Input Selection Technology of Neural Network and its Application for Hot Strip Mill

In this paper, the control system is proposed to obtain thedesirable width margin of a strip in a rolling process. Theneural network model is also suggested to improve the predictionperformance of the width spread. The selection method of inputparameter for the network using the hypothesis testing is proposedin this paper. The developed network model is based on themeasured data such as the entry, delivery width margin offinishing mill and process setup data such as unit tension betweenstands, roll force, temperature, etc. Moreover, an edger controlscheme is proposed to guarantee the desired strip width offinishing mill. It is shown through the field test of Pohang No.1hot strip mill of POSCO that the width margin is greatly improvedby the network model and the control scheme proposed in thispaper.