Research And Development Of Casting Grinding Robot System Technology Based On 3D Point Cloud

Casting technology is a common technology used to produce complex structural parts,which require a sprue connection between the casting and the mold,and subsequent grinding of the sprue by workers with hand-held tools.The manual polishing method is a harsh environment and has disadvantages such as low production efficiency and inconsistent product quality.In order to solve the above problems,this thesis designs a3 D point cloud-based grinding robot casting cleaning scheme with robot as the main body and 3D vision perception system as the auxiliary,to realize the intelligence of casting grinding and cleaning.The specific research contents are as follows:Firstly,according to the current situation of grinding process research at home and abroad and the difficulties of automated grinding,a set of 3D point cloud-based grinding robot system scheme is developed for the grinding robot which is difficult to precisely locate the area being polished.Then,Mech-vision and Mech-Mind 3D Pro S camera are used as the sensing system to acquire and process the data,and the UR robot combined with the end grinding tool is used as the execution part to grind the castings.The perception system is based on the K-D tree(k-dimensional tree)algorithm to pre-process the 3D point cloud data of the castings,such as denoising,mean filtering and non-uniform downsampling,and the object of study is the motorcycle engine head casting,which proves the effectiveness of the above algorithm in the implementation of practical application scenarios.Next,based on the original coarse alignment,a global model and local feature alignment localization method based on the generalized nearest point iterative algorithm is proposed,replacing the original two-stage coarse alignment and fine alignment with a three-step global coarse alignment,global fine alignment and local fine alignment localization method,avoiding the influence of the casting itself defects on the matching localization,while ensuring that the matching efficiency is not affected.Immediately afterwards,this thesis establishes the grinding robot system and embeds the algorithm into the system,establishes communication for the grinding robot,3D sensing device and six-dimensional force sensing device,and performs tool calibration for the grinding robot and hand-eye calibration for the robot and 3D sensing device,and uses virtual simulation technology to simulate the scheme.Finally,the simulation experiments of the grinding robot system and the perception system and the experiments of the physical feature localization of the grinding robot were conducted for the research object to verify the effectiveness of the scheme.The experiments obtained that the algorithm within the perception system timed about 4.3s from data acquisition to final output to the robot,and the error of the sanding positioning effect on the final sanded workpiece reached 0.5mm.The use of a robotic arm with a 3D perception system and the use of a three-stage alignment method combined with accurate local feature positioning of the casting is a reference for the improvement of the intelligence of the grinding process.