Minimizing order earliness and  lateness supports good customer delivery performance, but may hurt tundish  utilization and increase operating costs.   Maximizing tundish length reduces operating costs, but may result in  stock production and over-grading.  Maximizing hot charge rolling may reduce tundish utilization,  etc.
The complexity of balancing above-mentioned  goals taking into account specific technological constraints and rules  prevented for a long time companies from achievement of efficient results in  the area of melt shop and caster scheduling.   Existing methods usually targeted only one of the goals. There was no  method available for complex and efficient balancing of all the goals. The  necessity for development of such a method was obvious for specialists at the Czech  steel maker.

The key provisions of a new method

The complexity of melt shop and caster  scheduling problem defines necessity to apply a decomposition method to  it.  Using problem structure this scientific  method allows decomposing original problem into set of small interrelated and  more simple tasks.
The main idea of the new method is as  follows: a) proper decomposition of the problem into layers; b) application of  the most useful optimization techniques for each layer; and c) rapid iteration  between layers until the model converges on a solution that respects the  planner’s business goal priorities.    
The proposed method solves the problem using  three layers of optimization (see Figure 3).