基于STFT图像特征的天然气管道预警技术研究

Research on gas pipeline early warning technology based on STFT image features

摘要:

基于相位敏感光时域反射计(Φ-OTDR)的分布式光纤微振动传感系统,使用STFT算法对中石油西南某地20km天然气管道周围发生六类事件(噪声、机械破环、人工挖掘、榔头敲击、人跳跃以及人踏步通行)的振动数据进行时频域图像化,然后通过BP、SVM、GoogLeNet等算法进行六类事件的分类预测。该方法首先利用Φ-OTDR系统采集管道同沟敷设光缆一芯中瑞利后向散射光信号,从中提取出振动事件的振动信号,然后通过STFT对时间-空间图像的采集,最后对图像数据利用算法进行振动事件的识别。现场实时监测获取的数据表明：利用BP神经网络算法对振动事件识别,相较于传统的随机森林、K近邻、SVM算法以及深度学习GoogLeNet算法,机械破坏事件的误报率低至0%,人工挖掘事件的误报率低至0%,适合应用于实际的天然气管道监测。

Abstract:

Based on the distributed optical fiber micro-vibration sensing system of φ -OTDR, STFT algorithm is used to image the vibration data of six kinds of events (Noise, mechanical break, artificial dig, hammer, jump and one step) around a 20km natural gas pipeline in southwest China in time and frequency domain. Then six kinds of events are classified and predicted by BP, SVM, GoogLeNet and other algorithms. Firstly, the φ -OTDR system was used to collect Rayleigh backscattered light signals and extract vibration signals of vibration events. Then, the time-space images were collected by STFT. Finally, the vibration events were identified by the algorithm of images. The experimental results showed that compared with the traditional random forest, K-nearest neighbor, SVM and GoogLeNet. The false alarm rate of mechanical break and artificial dig are calculated to be as low as 0%, The proposed algorithm is suitable for applications in a complex field environment.