基于深度学习的航空铆钉分类及异常情况检测

Aviation rivet classification and abnormal situation detection based on deep learning

针对航空铆钉小目标检测准确率较低、速率较慢等问题,提出1种基于深度学习的航空铆钉分类及异常情况检测方法。首先,根据钉头外观对航空铆钉进行分类,制作航空铆钉数据集;然后,构建航空铆钉分类及异常情况检测模型;最后,从置信度、召回率r、精确率p、平均精度值AP、全类平均精度mAP等指标对检测结果进行评价,并将该算法与YOLOx-s、YOLOx-m、YOLOv5、YOLOv4检测结果进行对比。研究结果表明:该算法可以实现对航空铆钉的分类及异常情况检测,检测结果的置信度约为90%,精确率、召回率、AP值分别在95%,85%,88%以上;本文所涉及铆钉分类及异常情况检测效果由好到坏顺序依次为:开槽盘头自攻铆钉、十字槽头铆钉、半圆头铆钉、平头铆钉、沉头铆钉、抽芯铆钉、异常情况;与其他算法相比,基于深度学习的航空铆钉分类及异常情况的检测,其效果和速度均有一定优势。研究结果可为减小航空铆钉松动和脱落风险、减轻飞机重量、防止金属疲劳和提升飞行可靠性及稳定性提供参考。

Aiming at the low accuracy problem of small target detection for aviation rivet,a method of aviation rivet classification and abnormal situation detection based on deep learning was proposed.Firstly,the aviation rivet was classified according to the appearance of rivet head,and the data set of aviation rivet was made.Secondly,the aviation rivet classification and abnormal situation detection models were constructed.Finally,the detection results were evaluated from the indexes such as confidence,recall rate r,precision p,average precision AP and mean average precision mAP,and the algorithm was compared with the detection results of YOLOx-s,YOLOx-m,YOLOv5 and YOLOv4.The results showed that the algorithm can realize the classification and abnormal situation detection of aviation rivet,the confidence of the detection results is about 90%,and the precision,recall rate and AP value are above 95%,85%and 88%respectively.The order of rivet classification and abnormal situation detection effect from good to bad is:slotted pan head self-tapping rivet,cross slot head rivet,half-round head rivet,flat head rivet,countersunk head rivet,core pulling rivet and abnormal situation.Compared with other algorithms,the aerial rivet classification and abnormal situation detection based on deep learning have advantages in effect and speed.The research results can provide reference value and important significance for reducing the risk of aviation rivet loosening and detachment,reducing aircraft weight,preventing metal fatigue.