Multiscale modeling of textile-reinforced concrete shells

Licentiate thesis, 2025

In textile-reinforced concrete (TRC), the steel bars of conventional reinforced concrete (RC) are replaced by non-corroding fiber textiles. Compared to RC, TRC requires less concrete cover, eliminates corrosion-related maintenance, and provides higher specific strength and stiffness.

Analyzing and designing TRC structures, however, is challenging due to the complex interaction between the yarns and the surrounding concrete, as well as the brittle behavior of the yarns. The latter limits the possibility for stress redistribution and, consequently, the applicability of plastic limit analysis. Most existing studies use smeared representations that capture average effects but sacrifice explicit access to subscale behavior. In contrast, computational homogenization averages the subscale response while maintaining a direct link between the scales. In this framework, calibration is only required at the subscale, making it independent of the structural geometry. Despite its computational cost, this approach facilitates accurate predictions of structural response while providing access to subscale quantities such as crack widths, crack spacing, and reinforcement stresses. Moreover, it opens the door to holistic optimization, where both structural parameters (e.g., topology and shape) and material-level features (e.g., concrete properties and reinforcement layout) can be optimized within a consistent framework.

Building on these considerations, the present work introduces a multiscale framework for the design and analysis of TRC shells. A two-scale shell model is derived from the single-scale problem using Variationally Consistent Homogenization (VCH) with Kirchhoff–Love kinematics. The subscale simulations are performed on Representative Volume Elements (RVEs), where the effect of partial yarn activation is captured using efficiency factors. A second-order expansion of the out-of-plane displacement is also shown to be required for kinematic consistency. While suitable for generating offline data, upscaling through computational homogenization is often too time-consuming for online simulations. To overcome this, a surrogate model acting directly at the sectional level was introduced for use in the online stage of structural analysis. This model is thermodynamically consistent, accounts for tensile and compressive damage, and can be trained on multiscale simulation data. Further, it generalized well beyond the training data, accurately predicting responses under cyclic loading. Compared with a fully resolved single-scale simulation of a one-way slab, the proposed approach achieved comparable numerical results while being two orders of magnitude faster.

Together, these contributions represent significant progress toward the long-term goal of a practical multiscale framework for the design and optimization of TRC shells. The framework balances accuracy and computational efficiency and lays the foundation for integration into multiscale optimization workflows.