基于遗传算法的摩擦模型参数辨识研究

【作者】 刘红；

【导师】 高伟；

【作者基本信息】 中国科学院研究生院（西安光学精密机械研究所） ， 控制理论与控制工程， 2007， 硕士

【摘要】 摩擦是制约伺服系统控制精度提高的一个重要因素，要想实现伺服系统的低速高精度控制，就须对系统存在的摩擦进行补偿。基于摩擦模型的补偿方法更加具有针对性，如果能得到系统的比较准确的摩擦模型，一般都能得到很好的控制效果。摩擦模型参数辨识是本文讨论的主要问题。综述了国内外在摩擦模型参数辨识方面的研究进展情况，通过分析比较，提出了一种基于遗传算法进行摩擦模型参数辨识的方法。详细分析了摩擦力的产生机理、摩擦的动态现象、伺服系统中的摩擦现象以及几种常用的静态、动态摩擦模型。通过分析、比较，从而选择“库仑+粘性”摩擦模型和Tustin摩擦模型作为辨识对象进行辨识仿真。然后选择一个二阶系统为被控对象，在被控系统上加一个随机干扰信号来模拟系统中的摩擦现象，设计了一种遗传辨识方法，来辨识系统的摩擦模型参数，并通过设计Matlab程序实现这个辨识仿真，得到摩擦模型参数。最后利用“PID+前馈”摩擦补偿方法，通过设计Matlab／Simulink程序实现了基于摩擦模型的伺服系统摩擦补偿仿真，验证了辨识结果的有效性，也说明基于遗传算法的摩擦模型参数辨识方法是可行的。更多还原

【Abstract】 Friction is an impediment to improve the control accuracy of servo systems,before to achieve low-velocity and high precision control, we must compensate thefriction exist in the system. Model-based friction compensation is more pertinency.Generally, this method can achieve good control effect if we can gain an exactfriction model of the system. Parameters identification of friction model is the mainissue of this dissertation.The present research work of parameters identification of friction model athome and abroad is summarized firstly. After analyze and compare these methods, Aparameter identification method of friction model with Generic Algorithms is putforward.Then this dissertation analyzed the creation theory of friction, the dynamicphenomenon of friction, the friction phenomenon in servo sysytems and somestatic/dynamic friction models which are commonly used. Via analyze and compareof the characteristics of these models, I choose the "Coulomb&Viscous model" and"Tustin model" as the identification objects. I choose a two rank system as thecontrol system, a random distubance was plus to this system to simulate the frictionphenomenon in servo system, and use the Generitc Algorithms to identify theparameters of the "Coulomb&Viscous model" and "Tustin model". These are carriedout by some Maflab programs.Finally, using the Matlab/Simulink, a model-based "PID+feedforward" frictioncompensation method is carried out, and proved this parameter identificationmethod based on the Genetic Algorithms is effective.更多还原