Distributed Fiber Optic Warning Identification Algorithm for Oil and Gas Pipelines Based on the Inception-Dvs Model

Abstract

Distributed fiber optic vibration early warning technology, with its adequate monitoring range, non-electrical detection, and high sensitivity, is currently the most promising early warning technology for preventing third-party damage to long-distance oil and gas pipelines. To enable low-cost and universal applications of fiber optic vibration early warning devices, it is necessary to reduce the dependence of distributed fiber optic vibration early warning AI recognition models on hardware computing power and facilitate their integration with fiber optic vibration hosts and edge deployment. To address these concerns, this paper extends the benefits of the Inception network and constructs a novel and lightweight Inception-DVS model for fiber optic warning in complex environments along long-distance oil and gas pipelines. Experimental results show that the 3.519 MB model size reduces model computation and increases detection speed while maintaining detection accuracy. This model is more suitable for fiber optic warning and can detect various emergencies.

Keywords: Distributed fiber optic sensing, Oil and gas pipelines, convolutional neural networks, Safety warning