摘要
针对某顶置火炮高低向调炮控制系统存在的强非线性特征,提出了一种分数阶神经网络滑模控制(FNSMC)策略。引入分数阶微积分(FOC),设计了分数阶PID型滑模面,获得了用于火炮控制的具有分数阶动力学特征的等效控制量。采用饱和函数作为切换函数,基于RBF神经网络对其切换增益进行在线动态调节,以获得动态最优性能。通过数值仿真分析比较了引入FOC后滑模控制系统的动静态特性,结果表明:分数阶滑模控制(FSMC)系统能够更为快速平滑的趋近稳态,这将有效抑制抖振现象,减小控制系统响应时间。半实物仿真试验结果表明:所提出的FNSMC策略明显优于传统整数阶神经网络滑模控制(CNSMC),具有更强的鲁棒性及更高的控制精度,可以很好实现预期的快速、平稳和高精度调炮。
A novel fractional order neural sliding mode control (FNSMC) strategy is proposed for the nonlinearities of a gun control system (GCS) which is used to control the elevation of a certain top- mounted gun. A fractional order PID type sliding surface is especially designed by introducing the frac- tional order calculus, and an equivalent control discipline with fractional order dynamics is induced. The saturation function is employed as switch function. To achieve the best control performances, a dynamic adjustment approach of the switch gain is introduced based on RBF neural network. The dynamic and static characteristics of the fractional order sliding mode control (FSMC) system are analyzed by numeri- cal simulation, demonstrating that FSMC can reach up to the steady state more smoothly, which signifi- cantly suppresses the chatter effects and enhances the response rate of the control system. Finally, a se- ries of experiments on a semi-physics simulation platform are conducted to investigate the performances of control system. The results show that the proposed FNSMC is of more excellence than the conventional in- teger order neural SMC ( CNSMC). The FNSMC-based control system is of better tracking accuracy as well as high robustness, and the fast, smooth and accurate adjustments of the gun can be achieved.
出处
《兵工学报》
EI
CAS
CSCD
北大核心
2013年第10期1311-1317,共7页
Acta Armamentarii
基金
国家重点基础研究发展计划项目(61311603)
关键词
自动控制技术
炮控系统
滑模控制
分数阶微积分
RBF神经网络
automatic control technology
gun control system
sliding mode control
fractional order cal-culus
RBF neural network