Integrating System Identification and Blind Source Separation for Real-Time Pipeline Monitoring: A Field Study

This work introduces an innovative method for real-time pipeline monitoring using acoustic sensors. Our proposed algorithm is based on decomposing the acoustic measurements into ‘’sources’’ and monitoring the sources for changes that could be attributed to leaks. The source separation (SS) algorithm is implemented using tools from system identification. In contrast to past implementations, our method performs in real time and incorporates a verification step to improve the reliability of source estimations. We explicitly incorporate SS with a cross-correlation test to verify the algorithm's reliability in identifying mutually uncorrelated sources. Furthermore, the algorithm's adaptability improved with the regularized least squares (ReLS) technique and an automatic regularization factor computed from the measured data. This automation not only assures the algorithm's flexibility but also maintains real-time performance under various situations without requiring user intervention, which is critical for online monitoring systems. Real-time monitoring, robustness, and verifiability are the three main points of this paper to guarantee that the system can reliably detect abnormalities as they occur, perform continuously under changing conditions, and produce reliable outcomes. The approach was tested with field data from an operating pipeline, confirming its effectiveness in real-world scenarios. The findings demonstrate considerable increases in both detection accuracy and real-time performance, indicating a major advancement in pipeline monitoring technology.