A data-driven concurrent multiscale FE-FCA method for the elastoplastic nonlinear mechanical response prediction of large-scale composite structures

The hierarchical multiscale architecture of composite structures imposes severe computational demands. Accurately resolving microstructural heterogeneity across a component typically requires tracking hundreds of millions of degrees of freedom, performing coupled nonlinear constitutive updates, and transferring large volumes of data between scales, which together drive high memory consumption and huge computational cost. To address these challenges, we propose a data-driven concurrent multiscale reduced-order model method for predicting the elastoplastic nonlinear mechanical response of large-scale composite structures. The proposed method is divided into two phases. In the offline phase, the reduced-order model of the microscale RVE is constructed based on our previously proposed FEM–Cluster based Analysis (FCA, Comput. Methods Appl. Mech. Eng., 348:157–168, 2019). In the online phase, an incremental FE-FCA algorithm is proposed to enable effective prediction of the macroscale structure’s loading-unloading response and the local stress behavior within the microscale RVEs. The overall method, referred to as the FE-FCA method, is formulated within a unified finite element framework, which allows the use of irregular meshes and RVEs with high material-contrast, such as porous materials. Moreover, RVEs associated with all integration points are solved in parallel at each structural load step, and a load-step adaptivity strategy is proposed to further accelerate the RVE computations, which leads to a substantial increase in the overall computational efficiency. Compared to the classical FE 2 method, the proposed FE-FCA significantly reduces computational cost and memory requirements while maintaining comparable accuracy, offering a scalable, efficient route for virtual testing and microstructure‑informed design of composite structures.