摘要
电梯作为现代都市高层建筑中必不可少的重要设备,其中曳引钢丝绳是电梯最重要的零部件,在长时间的使用过程中会出现断丝、断股等情况,倘若不及时处理,会容易引起各种人员伤亡和财产损失的安全事故。现有钢丝绳检测方法,如目测法、电磁检测法容易受到外界环境的影响,导致检测结果稳定性不足。因此,提出使用人工智能算法(YOLOv5s)进行钢丝绳表面损伤检测。首先自制了钢丝绳表面损伤数据集;其次利用自制的数据集训练YOLOv5s网络模型;利用训练好的YOLOv5s网络提取不同种类钢丝绳图像的特征;最后实现钢丝绳表面损伤检测。实验结果表明,基于YOLOv5s网络的钢丝绳表面损伤检测模型,检测准确率高、鲁棒性好、计算速度快,平均检测速度达到了6.7 FPS,测试精度达到了95.3%,为电梯事故的预警和报警提供了有效的参考依据,有较强的实际意义。
As an essential and important equipment in modern urban high-rise buildings,elevators have brought great convenience to people's daily life.The traction wire rope is the most important part of the elevator,and the wire will be broken during long-term use.If not dealt with in time,it will easily lead to various safety accidents involving casualties and property losses.Existed wire rope detection methods,such as visual inspection methods and electromagnetic detection methods,are easily affected by the external environment,resulting in insufficient stability of the detected results.This paper proposes to use the artificial intelligence algorithm YOLOv5s to detect wire rope surface damage.First,a data set of wire rope surface damage is made.Secondly,the YOLOv5s network model is trained with the self-made data set,and then the trained YOLOv5s network is used to extract the characteristics of different types of wire rope images.The experimental results show that the wire rope surface damage detection model based on the YOLOv5s network has high detection accuracy,good robustness and fast calculation speed.It can provide an effective reference for the early warning and alarm of elevator accidents,and has strong practical significance.The average detection speed reached 6.7 FPS,and the test accuracy reached 95.3%.
作者
蔡林峰
汤斌
杨泞珲
徐艺菲
雷斯越
贺渝龙
张金富
龙邹荣
CAI Linfeng;TANG Bin;YANG Ninghui;XU Yifei;LEI Siyue;HE Yulong;ZHANG Jinfu;LONG Zourong(Chongqing Key Laboratory of Fiber Optic Sensor and Photodetector,Chongqing University of Technology,Chongqing 400054,China)
出处
《智能计算机与应用》
2023年第6期67-71,共5页
Intelligent Computer and Applications
基金
国家自然科学基金青年基金项目(61805029)
重庆市教委科研基金项目(KJQN201905605)。
