In Search for an Efficient Parameter Tuning Method for Steel Casting

We deal with the optimization of process parameters in industrial continuous casting of steel. The process requires fine-tuning of numerous parameters with respect to the metallurgical cooling criteria to achieve the highest possible quality of the cast steel. We tackle the problem with various optimization methods: local optimization, conjugate gradient, downhill simplex and several types of evolutionary algorithms. They search the parameter space and evaluate candidate settings using a numerical simulator of the process and a cost function defined over the metallurgical criteria. We analyze the performance of the methods with respect to effectiveness and efficiency, and compare the optimized parameter settings with the manual setting used previously at the steel plant. The best results are achieved by local optimization and conjugate gradient what suggests that the applied cost function is not complex.