A carregar...
Publicação
Gestão Avançada do Ciclo de Vida de Unidades de Produção
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.93 MB | Adobe PDF |
Orientador(es)
Resumo(s)
Esta dissertação foi desenvolvida no âmbito dos Sistemas de Manutenção e o seu objetivo foi
criar uma ferramenta que permita aferir sobre “Qual a melhor estratégia de manutenção a
utilizar de modo a obter o menor custo possível?”.
Para desenvolver o algoritmo era necessário ter uma amostra para testes e, por isso, foi
desenvolvido um simulador para gerar dados. Estes dados basearam-se na produtividade, custos
e nas falhas que iam ocorrendo ao longo do tempo. Foi necessário desenhar um modelo para as
máquinas e introduzir ruído de modo a torná-las o mais reais possível.
A ferramenta desenhada recebeu dados das máquinas e através de um Algoritmo
Genético gerou as estratégias a serem aplicadas durante os testes. Para aferir qual a estratégia
mínima associada ao menor custo, foi aplicado o Método dos Mínimos Quadrados.
Após obter o resultado da ferramenta criada, este foi comparado com outras duas
Estratégias de Manutenção, Run-To-Failure e Manutenção Baseada em Condições. Este passo
foi importante para verificar que o algoritmo estava a funcionar corretamente e quais os
benefícios de o utilizar. Foram comparados quatro Indicadores de Desempenho: Tempo Médio
entre Avarias (MTBF), o Tempo Médio entre Manutenções (MTBM), Taxa de Avaria (FR) e a
Disponibilidade (A), para além dos custos entre as três estratégias e verificou-se que o algoritmo
estava de acordo com as expectativas.
O projeto desenvolvido é útil não só a nível da ferramenta implementada que dá uma
estratégia aproximada do ideal, como o simulador desenhado permite fazer simulações
individuais para qualquer caso que se queira testar.
This dissertation was developed in the scope of Maintenance Systems. Its goal was to create a tool that could answer the question: “What is the best maintenance strategy to use that generates the lowest possible cost?”. A test dataset was required to build the algorithm, so a simulator that generates data was created. The data was based on productivity, costs and appearing failures over time. A model to represent the machines was built, white-noise was considered in the model to make it a trustworthy representation of reality. The conceived tool received data from the machines and through a Genetic Algorithm it presented strategies to be applied during the tests. The Least Squares Method was applied in order to find out which strategy has the lowest cost. After obtaining the result from the created tool, it was compared to two other maintenance strategies, Run-To-Failure and Condition-Based Maintenance. This step was important to validate the algorithm performance and the benefits of using the chosen algorithm. Four Performance Indicators: Mean Time Between Failere (MTBF), Mean Time Between Maintenance (MTBM), Taxe Failure (TF) and Availability (A), besides the cost, were compared among the three selected strategies and the algorithm performance was, once again, validated. This project is meaningful due to the created tool, that will generate an almost ideal strategy, but also because of the conceived simulator which allows to run individual simulations to any given case.
This dissertation was developed in the scope of Maintenance Systems. Its goal was to create a tool that could answer the question: “What is the best maintenance strategy to use that generates the lowest possible cost?”. A test dataset was required to build the algorithm, so a simulator that generates data was created. The data was based on productivity, costs and appearing failures over time. A model to represent the machines was built, white-noise was considered in the model to make it a trustworthy representation of reality. The conceived tool received data from the machines and through a Genetic Algorithm it presented strategies to be applied during the tests. The Least Squares Method was applied in order to find out which strategy has the lowest cost. After obtaining the result from the created tool, it was compared to two other maintenance strategies, Run-To-Failure and Condition-Based Maintenance. This step was important to validate the algorithm performance and the benefits of using the chosen algorithm. Four Performance Indicators: Mean Time Between Failere (MTBF), Mean Time Between Maintenance (MTBM), Taxe Failure (TF) and Availability (A), besides the cost, were compared among the three selected strategies and the algorithm performance was, once again, validated. This project is meaningful due to the created tool, that will generate an almost ideal strategy, but also because of the conceived simulator which allows to run individual simulations to any given case.
Descrição
Palavras-chave
Sistemas de Manutenção Estratégias de Manutenção Indicadores de Desempenho Sistemas Automatizados Algoritmo Genético Método dos Mínimos Quadrados