Trajectory tracking control of CNC system based on RBF neural network composite learning control

This paper addresses the high-precision control issues in CNC machine tool servo systems by proposing a feedforward compensation algorithm based on Radial Basis Function Neural Network (RBFNN) composite learning control. Unlike previous studies that updated neural networks solely based on tracking errors, this research prioritizes the accuracy of neural network learning. The paper employs the Selective Memory Recursive Least Squares (SMRLS) method to construct system information prediction errors, which, combined with tracking errors, update the neural network. This enables the neural network to learn the model of the CNC machine tool servo system more accurately, thereby achieving more precise feedforward compensation. Consequently, this method achieves exceptional tracking control performance. The stability of the closed-loop system and the boundedness of the errors are proven using the Lyapunov method. Experimental results on a three-axis CNC machine tool demonstrate that the proposed control algorithm effectively estimates system nonlinearity, thus enhancing tracking control precision.