Micromechanical modeling of transverse cracking with different methods for generating random fiber distributions

This study presents a modeling investigation into the initiation and propagation of transverse cracking in unidirectional fiber-reinforced composites, focusing on factors like fiber arrangement and matrix properties. Microscale computational models are created to address those challenges. Since existing fiber distribution generators are criticized for not representing features like resin-rich pockets, a deep learning approach is utilized to produce realistic microstructures for finite element modeling. A constitutive material model, incorporating epoxy plasticity and damage, successfully predicts the transverse crack formation and propagation under periodic boundary conditions. The study evaluates the impact of fiber arrangement on cracking behavior using microstructures generated by different methods, revealing similarities and differences in predicted outcomes. This research offers insights into enhancing the accuracy of computational models for simulating transverse cracking in composites.