利用深度神经网络实现分布式相干瑞利光纤振动事件分类

Deep neural network-based vibration signal classification in distributed coherent Reyleigh fiber sensor security pre-warning systems

该文利用分布式相干瑞利光纤传感系统,在西气东输一线无锡至苏州段开展现场测试,采集了光纤沿线车辆行走、机械挖掘、人工锄地、定向钻孔等8种振动作业产生的光纤信号,并提出了一种具有5层结构的全连接深度神经网络用于振动事件分类识别以实现不同振动作业的分级管理。振动作业产生的光纤信号能量集中在低频,该文利用梅尔对数频率的非均匀特性提取了25维单帧信号特征量,并将连续40帧信号特征量组合成高维向量作为网络输入特征向量,实现对不同振动作业时变特性的建模。分类识别结果表明,基于深度神经网络结构的振动信号分类识别器能够有效识别不同振动作业类型,实际线路实验验证了该文算法的有效性。

Based on the distributed coherent Rayleigh fiber sensor system,field tests are conducted on West East Gas Pipeline between Wuxi and Suzhou.Vibration fiber sensor signals excited by various classes of intrusion events,including car moving,mechanical excavation,artificial slapping,directional drilling,et al.are collected,and a five-layer neural network is used for vibration fiber sensor signal classification which is intended for hierarchical management.It is shown that the vibration signals are centered at low frequency regions,and the non-linear frequency characteristic of logarithm Mel-frequency is used to extract 25 dimensional features,and 40 successive frames features are combined as the input features for the network.Classification results show that the deep neural network-based vibration signal classification can achieve satisfied performance,and the realistic experiment validate the effectiveness of the proposed method.