動態系統之柴比雪夫遞迴式小波類神經模糊網路判別模型

Identification of Dynamical System Based on Chebyshev Recurrent Wavelet Neuro-Fuzzy Network

10.6989/JN.201112.0133

黃元瑞(Yuan-Ruey Huang)；童景賢(Jing-Syan Torng)

柴比雪夫 ； 小波 ； 類神經模糊網路 ； 遞迴式類神經網路 ； Chebyshev ； wavelet ； neuro-fuzzy network ； recurrent neural network

南亞學報

31期（2011 / 12 / 01）

133 - 146

繁體中文

本文提出柴比雪夫遞迴式小波類神經模糊網路模式（Chebyshev Recurrent Wavelet Neuro-Fuzzy network, CRWNF）以TSK型式的模糊網路模式為基礎、以柴比雪夫遞迴式小波類神經網路為模糊推論機構。小波函數能將處理的訊號在建構時頻表示時擁有良好的時域和頻域的定位。柴比雪夫遞迴式小波類神經網路利用柴比雪夫多項式與小波函數使其具有動態映射之功能，因此對具有非線性摩擦力與背隙之動態系統有較優之鑑別之功能。研究以單軸平台系統為非線性摩擦力之動態系統，經模擬結果可知柴比雪夫遞迴式小波類神經模糊網路較遞迴式類神經網路、適應性網路模糊推理系統等有較良好之非線性系統動態特性之鑑別。

In this study, Chebyshev Recurrent Wavelet Neuro-Fuzzy network, (CRWNF), based on the structure of TSK neuro-fuzzy model, uses Chebyshev Recurrent Wavelet Neural networks as inferring mechanisms. Wavelet basis functions have the ability to localize both in time and frequency domains. The outputs of Chebyshev Recurrent Wavelet Neural network have the previous outputs of its own which make it as a dynamical mapping mechanism. Hence, CRWNF with dynamical mapping functions can identify nonlinear dynamical system effectively. In this paper, the ball-screw-driving system is used as a nonlinear frictional dynamic system. According to the simulation results, the proposed CRWNF, comparing the effectiveness of the models ANFIS and RNN, has impressive generalization ability.