Predicting Mechanical Properties of Unidirectional Composites Using Machine Learning

The mechanical properties of unidirectional composites with different fiber arrays and volume fractions were predicted using machine learning. Repeating unit cell (RUC) containing randomly distributed fibers were used to represent the complex microstructures of the unidirectional composites. The effective elastic constants of the RUC with periodic boundary conditions were evaluated using high-fidelity generalized method of cells (HFGMC) micromechanical analysis. Data sets relating the microstructural images of the fiber composites to their corresponding effective properties were used to train a convolutional neural network (CNN) model. To validate the accuracy of the trained CNN model, the properties of the unidirectional composites with fiber volume fractions from 15 to 70% were modeled using HFGMC, and the results were compared with the CNN predictions. The differences were less than 3%, indicating that the machine learning network can accurately characterize the elastic constants of unidirectional composites with microstructural configurations.