摘要
基于神经网络算法的自学习非线性逼近能力提出了磁悬浮系统的参数自整定PID控制算法.为提高控制参数的优化速度,以权值梯度符号函数取代其变化值,构建了3层BP网络调节PID的比例积分和微分参数.在虚拟现实的辅助下搭建系统仿真结构,将算法应用到悬浮小球的位置控制之中.仿真结果显示,该算法可以在0.1 s内实现参数优化,位置跟踪速度快,超调量小,具有较强的抗干扰能力.
The BP neural network algorithm was adopted to design the PID regulator of a magnetic levitation system for the sake of its nonlinear approximation and generalization ability. To increase the speed of parameters optimization, a three layers BP neural net-work was introduced to obtain the main control parameters known as proportional integral derivative. The gradients were also replaced by their sign functions. Simulation structure was set up under virtual reality technology. The algorithm was applied to control the posi-tion of a iron ball. Simulation results show that the stable PID parameters can be given in 0. 1 second, then small overshoot, fast re-sponse and strong disturbance rejection ability are ensured.
出处
《河南工程学院学报(自然科学版)》
2016年第3期42-46,共5页
Journal of Henan University of Engineering:Natural Science Edition
基金
河南工程学院博士基金(D201213)
关键词
神经网络
磁悬浮
虚拟现实
自适应PID
neural network
magnetic levitation
virtual reality
adaptive PID