Measurement Of The Spatial Rotation Angle In The Precision Of Ball Joint Based On Artificial Neural Network

Precision spherical joint is a movement execution unit which can provide 3 –DOF rotation motion,which is widely applied in parallel mechanisms,robotics,and other areas,but its rotation orientation and angle cannot be obtained during its passive motion.If the spherical joint rotation angle can be measured accurately in real-time,it will be useful for compensating the motion clearance error and realizing the equipment motion control.The traditional optical and mechanical angle measurement methods are not suitable for precision spherical joint angle detection because of its simple construction and small size.On the premise of not affecting the accuracy and strength of the spherical joint,an embedded intelligent spherical joint that can detect its rotation angle based on magnetic effect is developed with embedding permanent magnet in the ball head,and some Hall sensors are embedded in the ball nest.When the permanent magnet rotates with the ball head in the space,the sensors detect the variation of magnetic induction intensity.Then,the rotation angle of the ball joint can be achieved by established model.A thorough and systematic study followed the previous results of the project team was proposed in this paper.Our destination is to improve measuring accuracy and resolution continuously.The work included: The magnetic model method on the inverse solution algorithm is time consuming.An artificial neural network is chosen to get the relationship between the measurement value of hall sensors and the rotation angle of the spherical joint instead.An improvement plan was presented for a selected permanent magnet and sensor on the basis of simulation analysis.The characteristic parameters of the spherical joint prototype were optimized.The calibration device was developed,and on the device the data training which neural networks needed was completed.Special measuring software was designed to match the neural network method.What's more,the accuracy test of the spherical joint prototype was completed with the help of the calibration device.Result of the experimental test and angle error calculation has shown: in the range of ±20°,the average errors of ? is about 1?51? while the maximum error is 23'56 " and the minimum error is-29'31".The average errors of ? is about 1?55? while the maximum error is 36?46? and the minimum error is-36?36?.Compared with the analytical method,after using the artificial neural network algorithm made the measurement error small in the small-angle range,the overall error distribution getting uniform,and there is no increasing error at the measurement boundary.Through the analysis of the test data,we have found the main source of error,and the measurement accuracy and resolution are expected to be further improved and this method will be used in precision engineering in the future.