摘要
航空发动机涡轮叶片(aero-engine turbine blades,ATB)表面缺陷直接关系发动机运行安全、服役寿命。传统人工检测方法存在准确性不稳定、效率低、劳动强度大等问题。该文提出一种结合SH(switchable atrous convolution+high-level screening-feature fusion pyramid networks)与YOLOv8的ATB表面缺陷检测识别方法。在骨干网络中引入SAConv以增强多尺度特征提取能力;在颈部网络中嵌入HS-FPN以提升特征融合与缺陷表征能力。基于自建涡轮叶片内窥图像数据集开展实验,结果显示改进后SH+YOLOv8模型在F1-score、mAP@0.5上分别达到0.93、0.948,其中mAP@0.5相较原始YOLOv8提升0.64%,能满足工程应用精度需求。研究表明该方法在复杂工况下具备良好适应性、鲁棒性,对航空发动机关键部件智能检测具有参考价值。
Surface defects of aero-engine turbine blades(ATB)directly affect the operational safety and service life of engines.Traditional manual inspection methods suffer from unstable accuracy,low efficiency,and high labor intensity.To address these issues,this paper proposes a detection and recognition method for ATB surface defects by combining SH(switchable atrous convolution+high-level screening-feature fusion pyramid networks)with YOLOv8.In the backbone network,the SAConv module is introduced to enhance multi-scale feature extraction,while the HS-FPN module is embedded in the neck network to improve feature fusion and defect representation.Experiments are conducted on a self-constructed borescope image dataset.The results show that the improved SH+YOLOv8 model achieves an F1-score of 0.93 and mAP@0.5 of 94.8%,with a 0.6% improvement in mAP@0.5 compared to the original YOLOv8,meeting the accuracy requirements of engineering applications.The findings demonstrate that the proposed method exhibits good adaptability and robustness under complex conditions,providing valuable insights for intelligent detection of key aero-engine components.
作者
邹佳明
刘桂雄
黎文富
ZOU Jiaming;LIU Guixiong;LI Wenfu(School of Mechanical and Automotive Engineering,South China University of Technology,Guangzhou 510640,China;Guangdong Photoelectric Jietai Technology Co.,LTD.,Dongguan 523000,China)
出处
《中国测试》
北大核心
2026年第2期121-127,共7页
China Measurement & Test
基金
广东省重点领域研发计划项目(2019B010154003)。
