類神經模糊推理網路用於溫度控制

Abstract

在非傳統適應性控制(adaptive control)的領域中，多層反向傳遞類神經 網路(BPNN)已證實具有高度潛在發展性，然而冗長的訓練時間卻阻礙了它 們在工業界中廣泛的應用。此外，當它們為了使自己能適應於受控體( plant)之變化而採取線上(on-line)訓練時，一過度調整(over-tuned)的 現象也經常發生。為了克服BPNN存在的缺點，在本篇論文中，我們提出一 類神經模糊推理網路(NFIN)以便適用於一般實際系統之適應性控制，尤其 是水槽系統之適應性溫度控制。NFIN是一具有類神經網路學習能力之改良 式TSK模型。若比較一般適應性類神經網路和NFIN的差異性，適應性類神 經網路的法則(rules)於參數學習之前需事先加以決定。然而對於NFIN， 起初沒有任何法則存在，當線上學習進行時，經由同時結構和參數的鑑別 才建構及調整NFIN的法則。在本篇論文中，NFIN被應用在一實際水槽溫度 控制系統上。當它和BPNN在相同的訓練程序之下比較時，模擬的結果證實 它不僅大大的降低訓練的時間及過度調整的現象，而且也擁有完美的調節 能力。此外，它的性能也與傳統的PID控制器以及模糊邏輯控制器(FLC)相 比，三個控制方法透過諸如定點調節、斜坡點追蹤、未知脈衝雜訊及水槽 溫度控制系統中大的參數變化的影響等實驗研究來互相比較。由實驗的結 果可看出，在這三個控制方法中，本篇論文所提出的NFIN控制方法也擁有 最好的控制性能。 Although multilayered backpropagation neural networks (BPNN) have demonstratedhigh potential in the nonconventional branch of adaptive control, its longtraining time usually discourages their applications in industry. Moreover,when they are trained on-line to adapt to plant variations, the over-tunedphenomenon usually occurs. To overcome the weakness of the BPNN, we proposea neural fuzzy inference network (NFIN) in this thesis suitable for adaptivecontrol of practical plant systems in general, and for adaptive t Moreover,control of a water bath system in particular. The NFIN is inherently amodified TSK-type fuzzy rule-based model possessing neural network's learningability. In contrast to the general adaptive neural fuzzy networks, whererules should be decided in advance before parameter learning is performed,there are no rules initially in the NFIN. The rules in the NFIN are createdand adapted as on-line learning proceeds via simultaneous structure andparameter identification. The NFIN has been applied to a practical water bathtemperature control system. As compared to the BPNN under the same trainingprocedure, the simulated results show that not only can the NFIN greatlyreduce the training time and avoid the over-tuned phenomenon, but the NFINalso has perfect regulation ability. The performance of the NFIN is alsocompared to that of the traditional PID controller and fuzzy logic controller(FLC) on the water bath temperature control system. The three control schemesare compared through experimental studies with respect to set-pointsregulation, ramp-points tracking, the influence of unknown impulse noiseand large parameter variationin in the temperature control system. It's foundthat the proposed NFIN control scheme has a best control performance amongthe three control schemes.