CFD and Machine Learning Approaches for Predicting Air Permeability in Technical Textiles

Predicting the thermo-physiological comfort of technical clothing requires an understanding of how microscopic textile structures influence macroscopic properties such as air, heat, and moisture permeability. This work represents the first step towards a multi-scale predictive tool capable of estimating key comfort-related properties from the geometrical features of woven fabrics. Focusing on air permeability, the effect of structural and design parameters was investigated while keeping the fibre material (cotton) constant. A computational framework that combines validated Computational Fluid Dynamics (CFD) simulations with a Fully Connected Neural Network (FCNN) was developed, enabling fast and accurate predictions before production. The CFD model accounts for both intra- and inter-yarn porosity, ensuring reliability across a wide range of fabric configurations. The FCNN, trained on simulation and literature data, achieved a mean absolute relative error of 2.01% and a maximum error of 7.72%, demonstrating excellent agreement with experimental results. The analysis highlights how weave type and yarn density govern airflow resistance, offering an efficient tool for the design and optimisation of breathable technical textiles.