斜楔固定机器人的视觉定位系统设计与研究

Design and Research of Vision Positioning System for Wedge Fixed Robot

目前铁路货车转向架检修作业时,减振装置(枕簧与斜楔)需要人工进行拆卸,劳动强度大、耗时长且危险,制约了检修线的工作效率。为推进铁路运维设备自动化智能化转型,基于转向架的拆装步骤以及结构特征,研发了一种转向架斜楔固定机器人,可自动穿过摇枕孔实现对斜楔的固定,代替了人工作业,并设计了机器视觉定位系统,完成转向架上两个摇枕孔的检测定位。用YOLOv3网络作为摇枕孔检测定位的模型,在自定义数据集上对网络模型进行训练,利用训练好的目标检测模型对转向架摇枕孔进行实时定位,实现对机器人的引导控制。在转向架智能拆装实验平台上进行了摇枕孔的定位试验,试验结果表明该方法能够高效、可靠实现摇枕孔的自动定位,对于不同型号转向架,形状不规则的摇枕孔都具有较强的鲁棒性。

At present,during the maintenance of railway freight car bogies,the vibration damping device(pillow spring and wedge)needs to be manually disassembled,which is labor-intensive,time-consuming and dangerous.These factors restrict the working efficiency of the maintenance line.Therefore,to promote the automation and intelligent transformation of railway operation and maintenance equipment,based on the disassembly and assembly steps and structural characteristics of the bogie,a bogie wedge fixing robot is developed,which can automatically pass through the bolster hole to fix the wedge and replace manual work,and a machine vision positioning system is designed to the detect and position the two bolster holes on the bogie.The YOLOv3 network is used as the bolster hole detection and positioning model,the network model is trained on the custom data set,and the trained target detection model is used to locate the bogie bolster hole in real time to realize the servo control of the robot.The positioning test of the bolster hole is carried out on the bogie intelligent disassembly test platform.The results show that the method can realize automatic positioning of the bolster hole efficiently and reliably.For different types of bogies,or irregular shape of the bolster hole,the method shows good robustness.