A High-Fidelity Modeling and Delamination Load Prediction Method for Composite Tenons Driven by Image Recognition

Abstract

Resin matrix composite fan blades exhibit complex internal geometries with intricate ply stacking sequences and manufacturing-induced defects. This study focuses on the tenon structure - a critical yet vulnerable component that sustains exceptionally high static loads during service. To accommodate thickness transition from blade root to airfoil, the tenon incorporates extensive ply drop-offs, creating pronounced stress concentrations. Furthermore, abrupt sectional changes during manufacturing often induce wrinkles and fiber misalignments, significantly compromising structural integrity and potentially leading to premature failure. These complexities pose significant challenges for accurate modeling and failure analysis. Conventional high-fidelity modeling approaches prove prohibitively time-consuming for such intricate structures, while simplified models inevitably overlook crucial geometric details, resulting in inadequate predictive capability. To address these limitations, this research introduces a high-fidelity modeling approach driven by image recognition and deep learning. A semantic segmentation network is developed, achieving a mean Intersection over Union (mIoU) of 89.20% on the validation set, enabling efficient and precise identification of ply contours. The reconstructed model demonstrates exceptional accuracy with a positional deviation below 0.62% when validated against CT scanning data. Based on this refined model, an interlaminar damage analysis combining progressive damage and cohesive zone models accurately predicts delamination onset loads within 6.4% of experimental results. More importantly, the refined model captures the influence of manufacturing defects on failure progression—revealing distinct failure mechanisms across different regions—which cannot be represented in conventional idealized models. This integrated framework significantly accelerates the high-fidelity model generation process, thereby facilitating the strength design and failure assessment of resin matrix composite fan blades.