In this study, a modelling approach is introduced that fuses the mechanistic model with an artificial neural network (ANN) during the cord steel manufacturing. This involves replacing certain mechanistic calculations with ANN representations, optimizing mechanistic parameters through ANN techniques, and implanting mechanistic equations into the ANN framework to achieve more accurate and efficient calculations of the cooling process. The mechanistic parameters come from the models including Natural Convection Heat Transfer, Forced Convection Heat Transfer, Radiation Heat Transfer and Phase Transition during the air-cooling process. The model demonstrated an average error of 8.6 °C and 5.4 °C, compared to measurement values from two temperature sensors on the air-cooling line, and an average error of 3.3 °C relative to manually obtained temperature.
Read full abstract