摘要
针对舰载火箭炮伺服系统中存在的非线性因素,提出了一种基于自构建神经网络的内模控制方法。该方法通过神经网络建立过程模型和过程逆模型,利用自构建算法调整神经网络的结构和参数。神经网络能依据给定的阈值自动增加和删除神经元节点,以实现结构学习;采用梯度下降法进行网络的参数学习。仿真结果表明,该控制方法有良好的跟踪性、鲁棒性和抗干扰性,能提高火箭炮的调炮精度。
In response to the nonlinear factors existing in ship-borne rocket servo system, an internal model control （IMC） method based on self-construction recurrent neural network was proposed. By means of neural network, the model and inverse model of the process were established. The structure and parameters of neural network were adjusted by self-constructing algorithm. The neural network was able to automatically add and delete neuron nodes according to the given threshold to achieve structural lear-ning. The gradient descent method was used to achieve parameter learning of network. The simu- lation results show that the control method has good tracking performance, robustness and anti-interfer- ence, and can improve the adjustment accuracy of rocket gun.
出处
《火炮发射与控制学报》
北大核心
2017年第4期40-44,共5页
Journal of Gun Launch & Control
关键词
舰载火箭炮
自构建神经网络
内模控制
伺服系统
ship-borne rocket
self-construction recurrent neural network
IMC
servo system
