Fibre phantom generation using FibreSimulator: an open-source Python tool

Area of Science:

• Materials Science
• Computational Imaging
• Polymer Composites

Background:

• Fibre-reinforced polymers (FRPs) are crucial engineering materials, requiring internal structure visualization for mechanical property assessment.
• Computed tomography (CT) is a key non-destructive imaging technique for FRPs, but machine learning methods need extensive, realistic datasets.
• Acquiring experimental CT datasets for composites is difficult due to high costs and lack of ground truth data.

Purpose of the Study:

• To develop an open-source Python tool, FibreSimulator, for generating labelled 3D synthetic phantoms of unidirectional fibre-reinforced polymers.
• To enable simulation of tomographic scans with customizable material properties, fibre orientations, and geometrical features.
• To facilitate systematic and quantitative investigation of how CT acquisition parameters influence reconstruction and segmentation.

Main Methods:

• Developed FibreSimulator, an open-source Python tool for creating 3D synthetic phantoms of FRPs.
• Integrated FibreSimulator with the ASTRA toolbox to simulate realistic CT scans of the generated phantoms.
• Conducted simulated experiments to evaluate the impact of varying X-ray projections and beam intensity on reconstruction and segmentation.

Main Results:

• Simulated CT scans showed that reduced projections and lower beam intensity blur fibre boundaries, overestimate diameters, and decrease fibre detection.
• FibreSimulator enabled quantitative analysis of artefacts like undersampling and fibre overlap, revealing nonlinear trends in detection accuracy.
• The tool demonstrated the ability to systematically study the effects of acquisition parameters and composite properties under controlled, ground-truth conditions.

Conclusions:

• FibreSimulator provides a valuable, controllable platform for generating synthetic data crucial for advancing CT-based analysis of fibre-reinforced polymers.
• The tool aids in optimizing CT parameters and guiding experimental design, overcoming limitations of purely experimental approaches.
• Enables quantitative investigation of imaging artefacts and their impact on material characterization, crucial for materials science research.