destructive testing for this application.