ITEMS

modern scholarly publishers in the finest tradition

References

	Search

ICHMT DIGITAL LIBRARY ONLINE

ISSN 961-91393-0-5

Individual price:

$246.00 (Must be sent to your home address)

You can order single issue or individual article. To purchase a single issue or an individual article as well as to view tables of contents and abstracts click on issue number.

Institutional price:

$518.00

Volumes per year:

various

For Online Access

Year 2001

Volume 2 - CHT'01 - Advances in Computational Heat Transfer II
Proceedings of a Second Symposium - Palm Cove, Queensland, Australia, May 20 - 25, 2001

773 pages

AN INNOVATIVE APPROACH TO THE MODELING OF NATURAL CIRCULATION

G. Desrayaud
Laboratoire MSNPT, INSSET, Université de Picardie Jules Verne, 48 rue Raspail BP 422, 02100, Saint-Quentin, FRANCE

A. Fichera
Dipartimento di Ingegneria Industriale e Meccanica, Università degli Studi di Catania, Viale A. Doria 6, 95125, Catania, Italy

A. Pagano
Dipartimento di Ingegneria Industriale e Meccanica, Università degli Studi di Catania, Viale A. Doria 6, 95125, Catania, Italy

ABSTRACT

In this paper the problem of modeling the dynamics of a natural circulation loop was addressed by means of neural networks. Determining such a model represents an important issue as it is the first step towards the design of a control system. In particular, the model aims to predict the temperature oscillations that characterize the system dynamics during unstable operations, which cause dangerous flow reversal leading to the system failure. Input-output measurements detected during an experimental phase were used to train and test the neural model. Unlike traditional discrete mathematical models, the neural model ensures good correspondence between experimental and simulated data. Finally, the neural model was used in a recursive scheme in order to perform long term predictions of the system dynamics, which may allow wider control opportunities.