| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 6 MB | Adobe PDF |
Advisor(s)
Abstract(s)
A atual exigência e instabilidade do mercado global provocam um grande impacto nos problemas industriais, podendo destacar-se o escalonamento e o sequenciamento de trabalhos. Esta situação implica que mais recursos, para além do fator tempo, sejam considerados críticos, como máquinas, mão-de-obra e instalações. O objetivo geral das organizações prende-se com a satisfação dos clientes, quer em termos de qualidade de produtos quer no cumprimento das datas estabelecidas. Por norma, os problemas de escalonamento são classificados como problemas de otimização combinatória sujeitos a restrições, com uma natureza dinâmica e de resolução bastante complexa, cujos elementos básicos são as máquinas e as tarefas. Uma maneira de promover a sobrevivência e a sustentabilidade das organizações é o uso eficiente dos seus recursos. Uma falha completa na preparação adequada das tarefas pode facilmente levar a um desperdício de tempo e recursos, o que pode resultar num baixo nível de produtividade e altas perdas monetárias. Diante do exposto, é essencial analisar e desenvolver continuamente novos modelos de programação de produção. O escalonamento da produção na presença de eventos em tempo real é de grande importância para a implementação bem-sucedida dos sistemas de escalonamento no mundo real. A maioria das empresas opera em ambientes dinâmicos, vulneráveis a vários eventos estocásticos em tempo real, o que força a contínua reconsideração e revisão de programas pré-estabelecidos. Num ambiente incerto, maneiras eficientes de adaptar as soluções atuais a eventos inesperados são preferíveis às soluções ótimas que logo se tornam obsoletas assim que são lançadas para o chão de fábrica. Tal realidade foi a principal motivação para o desenvolvimento de uma ferramenta de apoio ao escalonamento dinâmico, a qual tenta começar a preencher a lacuna entre a teoria e a prática do escalonamento. A presente dissertação pretende contribuir para facilitar a compreensão dos problemas e melhorar o processo de escalonamento na indústria, apresentando contribuições no domínio do escalonamento da produção em duas envolventes principais: uma a um nível teórico, conceptual e outra ao nível prático da resolução de problemas através do desenvolvimento de uma ferramenta dinâmica de apoio à decisão. Ao nível conceptual contribui para uma ontologia de problemas de escalonamento da produção e conceitos relacionados, tais como ambiente de escalonamento, características dos trabalhos e dos recursos, critérios de otimização, medidas de desempenho, ferramentas de escalonamento, tipos de escalonamento e técnicas de resolução de problemas combinatórios e sua parametrização, providenciando um enquadramento comum para a compreensão e partilha de conhecimento acerca destes conceitos. Ao nível da resolução de problemas, desenvolveu-se uma ferramenta simples, moderna e autónoma de apoio à decisão ao escalonamento dinâmico onde os critérios de desempenho são classificados através do modelo de Kano. Ou seja, o protótipo desenvolvido simula a sua conexão ao software MRP e usa meta heurísticas para gerar um escalonamento preditivo. Assim, sempre que ocorrem eventos não planeados, como a chegada de novas tarefas ou cancelamento de outras, a ferramenta começa a reagendar através de um módulo de eventos dinâmicos que combina regras de despacho que melhor se ajustam às medidas de desempenho pré-classificadas pelo modelo de Kano. A ferramenta proposta foi testada em um estudo computacional aprofundado com entradas dinâmicas de tarefas com tempos de execução estocásticos. Além disso, foi realizada uma análise mais robusta, que demonstrou que há evidência estatística de que o desempenho do protótipo proposto é melhor que o método único de programação e comprovou a eficácia do mesmo. O conceito da presente dissertação já deu origem a três publicações em revistas indexadas, o que motivou ainda mais ao desenvolvimento e aperfeiçoamento do sistema desenvolvido (L. Ferreirinha et al., 2019; Luis Ferreirinha et al., 2019; Luís Ferreirinha et al., 2020).
The current demand and instability of the global market have a major impact on industrial problems, including the staggering and sequencing of jobs. This situation means that more resources than time are considered critical, such as machines, labor and facilities. The overall goal of organizations is customer satisfaction, both in terms of product quality and meeting set dates. Usually, scheduling problems are classified as constrained combinatorial optimization problems, with a dynamic nature and very complex resolution, whose basic elements are machines and tasks. One way to promote the survival and sustainability of organizations is the efficient use of their resources. A complete failure to properly prepare tasks can easily lead to a waste of time and resources, which can result in low productivity and high monetary losses. Given the above, it is essential to continuously analyze and develop new production scheduling models. Production scheduling in the presence of real-time events is of great importance for the successful implementation of real-world scheduling systems. Most companies operate in dynamic environments that are vulnerable to multiple stochastic events in real time, which forces continuous reconsideration and review of pre-established programs. In an uncertain environment, efficient ways to adapt current solutions to unexpected events are preferable to optimal solutions that soon become obsolete as they are dropped to the shop floor. Such reality was the main motivation for the development of a dynamic scheduling support tool, which attempts to begin to bridge the gap between scheduling theory and practice. This dissertation aims to contribute to facilitate the understanding of the problems and improve the process of scheduling in the industry, presenting contributions in the field of production scheduling in two main environments: one at a theoretical, conceptual level, and another at the practical level of problem solving through the development of a dynamic decision support tool. At the conceptual level it contributes to an ontology of production scheduling problems and related concepts such as scheduling environment, job and resource characteristics, optimization criteria, performance measures, scheduling tools, scheduling types, and resolution techniques regarding combinatorial problems and their parameterization, providing a common framework for understanding and sharing knowledge about these concepts. In terms of problem solving, a simple, modern and autonomous dynamic scheduling decision support tool has been developed where performance criteria are classified through Kano’s model. This is, the developed prototype simulates its connection to MRP software and uses metaheuristics to generate predictive schedules. Thus, whenever unplanned events such as new tasks arrive or others are canceled, the tool begins to reschedule through a dynamic event module that combines dispatch rules that best fit the performance measures pre-classified by the Kano’s model. The proposed tool was tested in an in-depth computational study with dynamic task inputs with stochastic runtimes. In addition, a more robust analysis was performed, which showed that there is statistical evidence that the performance of the proposed prototype is better than the single programming method and proved its effectiveness. The concept of this dissertation has already given rise to three publications in indexed journals, which further motivated the development and improvement of the developed system (L. Ferreirinha et al., 2019; Luis Ferreirinha et al., 2019; Luís Ferreirinha et al., 2020).
The current demand and instability of the global market have a major impact on industrial problems, including the staggering and sequencing of jobs. This situation means that more resources than time are considered critical, such as machines, labor and facilities. The overall goal of organizations is customer satisfaction, both in terms of product quality and meeting set dates. Usually, scheduling problems are classified as constrained combinatorial optimization problems, with a dynamic nature and very complex resolution, whose basic elements are machines and tasks. One way to promote the survival and sustainability of organizations is the efficient use of their resources. A complete failure to properly prepare tasks can easily lead to a waste of time and resources, which can result in low productivity and high monetary losses. Given the above, it is essential to continuously analyze and develop new production scheduling models. Production scheduling in the presence of real-time events is of great importance for the successful implementation of real-world scheduling systems. Most companies operate in dynamic environments that are vulnerable to multiple stochastic events in real time, which forces continuous reconsideration and review of pre-established programs. In an uncertain environment, efficient ways to adapt current solutions to unexpected events are preferable to optimal solutions that soon become obsolete as they are dropped to the shop floor. Such reality was the main motivation for the development of a dynamic scheduling support tool, which attempts to begin to bridge the gap between scheduling theory and practice. This dissertation aims to contribute to facilitate the understanding of the problems and improve the process of scheduling in the industry, presenting contributions in the field of production scheduling in two main environments: one at a theoretical, conceptual level, and another at the practical level of problem solving through the development of a dynamic decision support tool. At the conceptual level it contributes to an ontology of production scheduling problems and related concepts such as scheduling environment, job and resource characteristics, optimization criteria, performance measures, scheduling tools, scheduling types, and resolution techniques regarding combinatorial problems and their parameterization, providing a common framework for understanding and sharing knowledge about these concepts. In terms of problem solving, a simple, modern and autonomous dynamic scheduling decision support tool has been developed where performance criteria are classified through Kano’s model. This is, the developed prototype simulates its connection to MRP software and uses metaheuristics to generate predictive schedules. Thus, whenever unplanned events such as new tasks arrive or others are canceled, the tool begins to reschedule through a dynamic event module that combines dispatch rules that best fit the performance measures pre-classified by the Kano’s model. The proposed tool was tested in an in-depth computational study with dynamic task inputs with stochastic runtimes. In addition, a more robust analysis was performed, which showed that there is statistical evidence that the performance of the proposed prototype is better than the single programming method and proved its effectiveness. The concept of this dissertation has already given rise to three publications in indexed journals, which further motivated the development and improvement of the developed system (L. Ferreirinha et al., 2019; Luis Ferreirinha et al., 2019; Luís Ferreirinha et al., 2020).
Description
Keywords
Escalonamento dinâmico da Produção Ferramenta de apoio à decisão MetaHeurísticas Planeamento de experiências de Taguchi Modelo de Kano Dynamic Production Scheduling Decision Support Tool Meta-Heuristics Design of Experiments of Taguchi Kano’s Model