A Fault Detection Method for Oil and Gas Pipelines Based on Distributed Acoustic Sensing

This study focuses on DAS (Distributed Acoustic Sensing) systems for oil and gas pipelines and proposes a fault detection method that integrates spatiotemporal frequency-domain alignment, residual signal enhancement, and multi-scale spatiotemporal feature extraction. By aligning frequency-domain signals and normalizing amplitudes, it effectively handles multi-source heterogeneity and noise interference. Meanwhile, a residual mechanism highlights subtle fault features. In the feature extraction stage, multi-scale dilated convolution is combined with spatiotemporal attention to adequately recognize and locate faults of varying frequencies and durations. Experimental results indicate that this method demonstrates robust performance and strong generalization in complex environments, delivering higher detection accuracy and lower false alarm rates compared to conventional deep learning models, thus showcasing its potential for pipeline health monitoring.