Evaluation of remaining flow capacity of pipe under ambient vibration measurement by computational modeling integration
Ahmad Braydi, Pascal Fossat, Mohsen Ardabilian, O. Bareille
Pipes play a crucial role in transporting essential resources such as water, oil, and gas across industrial, urban, and environmental infrastructures. Monitoring of the flow capacity in such extended structures has is a still persisting issue, potentially resulting in operational disruptions and serious safety risks. The current demand on these systems' reliability have been increasing maintenance costs. Therefore, any advance in diagnosis methods and tools is benefic. This study presents a prognostic and health monitoring approach that utilizes flow-induced acoustic emissions to detect and characterize pipeline blockages. An analytical model and a finite element is developed to capture the acoustic signatures flow-induced disturbances on the structure, and how the acoustic wave propagation is affected by the level of clogging. This reveals features highly sensitive to the changes of the flow rate. These insights drive the development of a machine learning-based predictive maintenance strategy, validated on real-case datasets. The results demonstrate exceptional accuracy, with most classifiers achieving 100% detection rates for clogging presence, shape, and severity. Additionally, model generalization tests show that machine learning algorithms adapt more effectively to varying clogging thickness than clogging shape. This research is the first step for more enhancing predictive maintenance and ensuring the reliability of industrial pipelines.