Abstract: The 3D stitching process enables the introduction of continuous reinforcing fibers between 2D layered fabrics, allowing the efficient fabrication of various complex-shaped fabric preforms. Composites reinforced with stitched fabrics overcome the weak interlaminar performance of 2D composite laminates and are widely used in the aerospace industry. However, during the forming process of 3D stitched composites, compaction deformation of the preform is inevitable, resulting in highly complex fiber architectures. This poses significant challenges for micro- and meso-scale modeling and mechanical property analysis of the composites. In this study, the "four-step" modeling technique is employed to simulate the stitching process of fabric layers and establish a fiber-level geometric twin model of the 3D stitched preform. Numerical simulations of the preform compaction process are conducted to analyze the evolution of the internal fiber architecture and reveal the compression deformation mechanisms. Micro-CT technology is used to scan the microstructure of the preform samples before and after compression, providing internal geometric information. Quantitative evaluation metrics for the stitching yarn path and cross-sectional areas of warp and weft yarns are proposed to validate the accuracy of the geometric twin model. The results demonstrate high consistency between the model and Micro-CT images before and after compression, with similarity rates of 88.1% and 95.0% for the stitching yarn paths, and relative errors within 5% for the cross-sectional areas of the warp and weft yarns in the layered fabric. The findings of this study provide a theoretical foundation for the structural design and mechanical performance simulation of 3D stitched composites.