柴比雪夫類神經模糊網路應用於伺服系統之判別模式

Identification of Servomechanism Model Based on a Chebyshev Recurrent Neuro-Fuzzy Network

黃元瑞(Yuan-Ruey Huang)

指導教授 ： 康 淵 ； 張永鵬

中原大學/工學院/機械工程研究所/博士(2008年)

https://doi.org/10.6840/cycu200800066

摘要

柴比雪夫類神經模糊網路(CRNF)是將柴比雪夫遞迴式類神經網路與模糊系統結合而成。柴比雪夫類神經模糊網路為九層之網路結構其中包含一個六層網路結構之柴比雪夫遞迴式類神經網路。柴比雪夫類神經模糊網路基於Takagi-Sugeno-Kang (TSK) 模糊模型之概念，改良後而形成，主要以柴比雪夫遞迴式類神經網路為模糊系統之模糊規則的推論機構。柴比雪夫遞迴式類神經模糊網路以柴比雪夫函數之展開功能，利用非線性近似的方式將原先在較低維的輸入空間擴展至高維度之超平面(hyperplane)以對輸入訊號具有更佳之判別能力。
本文以柴比雪夫類神經模糊網路法則作為平面皮帶輪伺服機構與單軸平台伺服機構之判別模型，並在Matlab環境下以柴比雪夫類神經模糊網路模擬平面皮帶輪伺服機構與單軸平台伺服機構並驗證其適應性與學習能力。另外以適應性類神經模糊推論系統(ANFIS)與遞迴式類神經網路(RNN)對相同伺服機構建立判別模型，經數值分析及探討分析比較得到柴比雪夫類神經模糊網路模型之收歛性與精確性優於其它兩種判別模型。
本文也以柴比雪夫類神經模糊網路建立之單軸平台伺服機構的模型，以模擬及實驗探討伺服機構定位之膠著滑動現象，並以模擬探討間接適應性控制法則，以CRNF為參考模型之伺服控制結果。

關鍵字

平面皮帶輪伺服機構 ； 柴比雪夫類神經模糊網路 ； 適應性類神經模糊推論系統 ； 摩擦力補償 ； Takagi-Sugeno-Kang模糊模型 ； 單軸平台伺服機構

並列摘要

A novel nine-layer neuro-fuzzy model, Chebyshev recurrent neuro-fuzzy (CRNF) network, is developed based on the combination of the Takagi-Sugeno-Kang (TSK) fuzzy model and the six-layer Chebyshev recurrent neural network (CRNN) for the modeling of nonlinear dynamic systems. System nonlinearity is approximated by enhancing the input dimensions of the consequent parts in the fuzzy rules due to functional expansion of a Chebyshev polynomial.
The modeling of a belt servomechanism and ball-screw-driven servomechanism, based on the CRNF network, is studied under the Matlab environment. In addition, the modeling of this servomechanism by using both identification methods of adaptive neuro fuzzy inference system and recurrent neural network is also studied. The analysis and comparison results indicate that CRNF makes identification of complex nonlinear dynamic systems easier. The identification results of a belt servomechanism verify that the accuracy and convergence of the CRNF are superior to those of ANFIS and RNN.
Also, the phenomena of stick-slip are studied by analysis and experiment on the ball-screw-driven servomechanism. Moreover, the adaptive friction compensation with indirect adaptive control is developed and simulated, based on the CRNF network. According to the result of simulation, the proposed controller shows good and robust tracking performances with respect to parameter variations.

並列關鍵字

Chebyshev recurrent neuro-fuzzy (CRNF) network ； ball-screw-driven servomechanism ； Takagi-Sugeno-Kang fuzzy model ； belt servomechanism ； friction compensation ； Adaptive network-based fuzzy inference system

參考文獻

[1] Narendra, K. S., Parthasarthy, K., Identification and control of dynamical systems using neural networks, IEEE Transactions on Neural Network, Vol. 1, No. 1, pp. 4-27, 1990.

[3] Wai, Rong-Jong, Lin, Faa-Jeng, “Adaptive recurrent-neural-network control for linear induction motor,” IEEE Transactions on Aerospace and Electronic Systems, Vol. 37, No. 4, pp. 1176 – 1192, 2001.

[4] Lin, Chih-Min, Hsu, Chun-Fei, “Recurrent-neural-network-based adaptive-backstepping control for induction servomotors,” IEEE Transactions on Industrial Electronics, Vol. 52, No. 6, pp. 1677 - 1684, 2005.

[5] Zadeh, L. A., “Fuzzy sets,” Inform. Control, Vol. 8, pp. 338-352, 1965.

[6] Takagi, T. and Sugeno, M., “Fuzzy Identification of systems and Its applications to modeling and control,” IEEE Transaction on Systems Man Cybernet. SMC-15, pp. 116-132, 1985.