From Fiber Architecture to Digital Twins: Characterization, Orientation Modeling, and Artificial Intelligence in Short Fiber-Reinforced Polymer Composites

Short fiber-reinforced polymer composites (SFRPCs) are widely used in lightweight components because they combine low density, good mechanical performance, and high-volume processability. Their local properties, however, are governed by a fiber architecture—fiber length, orientation, dispersion, and defects—that is created during processing and varies within each molded part. This review examines how this architecture can be measured, predicted, and carried into digital representations of the manufacturing process. It covers the mechanisms that form the orientation state during injection molding; two- and three-dimensional characterization, with emphasis on X-ray micro-computed tomography; orientation models from the Jeffery and Folgar–Tucker equations to anisotropic rotary diffusion and slow-kinetics formulations; and multiscale homogenization. Applications of artificial intelligence are then assessed for image segmentation, orientation prediction, micromechanical surrogates, and physics-informed learning. Finally, published Digital Twin implementations for composite and polymer manufacturing are classified by their level of data integration, and a layered architecture for SFRPCs is proposed. Most existing systems are digital models or digital shadows rather than bidirectionally coupled twins. The main obstacles are the scarcity of paired process–microstructure–property datasets, the lack of uncertainty propagation across scales, and the limited transferability of data-driven models.