智慧型B-spline類神經控制器設計於音圈馬達驅動電路

機構典藏 > College of Engineering > Graduate Institute & Department of Electrical Engineering > Thesis > Item 987654321/102677

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/102677

Title:	智慧型B-spline類神經控制器設計於音圈馬達驅動電路
Other Titles:	Design of an intelligent B-spline neural control system for a voice coil motor
Authors:	陳逸全;Chen, Yi-Chuan
Contributors:	淡江大學電機工程學系碩士班許駿飛
Keywords:	B-spline基底函數;B-spline類神經網路;全域滑動模式控制;遞迴步階控制;音圈馬達;B-spline;total sliding mode control;backstepping control;Voice Coil Motor
Date:	2014
Issue Date:	2015-05-04 10:01:40 (UTC+8)
Abstract:	由於音圈馬達具有高速動作特性，使得音圈馬達的應用範圍越來越廣，但其動態方程式卻時常難以正確取得。為了解決此困擾，眾多專家學者利用類神經網路，設計了多種不同的智慧型類神經控制器，其主要利用類神經網路線上自我學習能力來進行控制器設計。在常見的類神經網路隱藏層神經元一般都使用S型函數，其特性會使權重值沒有收斂的效果，則該類神經網路對於學習效果往往不盡理想。另一方面B-spline函數曲線擁有曲線收斂效果並具有局部調整能力，而且能迅速學習任意非線性函數的能力。基於此優點，本論文結合了B-spline基底函數提出了B-spline類神經網路，其網路架構中的隱藏層神經元激發函數即採用B-spline基底函數，如此，B-spline類神經網路具有良好的非線性未知方程式近似學習效果。接著，本論文提出了兩種智慧型控制方法，一為智慧型全域滑動模式控制，另一為智慧型遞迴步階控制，在上述兩種不同控制系統中均使用B-spline類神經網路線上學習近似音圈馬達的動態方程式，更利用泰勒線性展開技巧與李亞普諾夫穩定定理，推導出合適的參數學習法則。同時，我們設計了比例-積分形式的參數學習法則來加速網路學習速度使其具有良好的學習效果。另一方面，為了克服B-spline類神經網路學習受控系統動態方程式的誤差，利用模糊補償器來消除類神經網路學習近似誤差對控制響應的影響。最後，本論文將所提出的兩種智慧型控制方法應用到音圈馬達定位控制問題上，並在32位元微電腦單晶片上實現控制法則，經由實驗結果可以顯示，本論文所提出的兩種不同智慧型控制方法均可以獲得良好的控制效果。 Since the dynamic model of a voice coil motor (VCM) driver is difficult to obtain, the model-based control techniques are not easy to be used for a VCM driver. To overcome this drawback, this paper proposes two intelligent B-spline neural control system. One is intelligent total sliding mode control,and the other is intelligent backstepping control. In the proposed controler design, a B-spline neural network (BNN) is used to online approximate an unknown nonlinear term in the system dynamics of a VCM driver by tuning its interior parameters. The proposed intelligent control systems are composed of a computation controller and a fuzzy compensator. The computation control including a BNN approximator is the main control and the fuzzy compensator designed to eliminate the effect of the approximation error introduced by the BNN approximator. A parameter learning algorithm is desired to online tune the parameter of BNN approximator. Meanwhile, a proportional-integral parameter adaptation law is derived to speed up the convergence of tracking error. Finally, the proposed two intelligent control systems are implemented on a 32-bit microcontroller for possible low-cost and high-performance industrial applications. The experimental results show that two control methods can achieve high accuracy motion performance and is robust against payload variations of a VCM driver.
Appears in Collections:	[Graduate Institute & Department of Electrical Engineering] Thesis

Files in This Item:

File	Size	Format
index.html	0Kb	HTML	187	View/Open

Loading...