Detail publikace

Increasing the Adaptivity of a Self-Tuning PID Controller by Using Neural Network Based Identification

VELEBA, V. PIVOŇKA, P.

Originální název

Increasing the Adaptivity of a Self-Tuning PID Controller by Using Neural Network Based Identification

Anglický název

Increasing the Adaptivity of a Self-Tuning PID Controller by Using Neural Network Based Identification

Jazyk

en

Originální abstrakt

The new approach to analysis of on-line identification methods based on one-step-ahead prediction clears up their sensitivity to disturbances in control loop. It is obvious from the analysis that there is an upper bound of relative time constant caused by an existence of quantization interval in A/D conversion. The ability of three identification methods to the parameter estimation of the dynamic plant with great ratio of its time constant to sampling periods is compared. We concentrate our attention on dealing with adverse effects that work on real-time identification of process, especially quantization. It is shown, that a neural network applied to on-line identification process produces more stable solution in the rapid sampling domain. Taking advantage of this result, we propose here an adaptive controller with a neural network as on-line estimator. Simple heuristic synthesis based on modified Ziegler-Nichols open loop method (Z-N 1) are discussed, that deals with bad-estimated model of a plant and gives numerically stable parameters of the PID discrete controller.

Anglický abstrakt

The new approach to analysis of on-line identification methods based on one-step-ahead prediction clears up their sensitivity to disturbances in control loop. It is obvious from the analysis that there is an upper bound of relative time constant caused by an existence of quantization interval in A/D conversion. The ability of three identification methods to the parameter estimation of the dynamic plant with great ratio of its time constant to sampling periods is compared. We concentrate our attention on dealing with adverse effects that work on real-time identification of process, especially quantization. It is shown, that a neural network applied to on-line identification process produces more stable solution in the rapid sampling domain. Taking advantage of this result, we propose here an adaptive controller with a neural network as on-line estimator. Simple heuristic synthesis based on modified Ziegler-Nichols open loop method (Z-N 1) are discussed, that deals with bad-estimated model of a plant and gives numerically stable parameters of the PID discrete controller.

Dokumenty

BibTex


@inproceedings{BUT15250,
  author="Václav {Veleba} and Petr {Pivoňka}",
  title="Increasing the Adaptivity of a Self-Tuning PID Controller by Using Neural Network Based Identification",
  annote="The new approach to analysis of on-line identification methods based on one-step-ahead prediction clears up their sensitivity to disturbances in control loop. It is obvious from the analysis that there is an upper bound of relative time constant caused by an existence of quantization interval in A/D conversion. The ability of three identification methods to the parameter estimation of the dynamic plant with great ratio of its time constant to sampling periods is compared. We concentrate our attention on dealing with adverse effects that work on real-time identification of process, especially quantization. It is shown, that a neural network applied to on-line identification process produces more stable solution in the rapid sampling domain. Taking advantage of this result, we propose here an adaptive controller with a neural network as on-line estimator. Simple heuristic synthesis based on modified Ziegler-Nichols open loop method (Z-N 1) are discussed, that deals with bad-estimated model of a plant and gives numerically stable parameters of the PID discrete controller.",
  address="R. Hampel",
  booktitle="12th Zittau East-West Fuzzy Colloquium",
  chapter="15250",
  institution="R. Hampel",
  year="2005",
  month="september",
  pages="169",
  publisher="R. Hampel",
  type="conference paper"
}