Integrating Model-Centric and Data-Centric Techniques for Pipe System Prognostics and Health Management

• Published in: E-journal of Nondestructive Testing 2024-07, Vol.29 (7)
• Main Authors: Braydi, Ahmad, Fossat, Pascal, Casaburo, Alessandro, Pernet, Victore, Zwick, Cyril, Ardabilian, Mohsen, Bareille, Olivier
• Format: Article
• Language: English
• ISSN: 1435-4934; 1435-4934
• DOI: 10.58286/29631

In today's industrial landscape, the proactive implementation of predictive maintenance techniques is imperative, especially in the context of pipe systems, as companies increasingly embrace cutting-edge technologies such as artificial intelligence and the Internet of Things (IoT). Orano/La Hague, like many other industry leaders, recognizes the vital importance of integrating these technological advancements into their operations. One of the critical challenges they face relates to recurrent pipe-clogging incidents, leading to energy inefficiencies and financial losses. Addressing maintenance needs proactively is essential to mitigate risks and ensure the safety of both personnel and valuable assets. This research addresses these challenges by introducing an innovative hybrid approach that combines data-centric and model-centric methodologies for the continuous prognostics and monitoring of pipeline systems. Leveraging experimental passive acceleration measurements, this approach offers a reliable means to predict and assess the severity of clogs as they occur. To enhance the accuracy of predictions, a sliding window technique is employed to minimize noise and extract pertinent features from the data. The results of this study highlight the exceptional effectiveness of the proposed approach in accurately predicting clogging incidents and quantifying their severity, even in scenarios involving varying airflow rates within the pipes. This research marks a significant step forward in the domain of prognostics and health monitoring, with the potential for widespread applications across various industries. The integration of data-centric and model-centric approaches represents a promising solution to the complex challenge of predicting and preventing pipe-clogging incidents, ultimately contributing to enhanced operational efficiency and asset protection