采用遗传神经网络的直线伺服系统速度自适应控制

摘要

针对永磁直线同步电动机伺服系统多变量、强耦合、非线性,以及时变的特点,提出一种基于径向基函数(RBF)神经网络的永磁直线同步电动机速度控制方法,该方法根据RBF神经网络的结构特点,对其参数采用遗传算法进行离线训练,并通过在线训练调整隐层与输出层的连接权值,实现了电动机速度的自适应控制.实验结果表明,与常规PID控制相比,RBF神经网络控制器提高了直线伺服系统的静态和动态性能,使得电动机在启动和负载改变时转速超调量减小,且转速能更快地趋于稳定.%For the Permanent Magnet Linear Synchronous Motor (PMLSM) servo system which contains many variants,strong coupling,non-linear and time variable characteristics,a speed control method based on Radial Basis Function(RBF) neural network for PMLSM is proposed which achieves the adaptive control of the motor speed with the network parameters off-line trained by the Genetic Algorithm(GA) and connection weights of the hidden layer and output layer in-line trained.The experiment result shows that,comparing with traditional PID control ,the static and dynamic performances of the system are improved by using the RBF neural network controller, the overshoot of the speed decreases and the speed tends to be stable faster when motor starts or load changes.