Real-time prediction of geometrical distortion of hot-rolled steel rings during cooling

The paper deals with the application of neural network modelling to the real-time prediction of the geometrical distortion of hot rolled steel rings during cooling from rolling to room temperature. The neural network model was designed and developed to be part of a new modular system for the in-line monitoring and real-time control of the geometrical quality of rings, even those with a complex profile, during hot and warm ring rolling operations. The data utilised to train the neural network were generated by numerical simulations of the cooling phase. In order to do these simulations, an FIE model capable of coupling thermal, mechanical and metalllurgical events was accurately calibrated. The proposed model was then applied to an industrial case that is described in the paper.