Research On Grinding Control System Of Light-weight Manipulator And Method Of Measuring Surface Roughness

At present,in the aerospace,aviation,shipping,defense industry and other fields,a large number of complex free-form parts are used to meet the demand for special performance.These complex free-form parts are difficult to manufacture and require high machining accuracy,and some parts still rely on manual operation of the free-form grinding and machining process.Manual grinding has the problems of poor consistency and low efficiency,and long-term grinding work is very harmful to the body.How to use the robot to replace the manual grinding,while having the same grinding experience skills as human is one of the challenges of the robotic grinding research.In the grinding process,how to accurately control the grinding contact force and how to achieve the accurate prediction and measurement of the surface roughness of the workpiece are the bottleneck problems of robotic grinding control.In this paper,we designed a light-weight robotic arm based on EtherCAT bus,established a manual grinding experience skill model and given the robotic grinding "wisdom",improve the accuracy of grinding dynamic force tracking,and solved the problem of surface roughness prediction and measurement.The following research work is mainly carried out:In view of the high force control accuracy,fast force control response and large amount of communication data required for the robotic belt grinding,we completed the development of a flexible joint robot based on EtherCAT bus.The robot has a multisensing joint sensing system to achieve higher force tracking control accuracy.In this paper,the robot sensing system with richer external environment sensing capability was built,and the controller of the robot was built based on Xenomai real-time kernel in Linux system to ensure high real-time control system.Therefore,a high bandwidth and high real-time hardware platform was provided for the robotic belt grinding.To address the problem of complex grinding process of free-form surface workpieces,this paper proposed an imitative learning method based on Gaussian process regression to analyze the manual grinding skill experience.Then,the experience was transferred to the robotic grinding task operation.For the dynamic force tracking control problem of the light-weight robot in the process of grinding complex curved workpieces,this paper proposed an adaptive admittance hybrid force/position control algorithm based on dual nonlinear differential tracker.The accuracy of grinding contact force control was improved.For the problem that the rough parts of complex curved workpiece have poor consistency,the robot cannot follow a fixed position pattern for sanding processing.In this paper,the variance minimization point cloud matching algorithm based on tolerance constraint calculates the grinding margin distribution map of the surface of the workpiece to be polished.For the problem that the surface roughness of the workpiece cannot be accurately predicted during the grinding process,this paper proposed an RNN-based workpiece surface roughness prediction model.In order to improve the efficiency of workpiece surface roughness detection,this paper proposed a GAN-BPNN-based surface roughness measurement method.The influence of the difference of different curvature surface images on the accuracy of roughness prediction is avoided.In order to verify the effectiveness of the light-weight robotic grinding control system and the grinding force control algorithm,an experimental platform for lightweight robotic grinding was built.The force-controlled grinding experiments of the lightweight robot for free-form surface workpieces were carried out.Based on the kinematics and dynamics modeling of the light-weight robot and bus communication testing,experiments on the control performance of the light-weight robotic arm were conducted.Experiments were conducted on the transfer of manual grinding experience skills to robotic grinding operations.Experiments on adaptive guided force/position hybrid control algorithm based on dual nonlinear differential tracker.The accuracy of the recurrent neural network-based surface roughness prediction model and the GAN-BPNNbased surface roughness visual measurement method were verified.