非接触式螺栓松动在线检测方法研究

Research on Non-Contact Online Detection Method of Bolt Looseness

为了保证机械装备高质量的工作性能,需要对螺栓连接结合面在服役状态下的连接质量进行实时的监测。针对目前接触式螺栓松动监测方法的缺点,提出通过detectron2深度学习方法,对松动螺栓进行快速的识别和判断。通过imglab对现有的图片进行标注,确定目标检测的种类及角点位置;对标注数据集进行训练,最终可实现对松动的螺栓进行辨识。结果表明:相对于传统的螺栓松动监测方法,该方法不仅能快速识别和定位松动的螺栓,还能预测角点的位置,并计算出螺栓松动后旋转角度,为螺栓连接件的健康监测提供一种新的方法。

In order to ensure the high-quality performance of mechanical equipment,it is necessary to monitor the quality of bolted joint in service.In view of the shortcomings of the current contact monitoring method of bolt looseness,it puts forward the deep learning method of detectron2 to quickly identify and judge the loose bolts.The existing images are labeled by imglab to determine the type of target detection and corner position;the labeled data set is trained to identify the loose bolts.The results show that:compared with the traditional method of bolt looseness identification,this method can not only quickly identify and locate the loose bolts,but also predict the position of corner points,and calculate the rotation angle after bolt looseness,which provides a new method for the health monitoring of bolt structures.