Absolute Positioning Error Prediction And Compensation Method For Industrial Robot Based On WMPS

When an industrial robot was operational continuously in industrial field,the position of the end effector would drift,which significantly affected the machining accuracy of large complex structures.Establishing a spatial error database of end effector which data could be updated dynamically was an effective method of compensating the absolute positioning error of the robot.Due to the low sampling efficiency of the equipment,it was impossible to quickly and accurately obtain the overall distribution law of positioning error in the workspace,so it was difficult to accurately predict and compensate the error.For the shortcomings above,in this paper,the workspace measurement positioning system was used to improve the efficiency of establishing and updating the error database based on the high precision,multi-point parallel and dynamic measurement characteristics of the system.Depend on the spatial similarity of positioning errors,the grid linear interpolation method was studied to realize the prediction and compensation of absolute positioning errors at similar attitude and position points.Aiming at the weakening of similarity caused by the diversity of running attitudes,an all-attitude compensation method based on neural network was proposed to expand the application scenarios.Finally,an experimental system was built to verify the feasibility and compensation effect of the methods in this research.This thesis finished main work was as follows:(1)The error model was established based on MDH method and robot kinematics.The influence of minor perturbation of parameters in the model on the final absolute positioning error was analyzed.The similarity relationship of absolute positioning errors in the workspace was studied.According to the study,it is clear that attitude is the dominant factor which affects the similarity between absolute positioning errors of similar positions.(2)For industrial scenes with similar attitude of each point on the trajectory,the sufficient conditions for the absolute positioning error at the similar attitude position points to meet the linear similarity was derived.Based on dynamic measurement with high precision of the w MPS,a mesh dividing strategy with adjustable side length was designed.Thus,the spatial error database was established to realize the prediction and compensation of errors.(3)Considering that the large changes of attitude during the processing of large complex structures would weaken the linear similarity.A prediction and compensation method of absolute positioning error vector of all-attitude in space was proposed based on the multi attitude sampling information of w MPS omnidirectional target and BP neural network optimized by particle swarm optimization,the mapping relationship between attitude and position of points and absolute positioning error vector was established.The coordinates of the key nodes of the trajectory were optimized.(4)The experimental system of the robot under no-load and load conditions were built,and the compensation methods based on neural network was verified.By analyzing the experimental data of absolute positioning error compensation of different points with different attitudes,it could be seen that the proposed compensation method had good real-time performance and high efficiency,and could significantly improve the absolute positioning accuracy.