Research On Online Trajectory Prediction And Adaptive Constant Force Control For Robotic Grinding

In recent years,with the development of science and technology,robot automation production line technology prevails in manufacturing industry.Robot automation manufacturing,which is assistant to complete grinding processing operations,can save labor costs,liberate labor,avoid the health threat of grinding operations to the human body,and improve work efficiency.However,due to the high flexibility requirements of the robotic grinding,it’s difficult to replace sophisticated workers which can realize the fully automated grinding of workpiece.In addition,for meeting the grinding quality requirements,it’s necessary to generate the grinding trajectory planning under the premise of the accurate geometric surface prior model of the workpiece.Therefore,in order to improve the high automation and flexibility of robot grinding operation,this paper designs a robot compliant floating grinding force control end effector,and constructs a robot force control grinding system,focusing on the following three aspects for in-depth research:(1)In order to improve the flexibility of robotic polishing,a compliant floating grinding force-controlled end-effector is designed,a virtual prototype of the robotic grinding system is built.The impedance characteristics between the polishing system tools and the workpiece is analyzed,and then the constant-force controller for grinding based on admittance control is established.(2)Focusing on the polishing operation of the unknown surface workpiece without a prior model,in order to improve the real-time tracking performance of the curved surface normal,Also,an active adaptive online trajectory prediction method of robotic grinding for the unknown surface without a prior model is proposed to address the poor adaptability problem in robotic grinding trajectory prediction for this kind of surface.The orientation of the robotic endeffector spindle is real-time adaptively obtained by predicting and tracking the normal direction of the unknown surface without a prior model,which is based on the contact state between the grinding tool and the workpiece surface.Furthermore,the online trajectory prediction of robotic grinding for unknown surface without a prior model is implemented by solving for the directional intersection line of the directional feed plane and the tangent plane of surface.Through simulation and experiment analysis,the effectiveness of the method proposed by the method is verified.(3)Focusing on the problem of the poor adaptability of the industrial robotic compliant grinding for complex surface workpieces and variable grinding conditions in the unknown or less information contact environment,an active adaptive variable admittance control method for robotic grinding force is proposed,with reinforcement learning based on the ensemble Bayesian neural networks model.According to the contact state information of the robotic grinding,the multiple sampling samples from small amount of data is obtained by the Bootstrapping method,and the ensemble Bayesian neural networks model is trained to characterize interactions between the robotic grinding system and the variable grinding condition environment.Furthermore,the optimal admittance parameters are solved by the covariance matrix adaptation evolution strategy(CMA-ES),and the reinforcement learning based active adaptive robotic grinding force control for the complex surface workpiece under variable working conditions is realized to improve the adaptability of the industrial robot compliant grinding.A robotic grinding simulation experiment for a complex surface blade workpiece of the compressor turbine is conducted on the virtual prototype platform of the robotic grinding system,and the effectiveness of the proposed method is verified.