Development of an artificial neural network (ANN) model to predict the temperature of hot-rolled steel pipes

One important objective in steel pipe manufacturing is to avoid rejects. In order to adequately heat each individual pipe in the furnace, the surface temperature of all pipes after rolling must be predicted accurately. A fast model is needed that can provide this prediction quickly and repeatedly. To achieve this goal, artificial neural networks (ANN) were applied to the hot-rolling process used to create seamless steel pipes for the first time, and results are presented in this paper. Modelling the process is a complicated task, because a wide range of different geometries are manufactured, and the pipes can possibly be cooled after rolling. To address this issue, two ANN models were designed, with one model consisting of two coupled ANNs to increase its accuracy. This also represents a novel modelling approach. Both models were trained with data recorded during the production process. In general, the modelling results agree well with data collected by the in-plant measurement system for a wide range of different finished pipe geometries. The two models are compared, and differences in their behavior are discussed.