Vol. 56, Issue 1, pp. 19-34 (2026)

Abstract

To enhance the early warning rate of pipeline leak events and reduce the false alarm rate in Φ-OTDR-based DAS systems, this paper proposes a new method based on FMD. The introduced vibration event types include not only pipeline leaks but also hammer strikes, steel pipe hits, and noise. A 160-meter pipeline setup was constructed, and the sensing optical fiber wound around it was monitored to collect 2240 samples. The results after denoising demonstrate the effectiveness of the proposed denoising method. Subsequently, features were extracted using overlapping frames, dimensionality reduction was performed with t-SNE, and after training with NRBO-XGBoost, the leak event recognition rate reached 99.74%, the average classification accuracy reached 98.71%, and the false alarm rate was reduced to 0.19%.