摘要
提出了一种基于回归小波网络(recurren t w ave let netw orks)的多输入多输出系统(m u lti-inpu t m u lti-ou tpu t)动态解耦控制的新方法。该方法为分散式控制结构,采用回归小波网络作为解耦辨识器(decoup ling iden tifier),在线动态辨识、回馈对应输入输出的灵敏度信息(sens itiv ity in form ation),P ID神经网络控制器根据回馈信息实现自适应分散独立控制。小波函数的紧支性、波动性以及回归网络较强的动态非线性映射能力使得回归小波网络具有较好的综合性能。仿真结果表明,用该方法构成的控制系统解耦效果好,收敛速度快,且具有较好的鲁棒性。
A novel dynamic decoupling control method for MIMO nonlinear system based on recurrent wavelet networks is presented in this paper. It adopts dispersive neural networks control structure and utilizes recurrent wavelet networks as decoupling identifier and PID neural network as controller respectively. The decoupling identifier identifies the MIMO nonlinear system and feedback the sensitivity informaiton for PID controller through on-line learning. Simultaneously PID controller updates the weights and adjusts the system actively. Benefited from wavelet transform＇ being constrictive and fluctuant, it shows excellent temporal-frequency localization property,while it possesses such merits as powerful ability of mapping nonlinear systems, capturing the dynamic behavior of the system ect, thus, this network converges quickly with high precision and good robustness.
出处
《火力与指挥控制》
CSCD
北大核心
2006年第8期48-52,共5页
Fire Control & Command Control
作者简介
段翀（1981-），男，山西运城人，在读博士研究生，主要研究方向为航空发动机智能诊断与信息融合。
