Detection And Recognition Of Vibration Events Based On φ-OTDR

Phase-sensitive optical time domain reflectometer(φ-OTDR),as one of the most popular distributed fiber optic sensors for vibration signal detection and analysis in recent years,has the advantages of long-distance multi-point detection,high resolution,extensive dynamic response range,anti-electromagnetic interference,corrosion resistance,low energy loss,wide measurement band,anti-flammability and anti-explosion,etc.It can achieve long-distance distributed abnormal vibration detection in crucial areas such as perimeter security,oil transportation pipelines,cable detection,borderline intrusion detection,and building structure health monitoring.The optical fiber in φ-OTDR can be used as both a signal transmission medium and a vibration sensing medium to collect Rayleigh backscattered signals carrying vibration information generated by pulsed light in real-time and acquire sensing data continuously.However,the high sensitivity and long-range distributed detection characteristics of φ-OTDR result in the high false alarm rate that inevitably accompanies φ-OTDR in practical applications.The false alarms of the system usually originate from the frequency drift and linewidth range of the optics,the thermal and scattering noise of the electronics,and the incidental interference signals in the detection environment that do not require vibration alarms.The noise caused by optical or electronic devices can be postprocessed by the accepted optical signal and the collected vibration signal to improve the system signal-to-noise ratio,thereby reducing false alarms as much as possible.While the real external vibration source causes the interference signal in the detection environment,therefore,the postpattern recognition of the intrusion vibration signal becomes the key to distributed fiber optic sensing technology,which can reduce the system’s false alarm rate,improve the detection effect in practical applications and reduce the cost of security detection.This study investigates Rayleigh scattering in optical fiber and its sensing mechanism for vibration demodulation and graphic representation of distributed fiber optic sensing system.Based on this,the distributed fiber optic vibration sensing system of phase-sensitive optical time domain reflectometer(φ-OTDR)is developed independently,which can be applied to real-time vibration detection.Then,according to different vibration signal demodulation methods and the problem of multi-target recognition in the presence of interference signals arising from pattern recognition,the vibration recognition methods of φ-OTDR under different detection signal states are designed and discussed respectively,and a φ-OTDR vibration monitoring system based on pattern recognition is built to verify the feasibility of φ-OTDR pattern recognition applied to actual long-distance abnormal vibration detection.The combination of online demodulation and recognition of vibration signals is achieved while maintaining a lower cost,aiming to improve the classification accuracy of φ-OTDR and reduce the false alarm rate of interference signals.Finally,the model’s performance is evaluated regarding classification accuracy,precision,recall,F1 score,model size,and other metrics.Different design paths are provided to improve the vibration detection performance of φ-OTDR.The main work and findings of the dissertation are as follows:(1)A φ-OTDR vibration event recognition method based on a combination of φ-OTDR vibration spatio-temporal image segmentation pre-processing,texture,statistical and morphological feature extraction,and weighted support vector machines is proposed in the case of a φ-OTDR system collecting few non-cooperative intrusion vibration signals,which can effectively classify five types of vibration events in high-speed railway perimeter intrusion detection with small sample data and no parallel processing units.Erosion and dilation operations are applied to vibration signal image feature enhancement in image pre-processing.The vibration signal region and background are separated by the maximum inter-class variance method,then 35 features of the vibration signal region are calculated and finally employed to construct a WSVM.Experiments show that the method achieves 99 FPS and 98.8% accuracy on the test set with 330 vibration images as the training set to build the model in the presence of interference signals and with limited computing resources.It provides a generalized Φ-OTDR vibration event recognition method for small samples.(2)A real-time φ-OTDR multi-vibration event classification method based on the combination of convolutional neural network(CNN),bi-directional long short-term memory network(Bi-LSTM)and connectionist temporal classification(CTC)is proposed for the case where the spatio-temporal image detected by φ-OTDR contains multi-vibration events and no annotation tool is required,which can quickly and effectively identify the type and number of vibrations contained in the data image when multiple vibration signals are contained in a single image,and manual alignment is not required for model training.CNN is used to extract spatial dimensional features in spatio-temporal images,Bi-LSTM extracts temporal dimensional correlation features,and the hybrid features are automatically aligned with the labels by CTC.A dataset of 8,000 vibration images containing 17,589 abnormal vibration events is collected for model training validation and testing.Experiments show that the recognition model trained with this method can achieve 210 FPS and 99.62% F1 score on the test set.(3)An improved YOLO-A30 based on the YOLO target detection system is proposed to improve the system recognition performance when the detection distance of φ-OTDR is required to reach the hundred-kilometer level,which corresponds to a larger spatio-temporal image size in the distance dimension generated by a single φ-OTDR to be recognized.The model combines residual network,path aggregation network,and multi-head detection to extract multi-scale depth features and improve the accuracy of multi-target detection in terms of each aspect of target confidence loss,localization loss,and classification loss.Experiments show that using this method to process 8069 vibration data images generated from 5 abnormal vibration events for two types of fiber optic laying scenarios: buried underground or hung on razor barbed wire at the perimeter of high-speed rail,the system m AP@.5 is 0.995,555 frames per second(FPS),and can detect a theoretical maximum distance of 135.1 km per second.It can quickly and effectively identify abnormal vibration activities,reduce the false alarm rate of the system for hundred-kilometer level distance multi-vibration along high-speed rail lines,and significantly reduce the computational cost while maintaining accuracy.(4)For signal light phase demodulation of φ-OTDR,a fast hybrid digital demodulation method is proposed for demodulating the optical phase change due to vibration signals to obtain vibration time domain signals.The hybrid demodulation method divides the Rayleigh backscattered light into two paths via a coupler.One signal is demodulated using the direct detection method at the location of vibration occurrence;the other is demodulated using the local differential detection method for the vibration event time-domain variation signal.Based on this,after extracting the signal time-frequency features by the sliding window,CNN further extracts the signal features and analyzes the temporal relationship of each group of signal features using BiLSTM.Finally,the CTC is used to label the unsegmented sequence data to achieve single detection of multiple vibration targets.Experiments show that using this method to process the collected8563 data,containing five different frequency bands of multi-vibration acoustic sensing signal,the system F1 score is 99.49% with a single detection time of 2.2ms.The highest frequency response is 1k Hz.It is available to quickly and efficiently identify multiple abnormal vibration signals when a single demodulated acoustic sensing signal contains multiple vibration events.