Research On 3D Machining Target Detection And Motion Planning Of Ceramic Billet Grinding Robot

In recent years,with the transformation and upgrading of manufacturing industry and the rapid development of robot technology,more robots have been equipped to the industrial site.In order to ensure the surface quality of the ceramic billet after being one?time polished by the robot,it is necessary to pre-polish the machining target such as the parting line and the glue line which are discretely distributed on the surface of the workpiece.However,in actual production,on the one hand,the grinding operations of the same type of ceramic billet are different,mainly due to the poor dimensional consistency of the ceramic billet and the poor consistency of positioning on the workbench.Therefore,the fixed grinding trajectory obtained by teaching the robot will result in poor surface quality consistency.On the other hand,different types of ceramic billets require different grinding trajectories.However,the conventional programming method of obtaining grinding trajectories by teaching robot is inefficient,and it is difficult to meet the flexibility requirements of grinding operations.Therefore,in order to improve the automatic sensing and motion planning ability of the ceramic green grinding robot,it is necessary to study the on-line detection of non-reflective machining targets with discrete distribution of ceramic billet workpieces with poor dimensional consistency and the motion planning in the robotic grinding operations.This paper takes 3D point cloud analysis,high-precision 3D vision measurement system calibration and time-optimal motion planning as the breakthrough point.The research focus on the detection of 3D machining targets,the optimization of processing paths and the planning of collision avoidance trajectories for ceramic billet grinding robot,and carries out related experiments and applied research.The main research contents of this paper are summarized as follows:(1)In order to automatically detect the 3D machining target of the grinding robot,a new 3D vision system is proposed,which acquires point cloud data using a 3D scanner integrated on the fourth joint of the robot.A new method of point cloud segmentation based on minimum bounding box segmentation and sub-point cloud simplification is proposed.This method can effectively extract the required machining path points from the difference point cloud.A calibration method for 3D vision system including coarse calibration and accurate calibration is proposed.A new 3D visual measurement error compensation model is proposed in the accurate calibration and combined with multi-point constraint method to fit the error compensation.The experimental results show that the absolute average error of the 3D vision system after accurate calibration is 0.154mm.(2)When a plurality of discrete non-connected machining regions are distributed on the outer surface of the workpiece,a planning framework for the task-level time optimal processing path(TOPP)of the grinding robot is proposed for the automatic planning of the task-level processing path.Considering the requirements of actual robot motion and grinding process,the grinding tool have different speeds in the machining stroke and non-machining stroke during the whole grinding process.Therefore,the essence of the task-level TOPP is to search for the closed-loop machining path with the shortest weighted total length.In order to quickly obtain an approximate global optimal solution when the number of machining path points in the task-level TOPP is large,an improved parallel SA algorithm based on parallel sub-chain interaction(PSI)is proposed,which can update the current approximate global optimal solution in real time.The contrast experiments show that the proposed algorithm can quickly search for the approximate global optimal solution with good quality.(3)When a plurality of discrete non-connected machining regions are distributed on the inner surface of the workpiece,a planning framework for the task-level time optimal collision avoidance trajectory(TOCAT)of the grinding robot is proposed for the automatic planning of the task-level collision avoidance trajectory.Consider the actual robot movement and grinding process requirements.The duration of the machining stroke is constant throughout the grinding process,so the essence of the task-level TOCAT is to search for the trajectory ordering when the sum of the duration of the collision avoidance trajectories of the non-machining stroke is the shortest.The mathematical model of time optimal collision avoidance trajectory planning problem is constructed.A time-optimal collision avoidance trajectory planning method based on trajectory evaluation mechanism is proposed,and the function of evaluating the quality of collision avoidance trajectory is described.In order to obtain a better approximate global optimal solution for the task-level TOCAT planning problem,an improved SA algorithm for generating new solutions by combined stochastic perturbation method(CSPM)is proposed.The contrast experiments show that the proposed algorithm can search for the approximate global optimal solution with better quality.(4)A software and hardware platform for 3D visual inspection,task-level processing path planning and task-level collision avoidance trajectory planning of the grinding robot was built.The experimental results show that the proposed 3D vision system can effectively detect the machining information,and the absolute average error of repeated measurement is 0.028mm,and the composite detection error is 0.25mm.The proposed task-level TOPP planning framework can plan the task-time optimal processing path for the grinding and polishing robot.The acceleration ratio increases approximately linearly with the gradual increase in the number of machining path points,and the length of th'-obtained task-level TOPP is close to the approximate global optimal value.The proposed task-level TOCAT planning framework can plan the task-level time optimal collision avoidance trajectory for the grinding and polishing robot,and the optimization of the duration of the task-level collision avoidance trajectory is 63.33%.