This work presents a novel, non-destructive approach for determining fiber length distribution using CT scans, offering significant advantages over traditional experimental methods. Fiber length is a critical parameter in materials like composites, and conventional techniques are time-consuming and prone to bias. In contrast, the digital method described here provides accurate, reproducible results with minimal effort and allows for repeated analysis of materials over time to study aging effects.

The process begins with a CT scan of a sample, followed by image processing and segmentation to distinguish fibers, pores, and the matrix. Using AI, specifically a U-NET3D neural network, the centerlines of fibers are identified, enabling the calculation of fiber length and other key properties such as curvature and diameter. The method was applied to a GF/PA6 injection molding sample, and the digital fiber length distribution was compared with results from the traditional FASEP experimental apparatus.

The comparison shows a high correlation between the two methods, highlighting the effectiveness and accuracy of this digital approach. This technique offers a fast, reliable, and non-invasive way to assess fiber length distribution, with potential applications in monitoring material performance and aging in composite materials.