Design Of Autonomous Welding System In Tank Based On Welding Target Tracking Algorithm

Welding is a very important process in industry,in the container,shipbuilding industry,construction industry,aerospace and other fields are widely used.In recent years,welding robots have been widely introduced into industrial production,which has greatly improved welding productivity.Nowadays,the working scene of industrial welding robot is gradually transferred from outside the tank to inside the tank,in order to meet the new technological requirements and take into account the aesthetics at the same time,which poses a new challenge to the structural design of welding robot.At the same time,because of the frequent production safety problems caused by weld defects,welding quality detection has become very important.In order to solve the current problems such as the inability of the robot body to move effectively,low welding efficiency,insufficient welding accuracy and difficult welding quality detection,this thesis proposes an autonomous welding system in the tank based on directional feature driven welding target tracking algorithm.This thesis firstly designs the system scheme according to the functional requirements and performance requirements,and then designs and determines the control system scheme.Then,the weld capture and ranging unit,control unit and execution unit mentioned in the system scheme are designed and the hardware selection is carried out.Finally,according to the system scheme,the welding robot control software is built.In this thesis,image preprocessing,weld target tracking method introduction and improvement,weld defect target detection and implementation unit control for the context of the narrative.In order to make the target tracking of the weld obtain better results,this thesis firstly preprocessed the captured images of the weld.After experimental comparison,the pretreatment process most suitable for the weld target was obtained,and finally the weld image more suitable for Gabor feature extraction and subsequent HOG feature extraction was obtained.The control unit processes and calculates the data such as images collected by the welding seam capture and ranging unit by using the welding seam target tracking algorithm driven by the orientation feature proposed in this thesis,and then obtains the error of welding seam tracking and controls the execution unit to track the welding seam,which can complete the welding seam tracking more accurately and reliably.The algorithm proposed in this thesis adapts to the image features of V-groove under the irradiation of auxiliary light source,and enables the welding seam capturing and ranging unit to find the weld coordinates more accurately under the interference of arc,splash and other noises.The second industrial camera in the system uses the YOLOX-s target detection algorithm to detect the weld quality of the collected pictures after welding,and records the location of defects.Finally,a control algorithm was designed for the welding manipulator to move after receiving the deviation between the weld and the welding gun,so that the end of the welding manipulator could approach the weld with the most stable position.At the same time,the structure and motion control of the welding mobile platform are designed.The results of welding experiment in the tank show that the system has good stability,high accuracy and good weld forming quality,which is within 0.12 mm and meets the process requirements of pipeline welding.After the training and comparison of various target detection algorithms on the weld defect detection data set,the detection results of YOLOX-s show that the algorithm is more effective.A variety of weld defects can be accurately detected.