Neural Network and Observer Compensation Control for Virtual-Axis Machine Tool

This paper presents a novel method of neural network and observer compensation control for 6-dof virtual-axis machine tool driven by permanent magnetic linear synchronous motor (PMLSM) by keeping the 6 links in cooperating motion. The coupling 6-link system, in this paper, is divided into 6 separate systems, and each link´s load disturbance and the coupling disturbance are regarded as the dynamic parameters´ variation of the PMLSM. An observer is presented to identify the parameters´ variation and to compensate for propulsive current i_q of PMLSM through adjusting the weights of neural network IP position controller, which eliminates the coupling and obviates a lot of cumbersome dynamic calculations of the parallel mechanism. In each separated system, the position controller is adjusted on-line depending on the position error and the speed command to be appropriate for the time variation of the position and speed, and the speed controller is realized by one neuron, which can be rapidly trained on-line, the acceleration control is attained indirectly by the proposed observer and neuron, which improves tracking performance of the position and speed output. The results of simulation indicate that the presented control strategy used in the position, speed and acceleration of each link can be realized, and the system has a good robustness. Conclusions can be drawn that the new method can satisfy the control performance require of 6-dof virtual-axis machine tool, at the same time, this control strategy is reasonable and feasible