DATA-DRIVEN CHARACTERIZATION OF COMPOSITES BASED ON VIRTUAL DETERMINISTIC AND NOISY MULTIAXIAL DATA

摘要

This paper presents an inverse methodology capable of identifying the elastic moduli of laminated composites from both deterministic and noisy data originating from virtual multiaxial tests. Unlike the conventional uniaxial characterization of materials, the methodology exploits the energy balance between the increment of external work and the corresponding increment of strain energy. It then formulates an overdetermined system of linear equations that are solved using Singular Value Decomposition (SVD) to compute the associated pseudoinverse array. The proposed methodology further controls the multiaxial testing machine by utilizing performance measures of the SVD process to construct objective functions that are maximized in order to compute loading path design variables. Numerical examples investigate the significance, robustness and efficiency of the proposed methodology. Deterministic and noisy data are synthesized in order to demonstrate the applicability of the technique with respect to realistic characterization problems. The effect of noisy data in the characterization process has been examined in a manner that leads to a demonstration of the practicality of the approach.