A model for decision support in continuous steel casting

Abstract

Many of the difficulties that may be encountered during the continuous casting of steel are caused by an inappropriate liquid steel temperature in the steelmaking tundish. This paper presents a temperature prediction model which is used in a prototype advisory system for the casting process. Instead of modelling the phenomena involved in the conventional, deterministic manner, the method converts empirical expertise on the casting process into a model of qualitative nature. The advisory mechanism is linked with the prediction of the steel temperature evolution in the tundish. The simulation model is based on information of a qualitative nature about the castability of the steel. In the expertise used for developing the model, the possible outcome of the casting has been classified into one of four classes on the basis of the temperature evolution during the casting process. The temperature, in turn, is assumed to depend on a few central factors. Based on expert reasoning, the specific impact of each factor is considered and quantitative estimates of the cooling effects are obtained by formulating and solving an optimization problem. The model output - the steel temperature in the tundish at a few important time instants during the casting - has been found to be in relatively good agreement with measurements. At the end of the paper, some features of the advisory system are briefly described.