Publicação
Predictive models for equipment condition monitoring in real-time and in Industry 4.0 context
| Resumo: | The industry's pressing need to monitor equipment in real time stands as a crucial aspect today. Swift acquisition, storage, and effective management of the acquired data have become im- perative within a short timeframe. Addressing this need is the primary concern of this dissertation, focusing on data capture and the creation of predictive models capable of identifying potential failures or malfunctions dur- ing operational processes. Moreover, the aim is to validate the system's proper functionality. This process holds immense significance as it enables the assessment of equipment condition for maintenance or replacement needs and preemptively identifies failures through data- driven patterns. To facilitate these methodologies, expertise in neural network-based models is essential, as they facilitate the development of predictive models. Evaluating the performance of various models will guide the identification of the most suitable method for this process. The primary emphasis of this work was on creating multiple models proficient in predicting target signals. The outcomes extracted from these models' training were carefully assessed to determine the most effective model. Upon achieving a satisfactory model, it was deployed for daily forecasting, effectively identifying equipment failures or anomalies. Additionally, a real-time fault detection system was devised. This system operates by compar- ing valve opening and temperature values to swiftly identify faults as they occur. |
|---|---|
| Autores principais: | Macedo, Luís Pedro Gajeiro |
| Assunto: | Predictive Models Neuronal Networks Correlation Models Fuzzy Models Faults Data Analysis |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | The industry's pressing need to monitor equipment in real time stands as a crucial aspect today. Swift acquisition, storage, and effective management of the acquired data have become im- perative within a short timeframe. Addressing this need is the primary concern of this dissertation, focusing on data capture and the creation of predictive models capable of identifying potential failures or malfunctions dur- ing operational processes. Moreover, the aim is to validate the system's proper functionality. This process holds immense significance as it enables the assessment of equipment condition for maintenance or replacement needs and preemptively identifies failures through data- driven patterns. To facilitate these methodologies, expertise in neural network-based models is essential, as they facilitate the development of predictive models. Evaluating the performance of various models will guide the identification of the most suitable method for this process. The primary emphasis of this work was on creating multiple models proficient in predicting target signals. The outcomes extracted from these models' training were carefully assessed to determine the most effective model. Upon achieving a satisfactory model, it was deployed for daily forecasting, effectively identifying equipment failures or anomalies. Additionally, a real-time fault detection system was devised. This system operates by compar- ing valve opening and temperature values to swiftly identify faults as they occur. |
|---|
Atividades financiadas
Não foram encontrados projetos financiados.