Bounded Dynamic Programming Approach to minimize Makespan for the Blocking Flow Shop problem with Sequence Dependent Setup Times Constraints

• Ano de defesa: 2021
• Autor(a) principal: Souza, Edson Antonio Gonçalves de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Bounded Dynamic Programming; flow shop com bloqueio; makespan; B&B; blocking flow shop; MILP; setup times; tempos de setup
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/18/18156/tde-19072021-182420/

Throughout the decades, the flow shop scheduling problem has appeared in a variety of manufacturing scenarios and, as a consequence of that fact, several branches of it have taken place and are frequently studied by researchers so as to develop efficient methods to solve these problems that stem from the classic version. A problem that is frequently found on the shop floor is the blocking flow shop with sequence dependent setup times (BFSP-SDST) with makespan optimization, which can be found on electronics, metallurgy, chemical and food industries and it comes to show the importance of improving systems that can be modelled via this problem. However, when gathering information on stateof- art references, one will notice that little has been made regarding BFSP-SDST and, consequently, only few methods have been applied with the intent of optimizing makespan. Furthermore, research has shown that dynamic programming is a method that has not been given much attention with regard to the flow shop environment and therefore, its contribution has become, thus far, limited. Hence, this research aims at applying Bounded Dynamic Programming (BDP), which is a DP-based method, to solve the BFSP-SDST so as to enlarge the number of methods that are related to such problem. The BDP is employed in two sets of problems and for the first set, it is compared to a MILP and B&B methods, while the comparison occurs only between the BDP and MILP for the second set. The results have shown that BDP outperforms the MILP in both scenarios and the B&B in terms of CPU times and success rate. In addition, a trade-off analysis is provided in order to determine which method is a better choice for the scheduler in terms of solution quality and CPU times and, once again, BDP proves to be better for the job and machine sets considered. Sequentially, some suggestions are furnished to improve the method and develop further research considering this approach.