A design of multiloop PID controllers with neural-net based decoupler

Tokuda, Makoto; Yamamoto, Toru; Monden, Yoshimi

doi:10.1109/isic.2002.1157805

Cited by 6 publications

(5 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, artificial neural networks (ANNs) are popularly used in control engineering to improve the control ability. The effectiveness of ANNs is verified by some studies and applications [2][3][4][5]. ANNs achieve their control ability through their learning ability.…”

Section: Introductionmentioning

confidence: 84%

See 1 more Smart Citation

Design and Implementation of a Hierarchical‐Clustering CMAC PID Controller

Liao

Koiwai

Yamamoto

2018

Asian Journal of Control

View full text Add to dashboard Cite

In industrial control processes, proportional-integral-derivative (PID) control algorithm is widely employed. Therefore, it is meaningful to design advanced PID controllers, especially for nonlinear control objects. One of the advanced PID controllers is a cerebellar model articulation controller (CMAC) PID controller. In this controller, the PID control parameters are calculated and tuned. The CMAC achieves a higher accuracy by increasing the number of labels of each weight table; this requires a larger memory, and the generalization ability of the controller decreases. On the other hand, if the CMAC requires less memory, the generalization ability increases and accuracy decreases. Hence, in this paper, a novel CMAC in which the accuracy is compatible with the generalization ability is proposed in this paper. In the proposed CMAC, the number of labels of each weight table can be decided by using a hierarchical clustering technology. Moreover, the efficiency of the memory allocation is improved. The effectiveness of the proposed method is verified by experiments.

show abstract

Section: Introductionmentioning

confidence: 84%

“…Moreover, some input signals in the input space are mapped to the same weights as input signal S (6,3). With this ability, the CMAC can achieve the desired values in a shorter learning time compared to the conventional ANNs.…”

Section: Conventional Cmacmentioning

confidence: 99%

Design and Implementation of a Hierarchical‐Clustering CMAC PID Controller

Liao

Koiwai

Yamamoto

2018

Asian Journal of Control

View full text Add to dashboard Cite

show abstract

“…The inputs of modeling NN are chosen as the increment of control variables, and the outputs are designed to approximate the actual output of WWTPs. FNN has great ability as an approximator, which would ensure the modeling accuracy; meanwhile, the learning character makes FNN great decoupling ability . Thus, the output of modeling NN can be defined as:

{boldy}_{m} (k) = {boldW}_{2}^{m} (k) f ({boldW}_{1}^{m} (k) u (k))

…”

Section: The Neural Network On‐line Modeling and Controlling Methods Dmentioning

confidence: 99%

“…Sayed et al [12] proposed a multilayer FNN with two hidden layers, and the study results showed that this type of network has high accuracy in modeling when the number of hidden neurons is suitable. Moreover, the FNN is useful for multi-variable control decoupling [13,14].…”

Section: Introductionmentioning

confidence: 99%

Neural Network On‐Line Modeling and Controlling Method for Multi‐Variable Control of Wastewater Treatment Processes

Qiao¹,

Han²,

Han³

2013

Asian Journal of Control

View full text Add to dashboard Cite

A neural network on‐line modeling and controlling method (NNOMC) is proposed in this paper for multi‐variable control of wastewater treatment processes (WWTPs). According to the approximating character of the feedforward neural network (FNN), a modeling FNN is proposed to simulate and decouple WWTPs. Then, an FNN controller for multi‐variable control of WWTPs is designed. Moreover, the stability of the NNOMC method is proven in a general inference via the limitation of the learning rate of the FNN. Finally, this proposed NNOMC method is used in the international benchmark of WWTPs. The results show the NNOMC method owns both better approximating and controlling performance.

show abstract

“…Since most processes are considered as nonlinear MV systems with mutual interactions, the good control performance cannot be obtained by simple controllers. Makoto, Toru and Yoshimi have used a designed decoupler which has been generated by the sum of a static decoupler and a neural-net based decoupler [8]. Then, the former has been used so as to approximately decouple the controlled object.…”

Section: Introductionmentioning

confidence: 99%