Modeling of Fiber Structures

The Generate Nonwoven Statistical Digital Twin GeoApp was originally designed to create digital twins of nonwoven structures. Today, however, its capabilities extend far beyond this: it is equally well-suited to modelling a wide range of fiber-based materials, including fiber-reinforced composites, fiber-based gas diffusion layers in fuel cells, and various types of filter media.

Before generating a digital twin, the original sample is prepared in GeoDict. This process involves importing and segmenting the CT-scans of the original structure using ImportGeo-Vol, after which the sample is analyzed using either FiberFind-AI or by beginning the analysis of a pre-segmented sample directly.

Although creating structures with curved fibres is challenging, this GeoApp simplifies the process by automatically optimizing the digital twin to match the stochastic geometric properties of the original material.

The process begins with selecting the target's result file, IdentifyFibers.gdr, created by FiberFind-AI, and setting up the optimization parameters. While maintaining fixed values for key characteristics such as fiber diameter, length, orientation, and fiber volume fraction, the GeoApp dynamically adjusts fiber curvature and curliness. These adjustments ensure that the final digital twin accurately replicates the unique features of the original sample.

[1] Grießer A., Westerteiger R., Glatt E., Hagen H., Wiegmann A., Identification and analysis of fibers in ultra-large micro-CT scans of nonwoven textiles using deep learning, The Journal of The Textile Institute, 114(11), 1647–1657, 2023. https://www.tandfonline.com/doi/full/10.1080/00405000.2022.2145429

[2] Grießer A., Westerteiger R., Glatt E., Hagen H., Wiegmann A., Deep learning based segmentation of binder and fibers in gas diffusion layers, Next Materials, Volume 6, January 2025, 100411. https://www.sciencedirect.com/science/article/pii/S2949822824003083