Relax-and-Fix Strategies for Real-World Steel Production Scheduling Problems

The production scheduling problem in the steel industry involves complex operational constraints and conflicting objectives related to production efficiency and delivery performance. This work proposes a deterministic Mixed-Integer Programming (MIP) formulation for a multi-stage steel production scheduling problem, considering precedence relationships between stages as well as costs associated with tardiness and earliness. Due to the computational complexity of solving large industrial instances directly, three Relax-and-Fix strategies were investigated to improve the computational tractability of the MIP formulation, including an Instance Partitioning strategy designed to deal with the large number of precedence constraints. Computational experiments were conducted using industrial instances and compared with solutions generated by a commercial scheduling software. The results indicate that the proposed approaches are capable of obtaining better objective function values while maintaining computational times compatible with industrial decision-making requirements.