Optimization of Heavy Plate Production Sequencing in a Heavy Plate Rolling Mill

This work compares three optimization methodologies for heavy plate production sequencing. The problem involves selecting and ordering slabs across six stages, respecting constraints on capacity, dimensions, steel family, and unimodal temperature profiles. The goal is to maximize total tonnage allocated. The first approach is a Stage-Wise Model that combines Binary Integer Linear Programming (BILP) subproblems solved per stage with a global constructive heuristic. The second is a Monolithic Model that formulates the entire coffin as a single BILP optimizing all stages simultaneously. These are contrasted with a Multi-Stage Heuristic, a non-BILP alternative that assembles the coffin through a clustering-driven greedy procedure. Experiments on 20 real-world instances show the Stage-Wise Model consistently achieves the highest solution quality, whereas the Monolithic Model proves computationally intractable at scale. The Multi-Stage Heuristic emerges as a competitive alternative, offering rapid solutions with only modest quality loss compared to the Stage-Wise Model.