Friction coefficient of hot tandem finishing mill predicted by BP neural network

• Published in: 2010 3rd International Conference on Advanced Computer Theory and Engineering(ICACTE) Vol. 3; pp. V3-499 - V3-502
• Main Authors: Qiu Chunlin, Gao Xiuhua, Qi Kemin, Wen Jinglin
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.08.2010
• Subjects: artificial neural network; BP algorithm; friction coefficient; rolling pressure
• ISBN: 1424465397; 9781424465392
• ISSN: 2154-7491; 2154-7505
• DOI: 10.1109/ICACTE.2010.5579872

The friction coefficients of hot tandem finishing mill were used to be set as a fixed value, which lead to 2000~3000 kN rolling force deviation in the prediction. During the hot rolling, the friction coefficient is affected by lots of factors and the variation principle is relatively intricate. In this article, the BP neural network consisted of an input layer, an output layer and one or several hidden layers and there were several nodes in each layer. 7 input variables were investigated in the input layer. The output was friction coefficient. the friction coefficient was calculated according to Sims formula. 3090 sets of data from production were surveyed in this paper; the same amount of friction coefficient can also be calculated. 900 sets were used for training the network, after training, the network can obtain high accuracy in prediction. 40% of data were used for testing and the rest for verification. The results displayed the minimum error was only 0.00000193, the correlation coefficient reached 0.9977 and all data were located in 5% deviation. The achievements proved the BP neural network was an effective and reliable method to predict the friction coefficient during hot rolling. Furthermore, the BP neural network can enhance the precision of rolling force prediction considerably. With the BP neural network algorithm, the friction coefficients affected by multiple variables during hot rolling can be predicted correctly, which provides important information for improving the accuracy of rolling pressure prediction.