Two Hybrid Metaheuristic Algorithms for Hot Rolling Scheduling

Abstract

This paper presents a model for the hot rolling scheduling problem (HRSP), which is derived from the actual steel production. The model is characterized by some new features, such as the rolling length of the consecutive slabs with the same width, temperature jump between adjacent slabs. The warm up part is a minor part of a turn and is usually ignored, but it affects directly the product quality. Therefore, besides the slab sequence in the staple material section, we also consider the slab sequence in the warm up material section. These features make the solution methodology more difficult. Therefore, two hybrid strategies are proposed to determine good approximate solutions for this complicated problem. The first one (CT_ACO) is a hybrid strategy based on the solution construction mechanism of ant colony optimization (ACO) with cyclic transfers (CT). The second one (CT_SS) is to hybridize scatter search (SS) and CT neighborhood search. Moreover, we design a decision support system in which two algorithms have been embedded for the HRSP. The most popular feature of the system is the architecture of component management, which allows us to modify easily some components according to the practice situation. The computational experiments show that CT_ACO is superior to general ACO, and CT_SS is also better than SS in terms of solution quality. The CT_ACO method and CT_SS have more potential for improvement to solve the HRSP compared with the current scheduling method, and the CT_ACO generates slightly better quality solutions than CT_SS algorithm.