FibriPy: a software environment for fiber analysis from 3D micro-computed tomography data

T. Perciano, D. Ushizima, H. Krishnan, J. Sethian
Lawrence Berkeley National Laboratory,
United States

Keywords: 3D micro-computed tomography, fibers, manufacturing, FibriPy


Materials characterization using X-ray imaging has enabled development of advanced composites that will open up new opportunities to improve manufacturing. In order to deploy reinforced materials, such as ceramic matrix composites (CMC), several structures must be accounted for at all times during tensile or compressive loading experiments. In collaboration with DOE imaging facilities, we have investigated data generated at X-ray beamlines for non-destructive three-dimensional characterization of objects. Experiments using X-ray are widely performed in academia and industry including medical imaging, material science, electronics, geology, and others. As a result of joint data exploration, we have developed FibriPy, a computer-aided fiber detector, tracker and analyzer designed to recognize patterns from micro-computed tomography (microCT). FibriPy is a software environment that combines user-friendly dashboards with programmable functions to support the analysis automation of 3D microCT stacks. While our experiments showcase CMC samples and SiC fibers, FibriPy handles fiber detection and deformations tracking from other samples, such as concrete toughened with fibers. As scalability is one of the main issues in analyzing high-resolution microCT, FibriPy delivers process-based concurrency and scalable GPU-based visualization, with portability benefits brought by well-established Python packages. FibriPy provides tools for image enhancement, automatic detection of fiber cross-sections, interactive tools to improve fiber detection, automatic fiber tracking (connection of detected fiber cross-section), and 2D and 3D visualization. The main contributions are: (a) Python-centric multi-threading software architecture (b) advanced algorithms for image analysis and feature extraction, such as: structure-based image enhancement using nonlinear filtering, adaptive feature-based matching and learning, and motion tracking based on cross-correlation, and (c) GPU-accelerated visualization tools. Preliminary results show that FibriPy succeeds at detecting fibers from CMC samples (85% on average) automatically, and almost no false positive detections.