A data-driven self-consistent clustering analysis for the progressive damage behavior of 3D braided composites

Abstract

A data-driven self-consistent clustering analysis (SCA) method is applied to investigate the progressive damage behavior of 3D braided composites. The SCA-based method is split into the offline stage and the online stage. In the offline stage, the high fidelity RVE is compressed into a reduced RVE composed of several clusters. In the online stage, the mechanical responses are calculated by solving the discretized Lippmann-Schwinger integral equation. To validate the accuracy of proposed model, the SCA-based simulation is compared with the corresponding experiments and finite element analysis (FEA). The results show that the SCA method can accurately capture the stress and damage distribution, and the predictive stiffness and strength agree well with experimental data. More importantly, with the same constitutive laws and geometric model, SCA only takes a few hundred seconds, which is 1771 times faster than FEA. Because of the high efficiency, the SCA has the potential to be applied in concurrent multiscale analysis for braided composites.