Building Reliable Digital Nonwoven Models

Digital material models enable the virtual investigation and optimization of nonwoven materials, reducing the need for physical prototypes and allowing a much broader exploration of the design space. However, meaningful simulation results require digital models that reliably represent the relevant structural and physical properties of the real material. 

This presentation introduces a workflow for creating reliable stochastic digital twins of nonwoven materials with GeoDict. Starting from imaging data such as µCT scans, it discusses the transition from gray-value images to segmented material models and subsequently to stochastic models that describe the statistical properties of the fiber structure. Particular attention is given to parameters such as fiber diameter and orientation distributions, density variations, porosity, grammage, and structural inhomogeneity. 

The presentation also addresses practical challenges that strongly influence model quality, including sample preparation, segmentation artifacts, the trade-off between scan resolution and field of view, and the selection of a representative volume for analysis and simulation. 

Finally, strategies for verification and validation are presented. Generated structures are compared with the original material through visual inspection, geometric characterization, and numerical simulation of relevant physical properties. This iterative approach makes it possible to refine stochastic model parameters until the digital structure reproduces the properties that matter for the intended application. 

The resulting digital twins provide a foundation for studying material modifications virtually and for using simulation to support the development and optimization of nonwoven materials.