Multiscale modeling of 3D braided CMCs with void defects using unsupervised machine learning
     Topic(s) : Material science

    Co-authors​ :

     Xinyi SONG (CHINA) 

    Abstract :
    A novel multiscale damage analysis approach is proposed for 3D braided composite with void defects, utilizing unsupervised machine method. To implement the multiscale damage model, a user-defined material subroutine (UMAT) is developed. Experimental validation demonstrates the accuracy of this approach in capturing damage modes. Furthermore, the framework predicts the mechanical properties of braided composites with various porosities. The relationships between porosity, modulus reduction, and strength reduction are established through rigorous statistical analysis. This method provides a valuable tool for the efficient visualization of mechanical analyses of 3D braided composite materials with void defects, based on accurate geometric information. This approach offers significant advancements in the field of composite material damage analysis for further research in this area.