New Generation of AI techniques Applied to Third Party Interference and Leakage Detection on Pipelines
Proceedings Publication Date
Ana Paula Gomes
Ana Paula Gomes, Marco Marino, Massimilano Biagini, Fabio Chiappa
Part of the proceedings of

This paper introduces a pipeline monitoring system engineered to identify, locate, and categorize leakages and Third-Party Interference (TPI) incidents along a pipeline right-of-way, through the use of Vibroacoustic Technology. This method serves as a highly effective protective measure suitable for retrofitting onto existing pipelines. Its primary focus is real-time asset monitoring with an emphasis on identifying and locating various irregular activities, such as civil works, manual or mechanical digging, impacts, and other actions—whether fraudulent or unintentional—that could potentially lead to damage.

The Vibroacoustic Technology employed in this system is a reliable and robust pipeline monitoring solution. Strategically placed pressure sensors and accelerometers along the pipeline capture the vibroacoustic waves generated by anomalies, i.e., TPIs or leakages. The TPI physical propagation model is based on the mechanical energy produced by ground vibrations, which are then transferred to the soil. The resulting elastic wave is transmitted to the pipeline fluid and shell, undergoing an acoustic conversion that creates a signal propagating several kilometers inside the pipeline.

These captured data are transmitted to processing servers equipped with advanced algorithms for noise reduction, detection, and localization. Following detection and localization, this innovative approach incorporates machine learning and deep learning techniques to establish a data-driven system for the classification of which TPI event triggered the alarm.

In this paper we present a challenging solution that advances toward a more sophisticated approach through event classification based on deep learning and convolutional neural networks. Convolutional neural networks are currently considered to be one of the most powerful approaches for addressing classification problems. This novel approach features an AI engine trained to recognize and classify anomalous events, thereby minimizing false alarms. Such an enhanced system has the potential to become a versatile and powerful tool for ensuring the integrity of pipeline assets.

To view the video or download the paper please register here for free

You already have access? Sign in now.