Research On Constant Force Control Methods In Robot Grinding Process

When an industrial robot grinds a workpiece,the robot and the outside world are in contact with each other while working in a constrained space.Due to the time-varying characteristics,strong coupling and highly nonlinear complexity of the robot system,as well as environmental interference exists during the grinding process,the machining accuracy of the workpiece with the robot position control is usually not ideal.In order to obtain a good robot processing effect,it is necessary to plan the robot processing process and control the contact force between the robot and the environment.In this paper,the theoretical trajectory of the robotic grinding is planned to solve the problem that the theoretical trajectory and the actual trajectory are inconsistent during the grinding process of the robot.For the problem of the unstable force signal during the grinding process,intelligent constant grinding force controllers are designed to maintain contact state between robot and workpiece,thereby improving the accuracy of the robotic grinding process is improved.In order to solve the problem that the theoretical trajectory is difficult to plan when the robot is grinding,a theoretical trajectory of the grinding process is obtained by the robot constant force tracking.Aiming at the problem that the force signal fluctuates greatly during the robot constant force tracking process,the force analysis of the contact between the robot end effector and the curved surface is performed,and the mapping relationship between each coordinate system is obtained,and the normal force of the robot end and the sensor measurement force are constructed.In order to facilitate the constant force control of the tracking process,a stiffness model,a neural network contact model and a probabilistic dynamic model are constructed to simplify the tracking process.In order to maintain constant contact force during the tracking process,three different constant tracking force controllers based on fuzzy iterative algorithm,reinforcement learning based on probabilistic dynamic model and actor-critic algorithm are designed.For the control parameters are difficult to find,the intelligent algorithm with the powerful nonlinear fitting ability and the advantage in solving the optimal solution can be used.Aiming at the problem of instability of the force signal during the grinding process,the forces at robot end-effector are analyzed,and the stiffness and grinding process of the robot system are studied.For the complexity of the grinding process,contact models are proposed to construct the relationship between the forces and control parameters.Since the robot grinding process consists of three stages: impact stage,machining stage and leaving stage,and the force characteristics in different stages are different,different control methods are adopted in different stages of grinding.Aiming at the problem that overshoot is easy to occur in the impact stage,a squeeze-release model and the fastest feedback system are proposed to plan the speed of robot in the impact stage.For the problem that the grinding model in the machining stage is difficult to build,a model combines a stiffness model and the grinding experience formula model are proposed,meanwhile a BP neural network model and a RBF neural network model are proposed.According to the models,two constant force grinding controllers based on neural network model,reinforcement learning and grinding model and iterative learning are designed,which controlled the impact stage and processing stage respectively,by using reinforcement learning,neural network and iterative learning intelligent learning,the control parameters have been optimized.In order to verify the feasibility of the design algorithms,an industrial robot tracking and grinding experiment platform are built.The robotic curved-surface constant force tracking experiment is conducted to compare the effect between the designed robot intelligent constant force tracking algorithms and the traditional control algorithm,which verified designed algorithms.Meanwhile,the theoretical trajectory of robot polishing is obtained.Experiments are carried out on the robot grinding,and the effects of the robot position control and the robotic constant force grinding algorithm are compared.The experimental results show that the processing accuracy is improved under the designed robot grinding intelligent force control algorithm,and the accuracy of the proposed grinding model and the effectiveness of the above algorithms are verified.