Research On Weld Positioning Algorithm In 3D Laser Guided Robot Welding

With the continuous development of industrial automation,welding robots are widely used in the bicycle manufacturing industry and have gradually replaced manual welding.Currently,most welding robots still use the teaching reproduction mode,which cannot correct the deviation of the welding workpiece in real time,and it takes more time and effort when replacing different types of workpieces.Therefore,traditional welding robots have been difficult to meet the requirements of welding quality,efficiency and cost.They need to develop in the direction of intellectualization.The laser vision sensing technology can acquire the weld position of the workpiece to be welded in real time and realize the welding intellectualization.At present,lots of researches have been done on the weld positioning based on laser sensors at home and abroad,and many excellent results have been obtained.However,these algorithms are based on the weld of specific shapes and specific application requirements.They pay less attention to algorithm versatility,stability,real-time.The high requirements in welding applications are also ignored,including extracted weld precision,welding posture and etc.In this paper,a weld positioning algorithm combining two-dimensional depth image and three-dimensional point cloud data is proposed,based on summarizing the existing research results at home and abroad.It takes the common bicycle frame to be welded in bicycle production as the research object.At the same time,a robot welding system based on 3D laser guidance is built in the industrial field,and the algorithm designed in this paper is experimentally verified.Firstly,a weld positioning algorithm combining two-dimensional image and three-dimensional data is proposed,based on fully comparing the advantages and disadvantages of the two-dimensional and threedimensional data processing.The algorithm uses the two-dimensional image to quickly extract the weld information,and it obtains the three-dimensional weld according to the spatial transformation relationship.Moreover,the welding posture is calculated by three-dimensional data.The two-dimensional image processing is mainly divided into three parts.One is to complete the data preprocessing,which uses the method of traversing to find the largest inscribed rectangle to extract the image area containing the weld feature.The second is to extract the two-dimensional weld information initially,based on selecting the appropriate edge operator by the image gray histogram.The third is to position the two-dimensional weld accurately,which uses the nearest neighbor edge joint method and the Moving Least Square smoothing method.The three-dimensional data processing is mainly divided into two parts.One is to realize the mapping of two-dimensional welds based on the coordinate transformation relationship.The other is to calculate the welding posture of the corresponding weld points through the the local surface fitting method.Secondly,this paper designs a robot welding system based on 3D laser guidance.The workflow of the system is planned to realize welding.The system is divided into two subsystems,containing the weld positioning system and the robot control system.And what tasks need to be done in weld positioning system is clarified.Thirdly,data communication between various parts of the robot welding system is established,which realizes data collection.It is mainly divided into three parts.One is to realize the data interaction between the two subsystems through TCP-based Socket communication.The second is to realize the control of the 3D laser sensor by using the SmartRay API.The third is to collect the data of the workpiece to be welded after realizing the communication of each part.Fourthly,the versatility,stability and real-time performance of the weld positioning algorithm are verified.The versatility of the algorithm is verified by analyzing the extraction results of different types of weld.The stability is verified by analyzing the extraction results of the same type of weld under different external conditions.The real-time performance is verified by analyzing the time required to extract the weld information for each weld type.Finally,the weld positioning system and the robot control system are jointly debugged,and the bicycle frames are welded.The workflow of completing three types of welds in one workpiece is planned.Welding experiments are carried out on different types of workpieces.The results show that it is feasible to weld different types of bicycle frames by 3D laser guided robots in the industrial field.The effectiveness of the weld positioning algorithm proposed in this paper is further illustrated in the meantime.