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

Fibre-reinforced polymer composites are utilized across many industries for their stiffness and strength. Visualization of their internal structure is critical to understand their mechanical properties, with computed tomography serving as a popular non-destructive method. While classical tomographic reconstruction algorithms do not rely on training data, modern machine-learning-based methods require large datasets that realistically reflect experimental imaging conditions. However, acquiring such datasets for composites is challenging due to limited access to the ground truth and the high cost of repeated scans. There is therefore a need for realistic, controllable and labelled synthetic phantoms. To address this gap, we present FibreSimulator, an open-source Python tool that generates 3D synthetic phantoms of unidirectional fibre-reinforced polymers with customizable material properties, fibre orientations and geometrical features. The simulator is integrated with the ASTRA toolbox to simulate tomographic scans of the generated phantoms. We demonstrate its capacity through experiments evaluating how tomographic scan settings affect reconstruction and segmentation. Simulated tomographic experiments show that reducing the number of X-ray projections or lowering beam intensity (i.e. increasing noise) leads to blurred fibre boundaries, overestimated fibre diameters and fewer detectable fibres. While these qualitative effects are well known, FibreSimulator enables their systematic and quantitative investigation under fully controlled conditions, with access to the ground truth. In particular, by varying acquisition parameters and composite properties, we observe a nonlinear trend in detection accuracy arising from undersampling and fibre-overlap artefacts. Such controlled studies are difficult to perform experimentally due to cost, limited repeatability and absence of the ground truth. These results highlight FibreSimulator as a valuable tool for optimizing parameters and guiding experimental design under controllable conditions.