基于自构建神经网络高精度交流位置伺服系统研究

发布时间：2018-01-26 05:14

本文关键词： 舰载火箭炮交流位置伺服系统自构建神经网络算法自适应内模控制　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：现代战争中越来越强调火力打击的高精度性和稳定性,期望以最小的代价达到最好的打击效果。对于舰载火箭炮系统而言,影响其射击精度的因素很多,比如气象条件,舰船摇摆和调炮精度等,其中舰载火箭炮的调炮精度是其中最重要的因素之一。本文在某舰载火箭炮研制背景下,对其交流伺服系统建模和控制策略进行研究,提高其交流伺服系统控制性能,进而提高舰载火箭炮的调炮精度,最终实现提高舰载火箭炮射击精度的目的。本文主要研究了以下几个方面:(1)对某舰载火箭炮交流伺服系统基本构成进行介绍,分析交流伺服系统所采用的执行元件-永磁同步电机(Permanent Magnet Synchronous Motor,PMSM),并在矢量控制下,推导出PMSM模型。在PMSM模型的基础上,推导出舰载火箭炮交流伺服系统的数学模型。根据数学模型可知,其交流伺服系统是一个复杂非线性系统。(2)对系统进行了辨识研究。首先采用径向基函数(Radical Basis Function,RBF)神经网络对系统辨识。针对传统RBF神经网络辨识能力差,网络结构固定,节点数一旦确定就不能改变的不足,本文提出改进的自构建递归神经网络算法,其可以在线实现隐含层神经元节点数的增加和删除。仿真结果表明,基于自构建递归神经网络的系统辨识效果更好。(3)对系统的控制策略进行了研究。首先设计了内模控制应用于系统控制研究。针对内模控制的不足,设计了基于自构建递归神经网络的自适应内模控制应用于交流位置伺服系统控制。在MATLAB中搭建系统仿真模型,验证了所设计方案的有效性。(4)搭建了舰载火箭炮交流伺服系统半实物仿真平台,设计了硬件电路,上位机软件控制界面来验证所采用的基于自构建递归神经网络的自适应内模控制策略的控制性能。试验结果表明,该控制策略能够满足舰载火箭炮高精度控制要求。
[Abstract]:In modern warfare, more and more emphasis is placed on the high precision and stability of firepower strike, and the best strike effect is expected to be achieved at the lowest cost. For shipborne rocket launcher system, there are many factors that affect its firing accuracy. Such as weather conditions, warship swing and artillery accuracy, among which the accuracy of shipborne rocket launcher is one of the most important factors. The modeling and control strategy of AC servo system is studied to improve the control performance of AC servo system and improve the accuracy of shipborne rocket launcher. Finally, the purpose of improving the firing accuracy of shipborne rocket launchers is achieved. This paper mainly studies the following aspects: 1) the basic structure of AC servo system of a shipborne rocket launcher is introduced. The permanent Magnet Synchronous motor (PMSM), which is used in AC servo system, is analyzed. Under vector control, the PMSM model is derived. Based on the PMSM model, the mathematical model of shipborne rocket launcher AC servo system is derived. The AC servo system is a complex nonlinear system. Firstly, the radial basis function (radial basis function) is used to identify the system. The traditional RBF neural network has poor identification ability, fixed network structure, and can not change the number of nodes once determined. In this paper, an improved recursive neural network algorithm is proposed, which can increase and delete the number of neurons in the hidden layer on line. The control strategy of the system is studied based on the self-constructed recurrent neural network. Firstly, the internal model control is designed to study the system control, aiming at the shortcomings of the internal model control. Adaptive internal model control based on self-constructed recurrent neural network is designed for AC position servo system control. System simulation model is built in MATLAB. Verify the effectiveness of the design. 4) built a shipborne rocket launcher AC servo system hardware-in-the-loop simulation platform, designed the hardware circuit. The upper computer software control interface is used to verify the control performance of the adaptive internal model control strategy based on self-constructed recurrent neural network. The control strategy can meet the requirements of high precision control of shipborne rocket launchers.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TJ393;TP183

【相似文献】