Research On The Measurement Model Of Intelligent Ball Joint Rotation Angle Based On Gaussian Process Regression

As a moving part with three degrees of freedom of rotation in space,the intellige nt ball joint is flexible in movement,strong in bearing capacity and compact in structure.It is widely used in industrial robots,parallel mechanisms,medical instrume nts,aerospace and other fields,but the space rotation angle of intelligent ball joint in passive motion can not be automatically obtained.In order to solve this problem,the traditiona l methods mostly use optical and mechanical methods to measure the angle,but these methods increase the complexity of the ball joint system.Based on the defects of the above methods,the research group proposed two methods to detect the spatial rotation angle of the ball joint in the previous research: The research group developed an intelligent ball joint based on the magnetic effect,and used the magnetic field theory and RBF(Radial Basis Function)artificial neural network technology to model and reverse the angle of the ball;In addition,based on the eddy current effect,a new measurement method is proposed,and the feasibility verification and angle measurement are realized by GRNN(General Regression Neural Network)artific ia l neural network modeling technology.The above two methods realize the detection of the spatial rotation angle of the ball joint.However,there are still a series of problems,such as the time-consuming of inverse solution angle,the poor real-time display effect,the low measureme nt resolution,and the limited space for precision improvement.Aiming at the above problems,focusing on the measurement model of intelligent ball joint,this paper mainly carries out the following work: 1.This paper attempts to build measureme nt model of two modeling methods by Gauss process regression algorithm;2.Firstly,the theoretical data in the early stage of the research group are selected for the Gaussian process regression modeling,and it is found that the maximum angle error of the inverse solution can reach 5°18? and 4°48?.Then it is necessary to modify the parameters of the model through the calculated angular error,and establish the appropriate Gaussian process regression model according to different types of ball joint;3.The research group put different types of ball joint in the calibration experimental device and completed the data training of Gaussian process regression model,compiled a new upper computer interface and displayed the measured angle information on it;4.Finally,the accuracy test and error analysis of the whole new intelligent ball joint system are carried out.The simulation and experimental results show that the angle measurement accuracy and resolution of the intelligent ball joint are improved by using the Gaussian process regression modeling,and the angle measurement error distribution is relatively unifo r m in the whole working space of the ball joint.In the experiment of new intelligent ball joint based on magnetic effect sensor,the maximum error of rotation angle ? in X-axis direction is 6? and the minimum error is 33? in the measuring range of ±20°,the maximum error of rotation angle ? in Y-axis direction is 10?48? and the minimum error is 24?.In the experiment of new intelligent ball hinge based on eddy current sensor,the maximum error of rotation angle ? in X-axis direction is 5?24? and the minimum error is 29? in the measuring range of ±18°,the maximum error of rotation angle ? in Y-axis direction is 3? and the minimum error is 36?.Compared with the equivalent magnet ic charge method and the artificial neural network method,the Gauss process regression algorithm has some advantages.By improving the angle measurement accuracy and resolution of the new intelligent ball joint system,it lays a foundation for the wider application of the intelligent ball joint.