Metall. Res. Technol.
Volume 112, Number 2, 2015
|Number of page(s)||11|
|Published online||12 March 2015|
Self-learning factor prediction of the heat transfer coefficient based on a dynamic fuzzy neural network for ultra-fast cooling
1 Electronic Information and Engineering College, Taiyuan University of Science and Technology, Shanxi, Taiyuan, P.R. China
2 State Key Laboratory of Rolling and Automation, Northeastern University, Shenyang, Liaoning, P.R. China
Received: 8 June 2014
Accepted: 3 February 2015
The aim of this paper is to improve the plate mechanical properties by enhancing the control accuracy of the finish cooling temperature (FCT) during the ultra-fast cooling (UFC) process. In the mathematical model of ultra-fast cooling, the self-learning factor of the heat transfer coefficient is one of the important technological parameters for the finish cooling temperature. Some parameters can influence the self-learning factor, such as plate thickness, water flow and water temperature. In order to predict the self-learning factor of the heat transfer coefficient through these parameters, a dynamic fuzzy neural network (D-FNN) is introduced and combined with a traditional mathematical model. After training and prediction, it is shown that the D-FNN model has high prediction accuracy and can achieve predictive control of the mathematical model. Testing the BP neural network with the same data, the prediction accuracy of the D-FNN is higher than the BP neural network. In industrial production, FCT errors demonstrate that satisfactory performance can be achieved by the D-FNN.
Key words: Ultra-fast cooling / heat transfer coefficient / self-learning factor / dynamic fuzzy neural network
© EDP Sciences 2015
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.