Micro-CT based structure tensor analysis of fibre orientation in random fibre composites versus high-fidelity fibre identification methods

Abstract

The fibre orientation distribution controls the mechanical properties of random fibre composites. Generally accepted methods for its characterisation involve identification of fibres or their ellipsoidal cross sections as individual objects, requiring high image resolution and high computational resources. This paper investigates whether structure tensor analysis can be an alternative and whether it can work with lower resolution images. Micro-computed X-ray tomography images of random glass fibre/polypropylene injection moulded composites were processed using ellipsometry on 2D slices, 3D fibre identification (Avizo software) and analysis of the structure tensor (VoxTex software). The images had resolutions of 1.4, 3.2, 8 and 16 µm per pixel, compared to an average glass fibre diameter of 17 µm. All the methods yielded similar results for high-resolution images (1.4 and 3.2 µm). The high-fidelity, direct identification of fibres failed for low-resolution images, but the structure tensor analysis still yielded results close to the high-resolution scans.