Research On Modeling And Collision Process Of Grinding Robot System

The demand of industrial robots in machining tasks such as grinding is increasing.In such rigid contact tasks,it is important to study on positioning accuracy of robots and the control methods in the machining process.In order to improve the positioning accuracy of robots,a grinding robot test system is used for kinematic error analysis and compensation.During the process of robot grinding,contact force should be controlled in a reasonable range.For the purpose of realizing force control without external sensor,robot dynamics modeling and identification methods are studied firstly,and then establish the generalized force mapping relations from the joint space to the Cartesian space.The modified Coulomb-viscous friction model is proposed aiming at the problem that friction model used in dynamics model identification is not accurate enough.After that,utilizing substep identification methods to obtain model parameters.Finally,contact transition process in robot grinding is studied in this paper.A segmentation speed control strategy is proposed to reduce the contact force between workpiece and tool at collision moment.Furthermore a collision detection method using wavelet transform is proposed.The main work of this paper is as follows:(1)Chapter one studies the factor in kinematic error of a robot grinding test system.A fast calibration method is proposed for it.This method can compensate kinematic error directly with no need for calculating the specific value of error parameters to improve the positioning precision.A kinematic calibration experiment is performed on the grinding robot system.additionally,an experimental that verifying the compensation result by the magnitude of the contact force is designed in this paper.The experimente results show that the calibration method in this paper can effectively reduce the contact force caused by the error of the positioning accuracy of the robot.(2)The robot dynamics model is established based on Newton-Euler equation,and it is transformed into a minimum parameter set that every element is linear independent with each other.A Fourier excitation trajectory optimized by minimum condition number is used for parameter identification.To perceive the contact force that occur on robot terminal,this chapter studies the generalized force mapping relationship from the joint space to the Cartesian space by means of Jacobian matrix.Experiment results show that the dynamic modeling and identification method used in this paper has certain effects on predicting driving.The generalized force mapping experiment verifies that if external joint torque estimation is accurate enough,using generalized force mapping to estimating contact force is feasible.(3)Aiming at the shortcomings of existing friction model,chapter four studies on the friction of the robot joint with harmonic reducer and put forward the modified Coulomb-viscous friction model,besides,a substep identification method is proposed.In order to obtain a complete friction model,the parameters of Coulomb-viscous friction model is calculated by least squares method.After that,K-means and SVR are used to identified the nolinear part which is related to joint angle.The experimental results show that compared with the Coulomb-viscous model which commonly used in robot dynamics modeling,the friction model in this paper can effectively improve the calculation accuracy of joint friction.(4)Chapter five studies on the contact transition process in the grinding task,and come up with a segmentation control strategy.The strategy reduce robot's motion velocity from high speed to safe collision speed smoothly before contact occur so that contact force can be restricted in a safe range.The collision detection based on robot joint torque feedback is studied as well.A collision detection algorithm using wavelet transform is proposed.The experimental results show that the segmentation control strategy proposed in this paper can effectively reduce the contact force at collision moment.The collision detection using wavelet transform can determine the robot collision moment in time.