直接自适应动态递归模糊神经网络控制及其应用

Direct Adaptive Dynamic Recurrent Fuzzy Neural Network Control and Its Application

摘要: 针对某些仿射非线性系统中各状态变量间呈微分关系的特点，本文提出仅取某些可测状态变量作为动态递归模糊神经网络（dynamic recurrent fuzzy neural network，DRFNN）的输入，而由DRFNN的反馈矩阵描述系统内部动态关系的直接自适应DRFNN控制算法，克服了将系统所有变量作为输入的传统模糊神经网络（traditioanl fuzzy neural network，TFNN）因某些不可测状态变量所导致的不可实现问题．在电液伺服系统中的应用结果表明：直接自适应DRFNN控制算法相对于TFNN控制算法对系统稳态特性的改善具有较大的优越性．

Abstract: For the characteristics that derivative relations exist between states of some affine nonlinear systems,a direct adaptive dynamic recurrent fuzzy neural network(DRFNN)control algorithm is proposed,which takes some measurable state variables as the DRFNN inputs and describes the system inner dynamic relation by the DRFNN feedback matrix. The unrealizable problem caused by some system unmeasurable state variables in traditional fuzzy neural network(TFNN) which takes all the state variables as its inputs is overcome.The results of its application to electro-hydraulic servo system show that direct adaptive DRFNN control algorithm has an advantage over the TFNN control method in improving the steady state characteristics of the system.