基于神经网络的轮廓误差控制仿真研究

运动控制系统不同轴的动态特性的差异是造成系统轮廓误差较大的主要原因。为系统每个轴设计一个高性能的前馈补偿器，可以改善每个轴的动态特性，使不同轴相互匹配，从而残小系统的轮廓误差。由于前馈补偿器的传递函数通常具有二阶或二阶以上微分环节，物理实现较困难，本文利用神经网络辨识系统的控制对象逆模型作为前馈补偿器，很好地解决了该问题；而利用在遗传算法中加入BP算法形成的混合遗传算法作为神经网络的训练算法，此算法具有收敛速度快，不易陷入局部权小的特点。

The Simulation Research of Contour Error Control Methods Based on Neural Networks

In motion control system, the differences of the dynamic characteristics of each axes are the main reason of the contour error. We can design a feed forward module for each axes to improve the dynamic characteristics and reduce the contour error of the system. Because of the difficulty of getting the high rank differential signals of input, the compound control is hard to realize. To solve this problem we use two different kinds of Neural Networks to identify the contrary control object and use each Neural Networks as feed forward module. We also do some researches on the Genetic Algorithm. By adding BP algorithm into GA(Genetic Algorithm) we have a new algorithm called Hybrid Genetic Algorithm (HGA). It has the advantage of fast convergence and it will not converge at local nadir.