针对燃气轮机涡轮叶片涂层脱落形状不均、位置不一、叶片孔探图像角度不同、光线较暗等问题,本文提出了一种基于YOLOv10n的燃气轮机涡轮叶片涂层脱落检测模型。该模型旨在针对燃气轮机涡轮叶片的孔探图像或视频,静态或动态地检测叶片涂层的脱落情况,并对脱落位置进行精准定位。我们将该方法应用于MS9001FA重型燃气轮机的涡轮孔探图像集进行实验,以验证模型的精确性和快速性。实验结果表明,基于YOLOv10n的涡轮叶片涂层脱落检测模型的精确度相较于YOLOv5su和YOLOv8n分别提升了10.2%和4.5%。在识别精度相同的情况下,该模型的速度较YOLOv10b提升了23%。这些结果证明了基于YOLOv10n的涡轮叶片涂层脱落检测模型能够更好地平衡精确性和快速性,具有较高的工程应用价值。

To address issues such as irregular shapes of turbine blade coating detachment, varying locations, differing angles of borescope images, and low light conditions, this paper proposes a gas turbine turbine blade coating detachment detection model based on YOLOv10n. The model is designed to detect and precisely locate the coating detachment on turbine blades in borescope images or videos, whether in static or dynamic scenarios. We applied this method to the borescope image dataset of the MS9001FA heavy-duty gas turbine to evaluate the model’s accuracy and speed. The experimental results show that the detection accuracy of the gas turbine turbine blade coating detachment model based on YOLOv10n improved by 10.2% and 4.5% compared to YOLOv5su and YOLOv8n, respectively. Furthermore, with similar detection accuracy, the model"s speed increased by 23% compared to YOLOv10b. These results demonstrate that the proposed YOLOv10n-based detection model offers an effective balance between accuracy and speed, making it high

TK221

船舶动力基础科研（KY10300240082）