Recognition of aliasing fiber-optic vibration signals based on DCN network and FSWD algorithm

Abstract

Recognition methods for fiber-optic vibration signals are widely used in abnormal monitoring of oil lines, optical cables, railways, etc. However, existing methods are incapable of aliasing signals in actual situations. This study presents an aliasing signals recognition method, that can effectively recognize a signal aliased with single tone vibration signals and predict frequency of single tone vibration signals. Part of frequency spectrum of a aliased signal is input deep complex network (DCN) for recognition to reduce effect of single tone vibration signals, and frequency spectrum window-sliding detection (FSWD) algorithm is proposed to predict frequency of single tone vibration signals. To make predicted frequency close to true frequency, appropriate frequency calculation method and dataset generation criteria are designed. Experimental results show recognition accuracy of aliased signals can reach 0.989 when aliasing only happened in time domain, 0.969 when aliased with a single tone vibration signal in frequency domain. Frequency prediction accuracy can reach 0.975 when aliasing only happened in time domain, error ≤ 10 Hz.