A method and system for identifying abnormal disturbance of optical cable

Abstract

The present invention discloses a method and system for identifying abnormal disturbances in optical cables, the identification method comprising: collecting the backscattered Rayleigh scattering beat frequency signal of the optical cable; segmenting the backscattered Rayleigh scattering beat frequency signal, and processing it to obtain an amplitude grayscale image and a phase differential time domain signal; based on a pre-constructed recognition model, sequentially identifying the amplitude grayscale image and the phase differential time domain signal to be identified, and obtaining a disturbance identification result. The present invention can solve the problem of a large amount of data caused by a long sensor line in an actual optical cable monitoring scenario, reduce unnecessary data calculation and processing, and greatly improve the recognition efficiency and real-time performance.

Description

Technical Field

[0001] The present invention relates to the field of optical fiber sensing technology, and specifically to a method and system for identifying abnormal disturbance of an optical cable. Background Art

[0002] As a basic social industry, optical cables are widely used for signal transmission in various fields. However, due to their long-term outdoor deployment, they are easily damaged by environmental disturbances. At present, there is still a lack of intelligent and efficient monitoring methods for optical cable operation and maintenance, which leads to frequent accidents and huge economic losses. Phase-sensitive Optical Time Domain Reflectometry (Φ-OTDR), as a distributed fiber optic sensing disturbance detection technology, has the advantages of long monitoring distance, high spatial resolution and simple system composition. It is used to replace manual inspections in various safety monitoring fields. It is a suitable monitoring method for the operation status of optical cables.

[0003] The existing method for identifying disturbance events based on the Φ-OTDR distributed sensing system is usually to process the backscattered Rayleigh scattering amplitude signal on the entire optical fiber, which has a huge amount of data, low processing efficiency, and long-distance noise superposition seriously affects the effect of the algorithm model, resulting in a high false alarm rate. For example, CN116242470A discloses a method and system for monitoring vibration of power optical cables based on Φ-OTDR, characterized in that the method includes the following steps: Step 1, constructing a distributed optical fiber sensing detection system based on Φ-OTDR, and after applying external disturbance, analyzing and collecting the backscattered Rayleigh signal generated by the optical pulse input into the optical fiber under test; Step 2, pixel-level sampling of the collected backscattered Rayleigh signal, and based on the sampling results of the pixel-level sampling, respectively implementing horizontal sampling aggregation and vertical sampling aggregation to obtain horizontal aggregation results and vertical aggregation results of the sampling results; Step 3, respectively performing differential operations on adjacent horizontal aggregation results and adjacent vertical aggregation results, thereby realizing the detection of external disturbances of the optical fiber under test.

[0004] Therefore, how to improve the disturbance identification efficiency of Φ-OTDR is an urgent problem to be solved in practical applications. Summary of the invention

[0005] In view of the defects in the above-mentioned prior art, the present invention provides a method and system for identifying abnormal disturbances in optical cables, which can solve the problem of huge data volume caused by long sensor lines in actual optical cable monitoring scenarios, reduce unnecessary data calculation and processing, and greatly improve identification efficiency and real-time performance.

[0006] In a first aspect, the present invention provides a method for identifying abnormal disturbance of an optical cable, comprising:

[0007] Collect the backscattered Rayleigh frequency signal of the optical cable;

[0008] The backscattered Rayleigh scattering beat frequency signal is segmented and processed to obtain an amplitude grayscale image and a phase difference time domain signal;

[0009] Based on the pre-built recognition model, the amplitude grayscale image and phase difference time domain signal to be identified are identified in turn to obtain the disturbance recognition result.

[0010] Furthermore, the backscattered Rayleigh scattering beat frequency signal of the optical cable is collected, including:

[0011] The Φ-OTDR distributed optical fiber sensing system injects light pulses into the optical cable and collects the backscattered Rayleigh scattering beat frequency signal in the optical cable.

[0012] Furthermore, the pre-construction of the recognition model includes:

[0013] Obtain the historical backscattered Rayleigh scattering beat frequency signals corresponding to various disturbance types of the optical cable;

[0014] Divide the optical cable to obtain multiple sensing areas;

[0015] According to the multiple sensing regions, the historical Rayleigh backscattering beat signal is divided into the corresponding multiple historical Rayleigh backscattering beat sub-signals;

[0016] Preprocess each historical backscattered Rayleigh scattering beat sub-signal to obtain the corresponding historical amplitude grayscale image and historical differential phase time domain signal;

[0017] The historical amplitude grayscale image and the historical differential phase time domain signal are respectively labeled with disturbance labels and disturbance type labels;

[0018] The recognition model is obtained by training the convolutional neural network model with the labeled historical amplitude grayscale image and the historical differential phase time domain signal.

[0019] Furthermore, the optical cable is divided into multiple sensing areas, including:

[0020] The optical cable for collecting the historical backscattered Rayleigh beat frequency signal is segmented according to the preset minimum length threshold to obtain multiple sensing areas, including:

[0021] The optical cable for collecting historical backscattered Rayleigh beat signals is divided into a plurality of sensing areas, each of which has a length greater than a preset minimum length threshold.

[0022] The preset minimum length threshold is the sum of the theoretical spatial resolution of the Φ-OTDR distributed optical fiber sensing system and the optical fiber length of the disturbance effect.

[0023] Furthermore, the historical backscattered Rayleigh scattering beat frequency signals corresponding to various disturbance types of the optical cable are obtained, including:

[0024] Injecting optical pulses of predetermined parameters into optical cables of different disturbance types;

[0025] The historical backscattered Rayleigh scattering beat frequency signals in the optical cable with different disturbance types were collected multiple times.

[0026] Furthermore, the convolutional neural network model is trained by the labeled historical amplitude grayscale image and the historical differential phase time domain signal to obtain a recognition model, including:

[0027] The historical amplitude grayscale image and the historical differential phase time domain signal in the same historical backscattered Rayleigh scattering beat signal are divided into one group;

[0028] Divide all groups of historical amplitude grayscale images and historical differential phase time domain signals into training sets and test sets according to a predetermined ratio;

[0029] The first recognition model is obtained by training a two-dimensional convolutional network model through the historical amplitude grayscale image in the training set, and the second recognition model is obtained by training a one-dimensional convolutional network model through the historical differential phase time domain signal in the training set;

[0030] The first recognition model and the second recognition model are evaluated respectively by using the historical amplitude grayscale image and the historical differential phase time domain signal in the test set;

[0031] The first recognition model and the second recognition model that have passed the evaluation are combined to obtain a recognition model.

[0032] Furthermore, each historical backscattered Rayleigh scattering beat sub-signal is preprocessed to obtain the corresponding historical amplitude grayscale image and historical differential phase time domain signal, including:

[0033] The orthogonal demodulation algorithm is used to demodulate each historical backscattered Rayleigh scattering beat sub-signal to obtain the historical amplitude signal and the historical phase signal.

[0034] The historical amplitude signal is converted into a historical amplitude grayscale image, and the historical phase signal is processed to obtain a historical phase differential time domain signal.

[0035] Furthermore, after obtaining the historical amplitude signal and the historical phase signal, the method further includes:

[0036] Screen and remove historical amplitude signals and historical phase signals;

[0037] Among them, the historical amplitude signals are screened and eliminated, including:

[0038] Each historical amplitude signal is processed by forward difference to obtain a historical amplitude differential signal; specifically, the following steps are included:

[0039] Perform SG (Savitzky-Golay) smoothing filtering on the historical amplitude difference signal to obtain a smoothed historical amplitude difference signal;

[0040] The smoothed historical amplitude difference signal is used as the final historical amplitude signal.

[0041] Furthermore, the historical amplitude signal without forward difference is a matrix composed of amplitude vectors corresponding to different time points, specifically:

[0042] E={E 0 ,E 1 ,...,E i ,...,E N-1}

[0043] Where E is the historical amplitude signal, N is the time point information corresponding to the historical amplitude signal, that is, the number of optical pulses sent, and E i is the historical amplitude signal at the i-th time point;

[0044] Each historical amplitude signal is processed by forward difference to obtain a historical amplitude difference signal, which specifically includes:

[0045] The amplitude vector at the next time point in the historical amplitude signal is subtracted from the amplitude vector at the previous time point to obtain the historical amplitude difference signal, which is:

[0046] E diff ={ΔE 0 ,ΔE 1 ,...,ΔE i ,...,ΔE N-2}

[0047] In the formula, E diff is the historical amplitude differential signal, ΔE i is the historical amplitude difference signal at the i-th time point;

[0048] Perform SG smoothing filtering on the historical amplitude difference signal to obtain a smoothed historical amplitude difference signal; specifically including:

[0049] The weighted average of the n points adjacent to each sampling point replaces the initial value of the point.

[0050] The weighted average value of the n points adjacent to each acquisition point replaces the initial value of the point. Specifically, the weighted average value of the historical amplitude differential signal of the n acquisition points adjacent to the corresponding acquisition point of the historical amplitude differential signal is calculated, and the calculated weighted average value is used as the smoothing result of the historical amplitude differential signal.

[0051] Since the number of undisturbed signal samples is sufficient and the undisturbed signal only contains some background environmental noise and has similar characteristic components, when screening samples to construct a data set, undisturbed signal samples with a number equivalent to that of disturbed event samples can be selected from the undisturbed signal. Specifically, the historical phase signals are screened and eliminated, including:

[0052] All undisturbed historical phase signals are randomly shuffled to form a screening set;

[0053] The same number of undisturbed historical phase signals as the number of disturbed historical phase signals are selected from the screening set.

[0054] Furthermore, the historical phase signal is processed to obtain a historical phase differential time domain signal, including:

[0055] The historical phase signals at both ends of each sensing area are differentiated according to the spatial dimension to obtain the historical phase differential time domain signal, which specifically includes:

[0056] Δφ t =φ t,i -φ′ t,i

[0057] In the formula, Δφ t is the historical phase difference time domain signal of the i-th sensing area at time t, φ t,i is the historical phase signal of the i-th sensing area at the starting position, φ′ t,i is the historical phase signal of the i-th sensing area at the end position.

[0058] Furthermore, based on the pre-built recognition model, the amplitude grayscale image and the phase difference time domain signal to be recognized are recognized in turn to obtain the disturbance recognition result, including:

[0059] According to the pre-built recognition model, the amplitude grayscale image is recognized to obtain the result of whether there is disturbance or not;

[0060] According to the pre-built recognition model, the phase difference time domain signal corresponding to the amplitude grayscale image with disturbance is identified to obtain the type of disturbance.

[0061] Furthermore, the backscattered Rayleigh scattering beat frequency signal is segmented and processed to obtain an amplitude grayscale image and a phase difference time domain signal, including:

[0062] Segmenting the optical cable for collecting the backscattered Rayleigh beat frequency signal according to a preset minimum length threshold;

[0063] The Rayleigh backscattering beat signal is divided into a plurality of Rayleigh backscattering beat sub-signals according to the segmentation result of the optical cable;

[0064] The orthogonal demodulation algorithm is used to demodulate each segment of the backscattered Rayleigh scattering beat sub-signal to obtain the amplitude signal and the phase signal;

[0065] The amplitude signal is converted into an amplitude grayscale image, and the phase signal is processed to obtain a phase difference time domain signal.

[0066] Furthermore, each segment of backscattered Rayleigh scattering beat frequency sub-signal is expressed as:

[0067]

[0068] Among them, P AC is the backscattered Rayleigh scattering beat sub-signal, E represents the amplitude information of the backscattered Rayleigh scattering beat sub-signal, represents the phase information of the backscattered Rayleigh scattering beat sub-signal, Δf IF represents the frequency shift introduced by the AOM;

[0069] The orthogonal demodulation algorithm is used to demodulate each segment of the backscattered Rayleigh scattering beat frequency sub-signal, including:

[0070] Each segment of Rayleigh backscattering beat frequency sub-signal is respectively combined with sin(2πΔf IF t) and cos(2πΔf IF t) After multiplication, the amplitude signal is the root sum of the squares of the two, and the phase signal is the inverse tangent of the two.

[0071] Furthermore, the amplitude signal is converted into an amplitude grayscale image, including:

[0072] The amplitude signal is normalized by the maximum and minimum values, including:

[0073]

[0074] In the formula, x nom is the normalized result of the maximum and minimum values, x max is the maximum value of the amplitude signal, x min is the minimum value of the amplitude signal, and x is the signal value of the amplitude signal;

[0075] Map the normalized signal value to the grayscale value range to obtain an amplitude grayscale image.

[0076] Furthermore, the amplitude signal obtained by demodulation is a two-dimensional signal matrix containing time domain information and space domain information;

[0077] Map the normalized signal value to the grayscale value range to obtain the amplitude grayscale image, including:

[0078] The amplitude signal value at each spatiotemporal position corresponds to a grayscale pixel to convert the amplitude signal into an amplitude grayscale image.

[0079] Furthermore, the phase signal is processed to obtain a phase difference time domain signal, including:

[0080] The phase signals at both ends of each optical cable segment are differentiated in the spatial dimension to obtain a phase differential time domain signal.

[0081] In a second aspect, a system for identifying abnormal disturbance of an optical cable is also provided. The system adopts the above-mentioned method for identifying abnormal disturbance of an optical cable, and comprises:

[0082] A signal acquisition module, which is used to collect the backscattered Rayleigh beat frequency signal of the optical cable;

[0083] A signal processing module, which is used to segment the backscattered Rayleigh scattering beat frequency signal and process it to obtain an amplitude grayscale image and a phase difference time domain signal;

[0084] The disturbance recognition module is used to sequentially recognize the amplitude grayscale image and the phase difference time domain signal to be recognized based on a pre-built recognition model to obtain a disturbance recognition result.

[0085] Furthermore, the signal acquisition module is also used to inject light pulses into the optical cable through the Φ-OTDR distributed optical fiber sensing system, and to collect the backscattered Rayleigh scattering beat frequency signal in the optical cable.

[0086] Furthermore, the signal processing module is also used for:

[0087] Segmenting the optical cable for collecting the backscattered Rayleigh beat frequency signal according to a preset minimum length threshold;

[0088] The Rayleigh backscattering beat signal is divided into a plurality of Rayleigh backscattering beat sub-signals according to the segmentation result of the optical cable;

[0089] The orthogonal demodulation algorithm is used to demodulate each segment of the backscattered Rayleigh scattering beat sub-signal to obtain the amplitude signal and the phase signal;

[0090] The amplitude signal is converted into an amplitude grayscale image, and the phase signal is processed to obtain a phase difference time domain signal.

[0091] Furthermore, the signal processing module is also used for:

[0092] The phase signals at both ends of each optical cable segment are differentiated in the spatial dimension to obtain a phase differential time domain signal.

[0093] Furthermore, the disturbance identification module is also used to:

[0094] According to the pre-built recognition model, the amplitude grayscale image is recognized to obtain the result of whether there is disturbance or not;

[0095] According to the pre-built recognition model, the phase difference time domain signal corresponding to the amplitude grayscale image with disturbance is identified to obtain the type of disturbance.

[0096] Furthermore, the recognition system also includes a model building module, which is used to:

[0097] Obtain the historical backscattered Rayleigh scattering beat frequency signals corresponding to various disturbance types of the optical cable;

[0098] Divide the optical cable to obtain multiple sensing areas;

[0099] According to the multiple sensing regions, the historical Rayleigh backscattering beat signal is divided into the corresponding multiple historical Rayleigh backscattering beat sub-signals;

[0100] Preprocess each historical backscattered Rayleigh scattering beat sub-signal to obtain the corresponding historical amplitude grayscale image and historical differential phase time domain signal;

[0101] The historical amplitude grayscale image and the historical differential phase time domain signal are respectively labeled with disturbance labels and disturbance type labels;

[0102] The recognition model is obtained by training the convolutional neural network model with the labeled historical amplitude grayscale image and the historical differential phase time domain signal.

[0103] Furthermore, the model building module is also used to:

[0104] Injecting optical pulses of predetermined parameters into optical cables of different disturbance types;

[0105] The historical backscattered Rayleigh scattering beat frequency signals in the optical cable with different disturbance types were collected multiple times.

[0106] Furthermore, the model building module is also used to:

[0107] The historical amplitude grayscale image and the historical differential phase time domain signal in the same historical backscattered Rayleigh scattering beat signal are divided into one group;

[0108] Dividing all groups of historical amplitude grayscale images and historical differential phase time domain signals into training sets and test sets according to a predetermined ratio;

[0109] The first recognition model is obtained by training a two-dimensional convolutional network model through the historical amplitude grayscale image in the training set, and the second recognition model is obtained by training a one-dimensional convolutional network model through the historical differential phase time domain signal in the training set;

[0110] The first recognition model and the second recognition model are evaluated respectively by using the historical amplitude grayscale image and the historical differential phase time domain signal in the test set;

[0111] The first recognition model and the second recognition model that have passed the evaluation are combined to obtain a recognition model.

[0112] Furthermore, the model building module is also used to:

[0113] The orthogonal demodulation algorithm is used to demodulate each historical backscattered Rayleigh scattering beat sub-signal to obtain the historical amplitude signal and the historical phase signal.

[0114] The historical amplitude signal is converted into a historical amplitude grayscale image, and the historical phase signal is processed to obtain a historical phase differential time domain signal.

[0115] The present invention provides a method and system for identifying abnormal disturbance of optical cable, which at least have the following beneficial effects:

[0116] (1) Due to the system characteristics of the Φ-OTDR distributed optical fiber sensing system, the influence of the environmental noise in front of the optical cable line on the phase signal at that position will always exist in the phase signal at the rear position, resulting in low accuracy of far-end signal recognition. The present invention divides the entire optical cable link into regions and differentiates the phase signal according to the spatial dimension in each sensing area, thereby reducing the influence of the front noise, effectively avoiding the high false alarm rate caused by noise superposition and sensing environment changes caused by long-distance monitoring, and improving the accuracy of disturbance event recognition and classification.

[0117] (2) Area division and range through optical cables Figure 2 The rough screening mechanism of classification screens out the area with the highest probability of disturbance events, and then uses the phase signal of the corresponding area to perform specific event identification. Compared with the traditional method that needs to identify the data of all spatial sampling points on the entire optical cable, in real-time applications, the present invention only needs to perform specific event identification on individual areas, and the amount of data is greatly reduced. This effectively solves the problem of huge data volume caused by long sensor lines in actual communication optical cable monitoring scenarios, reduces unnecessary data calculation and processing, and greatly improves the recognition efficiency and real-time performance of the Φ-OTDR system. BRIEF DESCRIPTION OF THE DRAWINGS

[0118] Figure 1 A flow chart of a method for identifying abnormal disturbance of an optical cable provided by the present invention;

[0119] Figure 2 A flowchart of pre-building a recognition model for an embodiment of the present invention;

[0120] Figure 3 A flowchart of obtaining a recognition model according to an embodiment of the present invention;

[0121] Figure 4 A flowchart of processing a backscattered Rayleigh beat signal according to an embodiment of the present invention;

[0122] Figure 5 A schematic diagram of an amplitude signal and a phase signal obtained by demodulation in a certain embodiment of the present invention;

[0123] Figure 6 A schematic diagram of an amplitude grayscale image and a phase time domain signal of an embodiment provided by the present invention;

[0124] Figure 7 A flow chart for classification and identification according to an embodiment of the present invention;

[0125] Figure 8 A schematic diagram of a system for identifying abnormal disturbance of an optical cable provided by the present invention. DETAILED DESCRIPTION

[0126] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0127] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings, and "multiple" generally includes at least two.

[0128] It should also be noted that the term "includes", "comprising" or any other variation thereof is intended to cover non-exclusive inclusion, so that a commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprising a ..." do not exclude the existence of other identical elements in the commodity or device including the elements.

[0129] like Figure 1 As shown, the present invention provides a method for identifying abnormal disturbance of an optical cable, comprising:

[0130] Collect the backscattered Rayleigh frequency signal of the optical cable;

[0131] The backscattered Rayleigh scattering beat frequency signal is segmented and processed to obtain an amplitude grayscale image and a phase difference time domain signal;

[0132] Based on the pre-built recognition model, the amplitude grayscale image and phase difference time domain signal to be identified are identified in turn to obtain the disturbance recognition result.

[0133] See also Figure 2 As shown, the pre-construction of the recognition model of the present invention may include:

[0134] Obtain the historical backscattered Rayleigh scattering beat frequency signals corresponding to various disturbance types of the optical cable;

[0135] Divide the optical cable to obtain multiple sensing areas;

[0136] According to the multiple sensing regions, the historical Rayleigh backscattering beat signal is divided into the corresponding multiple historical Rayleigh backscattering beat sub-signals;

[0137] Preprocess each historical backscattered Rayleigh scattering beat sub-signal to obtain the corresponding historical amplitude grayscale image and historical differential phase time domain signal;

[0138] The historical amplitude grayscale image and the historical differential phase time domain signal are respectively labeled with disturbance labels and disturbance type labels;

[0139] The recognition model is obtained by training the convolutional neural network model with the labeled historical amplitude grayscale image and the historical differential phase time domain signal.

[0140] Wherein, obtaining the historical backscattered Rayleigh scattering beat frequency signals corresponding to various disturbance types of the optical cable may include:

[0141] Injecting optical pulses of predetermined parameters into optical cables of different disturbance types;

[0142] The historical backscattered Rayleigh scattering beat frequency signals in the optical cable with different disturbance types were collected multiple times.

[0143] Specifically, a Φ-OTDR distributed optical fiber sensing system is used to collect disturbance signals of various optical cable events (various disturbance types of abnormal optical cable disturbances). Each type of event is collected multiple times, with a total of M groups. Each group of data is composed of back-scattered Rayleigh beat signals obtained after N light pulses are injected into the optical cable, and is stored as a two-dimensional matrix Data(i,j), where i represents a spatial sampling point, corresponding to the distance between the optical cable position where the light pulse is injected and the received back-scattered Rayleigh beat signal, and j represents a time sampling point, corresponding to the number of light pulse injections.

[0144] In actual application scenarios, the emission frequency of optical pulses is 1 kHz. Each time, 200 optical pulses are injected into the optical cable and the backscattered Rayleigh scattering beat frequency signals are collected, corresponding to 200 ms. A total of 27,000 points of the signal returned by each optical pulse are collected, corresponding to a spatial distance of 11 km for the optical cable, totaling 600 sample groups containing different disturbance types.

[0145] The present invention divides the optical cable to obtain multiple sensing areas, including:

[0146] The optical cable for collecting the historical backscattered Rayleigh beat frequency signal is segmented according to the preset minimum length threshold to obtain multiple sensing areas, including:

[0147] The optical cable for collecting historical backscattered Rayleigh beat signals is divided into a plurality of sensing areas, each of which has a length greater than a preset minimum length threshold.

[0148] The preset minimum length threshold is the sum of the theoretical spatial resolution of the Φ-OTDR distributed optical fiber sensing system and the optical fiber length of the disturbance effect.

[0149] In actual application scenarios, the entire optical cable is divided into 100 sections, each section contains 270 points of sampling data, and only the signals at both ends of each sensing area are demodulated for amplitude and phase, and the signals within the section are only demodulated for amplitude.

[0150] Preprocess each historical backscattered Rayleigh scattering beat sub-signal to obtain the corresponding historical amplitude grayscale image and historical differential phase time domain signal, including:

[0151] The orthogonal demodulation algorithm is used to demodulate each historical backscattered Rayleigh scattering beat sub-signal to obtain the historical amplitude signal and the historical phase signal;

[0152] The demodulated historical amplitude signal is screened and eliminated: forward difference and SG smoothing filtering are used to obtain historical amplitude signal data with more prominent disturbance edges; specifically:

[0153] The historical amplitude signal without forward difference is a matrix composed of amplitude vectors corresponding to different time points, specifically:

[0154] E={E 0 ,E 1 ,...,E i ,...,E N-1}

[0155] Where E is the historical amplitude signal, N is the time point information corresponding to the historical amplitude signal, that is, the number of optical pulses sent, and E i is the historical amplitude signal at the i-th time point;

[0156] Each historical amplitude signal is processed by forward difference to obtain a historical amplitude difference signal, which specifically includes:

[0157] The amplitude vector at the next time point in the historical amplitude signal is subtracted from the amplitude vector at the previous time point to obtain the historical amplitude difference signal, which is:

[0158] E diff ={ΔE 0 ,ΔE 1,...,ΔE i ,...,ΔE N-2}

[0159] In the formula, E diff is the historical amplitude differential signal, ΔE i is the historical amplitude difference signal at the i-th time point;

[0160] Perform SG smoothing filtering on the historical amplitude difference signal to obtain a smoothed historical amplitude difference signal;

[0161] The smoothed historical amplitude difference signal is used as the final historical amplitude signal;

[0162] The final historical amplitude signal obtained by demodulation is converted into a historical amplitude grayscale image.

[0163] The demodulated historical phase signal is screened and eliminated, including:

[0164] All undisturbed historical phase signals are randomly shuffled to form a screening set;

[0165] The same number of undisturbed historical phase signals as the number of disturbed historical phase signals are selected from the screening set.

[0166] After filtering and eliminating the historical phase signals, the historical phase signals at both ends of each sensing area are taken and differentiated according to the spatial dimension to obtain the historical phase differential time domain signal that is linearly related to the external disturbance. The historical phase differential signal of the i-th sensing area at time t is expressed as is the phase difference time domain signal of the i-th sensing area at time t, is the historical phase signal of the i-th sensing area at the starting position, is the historical phase signal of the i-th sensing area at the end position.

[0167] In the actual application scenario, forward difference and SG smoothing filtering are used to process each segment of historical amplitude signal, and 600 sample groups are obtained, each of which contains historical amplitude data and historical phase data of 100 regional segments. After differential operation on the demodulated historical phase signal, a total of 600 phase sample groups are obtained, each of which contains 100 segments of one-dimensional phase change signals of the sensing area, and the corresponding duration of each signal is 200ms. The historical amplitude signal is processed using maximum and minimum value normalization to obtain a historical amplitude grayscale image, and the pixel size of each historical amplitude grayscale image is 270*200.

[0168] The historical amplitude grayscale image and the historical differential phase time domain signal are respectively labeled with disturbance labels and disturbance type labels, including:

[0169] Define the identification label according to whether there is disturbance or not, set the historical amplitude grayscale image label with disturbance to 1, and the historical amplitude grayscale image label without disturbance to 0, and get the label vector y corresponding to the historical amplitude signal A ;

[0170] Define labels for historical phase difference time domain signals and set labels according to different disturbance types j represents the disturbance type of the i-th sensing area, and the label vector corresponding to the historical phase signal is obtained

[0171] See also Figure 3 As shown, the recognition model is obtained by training the convolutional neural network model with the labeled historical amplitude grayscale image and the historical differential phase time domain signal, including:

[0172] The historical amplitude grayscale image and the historical differential phase time domain signal in the same historical backscattered Rayleigh scattering beat signal are divided into one group;

[0173] Divide all groups of historical amplitude grayscale images and historical differential phase time domain signals into training sets and test sets according to a predetermined ratio;

[0174] The first recognition model is obtained by training a two-dimensional convolutional network model (2-DCNN model) using the historical amplitude grayscale image in the training set, and the second recognition model is obtained by training a one-dimensional convolutional network model (2-D CNN) using the historical differential phase time domain signal in the training set;

[0175] The first recognition model and the second recognition model are evaluated respectively by using the historical amplitude grayscale image and the historical differential phase time domain signal in the test set;

[0176] The first recognition model and the second recognition model that have passed the evaluation are combined to obtain a recognition model.

[0177] Specifically, all historical amplitude grayscale images, historical phase difference time domain signals and label vectors y A , A sample data set is constructed in one-to-one correspondence. After the samples are balanced, shuffled and reordered, they are divided into training set and test set in the ratio of 7:3. The part used for training is input into the convolutional neural network model for training.

[0178] In this example, since the total number of sensing areas of the optical cable is large and disturbance events only occur in a few of them, the number of non-disturbance signals will be much larger than the number of disturbed signals. At the same time, the characteristics of non-disturbance signals are mostly similar. When constructing the sample data set, some non-disturbance signals are purposefully eliminated (i.e., the above-mentioned screening and elimination process of historical amplitude signals and historical phase signals) to achieve sample balance and reduce the amount of data training.

[0179] When collecting the backscattered Rayleigh scattering beat frequency signal of the optical cable, the present invention may include:

[0180] The Φ-OTDR distributed optical fiber sensing system injects light pulses into the optical cable and collects the backscattered Rayleigh scattering beat frequency signal in the optical cable.

[0181] See also Figure 4 As shown, the backscattered Rayleigh scattering beat frequency signal is segmented and processed to obtain an amplitude grayscale image and a phase difference time domain signal, including:

[0182] Segmenting the optical cable for collecting the backscattered Rayleigh beat frequency signal according to a preset minimum length threshold;

[0183] The Rayleigh backscattering beat signal is divided into a plurality of Rayleigh backscattering beat sub-signals according to the segmentation result of the optical cable;

[0184] The orthogonal demodulation algorithm is used to demodulate each segment of the backscattered Rayleigh scattering beat sub-signal to obtain the amplitude signal and phase signal (such as Figure 5 shown);

[0185] The amplitude signal is converted into an amplitude grayscale image, and the phase signal is processed to obtain a phase difference time domain signal (such as Figure 6 shown).

[0186] Among them, each segment of backscattered Rayleigh scattering beat frequency sub-signal is expressed as:

[0187]

[0188] Among them, P AC is the backscattered Rayleigh scattering beat sub-signal, E represents the amplitude information of the backscattered Rayleigh scattering beat sub-signal, represents the phase information of the backscattered Rayleigh scattering beat sub-signal, Δf IF represents the frequency shift introduced by the AOM;

[0189] The orthogonal demodulation algorithm is used to demodulate each segment of the backscattered Rayleigh scattering beat frequency sub-signal, including:

[0190] Each segment of Rayleigh backscattering beat frequency sub-signal is respectively combined with sin(2πΔf IF t) and cos(2πΔf IF t) After multiplication, the amplitude signal is the root sum of the squares of the two, and the phase signal is the inverse tangent of the two.

[0191] Specifically, the backscattered Rayleigh scattering beat signal is The backscattered Rayleigh scattering beat signal collected by the Φ-OTDR distributed optical fiber sensing system through the data acquisition card is a time-discrete sequence with an acquisition rate, that is, one acquisition point corresponds to every x seconds, and a backscattered Rayleigh scattering beat signal is collected at this time. Each segment of the backscattered Rayleigh scattering beat signal is respectively related to sin(2πΔf IF t) and cos(2πΔf IF t) multiplication means: take the corresponding Respectively with sin(2πΔf IF t) and cos(2πΔf IF t). In actual calculation, the multiplication is similar to the multiplication of the corresponding elements of the matrix. The backscattered Rayleigh beat signal and the backscattered Rayleigh beat sub-signal are M*N matrices, and sin (or cos) is a 1*N matrix. The multiplication is to multiply the M rows of the beat signal with the corresponding elements of sin (or cos) respectively, and finally obtain an M*N multiplication result. Each segment of the backscattered Rayleigh beat sub-signal is multiplied by sin (2πΔf IF t) and cos(2πΔf IF t) and then pass through the low-pass filter: each segment of the Rayleigh backscattering beat frequency signal is respectively multiplied by sin(2πΔf IF t) and cos(2πΔf IF t) The multiplied sequence is filtered, that is, each row of the matrix.

[0192] The backscattered Rayleigh beat signal is a cosine function, which is related to the two orthogonal signals (sin(2πΔf IF t) and cos(2πΔf IF After multiplying t)), the result will contain high-frequency and low-frequency components. Since only the low-frequency component is needed, a low-frequency filter is used to filter out the high-frequency component.

[0193] Furthermore, the amplitude signal is converted into an amplitude grayscale image, including:

[0194] The amplitude signal is normalized by the maximum and minimum values, including:

[0195]

[0196] In the formula, x nom is the normalized result of the maximum and minimum values, x max is the maximum value of the amplitude signal, x min is the minimum value of the amplitude signal, and x is the signal value of the amplitude signal;

[0197] Map the normalized signal value to the grayscale value range to obtain an amplitude grayscale image.

[0198] In addition, the backscattered Rayleigh scattering beat frequency signal of the optical cable is collected, and the same collection method is used in the process of building the recognition model and the specific recognition processing process; in the process of building the recognition model, the demodulation process and the process of converting the amplitude signal into an amplitude grayscale image are the same as the demodulation process and conversion process in the actual recognition process. And in the process of identifying the actual abnormal disturbance of the optical cable, the process of processing the phase signal to obtain the phase differential time domain signal is the same as the processing process in the process of building the recognition model. For the same steps, the present invention will not be repeated here.

[0199] Based on the pre-built recognition model, the amplitude grayscale image and phase difference time domain signal to be identified are identified in turn to obtain the disturbance recognition results, including:

[0200] According to the pre-built recognition model, the amplitude grayscale image is recognized to obtain the result of whether there is disturbance or not;

[0201] According to the pre-built recognition model, the phase difference time domain signal corresponding to the amplitude grayscale image with disturbance is identified to obtain the type of disturbance.

[0202] In actual application scenarios, when performing recognition based on a pre-built recognition model, known labels are not input into the recognition model, that is, the recognition model is in an unknown state with respect to the input data, and the labels are only used as a reference for the final measurement of the recognition performance of the recognition model.

[0203] The specific recognition process is to first use the trained 2-D CNN model (the first recognition model) to determine whether the amplitude grayscale image is disturbed. If it is judged to be without disturbance, no further processing is performed. If it is judged to be disturbed, the phase difference time domain signal corresponding to the amplitude signal grayscale image (i.e., the same sensing area) is extracted from the data set and sent to the trained 1-D CNN model (the second recognition model) for classification and recognition (such as Figure 7 As shown), to obtain the disturbance category occurring in the sensing area.

[0204] The present invention only needs to perform recognition model training at the beginning, and save the recognition model parameters after the training is completed. In subsequent use, it is only necessary to use the Φ-OTDR distributed optical fiber sensing system to collect the back Rayleigh scattering beat frequency signal of the optical cable, and directly use the trained recognition model after processing to obtain the amplitude grayscale image and the phase difference time domain signal to realize classification and recognition, without adding operations such as setting labels, sample balancing and building data sets.

[0205] The present invention first divides the area, and performs rough screening of disturbances through amplitude binary classification and fine classification of events through phase classification, which achieves the purpose of reducing the amount of real-time data processing and improving real-time performance. At the same time, the processing method of dividing the optical cable into areas can also reduce the influence of noise superposition under long distances. If the division is not performed, the noise at the front end of the optical cable line will always exist in the signal at the rear position. After the division, it is equivalent to independent processing of each area.

[0206] See also Figure 8 As shown, the present invention also provides a system for identifying abnormal disturbance of optical cable, adopting the above-mentioned method for identifying abnormal disturbance of optical cable, the identification system comprises:

[0207] A signal acquisition module, which is used to collect the backscattered Rayleigh beat frequency signal of the optical cable;

[0208] A signal processing module, which is used to segment the backscattered Rayleigh scattering beat frequency signal and process it to obtain an amplitude grayscale image and a phase difference time domain signal;

[0209] The disturbance recognition module is used to sequentially recognize the amplitude grayscale image and the phase difference time domain signal to be recognized based on a pre-built recognition model to obtain a disturbance recognition result.

[0210] The signal acquisition module is also used to inject optical pulses into the optical cable through the Φ-OTDR distributed optical fiber sensing system and to collect the backscattered Rayleigh scattering beat frequency signal in the optical cable.

[0211] The signal processing module is also used to:

[0212] Segmenting the optical cable for collecting the backscattered Rayleigh beat frequency signal according to a preset minimum length threshold;

[0213] The Rayleigh backscattering beat signal is divided into a plurality of Rayleigh backscattering beat sub-signals according to the segmentation result of the optical cable;

[0214] The orthogonal demodulation algorithm is used to demodulate each segment of the backscattered Rayleigh scattering beat sub-signal to obtain the amplitude signal and the phase signal;

[0215] The amplitude signal is converted into an amplitude grayscale image, and the phase signal is processed to obtain a phase difference time domain signal.

[0216] The signal processing module is also used to:

[0217] The phase signals at both ends of each optical cable segment are differentiated in the spatial dimension to obtain a phase differential time domain signal.

[0218] The disturbance identification module is also used to:

[0219] According to the pre-built recognition model, the amplitude grayscale image is recognized to obtain the result of whether there is disturbance or not;

[0220] According to the pre-built recognition model, the phase difference time domain signal corresponding to the amplitude grayscale image with disturbance is identified to obtain the type of disturbance.

[0221] The recognition system also includes a model building module, which is used to:

[0222] Obtain the historical backscattered Rayleigh scattering beat frequency signals corresponding to various disturbance types of the optical cable;

[0223] Divide the optical cable to obtain multiple sensing areas;

[0224] According to the multiple sensing regions, the historical Rayleigh backscattering beat signal is divided into the corresponding multiple historical Rayleigh backscattering beat sub-signals;

[0225] Preprocess each historical backscattered Rayleigh scattering beat sub-signal to obtain the corresponding historical amplitude grayscale image and historical differential phase time domain signal;

[0226] The historical amplitude grayscale image and the historical differential phase time domain signal are respectively labeled with disturbance labels and disturbance type labels;

[0227] The recognition model is obtained by training the convolutional neural network model with the labeled historical amplitude grayscale image and the historical differential phase time domain signal.

[0228] The model building module is also used to:

[0229] Injecting optical pulses of predetermined parameters into optical cables of different disturbance types;

[0230] The historical backscattered Rayleigh scattering beat frequency signals in the optical cable with different disturbance types were collected multiple times.

[0231] The model building module is also used to:

[0232] The historical amplitude grayscale image and the historical differential phase time domain signal in the same historical backscattered Rayleigh scattering beat signal are divided into one group;

[0233] Divide all groups of historical amplitude grayscale images and historical differential phase time domain signals into training sets and test sets according to a predetermined ratio;

[0234] The first recognition model is obtained by training a two-dimensional convolutional network model through the historical amplitude grayscale image in the training set, and the second recognition model is obtained by training a one-dimensional convolutional network model through the historical differential phase time domain signal in the training set;

[0235] The first recognition model and the second recognition model are evaluated respectively by using the historical amplitude grayscale image and the historical differential phase time domain signal in the test set;

[0236] The first recognition model and the second recognition model that have passed the evaluation are combined to obtain a recognition model.

[0237] Furthermore, the model building module is also used to:

[0238] The orthogonal demodulation algorithm is used to demodulate each historical backscattered Rayleigh scattering beat sub-signal to obtain the historical amplitude signal and the historical phase signal.

[0239] The historical amplitude signal is converted into a historical amplitude grayscale image, and the historical phase signal is processed to obtain a historical phase differential time domain signal.

[0240] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic inventive concepts are known. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for identifying abnormal disturbance of an optical cable, characterized in that: include: Collect the backscattered Rayleigh frequency signal of the optical cable; The backscattered Rayleigh scattering beat frequency signal is segmented and processed to obtain an amplitude grayscale image and a phase difference time domain signal; Based on the pre-built recognition model, the amplitude grayscale image and phase difference time domain signal to be recognized are recognized in turn to obtain the disturbance recognition result; Among them, the pre-construction of the recognition model includes: Obtain the historical backscattered Rayleigh scattering beat frequency signals corresponding to various disturbance types of the optical cable; Divide the optical cable to obtain multiple sensing areas; According to the multiple sensing regions, the historical Rayleigh backscattering beat signal is divided into the corresponding multiple historical Rayleigh backscattering beat sub-signals; Preprocess each historical backscattered Rayleigh scattering beat sub-signal to obtain the corresponding historical amplitude grayscale image and historical differential phase time domain signal; The historical amplitude grayscale image and the historical differential phase time domain signal are respectively labeled with disturbance labels and disturbance type labels; The recognition model is obtained by training a convolutional neural network model with labeled historical amplitude grayscale images and historical differential phase time domain signals, including: dividing the historical amplitude grayscale images and historical differential phase time domain signals in the same historical back-scattered Rayleigh scattering beat signal into a group; dividing all groups of historical amplitude grayscale images and historical differential phase time domain signals into training sets and test sets according to a predetermined ratio; training a two-dimensional convolutional network model with the historical amplitude grayscale images in the training set to obtain a first recognition model, and training a one-dimensional convolutional network model with the historical differential phase time domain signals in the training set to obtain a second recognition model; respectively evaluating the first recognition model and the second recognition model with the historical amplitude grayscale images and the historical differential phase time domain signals in the test set; and obtaining the recognition model by combining the first recognition model and the second recognition model that have passed the evaluation.

2. The identification method according to claim 1, characterized in that: Collect the backscattered Rayleigh beat frequency signal of the optical cable, including: The Φ-OTDR distributed optical fiber sensing system injects light pulses into the optical cable and collects the backscattered Rayleigh scattering beat frequency signal in the optical cable.

3. The identification method according to claim 1, characterized in that: Obtain the historical backscattered Rayleigh scattering beat frequency signals corresponding to various disturbance types of the optical cable, including: Injecting optical pulses of predetermined parameters into optical cables of different disturbance types; The historical backscattered Rayleigh scattering beat frequency signals in the optical cable with different disturbance types were collected multiple times.

4. The identification method according to claim 1, characterized in that: Preprocess each historical backscattered Rayleigh scattering beat sub-signal to obtain the corresponding historical amplitude grayscale image and historical differential phase time domain signal, including: The orthogonal demodulation algorithm is used to demodulate each historical backscattered Rayleigh scattering beat sub-signal to obtain the historical amplitude signal and the historical phase signal. The historical amplitude signal is converted into a historical amplitude grayscale image, and the historical phase signal is processed to obtain a historical phase differential time domain signal.

5. The identification method according to claim 4, characterized in that: After obtaining the historical amplitude signal and the historical phase signal, it also includes: Screen and remove historical amplitude signals and historical phase signals; Among them, the historical amplitude signals are screened and eliminated, including: Each historical amplitude signal is processed by forward difference to obtain a historical amplitude differential signal; Perform SG smoothing filtering on the historical amplitude difference signal to obtain a smoothed historical amplitude difference signal; The smoothed historical amplitude difference signal is used as the final historical amplitude signal; Filter and remove historical phase signals, including: All undisturbed historical phase signals are randomly shuffled to form a screening set; The same number of undisturbed historical phase signals as the number of disturbed historical phase signals are selected from the screening set.

6. The identification method according to claim 1, characterized in that: The backscattered Rayleigh scattering beat frequency signal is segmented and processed to obtain the amplitude grayscale image and the phase difference time domain signal, including: Segmenting the optical cable for collecting the backscattered Rayleigh beat frequency signal according to a preset minimum length threshold; The Rayleigh backscattering beat signal is divided into a plurality of Rayleigh backscattering beat sub-signals according to the segmentation result of the optical cable; The orthogonal demodulation algorithm is used to demodulate each segment of the backscattered Rayleigh scattering beat sub-signal to obtain the amplitude signal and the phase signal; The amplitude signal is converted into an amplitude grayscale image, and the phase signal is processed to obtain a phase difference time domain signal.

7. The identification method according to claim 6, characterized in that: Based on the pre-built recognition model, the amplitude grayscale image and phase difference time domain signal to be identified are identified in turn to obtain the disturbance recognition results, including: According to the pre-built recognition model, the amplitude grayscale image is recognized to obtain the result of whether there is disturbance or not; According to the pre-built recognition model, the phase difference time domain signal corresponding to the amplitude grayscale image with disturbance is identified to obtain the type of disturbance.

8. A system for identifying abnormal disturbance of optical cable, using the method for identifying abnormal disturbance of optical cable according to any one of claims 1 to 7, characterized in that: The identification system comprises: A signal acquisition module, which is used to collect the backscattered Rayleigh beat frequency signal of the optical cable; A signal processing module, which is used to segment the backscattered Rayleigh scattering beat frequency signal and process it to obtain an amplitude grayscale image and a phase difference time domain signal; The disturbance recognition module is used to sequentially recognize the amplitude grayscale image and the phase difference time domain signal to be recognized based on a pre-built recognition model to obtain a disturbance recognition result.

Images