An Artificial Intelligence and Image Processing based approchs for evaluating Micro-scale Mechanical Properties Variations in Composite fibre bundle

The variations in microstructural characteristics of composite fiber bundles significantly affect their mechanical performance. In this study, we propose an approach integrating image processing techniques and artificial intelligence (AI) to evaluate the micro-scale variations in the mechanical properties of composite fiber bundles. Microstructure image cross-sections from different positions in the fiber bundle are binarized and analyzed using image segmentation methods such as Otsu thresholding and watershed segmentation to extract key characteristics, including the local volume fraction (V<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">f</inf>) and average fiber radius (R<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">f</inf>). The Random Sequential Absorption (RSA) algorithm is applied, with the extracted characteristics as inputs, to model the 3D Representative Volume Element (RVE) in Abaqus. Homogenization techniques using the Finite Element Method (FEM) are employed to compute the local mechanical properties. Three machine learning models—Neural Network (NN), Random Forest (RF), and Support Vector Regression (SVR)—are trained on the V<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">f</inf> values (ranging from 0.14 to 0.46) to predict the local effective mechanical properties, including longitudinal and transverse elastic moduli (E<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf>, E<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>). The results demonstrate that the Random Forest Model outperforms the other models in terms of prediction accuracy, achieving R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> scores of 0.99 and 0.98 for E<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> and E<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> respectively. This highlights the effectiveness of the proposed method in capturing variations within fiber bundles and accurately predicting their associated mechanical properties while significantly reducing the computational cost associated with extensive FEM simulations.