An improved Pareto local search for solving bi-objective scheduling problems in hot rolling mills

Abstract

Hot rolling scheduling is a complex decision-making optimization problem in the production management of the iron and steel industry. Comprehensively considering temporal and technical constraints of hot rolling production, we formulate a prize-collecting capacitated vehicle routing problem with time windows (PC-CVRPTW) involving special constraints and multiple objectives. Furthermore, this paper develops a Pareto local search (PLS)-based solving algorithm via leveraging problem-specific properties and implements three improvement mechanisms: (1) solution initialization with a greedy strategy; (2) promising solution selection with optimistic hypervolume contribution; (3) hybrid neighborhood exploration integrated variable neighborhood search with constraint programming. To validate the proposed method, computational experiments are conducted on a set of randomly synthetic and realistic industrial instances. The results compared with other variants of PLS and state-of-art algorithms demonstrate that the improved PLS algorithm is effective and can provide promising solutions for bi-objective hot rolling scheduling problems.