Forming prediction of woven composites is crucial for optimizing manufacturing parameters in the fabrication of large-scale and high-quality equipment across modern automotive, aerospace and marine industries. This work presents a systematic combination of multiscale forming simulation framework and the high-fidelity finite element (FE) model to predict non-uniform yarn orientation and laminate thickness distributions. Specifically, the high-fidelity FE model based on the unit-cell reconstructed directly via a deep-learning algorithm compatible with low-resolution CT images without manual adjustment is incorporated into the integrated mesoscopic simulation model previously developed for elucidating deformation mechanisms. The indispensable characterization experiments are simplified to only uniaxial tension and consolidation tests on orthogonal composites, minimizing experimental redundancy while preserving essential data integrity. Concurrently, the coupling deformation modes impervious to direct measurement can be reliably captured through virtual characterization based on the high-fidelity FE model, leading to exceptional predictive accuracy for stress-strain relationships with 2.1% average error compared to the idealized unit-cell model established before with 15.2% average error. Subsequently, these homogenized relationships were fed into the macroscale forming simulation, which was then validated through prepreg compression molding (PCM), resulting in strong consistency between experimental and simulation data on the deformed geometry, yarn angle and thickness distribution, exhibiting respective errors of 7.59%, 4.43° and 14.7%. Therefore, the proposed multiscale simulation framework not only mitigates experimental time costs and operational complexity but also augments forming process precision by accounting for the coupling effects between distinct deformation modes, addressing a critical gap in the efficient fabrication of high-performance woven composite components.
Sun et al. (Sun,) studied this question.