一类交流伺服系统RBF神经网络与PID复合控制策略研究

【作者】 牛姗姗；

【导师】 陈庆伟； 樊卫华；

【作者基本信息】 南京理工大学 ， 系统工程， 2007， 硕士

【摘要】 随着现代科技的发展，交流伺服技术的应用领域更为广泛，所发挥的作用越来越突出，对其性能指标的要求也越来越高。交流伺服系统是一个具有非线性、耦合性和时变性的复杂系统，常规的控制策略难以取得理想的控制效果，本文针对交流伺服系统高品质的控制要求以及其存在的摩擦非线性、系统参数变化、负载扰动等问题，研究了一种RBF神经网络与PID控制相结合的复合控制策略。本文首先建立了包括摩擦非线性在内的交流伺服系统动力学模型；在论证了继电反馈技术应用在交流伺服系统中可行性的基础上，采用一种利用系统的多点频率信息及根据系统特性来拟合理想闭环响应的继电反馈PID控制器设计方法；针对系统中存在的摩擦非线性以及不确定性，提出了一种RBF神经网络与PID复合控制策略，并证明了该复合控制策略下闭环系统的稳定性；对RBF网络的最近邻聚类学习算法进行了改进；基于改进的最近邻聚类法和梯度下降法的混合学习算法，设计了系统的RBF网络控制器。仿真结果表明，应用本文所设计的RBF神经网络与PID复合控制器能获得较高的动态和稳态性能，并对参数变化和负载扰动具有良好的鲁棒性，可很好地满足系统的高性能指标要求。更多还原

【Abstract】 With the development of modern science, AC servo technique has been widely used. AC servo system is a high order, nonlinear, coupling system. A single control strategy can hardly get ideal control effect. The performance of system is highly influenced by the friction, parameter variations and external load disturbances. In order to enhance static and dynamic performances of the system, this dissertation is to study a combined control strategy—PID control and RBF neural network control.In this dissertation, the linear mathematical model and friction nonlinear model of AC servo system are presented. Relay feedback’s results on the existence and stability of the limit cycles are established by the servo system’s transformatiOn function. A method based on relay feedback technique is presented to identify multiple ,points on the process frequency response, then to pattern ideal closed loop characteristic of AC servo systemt. To overcome the friction and variations in the system, a RBF neural network. controller is designed as a parallel controller, and the stability of system is demonstrated. Meanwhile, a hybrid learning algorithm for RBF network based on improved nearest neighbor-clustering algorithm and gradient descent training, is proposed to identify system model and design the RBF controller.Simulation results show that AC servo system based on proposed mixed control strategy can satisfy the system’s need of speediness、Stability and robustness preferably.更多还原