Composite Structures 382 (2026) 120107

Keywords: Ceramic fiber fabric, Unit cell,   Artificial neural network

Ceramic fiber fabrics are vital for   high-temperature morphing skins due to their exceptional thermal stability   and structural adaptability. However, their mechanical properties are   strongly influenced by weave architecture, necessitating detailed and systematic   characterization. Current challenges include the lack of robust predictive   theoretical models and the inefficiencies of experimental methods. This study   tackles these issues by developing a parametric modeling framework for 2D   woven fabrics using three topological parameters, combined with an automated   simulation system to evaluate tensile and shear properties through   Python-driven numerical analysis. The framework demonstrates high predictive   accuracy, validated by experimental data. Additionally, an artificial neural   network (ANN) surrogate model employs the resulting property database to   reveal correlations between weave architecture and mechanical properties. A   novel integrated resistance factor is introduced to comprehensively assess   mechanical performance, identifying plain weave architectures as optimal for   combined tensile and shear resistance. This ANN-based surrogate model   approach significantly improves efficiency in material design and performance   prediction.

Zhou Jiang, Mingming Xu, Jian Sun,   Jinsong Leng

Rapid mechanical prediction of woven ceramic fabrics via a neural networksurrogate model based on the parameterized unit cell.pdf