ISSN: 1304-7191 | E-ISSN: 1304-7205
Evaluating the performance of Iterated Greedy methods for the distributed permutation flowshop with sequence-independent setup times
1Department of Production Engineering, University of São Paulo, São Paulo, 05508-010, Brazil
2Department of Management, Federal Institute of São Paulo, São Paulo, 01109-010, Brazil
Sigma J Eng Nat Sci 2026; 44(4): 2626-2640 DOI: 10.14744/sigma.2026.2119
Full Text PDF

Abstract

Given the strategic needs of meeting demand and minimizing production costs to remain competitive, companies have invested in more than one plant in different geographical spaces. Therefore, in addition to scheduling, organizations also deal with the designation of which factory will produce certain products. In more recent studies, the integration of distribution and scheduling has mostly driven the advancement of strategies to improve the overall performance of factory operations. The research addresses the distributed flowshop scheduling problem under sequence-independent setup times, a constraint that has not been widely studied in the literature. Sixteen Iterated Greedy methods were developed, combining multiple initialization, destruction, and reconstruction approaches, which makes it a comprehensive analysis of IG performance for this problem. The IG methods combine four initialization approaches, two destruction and two construction procedures. Experiments were conducted to examine the impact of two parameter settings in order to determine the optimal configuration for each method. The adopted performance objective is the total duration of production, commonly referred to as makespan. Computational experiments conducted on benchmark datasets revealed that IG7, which combines DESTR_2, NEH_3, and RECONST_1, consistently outperformed other variants, achieving an average ARPD improvement of 0.938. Results from all computational experiments suggest that the novel destruction procedure significantly influences the outcomes, with methods outperforming others across all analyzed instances.