Detail publikace

Sliding Window Recursive Neural Networks Learning Algorithm and its Application on the Identification in Adaptive PID

Originální název

Sliding Window Recursive Neural Networks Learning Algorithm and its Application on the Identification in Adaptive PID

Anglický název

Sliding Window Recursive Neural Networks Learning Algorithm and its Application on the Identification in Adaptive PID

Jazyk

en

Originální abstrakt

This article deals with the implementation of the adaptive PID controller based on the principle of forced separation imposed on the identification and system control. Original implementation of both Gauss-Newton (GN) and Levenberg-Marquardt (LM) algorithms operating in recursive learning mode over exponential-sliding finite data window for modelling of nonlinear dynamic systems is suggested. Their dynamics can be represented by a feed forward neural network. Synthesis of the PID controller is achieved using the Ziegler-Nichols method which utilizes the linearized ARX model of the neural network at the working point of the process. Benefits of the suggested algorithms are illustrated in the example simulations on the mathematical model.

Anglický abstrakt

This article deals with the implementation of the adaptive PID controller based on the principle of forced separation imposed on the identification and system control. Original implementation of both Gauss-Newton (GN) and Levenberg-Marquardt (LM) algorithms operating in recursive learning mode over exponential-sliding finite data window for modelling of nonlinear dynamic systems is suggested. Their dynamics can be represented by a feed forward neural network. Synthesis of the PID controller is achieved using the Ziegler-Nichols method which utilizes the linearized ARX model of the neural network at the working point of the process. Benefits of the suggested algorithms are illustrated in the example simulations on the mathematical model.

BibTex


@inproceedings{BUT34626,
  author="Petr {Pivoňka} and Jakub {Dokoupil}",
  title="Sliding Window Recursive Neural Networks Learning Algorithm and its Application on the Identification in Adaptive PID",
  annote="This article deals with the implementation of the adaptive PID controller based on the principle of forced separation imposed on the identification and system control. Original implementation of both Gauss-Newton (GN) and Levenberg-Marquardt (LM) algorithms operating in recursive learning mode over exponential-sliding finite data window for modelling of nonlinear dynamic systems is suggested. Their dynamics can be represented by a feed forward neural network. Synthesis of the PID controller is achieved using the Ziegler-Nichols method which utilizes the linearized ARX model of the neural network at the working point of the process. Benefits of the suggested algorithms are illustrated in the example simulations on the mathematical model.",
  booktitle="17th Zittau East-West Fuzzy Colloquium",
  chapter="34626",
  howpublished="print",
  year="2010",
  month="september",
  pages="55--62",
  type="conference paper"
}