Research On High Precision Control Of Trajectory For Robotic Machining

Large complex curved surface parts are widely used in aerospace industry,and the manufacturing accuracy and quality affect the performance and safety of aerospace equipment directly.At present,the large complex curved surface parts are mostly processed by multi-axis CNC machine tools.And there are some problems,such as expensive equipment,complex configuration,and so on.With the development of robotic machining technology,the advantages of flexible processing and flexible configuration are becoming more and more obvious.It has become a new research topic to use robots to achieve high-efficiency and precision machining of large complex curved surface parts.However,the robot has some problems,such as low stiffness,poor absolute accuracy,strong non-linear coupling effect,which makes it difficult to guarantee the processing accuracy of the robot.Therefore,aiming at the problem of poor machining accuracy of robot for large complex curved surface parts in this paper,the research of high precision control algorithm for robotic machining is carried out.In order to improve the absolute tracking accuracy of lager complex curved surface parts,the contour error is used to describe it and the research on motion control in joint space,trajectory tracking in task space,contour control in task space and contour error compensation in task space are carried out.Firstly,a fractional sliding mode control strategy is proposed for robot joints.Secondly,a linear model predictive controller is designed based on task trajectory.Then,contour error is introduced to design controller.Finally,the contour error compensation strategy based on data-driven is studied.And the experiment is validated on platform one by one.The main contents and innovative achievements of this paper are as follows:1.Motion control in robot space.The kinematics and differential kinematics models of six-joint serial robot are established,and the dynamic model parameters of the robot are identified.In order to achieve high tracking performance of each joint,the fractional sliding mode controller is designed based on the dynamic model,fractional calculus theory and sliding mode control theory.The controller is validated on the six-joint industrial robot.The results show that the proposed method can effectively improve the joint tracking accuracy to(±0.003rad)compared with the sliding mode controller.2.Trajectory tracking in task space.The limitation of joint space motion control is analyzed: the coupling effect between joints is not considered.To deal with this problem,trajectory tracking in task space is studied,and linearized model predictive control algorithm is designed based on the principle of model predictive control.The controller is validated in MATLAB,and the results show that the proposed method can improve the tracking accuracy of 74.82% in X direction,72.58% in Y direction and 27.78% in Z direction compared with the PID controller.3.Contour control in task space.There are two problems in the process of introducing the estimated components of contour errors into the traditional contour control framework by weighted sum: the reduction of contour errors is inconsistent in the whole processing area;the reduction of contour errors is limited,up to 37.5%.Aiming at these two problems,a weighted sum algorithm of error vector norm and a new contour control framework are proposed.The experimental results show that the proposed strategy can consistently reduce 85.4% of the position contour error and 88.89% of the orientation contour error in the whole processing area compared with the traditional strategy.4.Compensation of contour error in task space.The existing problems in contour control process are analyzed: on-line estimation of contour error or complex controller design.To solve this problem,a complete mapping model of robot input and output based on deep neural network is proposed,and a better reference input is matched based on model modification.The experimental results show that the proposed compensation strategy can effectively reduce the position contour error by 43.84% and the orientation contour error by 52.14%.