Modelling of lateral flow in a Hot Strip Mill (HSM) using adaptive techniques

During the last years, data mining models have proven to be a promising approach to improve hot rolling processes. In the present research we propose a model for prediction of lateral flow. In hot rolling mills this will lead to exact predictions of the strip width after rolling, which reduces cut-offs and scrapped material. Any reduction of the cut-offs implies important economical and environmental benefits. Physically based models were developed some years ago, but they require simplifications, need data that is difficult to achieve online or include experimental parameters that have to be optimized. Adaptive techniques can contribute widely to the improvement of the diagnostics. For this work, production data was gathered from a Hot Strip Mill (HSM) and a nonlinear model was built using a data-mining methodology based on multivariate adaptive regression splines (MARS). The agreement of the MARS model with observed data confirmed its good performance.