The main purpose of this paper is to propose a new two-stage model for cupola furnance melt control. Approaches validated by the historical experiences are considered and their disadvantages are listed. Results obtained for different models are discussed.