摘要
为了减小压电陶瓷固有的迟滞非线性特点对快速伺服刀架(FTS)控制精度的影响,提出了一种基于RBF神经网络的快速伺服刀架迟滞特性建模方法.利用拓展输入空间法建立了FTS迟滞系统的RBF神经网络模型,通过引入指数型迟滞算子,将FTS系统的输入与迟滞算子的输出一起作为RBF神经网络的输入向量,实现了FTS迟滞系统由多值映射到单值映射的转换,进而利用神经网络对其进行建模.为了更精确地跟踪快速伺服刀架的迟滞位移曲线,通过增加调整系数σ来对迟滞算子进行改进.实验表明,该迟滞模型可以很好地预测快速伺服刀架的迟滞位移曲线,模型的验证均方差MSE=5.163 3×10-6.
A hysteresis model based on RBF neural networks is proposed for reducing the nonlinearity of piezoelectric actuator which leads to the inaccuracy of fast tool servo system （FFS）. The hysteresis model using RBF （radical basis function） neural networks of FFS is established based on the expanded input space method. An exponential hysteresis operator is proposed to combine the input of the FFS system and the out put of the hysteresis operator together as the input of the RBF neural network and construct an expanded input space, so as to transform the multivalued mapping into a onetoone mapping which enables neural networks to establish the model for the behavior of hysteresis of FFS. An adjustment coefficient was added to improve hysteresis curve tracing accuracy of the hysteresis oper ator. Experimental results indicate that the measured displacement curves obtained from the hysteresis model are in good agreement with the predicted curves, and the mean square error is 5. 163 3 x 106.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2012年第A01期217-220,共4页
Journal of Southeast University：Natural Science Edition
关键词
快速伺服刀架
迟滞算子
RBF神经网络迟滞模型
fast tool servo
hysteresis operator
RBF（radical basis function） neural network hyster-esis model
