ESTG - Mestrado em Engenharia Eletrotécnica - Energia e Automação
Permanent URI for this collection
Browse
Browsing ESTG - Mestrado em Engenharia Eletrotécnica - Energia e Automação by advisor "Bento, Luís Manuel Conde"
Now showing 1 - 3 of 3
Results Per Page
Sort Options
- COMPARAÇÃO DE MÉTODOS DE APRENDIZAGEM POR REFORÇO EM PROCESSOS INDUSTRIAIS DISCRETOS SEQUENCIAISPublication . Silva, Tiago Resende da; Bento, Luís Manuel Conde; Perdigoto, Luís Miguel RamosO processo de comissionamento de equipamentos durante a implementação de um novo sistema, ou na reconfiguração de um já existente, é uma etapa em que as empresas gastam dinheiro e tempo antes da entrada em operação. Baseando-se nesse problema, este trabalho analisa a utilização de Gémeos Digitais, que simulem o ambiente fabril, em conjunto com a utilização de técnicas de Aprendizagem por Reforço para permitir que o sistema se reconfigure e se programe de forma automática. Diferentemente das técnicas tradicionais de controlo, a Aprendizagem por Reforço encara o sistema como uma caixa negra, em que a interação entre o agente e o ambiente promove a sintonização dos parâmetros necessários para o funcionamento correto do processo industrial. Isso resulta na economia de dinheiro, na diminuição de tempo de produção e na busca de inúmeras possibilidades de operação até que se encontre a mais eficiente. Acrescenta-se a tudo isso a diminuição do tempo de exposição de pessoas ao processo de implementação inicial e consequente diminuição de acidentes, uma vez que o comissionamento ocorre no ambiente virtual. Dessa forma, este trabalho desenvolve e aplica alguns dos diferentes algoritmos de Deep Reinforcement Learning a um sistema de empacotamente de latas. O principal objetivo é avaliar a viabilidade e o desempenho de utilizar estes tipos de algoritmos na aprendizagem e otimização automática das sequências de controlo em processos industriais de natureza sequencial e discreta. Dada a natureza sequencial dos processos, com necessidade inerente de efeito de memória, foram experimentadas diferentes arquiteturas de redes neuronais, realizando um estudo comparativo sobre a performance das redes neuronais LSTM (Long Short-Term Memory) frente à utilização de buffers de memória de estados anteriores de diferentes tamanhos. Por fim, os modelos com melhores resultados e maior estabilidade foram aplicados aos Gémeos Digitais, mostrando a capacidade que estes tipos de algoritmos de aprendizagem automática têm para ser aplicados no controlo de sistemas industriais.
- HAPTIC DEVICES FOR ADVANCED INTERACTION WITH MULTIMEDIA SYSTEMSPublication . Silva, Bruno Miguel Lopes e; Castro, Hugo Filipe Costelha de; Bento, Luís Manuel Conde; Assunção, Pedro António Amado deThis project report describes the work carried out to design and evaluate new interfaces for media consumption in Smart TVs. The main goal is the study of the different technologies that can integrate a new generation of remote controls, and the development of prototypes (interfaces) that allow to connect all selected technologies in order to find a combination that improves the users quality of experience (QoE). Traditionally, the use of remote controls for television control uses only mechanical buttons to execute different tasks but, due to constant technological progress, an evolution is underway in the way users consume television content, moving from traditional passive visualization of contents to interactive forms. In this new interactive forms the user increasingly uses actions, adapts and interacts with multimedia content, services on demand and applications that support multiple users (games, social networks, etc). To keep up with these developments, it is necessary to add features to the conventional remote controls to enable a dynamic control, fast and accurate text input methods, and haptic feedback that is appropriate for the tasks that users need to perform. Two interfaces were developed integrating different haptic technologies that would provide the features mentioned above (dynamic control, text input methods, and haptic feedback). To assess the operation of the developed interfaces, tests were performed with different users to obtain objective and subjective classifications. The contribution of this work, was the integration of technologies that can improve the users QoE when using a Smart Tvs/Boxes and collecting users satisfaction levels regarding to the use of these technologies in remote controls. The tests show that despite the large number of different haptic effects users preferred one for each type of different technologies. Based on user rating, for the dynamic interaction case, one technology has proved to be preferred. In the case of text input/editing, among the various technologies, one of the technologies allows a fast text input with the highest level of user satisfaction.
- INDUSTRIAL DEVICE INTEGRATION AND VIRTUALIZATION FOR SMART FACTORIESPublication . Martins, André Fagundes; Castro, Hugo Filipe Costelha de; Neves, Carlos Fernando Couceiro de Sousa; Bento, Luís Manuel CondeGiven the constant industry growth and modernization, several technologies have been introduced in the shop floor, in particular regarding industrial devices. Each device brand and model usually requires different interfaces and communication protocols, a technological diversity which renders the automatic interconnection with production management software extremely challenging. However, combining key technologies such as machine monitoring, digital twin and virtual commissioning, along with a complete communication protocol like OPC UA, it is possible to contribute towards industrial device integration on a Smart Factory environment. To achieve this goal, several methodologies and a set of tools were defined. This set of tools, as well as facilitating the integration tasks, should also be part of a virtual engineering environment, sharing the same virtual model, the digital twin, through the complete lifecycle of the industrial device, namely the project, simulation, implementation and execution/monitoring/supervision and, eventually, decommissioning phases. A key result of this work is the development of a set of virtual engineering tools and methodologies based on OPC UA communication, with the digital twin implemented using RobotStudio, in order to accomplish the complete lifecycle support of an industrial device, from the project and simulation phases, to monitoring and supervision, suitable for integration in Industry 4.0 factories. To evaluate the operation of the developed set of tools, experiments were performed for a test scenario with different devices. Other relevant result is related with the integration of a specific industrial device – CNC machining equipment. Given the variety of monitoring systems and communication protocols, an approach where various solutions available on the market are combined on a single system is followed. These kinds of all-in-one solutions would give production managers access to the information necessary for a continuous monitoring and improvement of the entire production process.