Loading...
Publication
Simulação de um sistema de fabrico usando modelação baseada em agente
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 2.94 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A evolução tecnológica tem impulsionado mudanças significativas nos processos produtivos, elevando-os a patamares cada vez mais automatizados, assim o presente trabalho
propõe o desenvolvimento e a avaliação de um modelo de simulação baseado em agentes para representar o funcionamento de um sistema ciber-físico de manufatura inteligente, com foco na descentralização da tomada de decisão e na auto-organização dos processos produtivos. O modelo foi implementado na plataforma NetLogo e utiliza uma lógica de alocação inspirada em aprendizado por reforço, baseada em colônias de formigas, para otimizar a escolha de máquinas de acordo com desempenho histórico e nível de ocupação. A modelagem considera agentes de produto e de recurso, com diferentes habilidades, tempos de processamento e buffers limitados, permitindo simular fluxos produtivos complexos e avaliar seu desempenho sob diferentes condições operacionais. Foram realizados testes com cinco cenários distintos, incluindo variações na taxa de chegada de produtos, tamanho dos buffers e ocorrência de falhas em máquinas. Os resultados demonstram que o sistema é capaz de se adaptar a restrições e mudanças no ambiente, promovendo alocação mais eficiente das tarefas e redução de gargalos. Conclui-se que a abordagem baseada em agentes, aliada a mecanismos simples de reforço, permite criar modelos flexíveis, escaláveis e alinhados aos princípios da Indústria 4.0. A plataforma NetLogo mostrou-se adequada para prototipagem e experimentação iterativa, permitindo futuras extensões do modelo para sistemas mais complexos.
Technological advancements have driven significant changes in production processes, making them increasingly automated. This study proposes the development and evaluation of an agent-based simulation model to represent the operation of a cyber-physical system for smart manufacturing, focusing on decentralized decision-making and self-organization of production processes. The model was implemented on the NetLogo platform and employs a reinforcement learning-inspired allocation logic, based on ant colony behavior, to optimize machine selection according to historical performance and occupancy level. The modeling includes product and resource agents with varying skills, processing times, and limited buffers, enabling the simulation of complex production flows and the evaluation of system performance under different operational conditions. Tests were conducted across five distinct scenarios, including variations in product arrival rates, buffer sizes, and machine failures. The results show that the system can adapt to constraints and environmental changes, promoting more efficient task allocation and reducing bottlenecks. It is concluded that the agent-based approach, combined with simple reinforcement mechanisms, enables the creation of flexible and scalable models aligned with the principles of Industry 4.0. The NetLogo platform proved suitable for prototyping and iterative experimentation, allowing future extensions of the model to more complex systems.
Technological advancements have driven significant changes in production processes, making them increasingly automated. This study proposes the development and evaluation of an agent-based simulation model to represent the operation of a cyber-physical system for smart manufacturing, focusing on decentralized decision-making and self-organization of production processes. The model was implemented on the NetLogo platform and employs a reinforcement learning-inspired allocation logic, based on ant colony behavior, to optimize machine selection according to historical performance and occupancy level. The modeling includes product and resource agents with varying skills, processing times, and limited buffers, enabling the simulation of complex production flows and the evaluation of system performance under different operational conditions. Tests were conducted across five distinct scenarios, including variations in product arrival rates, buffer sizes, and machine failures. The results show that the system can adapt to constraints and environmental changes, promoting more efficient task allocation and reducing bottlenecks. It is concluded that the agent-based approach, combined with simple reinforcement mechanisms, enables the creation of flexible and scalable models aligned with the principles of Industry 4.0. The NetLogo platform proved suitable for prototyping and iterative experimentation, allowing future extensions of the model to more complex systems.
Description
Mestrado de dupla diplomação com a UTFPR - Universidade Tecnológica Federal do Paraná
Keywords
Modelagem baseada em agentes Indústria 4.0 Aprendizado por Re- forço NetLogo