Answers are grounded in this paper and cite their sources.
Accurate prediction for curing-induced deformation (CID) of preformed woven carbon fiber reinforced polymer (CFRP) structures with complex geometry remains challenging due to inherent non-uniformity, anisotropy, and strong chemo-thermo-mechanical coupling. Conventional multiscale FE 2 methods relying on concurrent finite element analysis (FEA) at both material- and part-levels suffer from prohibitive computational costs, limiting their industrial applicability. To address this, a novel multiscale Long Short-Term Memory (LSTM)-FEA integrated modeling framework is established to seamlessly bridge physics-informed artificial intelligence (AI)-based material modeling and commercial finite element solvers. Unlike some AI-driven approaches that directly rely on data from processing of entire parts at macroscale, the reported method trains an LSTM network using data generated from validated mesoscopic representative volume element (RVE) simulations at the material level under diverse preforming and curing conditions. This characterizes the complex, process-dependent viscoelastic behavior of the composite material. Subsequently, the trained LSTM was decomposed into mathematical formulations and parameters, embedded into ABAQUS via UEXTERNALDB and UMAT subroutines, enabling uninterrupted updates of relaxation moduli depending on time, temperature and degree of curing of the woven CFRP during macroscopic FEA curing modeling with arbitrary part design and boundary conditions. Validation on CID of non-uniformly preformed 2D woven CFRP parts with 3D doubly-curved geometries woven CFRP structures demonstrates that the developed integrated method can realize prediction accuracy above 94.1%, computation efficiency improved by 2∼4 orders of magnitude compared to the FE 2 methods. This study bridges the gap between AI tools and FEA solvers arising from disparate operational environments and offers a generalizable geometry- and loading-condition-adaptive multiscale strategy for CID prediction that can advance practical composite manufacturing process design.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: