A Parallel Tempering-Based Approach to the Problem of Minimizing Open Cells

The Open Stack Minimization Problem (MOSP) consists of determining the order of processing cut patterns that minimizes the maximum number of simultaneously open stacks in the intermediate stock. This study investigates its resolution through PT-MOSP, an adaptation of the Parallel Tempering framework to the problem. The method was evaluated in 1330 instances of the literature and compared to the heuristic reference methods PieceRank, Biased Random-Key Genetic Algorithm and Variable Neighborhood Descent. The results indicated consistent performance of PT-MOSP in the classic set and statistically significant superiority in large instances, with more expressive gains in sparse and medium-density instances. These results indicate that PT-MOSP is a robust heuristic alternative for MOSP resolution.