Making steel cool – an intelligent search strategy for optimising water spray settings on a continuous caster

Optimising the water spray settings on a continuous casting machine in the steel industry can be formulated as a non-linear programming problem. For specified water spray settings, a computer-based thermal model computes the temperature and solidification conditions throughout the entire continuous casting machine. Both gradient-based and pattern search algorithms require that the thermal model be executed often to obtain optimal water spray settings. In this paper, we discuss an intelligent search stategy that makes extensive use of domain knowledge in order to obtain a near-optimal result in less thermal model executions than required to calculate one gradient.