Research on Measurement Method of Spherical Joint Rotation Angle Based on ELM Artificial Neural Network and Eddy Current Sensor

This paper proposes a measurement method for the rotation angle of the spherical joint based on the extreme learning machine (ELM) artificial neural network and four eddy current sensors. Aiming at the problems of small range and low accuracy in the early three-eddy-current angle measurement prototype, the position matching scheme of four eddy current sensors is researched, a new prototype is developed through simulation analysis, and ELM neural network substitutes the previous generalized regression neural network (GRNN) for building a new measurement model. The modelling training and comparison test are completed in the self-developed high-precision angle calibration device. Experimental results show that the new prototype not only covers a +/- 20 degrees measurement range but also promotes measurement accuracy, and the standard deviation of the single-axis measurement drops to 3' within the range of 5 degrees-15 degrees. It provides a relatively high-precision measurement method for real-time, multi-axis active detection of spherical joint space rotation angle error.