Characterizing failure morphologies in fiber-reinforced composites via k-means clustering based multiscale framework

Abstract: A novel homogenization methodology is proposed for analyzing the failure of fiber-reinforced composite materials, utilizing elastic and eigen influence tensors within a damage informed transformation field analysis (D-TFA) framework. This approach includes a technique for calculating macroscopic damage under uniform stress and strain conditions, offering more realistic simulations. Computational efficiency is enhanced through a reduced-order modeling strategy, while elastic and eigen strain distribution driven k-means clustering methods are employed to partition the microscale domain. The model's performance is assessed by simulating the response of a representative volume element (RVE) treated as a homogenized continuum. Subsequently, a comparative assessment is carried out to check the efficacy of two clustering schemes. Damage morphologies are calculated using proposed framework and compared with predictions obtained using finite element method. Furthermore, open-hole specimen tests are simulated and failure paths are predicted for the domains with different fiber layups. Ultimately, we show that D-TFA can accurately capture damage patterns and directional strengths, providing improved predictions of the mechanical behavior of composite materials. It has been demonstrated that higher cluster counts are crucial for capturing a more accurate stress-strain response, especially for complex microstructures.