Abstract
Performance breakthroughs and safety assurance of aerospace equipment are critical to the advancement of modern aerospace technology. As a key protective system for the hot-end components of aeroengines, thermal barrier coatings (TBCs) play a vital role in ensuring the safe operation of aeroengines and overall flight safety. To address the core detection technology challenge for micro-nano damage precursors in aerospace TBCs, this study proposes an enhanced detection framework, namely YOLOv11-LADC. Specifically, the framework integrates the LSKA attention mechanism to construct the C2PSA-LA module, thereby enhancing the detection capability for micro-nano damage precursors and adaptability to complex small-sample datasets. Additionally, it introduces deformable convolutions (DeformConv) to build the C3k2-DeformCSP module, which dynamically adapts to the irregular deformations of micro-nano damage precursors while reducing computational complexity. A data augmentation strategy incorporating 19 transformations is employed to expand the dataset to 5140 images. A series of experimental results demonstrates that, compared with the YOLOv11 baseline model, the proposed model achieves a 1.6% improvement in precision (P) and a 2.0% increase in recall (R), while maintaining mAP50 and mAP50-95 at near-constant levels. Meanwhile, the computational complexity (GFLOPs) is reduced to 6.2, validating the superiority of the enhanced framework in terms of detection accuracy and training efficiency. This further confirms the feasibility and practicality of the YOLOv11-LADC algorithm for detecting multi-scale micro-nano damage precursors in aerospace TBCs. Overall, this study provides an effective solution for the intelligent, high-precision, and real-time detection of multi-scale micro-nano damage precursors in aerospace TBCs.