Recognition and suppression of phase-sensitive OTDR interference fading based on machine learning

In phase-sensitive optical time domain reflectometer (Φ-OTDR), the existence of interference fading signal leads to misjudgment and missing judgment in vibration sensing. In this paper, a fading recognition model of back propagation (BP) neural network is constructed, features are extracted from beat frequency generated based on frequency shift delay loop. A fading recognition accuracy of 91.82% is achieved. Rotated-vector-sum algorithm is introduced to aggregate signal labelled non-fading. Vibration location and phase demodulation is realized and on this basis, this method improves the interference fading suppression probability.