Hybrid Feature Extraction of Flow-Induced Vibration for Nonintrusive Pipeline Flow Velocity Monitoring via a Fiber-Optic Acoustic Sensor

Although flow-induced vibration (FIV) has been widely used for pipeline flow velocity measurement, accurate low-velocity measurement remains challenging due to the inherently weak FIV signal and strong background noise, which often lead to significant errors. This study proposes a non-intrusive pipeline flow velocity monitoring method based on hybrid feature extraction from FIV signals acquired via a fiber optic acoustic sensor. The hybrid feature vector comprises standard deviation, energy, frequency band energy, and zero-crossing rate. Among these, the zero-crossing rate demonstrates excellent performance in predicting low flow velocity. Moreover, owing to the complementary characteristics among the different feature parameters, the average measurement error is significantly reduced across both low and high flow velocity ranges. Experimental results show that the relative error at a low flow velocity of 0.5 m/s is as low as 0.39%, and the average measurement error over a range of 0.5–4.5 m/s is reduced to 4.04%. These findings underscore the strong potential of the proposed approach for high-accuracy pipeline flow velocity monitoring in future applications.