Um algoritmo de busca híbrido para o problema de roteamento de embarcações de suprimento periódico

• Ano de defesa: 2018
• Autor(a) principal: Boucher, Juliana Beatriz Carvalho de Oliveira Soares
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia de Produção; UFRJ
• Palavras-chave em Português: Logística offshore; roteamento de veículos; CNPQ::ENGENHARIAS::ENGENHARIA DE PRODUCAO
• Link de acesso: http://hdl.handle.net/11422/12255

High quality maritime petroleum transportation is critical to ensure timely flow of goods, reducing the total logistics cost and guaranteeing an efficient process. In this thesis, we focus on the transportation of deck cargo to offshore units as observed in the operations of our industrial partner in Rio de Janeiro, Brazil. The main objective of this research is to define the maritime routes to solve a periodic supply vessels routing problem, taking into account the port departure and opening hours at the offshore facilities. We describe the solution procedures currently used by the company, and we formally formulate the problem mathematically. Given that the sizes of the instances are too large to be solved exactly, we propose different methods to achieve better solutions with a reduced computational time. The first method, composed of three phases, uses a clustering heuristic combined with an exact method in order to perform the routing and ends with a resenquencing according to the operating hours, port departures and time between services constraits. The second method uses the adaptive large neighborhood search (ALNS) heuristic in an attempt to reduce the number of operations performed by the previous heuristic and the computational time. Finally, a hybrid method is proposed based on the ALNS heuristic and innovates with the concepts of the clustering search algorithm that proposes a detection of promising regions of the search space. The computational results indicate that the hybrid heuristic brings benefits to the ALNS by finding better solutions in less time and still reduces the coefficient of variation within a sample of solutions in different executions.