Research On Location And Identification Of Underground Disturbance Sources Based On Distributed Optical Fiber Field Data

With the rapid advancement of China's urbanization process,the scale of urban underground pipeline construction continues to increase.Urban underground pipelines are crisscrossed,the environment is complex,and the leakage of urban underground pipelines caused by corrosion,aging,construction damage and other reasons are frequent.If it is not discovered in time,it will lead to large-scale accidents such as building subsidence,collapse,explosion,and bring immeasurable loss of life and property.Therefore,it is necessary to monitor the safety status of urban underground pipelines in real time to avoid major accidents.Phase Optical Time-Domain Reflectometer(?-OTDR)is a distributed optical fiber sensor that monitors vibration changes,which can realize the continuous spatial distribution and temporal changes of vibration events around the optical fiber transmission path compared with traditional monitoring methods,monitoring has the advantages of corrosion resistance,anti-electromagnetic interference,continuous distributed monitoring,high detection sensitivity,and fast response speed.It meets the requirements of pipeline monitoring and has become a frontier issue and research hotspot in the field of urban underground pipeline safety monitoring.Due to the influence of urban underground complex environment,the application of ?-OTDR sensing technology still needs more in-depth research in data processing,feature extraction and recognition.There are three main problems affecting the leakage monitoring of urban underground pipelines(1)In the early stage of pipeline leakage,the leakage signal strength is weak and the signal-to-noise ratio is low;(2)Affected by the environmental interference factors and the amount of monitoring data,the space-time positioning accuracy is poor and the real-time performance is low;(3)Under the coupling interference of many factors,the key information of the event is difficult to represent,which leads to the low recognition rate of urban underground pipeline leakage.In response to the above problems,research on the location and identification of underground disturbance sources based on distributed optical fiber field data is carried out.There are two points to be explained here:Firstly,the ?-OTDR monitoring data has field characteristics,and the data also contains three-dimensional properties of time,distance and phase;Secondly,the disturbance source refers to the vibration that acts on the optical fiber and disturbs its propagation signal event.Underground disturbance sources include bursts and leaks of underground pipelines.This paper focuses on the early leakage of urban underground pipelines.First,we study weak signal enhancement methods,then locate the event under multi-interference coupling,accurately determine the location information and action time of the disturbance source,and then study the relationship between multi-dimensional features and pipeline leakage,and extract a few key points.Hybrid features combined with weighted random forest algorithm to achieve accurate identification of pipeline leakage events.The main contents of this paper are as follows:1.Research on disturbance source signal enhancement method based on empirical mode decomposition.Aiming at the problem of non-stationary monitoring signals and environmental noise,a complete empirical mode decomposition(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise,CEEMDAN)and distribution difference measurement(Kullback-Leibler,KL)based on adaptive noise are proposed.The disturbance source signal enhancement method.Firstly,the method performs empirical mode decomposition on the monitoring signal,and calculates the distribution difference between each Intrinsic Mode Function(IMF)and the original signal.Secondly,the characteristic information in the reconstructed signal is improved through the distribution difference measurement,the main characteristics of the signal are improved,the secondary characteristics are reduced,and the signal noise reduction and reconstruction are realized to achieve the purpose of enhancing the signal.Finally,a comparison analysis of disturbance source signal enhancement is carried out through simulation experiments.The experimental results show that the method can effectively remove the interference of environmental noise,improve the quality of the disturbance source signal,and the signal-to-noise ratio is increased by 9.2dB on average.2.Research on disturbance source location method based on adaptive nearest neighbor binary mode.Aiming at the problem of poor temporal and spatial positioning accuracy and low real-time performance of the disturbance source,a disturbance source location method based on Adaptive Moving Neighbor Binary Pattern(AMNBP)is proposed.According to the characteristics of the vibration propagation of the disturbance source,AMNBP realizes the location of the disturbance source through the adaptive close proximity binary mode.First,the signal is divided into windows,and each window corresponds to a four-bit binary code.Then,the average value of the window and the average value of each group of signals are compared to obtain the binary coded result,which reduces the influence of local noise.Then through the non-leakage signal for adaptive gain adjustment,the key information in the encoding result is improved,and the space-time boundary of the pipeline leakage event is quickly located.It not only solves the problem of accurate event location,but also overcomes the problem of huge data volume and untimely location of the ?-OTDR sensing system,and reduces the data scale used for feature extraction and event recognition.Finally,the effectiveness of the positioning method is proved through experiments.The spatio-temporal positioning accuracy reaches 94.35%,and the average efficiency is increased by 27.32%.3.Research on Multi-dimensional Feature Extraction and Selection Method of Disturbance Source.Feature extraction research is a prerequisite for accurate identification of pipeline leakage events.There are many coupling factors in the monitoring environment of urban underground pipelines,a single feature is difficult to accurately represent the key information of the event,and there is a lack of description of the core characteristics of the disturbance source.Based on a large amount of literature research,this paper sorts out 20 common feature extraction methods covering the time domain,frequency domain,signal processing and other fields,compares the differences in the representation of the disturbance source from multiple dimensions,and adopts a random forest-based classifier the wrapping method realizes feature selection.Among them,according to the characteristics of ?-OTDR data,improved the peak-to-average ratio and zero-crossing rate feature extraction methods;introduced features such as average amplitude difference,signal duty cycle,and cepstrum feature coefficients in the voice processing field,and peak coefficients and trough coefficients in the statistical field,Skewness and kurtosis.Finally,experiments are carried out to sort out and compare the differences of various features in pipeline leakage and interference events,and obtain the key feature combination of pipeline leakage by calculating the data errors of the features outside the bag.4.Research on the method of perturbation source identification based on mixed features and weighted random forest.In view of the unclear correspondence between different features and disturbance sources,and the low recognition rate of urban underground pipeline leakage events,a disturbance source identification method(Hybrid Features and Weighted Random Forest,HF-WRF)based on hybrid features and weighted random forest is proposed.Based on the result of feature importance screening that affects pipeline leakage identification,the weighted random forest algorithm based on feature combination is studied.According to the feature selection difference results,increase the weight value of the decision tree containing the key feature of pipeline leakage identification,reduce the weight value of other feature decision trees,and improve the efficiency of random forest identification.Finally,a pipeline leakage experiment under various pressures was carried out.The results showed that based on the peak-to-average ratio,short-time interval zero-crossing rate,and frequency band width hybrid characteristics and the weighted random forest identification method,the pipeline leakage event under the influence of multiple interference factors can be accurately identified.The average recognition accuracy of underground pipeline leakage events reached 98.16%.According to the above research content,the innovation of this paper has the following three points(1)A ?-OTDR monitoring signal enhancement method based on ceemdan-kl is proposed.Aiming at the problem of weak and non-stationary monitoring signal in the actual environment,the empirical mode decomposition(EMD)of complete Gaussian white noise is added to solve the problem of component mode aliasing.The distribution difference between the empirical mode component and the original signal is analyzed and measured to improve the effect of signal reconstruction and de-noising,retain the main features of the signal,solve the problem of non-stationary signal reconstruction and enhancement,and improve the performance The signal-to-noise ratio is improved,and the de-noising of monitoring signal and the enhancement of data quality in actual environment are realized.(2)A disturbance source localization algorithm based on amnbp is proposed.Aiming at the problem of poor accuracy and low real-time of event location under multi factor coupling,the projection law of monitoring event in sensor field data is used to reduce the dimension and energy density of signal by using neighborhood binary mode,and the gain of coding result is adjusted by no event signal,which realizes the fast location of monitoring event in time and space,and reduces the large amount and quality of monitoring data The influence of the magnitude difference improves the efficiency of event location,overcomes the influence of noise interference on location accuracy,and solves the problem of accurate and efficient location of weak signal in actual monitoring.(3)A disturbance source identification algorithm based on hf-wrf is proposed.Aiming at the problem that single feature is difficult to accurately represent the key information of urban underground pipeline leakage and the recognition rate is low in the actual environment,the relationship between multi-dimensional feature extraction methods such as time domain,frequency domain and statistics and the characterization of disturbance source is studied.The peak to average ratio,short time zero crossing rate and average amplitude difference mixed feature combination are selected by correlation analysis and wrapping method to describe urban underground pipeline leakage Through feature importance analysis,the weight coefficient of decision tree is adjusted to reduce the influence of various interference factors on the final recognition,and effectively improve the accuracy of pipeline leakage event recognition under multi factor coupling.