Identification and Prediction of Low Flow Rate Based on Time-Shift Statistical Feature Enhancement With DAS

In recent years, distributed optical fiber acoustic sensing (DAS) technology has revolutionized the monitoring of long-distance pipelines, offering full-line, real-time, and high-precision velocity measurements. Nonetheless, achieving high resolution at low flow rate remains a pivotal challenge confronting researchers in this field. This article delves into the mechanics of vortex generation and shedding on the internal fluid surface within pipelines, uncovering the intricate relationship between flow rate and the variance characteristics of pipeline surface displacement. We propose time-shift statistical feature enhancement (TSSFE) combined with deep learning techniques. This method enables us to not only accurately identify but also predict flow rates using DAS data. Remarkably, our noninvasive detection method achieves an identification accuracy of 94.31% and a prediction accuracy of 92.3%, while maintaining an exceptional average resolution of 0.015 m/s. This breakthrough paves a novel path for enhancing the resolution of low-flow rate detection, offering significant potential for applications in various industries where precise monitoring of pipeline flow is crucial.