A Novel Framework for Industrial Steam Turbine Generator Start-Up Predictive Monitoring Considering Data Limitations
[EN] This work investigates the application of predictive diagnostics for large steam turbine generators, focusing on operational challenges arising from journal bearing imbalance during start-up procedures. Utilizing industrial low-sampling frequency monitoring data, the study addresses the inheren...
| Autores: | , , |
|---|---|
| Formato: | artículo |
| Fecha de publicación: | 2025 |
| País: | España |
| Recursos: | Universitat Politècnica de València (UPV) |
| Repositorio: | RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia |
| Idioma: | inglés |
| OAI Identifier: | oai:riunet.upv.es:10251/227127 |
| Acesso em linha: | https://riunet.upv.es/handle/10251/227127 |
| Access Level: | acceso abierto |
| Palavra-chave: | Time series analysis Monitoring Artificial intelligence Training Shafts Predictive models Power generation Generators Displacement Electrical machine Journal bearing Long short term memory Predictive maintenance Steam turbine Synchronous generator Vibration |
| Resumo: | [EN] This work investigates the application of predictive diagnostics for large steam turbine generators, focusing on operational challenges arising from journal bearing imbalance during start-up procedures. Utilizing industrial low-sampling frequency monitoring data, the study addresses the inherent difficulties of real-time monitoring and data storage in industrial settings. The research includes an extensive review of state-of-the-art techniques in pertinent predictive maintenance and time-series forecasting, with a particular emphasis on artificial intelligence-based methods. A novel diagnostic framework is developed to leverage multivariate time-series data, incorporating critical speed zones, nonlinear feature correlations, and data augmentation through time warping techniques to enhance model robustness. The work also highlights the use of tailored development and normalization strategies to adapt artificial intelligence models for industrial applicability. Results demonstrate the model's ability to predict start-up behavior, providing actionable insights to optimize operational procedures. This study lays the groundwork for further integrating advanced artificial intelligence driven diagnostics into industrial systems, offering a scalable and effective solution for improving system reliability and reducing operational costs. |
|---|