Research And Implementation Of Classification Algorithm Based On ?-OTDR Optical Fiber Intrusion

With the rapid development of Chinese pipeline transportation industry,the traditional security early warning system is facing a greater challenge.Crude oil and refined oil pipeline transport was rapidly developing,and it has become a major underground energy artery,so the protections of pipeline transport normal operation and the national energy security and economic development are very important.The pipeline transportation safety warning system is a large-scale safety early warning system,which is widely distributed and susceptible to electromagnetic interference in the outside world,so it couldn't play the role of monitoring in traditional way.In order to indentify and warning the harmful vibration events which may appear on the pipeline transportation in China.Based on fiber optic sensing technology,we used the fiber vibration technology to distribute optical fiber sensing technology,and applied it to long-distance safety early warning and monitoring system.In this paper,a classification algorithm based on ?-OTDR fiber intrusion is proposed,which aims to accurately identify two kinds of common external intrusion signals in the case of low SNR.The proposed algorithm includes three parts:signal correlation processing,vibration signal feature extraction and RVFL neural network to classify and identify vibration signals.First of all,use correlation arithmetic to process the acquisition signal,use the energy comparison method to confirm the template signal,and conduct the related operations between the template signal and the acquisition signal.Compared with the conventional signal filtering method,this method can get good results in restraining noise and retain the details of the signal as much as possible to avoid the problem that the system identification rate is reduced due to the loss of the vibration signal.Secondly,using the wavelet decomposition principle,the correlation coefficients are decomposed into five layers,we calculated the average energy of each band and selected five bands coefficient energy ratio as vibration signal characteristics.Then,the vibration signal characteristic samples are sent into the RVFL neural network for training,so as to complete the classification and identification of vibration signals.Finally,the effectiveness of the proposed algorithm is verified by real field test.