Towards Identification of Lower Scale Composite Material Properties From Higher Scale Experimental Data via Inverse Analysis of Coupled Multiscale Models

In the present paper we demonstrate the application of a multiscale inverse methodology for identifying material properties of the constituents of a selected composite material with long fibers embedded in a polymer matrix by utilizing macro-scale experimental data. Taking advantage of a computational homogenization technique for periodic microstructures, the proposed optimization methodology allows, for the determination of a considerable number of the elastic properties of the composite material at the micro-scale of the constituents and their interface zone. Our approach describes the theoretical development and numerical implementation of a multi-scale modeling chain of the composite, extending from the periodic microstructure represented by a suitable unit cell and subjected to appropriate periodic boundary conditions at the micro scale, to the composite lamina at the meso-scale, to the laminated, multi-axially loaded material at the macro-scale. By applying the proposed methodology, we have been able to accurately calculate several fiber, matrix properties by utilizing properly generated synthetic data of the macro-scale behavior of the composite laminate. Furthermore, in an effort to explore the potential of our method for identifying quantities that manifest only after manufacturing including damage quantities at the micro-scale, we have initiated an effort to explore the capability of determining fiber-matrix interfacial properties and have demonstrated initial success.