A HYBRID NEURAL-NETWORK MATHEMATICAL PREDICTION MODEL FOR TANDEM COLDMILL

In tandem cold mill, a strip is flattened by stands of rolls to a desi red thickness. At Pohang Iron and Steel Company (POSCO) in Pohang, Kor ea, precalculation determines the mill settings before a strip actuall y enters the mill and is done by an outdated mathematical model. A cor rective neural network model is proposed to improve the accuracy of th e roll force prediction. The network is fed not only the usual mathema tical model's input but also a set of additional inputs such as the ch emical composition of the coil, its coiling temperature and the aggreg ated amount of processed strips of each roll. The network was trained using a standard backpropagation with 4,944 process data collected at POSCO from March 1995 through December 1995, then was tested on the un seen 1,586 data from February 1996 through April 1996. The combined mo del reduced the prediction error by 33.88% on average. (C) 1997 Elsevi er Science Ltd.