Dynamic Scheduling for Maintenance Tasks Allocation supported by Genetic Algorithms

• Autor(a) principal: Alemão, Duarte José Marques
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/30816
• topic: Job-Shop Scheduling; Task Allocation; Genetic Algorithms; Manufacturing Systems; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

Since the first factories were created, man has always tried to maximize its production and, consequently, his profits. However, the market demands have changed and nowadays is not so easy to get the maximum yield of it. The production lines are becoming more flexible and dynamic and the amount of information going through the factory is growing more and more. This leads to a scenario where errors in the production scheduling may occur often. Several approaches have been used over the time to plan and schedule the shop-floor’s production. However, some of them do not consider some factors present in real environments, such as the fact that the machines are not available all the time and need maintenance sometimes. This increases the complexity of the system and makes it harder to allocate the tasks competently. So, more dynamic approaches should be used to explore the large search spaces more efficiently. In this work is proposed an architecture and respective implementation to get a schedule including both production and maintenance tasks, which are often ignored on the related works. It considers the maintenance shifts available. The proposed architecture was implemented using genetic algorithms, which already proved to be good solving combinatorial problems such as the Job-Shop Scheduling problem. The architecture considers the precedence order between the tasks of a same product and the maintenance shifts available on the factory. The architecture was tested on a simulated environment to check the algorithm behavior. However, it was used a real data set of production tasks and working stations.