含齿隙转台伺服系统的RBF神经网络反步控制  

RBF neural network backstepping control of turntable servo system with backlash

作　　者：王世剑 孙红江 王佳伟 董光能[1] WANG Shijian;SUN Hongjiang;WANG Jiawei;DONG Guangneng(School of Mechanical Engineering,Xi’an Jiaotong University,Xi’an 710049,China)

机构地区：[1]西安交通大学机械工程学院,西安710049

出　　处：《现代制造工程》2021年第5期39-46,13,共9页Modern Manufacturing Engineering

基　　金：国家重点研发计划“智能机器人”重点专项项目(2018YFB1306100)。

摘　　要：针对工业机器人转台伺服系统中齿轮齿隙导致的驱动延时和响应迟滞,提出了一种基于径向基函数(Radial Basis Function, RBF)神经网络的反步控制方法。首先,建立工业机器人转台伺服系统的状态空间方程,运用死区函数描述了齿隙非线性;其次,设计了速度观测器,对速度信息进行估计,解决了工业机器人转台伺服系统负载的速度信息难以测量的问题,同时应用反步法,引入虚拟控制量,选择Lyapunov函数反步递推,RBF神经网络用于逼近系统中的非线性部分,设计出RBF神经网络反步控制器,并通过李雅普诺夫稳定性原理证明了系统的稳定性;最后,搭建了仿真系统和实验平台,并且和单一反步控制策略仿真结果进行了对比,从系统对不同信号的跟踪精度、系统的鲁棒性以及自适应能力三方面验证了所提控制策略的优越性,在系统受到干扰时,对正弦信号的跟踪误差约为单一反步控制策略的50%;在方波信号跟踪下,响应时间降低了约40%,跟踪精度提升了50%。A control method combining Radial Basis Function(RBF)neural network and backstepping control was proposed to solve the problem of drive delay and response delay caused by backlash in turntable servo system.Firstly,the state space model of the servo system was established and the dead zone function was introduced to describe the nonlinearity of the backlash.Secondly,speed observer was designed to estimate the speed of the servo system,which is difficult to acquire.Meanwhile,virtual control quantity was introduced.Using the backstepping control method and recursive method with Lyapunov function step by step,the non-linearity in the system was approximated by RBF neural network.Then a RBF neural network backstepping controller was designed and the stability of the system was proved by Lyapunov stability theory.Finally,a simulation system and an experimental platform were established.Compared with the results of single backstepping control strategy,the advantages of the proposed control strategy were verified from three aspects:the tracking accuracy of the system to different signals,the robustness of the system and the adaptive ability.The tracking error of the sinusoidal signal is about 50%of the single backstepping control strategy when the system is disturbed.Under square wave signal tracking,the response time is reduced by 40%and the tracking accuracy is improved by 50%.

关 键 词：伺服系统 齿隙 反步控制 径向基函数神经网络反步控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]