適應性類神經網路的強健性探討

Robustness Investigation of Adaptive Neural Networks

黃臺民(Tai-Min Huang)

指導教授 ： 賴玲瑩

中原大學/工學院/電機工程研究所/碩士(2002年)

https://doi.org/10.6840/cycu200200166

摘要

摘 要

本文以模型參考適應性控制系統架構來探討強健性控制器。控制器及鑑別器以類神經網路設計，參考模型使用一階轉移函數可視為監督者，命令信號使用步階函數。類神經網路主要是以倒傳遞演算法線上調整參數值，並輔以比例控制器調整偏壓。本控制器特點，以類神經網路為控制器，可減少困難的數學推導，並可線上調整控制器參數值，使其更能適應受控體參數之變動及干擾問題。

本文以適應性類神經網路所設計的控制器分別模擬受控體在16種不同參數變動及瞬間干擾下，使用步階響應顯示此方法確實滿足強健性控制器特性，使強健性控制器設計多一種選擇。

關鍵字

類神經網路 ； 模型參考適應性控制系統 ； 倒傳遞演算法 ； 強健性 ； 學習速率 ； 靈敏度函數

並列摘要

Abstract

In the thesis, it is discuss about adaptive neural network and its robustness. A control strategy well suited for the use of neural networks is the model reference adaptive control system (MRAS) where the system is supervised learning. The parameters of the controller are adjusted by back-propagating the error between the plant and model reference outputs. An identification model is used to train a neural network to identify the plant. The proposed MRAS can be achieved with on-line to adjust the parameters of the controller with robustness against plant parameters variation than the conventional control systems. Computer simulations are given to highlight the feasibility, the simplicity, and the robustness of the proposed method.

並列關鍵字

model reference adaptive control system (MRAS) ； learning rate ； Robust ； sensitivity function ； back-propagating algorithm ； NN

參考文獻

[1] S. Haykin, Neural Networks A Comprehensive Foundation , 2nd ed , Prentice Hall International Inc., 1999, ch. 15.

[2] S. Buso, “Design of a Robust Voltage Controller for a Buck-Boost Converter Using μ-Synthesis,” IEEE Trans. on Control Systems Technology, vol. 7, No. 2, March 1999.

[3] W. S. McCulloch and W. Pitts, “A logical calculus of the ideal immanent in neurons activity,” Bull. Math. Biophy., vol. 5, pp. 115-133, 1943.

[4] F. Rosenblatt, Principles of Neurodynamics, New York:Spartan, 1962.

[6] T. Kohonen, “Self-organized formation of topologically correct feature maps,” Biol. Cybern., no. 43, pp. 59-69, 1982.