Development of the Metal Temperature Prediction Model for Steel-pouring and Tundish Ladles Used at the Casting and Rolling Complex

Abstract

A model has been developed for predicting the metal temperature in the specified superheat range above the liquidus temperature inside the tundish of a continuous casting machine (CCM) depending on the last temperature measurement in the steel-pouring ladle (prior to casting), as well as ladle history and secondary steelmaking process under the conditions of the Casting and Rolling Complex (CRC) of the JSC “Vyksa Metallurgical Plant” (VMZ). It is suggested to predict the superheat temperature of steel at the CRC by using statistical models considering the specifics of technological process. During the first stage, the model itself has been developed by integrating the following two approaches: a machinelearning algorithm and probabilistic graphical model (PGM and Bayesian networks). The first approach provides the point estimate of the intermediate and final temperatures of the specific melting, while the second approach allows evaluating the probability distribution. The PGM-approach is very promising in the situations involving missing, incorrect, or undefined input data, which are quite common at the metallurgical plants. During the first stage, the achieved level of accuracy of tundish temperature prediction was 6.0°C. During the second stage, the model was integrated into the shop automated process control system, which allowed increasing the prediction accuracy and ensuring a real-time control of the process parameters. The model is used as a guidance (master) tool for the process personnel operating the ladlefurnace and vacuum degasser equipment.