Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/7723
標題: 遞迴類神經模糊網路於橡膠射出成型機模具加溫之應用
Application Of A Recurrent Neural Fuzzy Network To Mold Temperature Control Of A Rubber Shoot Shaping Machine
作者: 黃水田
Huang, shui-tien
關鍵字: TRNFN;類神經模糊網路;direct inverse control;mold;parameter learning;structure learning;直接反模型;溫度控制;橡膠射出成型機;架構學習;參數學習
出版社: 電機工程學系
摘要: 
摘要
本論文將提出一個具有學習能力且以直接反模型來作溫度控制的TSK型遞迴類神經模糊網路( TRNFN )控制器 ,並將它實現於一真實的橡膠射出成型機模具溫度控制。TRNFN一開始並無任何規則存在,網路架構主要透過架構學習( structure learning )和參數學習( parameter learning )來達成。TRNFN控制器本身具有下列優點:
(1) 不要求受控體數學模型及階數。
(2) 強有力的學習能力特性,能準確塑造出受控體的反模型。
(3)具有線上學習能力,藉由不斷自我調整適應控制環境內不可預測的變化。
在實驗過程中,我們將表現出即使改變取樣週期而沒有在TRNFN新加入任何變數的情況下也一樣能得到不錯的性能表現。此外,TRNFN的控制性能也將與傳統的PI控制器以及模糊邏輯控制器( FLC ) 透過實驗研究相互比較來驗證。

Abstract
Based on the direct inverse control approach, temperature control by TSK-type recurrent neural fuzzy network (TRNFN) controller is proposed in this thesis.
The mold temperature control of a shoot shaping machine system is experimented by using TRNFN controller based direct inverse control. There are no rules initially in TRNFN; they are constructed by concurrent structure and parameter learning. The TRNFN has the following advantages:
(1) mathematical model and system order are not required. (2) powerful learning ability, which can model the inverse of the plant accurately. (3) on-line learning ability, i.e. the TRNFN can deal with unpredictable changes in the control environment.
In the experiment, we will show that without reassigning the input variables in TRNFN, a good control performance is achieved even when the sampling period changes. The same experiment is also performed by PI and fuzzy controllers. From comparisons, the aforementioned advantages of TRNFN have been verified.
URI: http://hdl.handle.net/11455/7723
Appears in Collections:電機工程學系所

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.