Detail publikace

Closed Loop On-Line Identification Based on Neural Networks in Adaptive Optimal Controller

Originální název

Closed Loop On-Line Identification Based on Neural Networks in Adaptive Optimal Controller

Anglický název

Closed Loop On-Line Identification Based on Neural Networks in Adaptive Optimal Controller

Jazyk

en

Originální abstrakt

Last ten years, algorithms based on Neural Networks were used successfully for the pattern recognition, process control and system identification. Artificial Neural Networks applied this way have been strongly developing together with the classical control. It is mainly because of their self-learning property and wide-range of easy algorithm designs. Using Neural Networks for identification is well-known strategy where the process is observed usually through its input and output only. The real process is often influenced by disturbances. In this case, the more identified transfer function is inaccurate the more as disturbance influences IO of the measured process. This paper shows a comparison between on-line identification (in the real time) based on Neural Networks and a classical identification implemented in adaptive optimal controller. The setting of the sampling period for the both identification methods is investigated.

Anglický abstrakt

Last ten years, algorithms based on Neural Networks were used successfully for the pattern recognition, process control and system identification. Artificial Neural Networks applied this way have been strongly developing together with the classical control. It is mainly because of their self-learning property and wide-range of easy algorithm designs. Using Neural Networks for identification is well-known strategy where the process is observed usually through its input and output only. The real process is often influenced by disturbances. In this case, the more identified transfer function is inaccurate the more as disturbance influences IO of the measured process. This paper shows a comparison between on-line identification (in the real time) based on Neural Networks and a classical identification implemented in adaptive optimal controller. The setting of the sampling period for the both identification methods is investigated.

BibTex


@inproceedings{BUT4918,
  author="Kamil {Švancara} and Petr {Pivoňka}",
  title="Closed Loop On-Line Identification Based on Neural Networks in Adaptive Optimal Controller",
  annote="Last ten years, algorithms based on Neural Networks were used successfully for the pattern recognition, process control and system identification. Artificial Neural Networks applied this way have been strongly developing together with the classical control. It is mainly because of their self-learning property and wide-range of easy algorithm designs. Using Neural Networks for identification is well-known strategy where the process is observed usually through its input and output only. The real process is often influenced by disturbances. In this case, the more identified transfer function is inaccurate the more as disturbance influences IO of the measured process. This paper shows a comparison between on-line identification (in the real time) based on Neural Networks and a classical identification implemented in adaptive optimal controller. The setting of the sampling period for the both identification methods is investigated.",
  address="Rektor der Hochschule Zittau/Görlitz",
  booktitle="Proceedings East West Fuzzy Colloquium 2002",
  chapter="4918",
  institution="Rektor der Hochschule Zittau/Görlitz",
  year="2002",
  month="september",
  pages="218",
  publisher="Rektor der Hochschule Zittau/Görlitz",
  type="conference paper"
}