Inverse design of 2.5D interlock composite architectures using a constrained stochastic framework

The mechanical performance of textile composite materials strongly depends on the architecture of the reinforcement, particularly in complex structures such as 2.5D interlock fabrics. While numerous optimization approaches have been developed for composite materials, most studies focus on parameter optimization within predefined architectures, such as laminate stacking sequences, fiber orientations, or structural layouts. The topology of textile reinforcement architectures themselves is rarely considered as a design variable. This work introduces a topology-driven optimization framework for 2.5D interlock composite architectures in which the textile architecture itself is optimized. The interlock reinforcement is represented using a motif-based discretization of the representative volume element (RVE), allowing the architecture to be described through a compact topology matrix. Geometric consistency rules derived from interlock construction principles are applied to ensure the validity of the generated architectures. The mechanical response of each candidate architecture is then evaluated through a micromechanical stiffness prediction model based on the Chamis formulation, enabling the computation of the effective stiffness matrix of the composite. The proposed framework is intended as a computational design and screening tool for identifying manufacturable textile architectures prior to high-fidelity numerical or experimental validation. This framework enables the systematic exploration of admissible textile architectures and the identification of configurations capable of achieving targeted stiffness characteristics. The results demonstrate that treating the textile architecture as a topological design variable provides a new pathway for the optimization of textile composite materials.