| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 7.19 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Atualmente, as organizações enfrentam desafios complexos, onde a transição para a Indústria
4.0 oferece uma oportunidade única para abordá-los de maneira inovadora e sustentável. A
integração de novas tecnologias avançadas como a inteligência artificial, Internet das Coisas
(IoT) e Big Data Analytics, não só impulsiona a eficiência e competitividade das organizações,
mas também garante soluções para os desafios globais, como segurança e sustentabilidade.
Neste contexto, muitas organizações estão a passar por um processo de adaptação e
digitalização das suas operações e gestão. A implementação de ferramentas como o SCADA
(Sistema de Supervisão e Aquisição de Dados) é essencial para adaptar processos já
estabelecidos à nova realidade industrial. O SCADA tem como objetivo monitorizar e controlar
sistemas industriais em tempo real, com a capacidade de recolher, armazenar e analisar os
dados resultantes das operações.
A presente dissertação foi desenvolvida numa empresa que integra um grupo líder mundial
destacado no setor da cortiça, atualmente em processo de modernização das suas operações.
O trabalho teve como objetivo a melhoria de um sistema SCADA já implementado, mas com
várias lacunas identificadas. O projeto desenvolveu-se em três fases distintas, duas com
resultados concretos no presente e a última projetando o potencial futuro do sistema.
Em primeiro lugar, foram identificados os pontos a melhorar no sistema já implementado, que
estava restrito a uma parte da unidade industrial. Além disso, foi também desenvolvida uma
metodologia de trabalho a seguir para as restantes etapas de desenvolvimento do sistema. De
seguida, foi realizada uma análise do histórico de dados das operações e dos alarmes críticos,
na qual se identificaram oportunidades de melhoria nos processos físicos. Neste contexto, foi
constatada a ausência de comunicação do sistema com parte dos equipamentos e propôs-se
um novo método de abastecimento de material a um dos moinhos. Por último, após a
transformação da base de dados original, foram aplicados três modelos de machine learning
com o objetivo de prever o comportamento dos alarmes. Esta fase funcionou como uma prova
de conceito, uma vez que a implementação completa do sistema não estava concluída e os
dados extraídos não eram completamente fiáveis.
O projeto enfrentou diversas limitações que condicionaram os resultados obtidos,
nomeadamente o atraso no arranque da implementação do SCADA. Assim, foi apenas possível
atualizar e corrigir a parte do sistema já implementada. Relativamente ao processo físico, foi
assegurada a comunicação de todos os equipamentos com o SCADA, bem como um novo
sistema de alimentação para corrigir um dos alarmes críticos identificados, que resultou numa
diminuição significativa no disparo de um alarme crítico. Os algoritmos de machine learning
aplicados a uma base de dados de treino demonstraram a sua eficácia, ao preverem
corretamente o comportamento dos alarmes numa base de dados de teste. Este trabalho
aumentou a visibilidade dos processos produtivos com o SCADA, identificou ineficiências e
propôs melhorias, contribuindo para a maior eficiência operacional da unidade industrial.
Currently, organizations face complex challenges where the transition to Industry 4.0 offers a unique opportunity to address them in an innovative and sustainable way. The integration of advanced technologies such as artificial intelligence, the Internet of Things (IoT), and Big Data Analytics not only enhances the efficiency and competitiveness of organizations but also provides solutions to global challenges such as security and sustainability. In this context, many organizations are undergoing a process of adaptation and digitalization of their operations and management. The implementation of tools such as SCADA (Supervisory Control and Data Acquisition System) is essential to adapt established processes to the new industrial reality. SCADA's primary objective is to monitor and control industrial systems in real-time, with the capability to collect, store, and analyze data resulting from operations. This dissertation was developed within a company that is part of a globally leading group in the cork sector, currently undergoing modernization of its operations. The main objective of the work was to improve an already implemented SCADA system, though with several identified faults. The project was carried out in three distinct phases, two with tangible results in the present, and the last one projecting the system's potential for the future. First, the points for improvement in the existing system, which was limited to a portion of the industrial unit, were identified. Additionally, a working methodology was developed to guide the subsequent stages of system development. Next, an analysis of the historical operations data and critical alarms was conducted, identifying opportunities for improvement in the physical processes. In this context, the absence of communication between the system and some of the equipment was noted, and a new material supply method for one of the mills was proposed. Finally, after transforming the original database, three machine learning models were applied to predict alarm behavior. This phase served as a proof of concept, as the complete system implementation was not yet finalized, and the extracted data was not entirely reliable. The project faced several limitations that impacted the results obtained, particularly the delay in the implementation of the SCADA system. As a result, it was only possible to update and correct the part of the system that had already been implemented. Regarding the physical process, communication between all equipment and the SCADA system was ensured, and a new material supply system was implemented to address one of the critical alarms identified, which led to a significant reduction in the triggering of a critical alarm. The machine learning algorithms applied to a training dataset proved effective, as they successfully predicted the alarm behavior in a test dataset. This work enhanced the visibility of production processes through SCADA, identified inefficiencies, and proposed improvements, contributing to increased operational efficiency of the industrial unit.
Currently, organizations face complex challenges where the transition to Industry 4.0 offers a unique opportunity to address them in an innovative and sustainable way. The integration of advanced technologies such as artificial intelligence, the Internet of Things (IoT), and Big Data Analytics not only enhances the efficiency and competitiveness of organizations but also provides solutions to global challenges such as security and sustainability. In this context, many organizations are undergoing a process of adaptation and digitalization of their operations and management. The implementation of tools such as SCADA (Supervisory Control and Data Acquisition System) is essential to adapt established processes to the new industrial reality. SCADA's primary objective is to monitor and control industrial systems in real-time, with the capability to collect, store, and analyze data resulting from operations. This dissertation was developed within a company that is part of a globally leading group in the cork sector, currently undergoing modernization of its operations. The main objective of the work was to improve an already implemented SCADA system, though with several identified faults. The project was carried out in three distinct phases, two with tangible results in the present, and the last one projecting the system's potential for the future. First, the points for improvement in the existing system, which was limited to a portion of the industrial unit, were identified. Additionally, a working methodology was developed to guide the subsequent stages of system development. Next, an analysis of the historical operations data and critical alarms was conducted, identifying opportunities for improvement in the physical processes. In this context, the absence of communication between the system and some of the equipment was noted, and a new material supply method for one of the mills was proposed. Finally, after transforming the original database, three machine learning models were applied to predict alarm behavior. This phase served as a proof of concept, as the complete system implementation was not yet finalized, and the extracted data was not entirely reliable. The project faced several limitations that impacted the results obtained, particularly the delay in the implementation of the SCADA system. As a result, it was only possible to update and correct the part of the system that had already been implemented. Regarding the physical process, communication between all equipment and the SCADA system was ensured, and a new material supply system was implemented to address one of the critical alarms identified, which led to a significant reduction in the triggering of a critical alarm. The machine learning algorithms applied to a training dataset proved effective, as they successfully predicted the alarm behavior in a test dataset. This work enhanced the visibility of production processes through SCADA, identified inefficiencies, and proposed improvements, contributing to increased operational efficiency of the industrial unit.
Description
Keywords
Cork Industry 4.0 Cortiça SCADA Alarms Machine learning Alarmes Indústria 4.0