Railway traction return current monitoring and analysis method and system based on multi-path current fusion

Abstract

The application relates to the technical field of traction return current monitoring, and provides a railway traction return current monitoring and analyzing method and system based on multi-path current fusion. The method comprises the following steps: generating multi-path current waveform data based on the mapping relationship between phase variation and traction return current intensity, performing time-frequency domain joint noise reduction, aligning with a time axis, and obtaining aligned multi-path current waveform data; performing cross-path cooperativity analysis, extracting the synchronization deviation characteristics of the multi-path current, and matching and verifying the preset traction load dynamic change interval, so that abnormal current events caused by harmonic distortion or transient impact and the positioning information of the abnormal current events in the longitudinal direction of the steel rail are determined. The application can realize high-precision detection of railway traction return current, positioning and identification of harmonics and transient impact, and solves the problem of low-precision detection of railway traction return current caused by weak anti-interference ability, insufficient spatial resolution and the absence of multi-path current cooperativity.

Description

Technical Field

[0001] The present application relates to the technical field of traction return current monitoring, and in particular to a railway traction return current monitoring and analysis method and system based on multi-path current fusion. Background Art

[0002] In railway traction power supply systems, abnormal current events in the traction return current can cause abnormal increases in rail potential, potentially causing equipment damage or safety hazards. To ensure safe railway operations, continuous monitoring of the longitudinal rail current path is required to detect even the smallest abnormal signals.

[0003] The mainstream anomaly detection solution for railway traction backflow is axlebox acceleration monitoring technology. This technology installs accelerometers in the train's axlebox to collect vibration signals from wheel-rail contact. Using machine learning algorithms, it analyzes the vibration spectrum and indirectly infers abnormal current events. The core logic uses the correlation between vibration signals and current distortion to establish a threshold model and trigger an anomaly alarm.

[0004] However, in the existing scheme, the mechanical vibration noise during train operation will seriously contaminate the vibration signal, resulting in an increased false alarm rate; it can only achieve meter-level positioning accuracy and cannot meet the precise positioning requirements of sub-meter-level anomaly sources; single-point vibration signals are difficult to characterize the dynamic correlation characteristics of the longitudinal multi-path current of the rail, and cannot effectively extract the synchronization deviation characteristics. Summary of the Invention

[0005] The present application provides a railway traction return current monitoring and analysis method and system based on multi-path current fusion, which is used to solve the problem of low-precision detection of railway traction return current caused by weak anti-interference ability, insufficient spatial resolution and lack of multi-path current coordination in the existing technology, and realizes high-precision detection, positioning and identification of harmonics and transient shocks of railway traction return current.

[0006] In a first aspect, the present application provides a railway traction return current monitoring and analysis method based on multi-path current fusion, comprising:

[0007] The distributed optical fiber sensing devices arranged at intervals along the longitudinal direction of the rail are used to detect the phase change of the optical signal, and multi-path current waveform data is generated based on the mapping relationship between the phase change and the traction return current intensity;

[0008] Performing a time-frequency domain joint noise reduction process on the multi-path current waveform data, aligning the processed multi-path current waveform data with a time axis, and obtaining time-axis aligned multi-path current waveform data;

[0009] Performing cross-path synergy analysis on the time-axis aligned multi-path current waveform data using a dynamic correlation model of multi-path current to extract synchronization deviation characteristics of the multi-path current, wherein the dynamic correlation model is constructed based on spatial gradient constraints on rail longitudinal current propagation;

[0010] By matching and verifying the synchronization deviation characteristics with a preset dynamic change interval of the traction load, the abnormal current event caused by the harmonic distortion or transient impact and the positioning information of the abnormal current event in the longitudinal direction of the rail are determined.

[0011] Optionally, performing cross-path synergy analysis on the time-axis aligned multi-path current waveform data using a dynamic correlation model of the multi-path current to extract synchronization deviation features of the multi-path current includes:

[0012] The multi-path current waveform data after time axis alignment is divided into multiple analysis windows according to a fixed time length, and the peak point and corresponding time stamp of each path current waveform are extracted in each analysis window;

[0013] Within each analysis window, based on the peak points and corresponding timestamps of the current waveforms of each path, bidirectional timing matching is performed on the peak point timestamps of adjacent paths. If the forward time difference and the reverse time difference in the bidirectional timing matching result are equal, the forward time difference is used as the timestamp difference of the peak points between the adjacent paths.

[0014] Based on the timestamp difference, combined with the rail longitudinal current propagation velocity range and rail longitudinal spacing parameters preset in the dynamic association model, an actual current propagation velocity is generated;

[0015] The actual current propagation speed exceeding the rail longitudinal current propagation speed range, the corresponding path identifier and the amplitude of the corresponding peak point are stored in a cross-path abnormal time offset set;

[0016] Generating a dynamic impact weight corresponding to each path according to the abnormal frequency of each path in the cross-path abnormal time offset set and the amplitude of the corresponding peak point;

[0017] A collaborative analysis is performed based on the dynamic impact weights corresponding to the paths to generate synchronization deviation features.

[0018] Optionally, performing collaborative analysis based on the dynamic impact weights corresponding to the paths to generate synchronization deviation features includes:

[0019] Performing weighted superposition on the multi-path current waveform data using the dynamic impact weights corresponding to all the paths to generate cross-path synergy reference waveform data;

[0020] Calculating the phase difference and amplitude deviation between the multipath current waveform and the cross-path cooperativity reference waveform point by point based on the cross-path cooperativity reference waveform data and the time-axis aligned multipath current waveform data to form a synchronization deviation measurement sequence;

[0021] The analysis window is slid along the time axis, and the synchronization deviation measurement sequence is accumulated and summed. When the accumulated value exceeds the distortion threshold preset in the dynamic association model, a synchronization deviation feature is generated, and the synchronization deviation feature includes a path identifier with the largest amplitude deviation.

[0022] Optionally, performing time-frequency domain joint noise reduction processing on the multi-path current waveform data, aligning the processed multi-path current waveform data with a time axis, and obtaining time-axis aligned multi-path current waveform data, includes:

[0023] Decomposing the multi-path current waveform data into frequency band component signals at multiple levels by a multi-scale decomposition method;

[0024] For the frequency band component signals at each level, based on the statistical distribution difference of the frequency band component signals at adjacent levels, calculating a dynamic noise threshold of the current level;

[0025] Filtering frequency band component signals whose amplitudes are lower than the dynamic noise threshold of the corresponding level, and cross-level fusing the filtered frequency band component signals of all levels to generate processed multi-path current waveform data;

[0026] Extracting a phase principal component sequence of each path from the processed multi-path current waveform data, and calculating a relative time offset of the current waveform data of each path relative to a time axis reference time based on a starting phase mutation point of the phase principal component sequence of each path;

[0027] Based on the relative time offset, the current waveform data of each path is subjected to reverse time shift correction so that the starting point of the phase principal component sequence of the corrected current waveform data of each path is aligned with the time axis reference time, thereby obtaining the multi-path current waveform data after time axis alignment.

[0028] Optionally, the calculating, based on the starting phase mutation point of the phase principal component sequence of each path, the relative time offset of the current waveform data of each path relative to the time axis reference moment includes:

[0029] performing differential processing on the phase principal component sequences of each path to generate corresponding phase difference sequences representing phase changes;

[0030] In the phase difference sequence, detecting a phase mutation interval that exceeds a dynamic mutation threshold for N consecutive times, where the dynamic mutation threshold is calculated by calculating the statistical variance of the phase principal component sequence in a non-phase mutation interval, and N is greater than or equal to 2;

[0031] Marking the moment when the phase mutation threshold is exceeded for the first time within the phase mutation interval as the starting phase mutation point;

[0032] Taking the time axis reference time as the reference zero point, the relative time offset between the starting phase mutation point corresponding to each path current waveform data and the reference zero point is calculated.

[0033] Optionally, the calculating the dynamic noise threshold of the current level based on the statistical distribution difference of the frequency band component signals of adjacent levels includes:

[0034] Constructing a probability density distribution function of the current layer and a probability density distribution function of the adjacent layer for the frequency band component signal of the current layer and the frequency band component signal of the adjacent layer respectively;

[0035] Calculating the cumulative probability difference between the probability density distribution function of the current level and the probability density distribution function of the adjacent level within a preset amplitude range, and using the cumulative probability difference as an indicator of the statistical distribution difference between the levels;

[0036] Constructing a dynamic noise threshold calculation function according to the statistical distribution difference index and the energy proportion of the frequency band component signal at the current level;

[0037] Inputting the amplitude median of the frequency band component signal of the current layer into the dynamic noise threshold calculation function, and outputting the dynamic noise threshold of the current layer;

[0038] According to the amplitude extreme difference of the frequency band component signals of adjacent levels, a cross-level smoothing constraint is performed on the dynamic noise threshold to generate a dynamic noise threshold of the current level.

[0039] Optionally, the matching verification between the synchronization deviation feature and a preset dynamic change interval of the traction load to determine the abnormal current event caused by the harmonic distortion or transient impact and the positioning information of the abnormal current event in the longitudinal direction of the rail includes:

[0040] In combination with the preset rail longitudinal current propagation velocity range and rail longitudinal spacing parameter in the dynamic correlation model, the equivalent time delay between the current propagation direction corresponding to the path with the maximum amplitude deviation in the synchronization deviation feature and the adjacent path is calculated, where N is greater than or equal to 2;

[0041] Matching the amplitude deviation of the maximum amplitude deviation path with the upper and lower bounds of the dynamic change interval of the traction load point by point;

[0042] If the amplitude deviation exceeds the upper bound N times in a row and the phase difference fluctuates periodically, a harmonic distortion determination index is generated; if the amplitude deviation is lower than the lower bound and the equivalent time delay exceeds the preset transient propagation delay threshold, a transient impact determination index is generated;

[0043] Based on the harmonic distortion determination index or the transient impulse determination index, the equivalent propagation path length of the abnormal current event between adjacent paths is calculated, and an initial positioning interval is generated according to the longitudinal coordinate of the path identifier;

[0044] Performing weighted correction on the initial positioning interval to generate an optimized positioning interval in the longitudinal direction of the rail, wherein the boundary value of the optimized positioning interval is dynamically adjusted by multiplying the equivalent propagation path length by a dynamic influence weight;

[0045] The abnormality confidence of the optimized positioning interval is calculated, and the reference coordinates of the abnormal current event in the longitudinal direction of the rail are calibrated based on the path identifier of the path with the maximum amplitude deviation, so as to obtain positioning information including the abnormality type, optimized positioning interval, abnormality confidence and reference coordinates.

[0046] In a second aspect, the present application provides a railway traction return current monitoring and analysis system based on multi-path current fusion, comprising:

[0047] a generation module configured to detect a phase change of an optical signal through distributed optical fiber sensing devices spaced apart along the longitudinal direction of the rail, and to generate multipath current waveform data based on a mapping relationship between the phase change and the traction return current intensity;

[0048] a processing module, configured to perform time-frequency domain joint noise reduction processing on the multi-path current waveform data, align the processed multi-path current waveform data with a time axis, and obtain time-axis aligned multi-path current waveform data;

[0049] an extraction module that performs cross-path synergy analysis on the time-axis aligned multi-path current waveform data using a dynamic correlation model of multi-path current, and extracts synchronization deviation characteristics of the multi-path current, wherein the dynamic correlation model is constructed based on spatial gradient constraints on the longitudinal current propagation of the rail;

[0050] The verification module is used to match and verify the synchronization deviation characteristics with the preset dynamic change interval of the traction load to determine the abnormal current event caused by harmonic distortion or transient impact and the positioning information of the abnormal current event in the longitudinal direction of the rail.

[0051] In a third aspect, the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a railway traction return current monitoring and analysis method based on multi-path current fusion as described in any one of the first aspects.

[0052] In a fourth aspect, the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements a railway traction return current monitoring and analysis method based on multi-path current fusion as described in any one of the first aspects.

[0053] In an embodiment of the present application, a railway traction return current monitoring and analysis method based on multi-path current fusion is provided, the method comprising: detecting the phase change of the optical signal through distributed optical fiber sensing devices arranged at intervals along the longitudinal direction of the rail, and generating multi-path current waveform data based on the mapping relationship between the phase change and the traction return current intensity; performing time-frequency domain joint noise reduction processing on the multi-path current waveform data, aligning the processed multi-path current waveform data with the time axis, and obtaining time-axis aligned multi-path current waveform data; performing cross-path synergy analysis on the time-axis aligned multi-path current waveform data through a dynamic correlation model of the multi-path current, and extracting the synchronization deviation characteristics of the multi-path current, the dynamic correlation model being constructed based on the spatial gradient constraint conditions of the longitudinal current propagation of the rail; determining the abnormal current events caused by harmonic distortion or transient shocks and the positioning information of the abnormal current events in the longitudinal direction of the rail by matching and verifying the synchronization deviation characteristics with the preset dynamic change interval of the traction load.

[0054] This application is based on the mapping relationship between optical fiber phase change and current intensity to achieve real-time capture of small anomalies in traction return current, with a sensitivity of sub-ampere level; through joint noise reduction and dynamic correlation model in time-frequency domain, it effectively suppresses noise interference, extracts the synchronization deviation characteristics of multi-path current, and accurately distinguishes harmonic distortion from transient impact events; combined with cross-path synergy analysis under spatial gradient constraints, it achieves accurate positioning of abnormal current events in the longitudinal direction of the rail, reducing positioning errors. Furthermore, based on the multi-path current waveform data aligned with the time axis, the peak point timestamps are extracted by dividing the analysis window, and bidirectional timing matching of adjacent paths is performed to calculate the timestamp difference. The actual propagation speed is generated by combining the preset current propagation speed range and spacing parameters, abnormal speed is screened, and a set of cross-path abnormal time offsets is constructed; further, based on the dynamic impact weight, a cross-path synergy reference waveform is generated, and the phase difference and amplitude deviation are calculated point by point to form a synchronization deviation measurement sequence. The threshold is determined by cumulative summation, and the synchronization deviation characteristics of the path with the maximum amplitude deviation are finally output. Through bidirectional timing matching and dynamic weight collaborative analysis, the detection accuracy of multi-path current anomaly time offset is improved, and mechanical vibration noise and real current distortion events are effectively distinguished. Combined with the cumulative threshold judgment mechanism, sub-meter positioning accuracy is achieved. At the same time, the fault source is quantitatively locked based on the path amplitude deviation, which enhances the classification ability of harmonic distortion and transient shock and reduces the false alarm rate.

[0055] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0057] Figure 1 A flow chart of a railway traction return current monitoring and analysis method based on multi-path current fusion provided in an embodiment of the present application;

[0058] Figure 2 A schematic diagram of the structure of a railway traction return current monitoring and analysis system based on multi-path current fusion provided in an embodiment of the present application;

[0059] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0061] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 11, 12, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0062] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0063] In order to solve the problem of low-precision detection of railway traction return caused by weak anti-interference ability, insufficient spatial resolution and lack of multi-path current synergy in the existing technology, the embodiment of the present application provides a railway traction return monitoring and analysis method based on multi-path current fusion. The method adopts the following ideas: distributed optical fiber sensing devices are arranged along the longitudinal direction of the rails to capture the phase changes of optical signals in real time and convert them into multi-path current waveform data; a time-frequency domain joint noise reduction algorithm is designed to eliminate environmental noise interference and complete the timing alignment of multi-path data; a dynamic correlation model is constructed based on the physical laws of rail current propagation, and synchronization deviation features are extracted through cross-path waveform synergy analysis; finally, the preset load dynamic interval is combined to realize the abnormal current event type determination and spatial positioning, forming a closed-loop detection system of "signal perception, data cleaning, feature mining, and intelligent decision-making".

[0064] Figure 1 A flow chart of a railway traction return current monitoring and analysis method based on multi-path current fusion provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes:

[0065] S11. Detect the phase change of the optical signal through distributed optical fiber sensing devices arranged at intervals along the longitudinal direction of the rail, and generate multi-path current waveform data based on the mapping relationship between the phase change and the traction return current intensity.

[0066] Among them, the distributed fiber optic sensing device is a sensing system based on the Rayleigh scattering effect of optical fibers, which continuously monitors strain or temperature changes along the optical fiber. It includes a laser light source, a photodetector, and a signal processing unit. The phase change refers to the relative phase offset caused by external disturbances when the optical signal is transmitted in the optical fiber, and the unit is radians. The traction return current intensity refers to the current intensity returned to the substation through the rails during operation of the electric locomotive, and the unit is amperes. Multipath current waveform data refers to the collection of current time series data with temporal and spatial correlation collected from different sections of the rail.

[0067] In this embodiment, distributed fiber optic sensors are spaced longitudinally along the rails, using optical time-domain reflectometry to acquire Rayleigh scattered light signals. The continuously acquired phase change data is converted into multipath current waveform data by detecting the linear mapping relationship between the optical signal phase change Δφ and the traction return current intensity I (Δφ = K·ΔI), where K is the fiber strain sensitivity coefficient. This multipath current waveform data contains the amplitude, frequency, and time series characteristics of the current in different rail sections, providing data support for subsequent analysis.

[0068] S12, performing time-frequency domain joint noise reduction processing on the multi-path current waveform data, aligning the processed multi-path current waveform data with the time axis, and obtaining time-axis aligned multi-path current waveform data.

[0069] Among them, time-frequency domain joint noise reduction is a hybrid signal processing technology that uses both time domain and frequency domain.

[0070] As a possible implementation approach, a time-frequency domain joint denoising method combining wavelet threshold denoising and Fourier frequency-domain filtering is first used to decompose and reconstruct the multipath current waveform data. Wavelet packet decomposition is used to extract sub-signals in each frequency band, and an improved unbiased risk threshold is used to suppress high-frequency noise. Frequency-domain bandpass filtering is also used to eliminate power frequency interference. A dynamic time warping algorithm is then applied to align the time axis of the denoised multipath current waveform data. By calculating the dynamic bending paths of each path signal relative to a reference time base, the transmission delay differences caused by sensor spacing are eliminated, resulting in multipath current waveform data with unified time tags.

[0071] As another possible implementation method, the specific description of S12 can refer to the following steps 121 to 125, which will not be repeated here.

[0072] S13. A dynamic correlation model of multi-path current is used to perform cross-path synergy analysis on the multi-path current waveform data after time axis alignment to extract the synchronization deviation characteristics of the multi-path current. The dynamic correlation model is constructed based on the spatial gradient constraint conditions of the longitudinal current propagation of the rail.

[0073] Among them, the dynamic correlation model can be a mathematical model that describes the spatiotemporal evolution of multi-path currents, including electromagnetic propagation equations and statistical correlation analysis. The spatial gradient constraint condition is a differential relationship between rail current and distance derived from electromagnetic field theory. This embodiment does not specifically limit the expression of this relationship. The synchronization deviation feature is a set of characteristic parameters that reflect the spatiotemporal differences in multi-path current waveforms. The characteristic parameter set can be a sequence of phase difference and amplitude deviation. Therefore, the synchronization deviation feature involves phase difference, amplitude difference, etc. In other examples, the synchronization deviation feature can also include a path identifier with the largest amplitude deviation.

[0074] In this embodiment of the present application, step S13 can construct a dynamic correlation model based on the spatial gradient constraint of the longitudinal current propagation of the rail by deriving the rail's resistance per unit length R and inductance per unit length L through Maxwell's equations, and establishing the current gradient equation ∂I / ∂x=-(R+jωL)I, where R is the resistance per unit length, L is the inductance per unit length, ∂I / ∂x is the gradient of the current I along the length of the conductor, j is the imaginary unit, ω is the angular frequency of the current, and I is the complex current propagating in the rail. Based on this physical constraint, multivariate covariance analysis is used to calculate the synchronization index of the current waveforms of each path, and principal component analysis is used to extract synchronization deviation characteristics such as phase difference and amplitude difference between the waveforms.

[0075] S14. By matching and verifying the synchronous deviation characteristics with the preset dynamic change range of the traction load, the abnormal current event caused by the harmonic distortion or transient impact and the positioning information of the abnormal current event in the longitudinal direction of the rail are determined.

[0076] Among them, the dynamic change range of traction load refers to the allowable range threshold of current amplitude, harmonic content, and change rate under normal operating conditions.

[0077] In an embodiment of the present application, step S14 can be implemented in the following manner, for example: input the synchronization deviation feature into a preset traction load dynamic change interval classifier, and perform matching verification by calculating the Mahalanobis distance between the synchronization deviation feature and the center of each category. When the feature deviation exceeds the threshold, the harmonic distortion is detected in combination with the wavelet energy entropy, and the transient shock is identified by the mutation point detection algorithm. Anomaly positioning can be achieved by using the cross-correlation function method, and the position of the anomaly point is calculated by the multi-path signal peak delay difference and the sensor spacing: L=Δx·c / (2Δt); wherein c is the current wave velocity, Δx is the sensor spacing, and Δt is the multi-path signal peak delay difference. The above position calculation formula can achieve longitudinal positioning with meter-level accuracy.

[0078] The following is a specific example: distributed fiber optic sensors were deployed at 200-meter intervals along a high-speed rail line. During monitoring, the sensor array detected abnormal fluctuations in the optical signal phase change while a train passed through the section between K325+400 and K325+600. The amplitude exceeded 3 rad, while the normal value is typically less than 1 rad. Based on the linear mapping relationship between phase change and traction return current intensity, and after conversion using a calibration factor of 0.05 rad / A, the multipath current waveform data for sensors 3, 4, and 5 generated during the S11 phase showed that the raw waveform from sensor 3 showed an amplitude of 25A for the 1150Hz high-frequency component and 150A for the 50Hz power frequency component. Next, joint time-frequency domain denoising was performed on the data: a five-layer wavelet decomposition was performed in the time domain, and high-frequency random noise was filtered out using the Stein's Unbiased Risk Estimate (SURE) threshold method. In the frequency domain, a 50Hz±2Hz band-stop filter was used to eliminate power frequency interference, retaining the 1150Hz harmonic component. Simultaneously, using sensor 4 as a benchmark, a dynamic time warping algorithm was used to calibrate the 0.6μs and 1.2μs propagation delays of sensors 3 and 5. After this process, the three sensors still exhibited persistent oscillations of 18-22A in the 1150Hz frequency band. A dynamic correlation model was then used, combining the resistance and inductance parameters in the longitudinal rail current propagation equation to extract features such as a 15-degree phase difference between paths, an amplitude fluctuation rate of 1.22, and an abnormal speed ratio of 85%. The resulting overall synchronization deviation score of 0.78 far exceeded the 0.3 threshold. Finally, the Support Vector Machine (SVM) classifier verified that the anomaly was due to 23rd harmonic distortion. Based on the 0.6μs time delay difference between sensors 3 and 5 and the rail current wave velocity of 250m / μs, the anomaly was calculated to be located at K325+450 mileage. On-site verification confirmed that the impedance drop was caused by damage to the rail insulation gasket at this location, which was completely consistent with the test results.

[0079] By executing S11 to S14, the embodiment of the present application realizes full-line monitoring of rail current through a distributed fiber optic sensing network, combines time-frequency noise reduction to improve the signal-to-noise ratio, accurately identifies abnormal current events through a dynamic correlation model, and reduces positioning errors through spatial gradient constraints. Compared with traditional point detection methods, the detection sensitivity is improved, and complex fault types such as harmonic distortion and transient shocks can be effectively distinguished.

[0080] In a possible embodiment, S13, performing cross-path synergy analysis on the time-axis aligned multi-path current waveform data using a dynamic correlation model of the multi-path current to extract synchronization deviation features of the multi-path current includes:

[0081] Step 131 : Divide the multi-path current waveform data after time axis alignment into multiple analysis windows according to fixed time lengths, and extract the peak points and corresponding timestamps of the current waveforms of each path in each analysis window.

[0082] The analysis window is a data segment divided by a fixed duration, which is used to limit the spatiotemporal range of local feature extraction. Its duration must be at least 100 times the current propagation delay to ensure data integrity. The peak point can be the point of maximum amplitude.

[0083] In this embodiment, the time-aligned multipath current waveform data is divided into multiple analysis windows of fixed duration. Within each analysis window, a sliding window local maximum detection algorithm is used to extract the peak points and corresponding timestamps of the current waveforms of each path. The specific process is as follows: After applying a Savitzky-Golay filter to the current waveform of each path, a search is performed with a sliding step of 50ms to find the amplitude maximum point. If the current point has an amplitude greater than that of the 10 preceding and following sampling points, it is marked as a peak point and its precise timestamp is recorded.

[0084] For example, path 1 detects a peak timestamp sequence of t1=1000.5μs, t2=1500.2μs within a certain analysis window, and path 2 detects a peak timestamp sequence of t1′=1001.0μs, t2′=1501.8μs within a certain analysis window, thereby forming the peak spatiotemporal distribution matrix of each path.

[0085] Step 132: Within each analysis window, based on the peak points and corresponding timestamps of the current waveforms of each path, perform bidirectional timing matching on the peak point timestamps of adjacent paths. If the forward time difference and the reverse time difference in the bidirectional timing matching results are equal, the forward time difference is used as the timestamp difference of the peak points between adjacent paths.

[0086] Bidirectional time series matching involves verifying the reliability of peak point correspondences through forward and reverse matching, avoiding the error accumulation caused by unidirectional matching. Its core is the time series alignment capability of the dynamic time warping algorithm. Optionally, embodiments of the present application can incorporate a fault-tolerance mechanism into the bidirectional time series matching process to address potential mismatches in peak points in real-world scenarios.

[0087] In this embodiment of the present application, forward matching is performed by subtracting the timestamp of each peak point in path A from the timestamp of the nearest subsequent peak point in path B to calculate a forward time difference set. Reverse matching is performed by subtracting the timestamp of each peak point in path B from the timestamp of the nearest subsequent peak point in path A to calculate a reverse time difference set. If the median or mean of the forward and reverse time difference sets are equal, the successfully matched difference is used as the timestamp difference of the peak points between adjacent paths; otherwise, the window of data is discarded. This process can achieve timing alignment using a dynamic time warping algorithm or correlation coefficient analysis method.

[0088] Step 133: Based on the timestamp difference, combined with the rail longitudinal current propagation velocity range and rail longitudinal spacing parameters preset in the dynamic association model, generate the actual current propagation velocity.

[0089] The longitudinal rail spacing parameter refers to the distance between adjacent distributed fiber optic sensors, calculated based on the rail impedance characteristics and detection accuracy requirements, and is typically 50-200 meters. The actual current propagation velocity refers to the propagation rate of the current wave in the rail, calculated from the measured data of the multipath sensor.

[0090] In the embodiment of the present application, based on the obtained timestamp difference Δt, combined with the preset rail longitudinal current propagation velocity range [v_min, v_max] and the rail longitudinal spacing parameter L, the actual current propagation velocity is calculated using the formula v = L / Δt.

[0091] Step 134: Store the actual current propagation velocity exceeding the rail longitudinal current propagation velocity range, the corresponding path identifier, and the amplitude of the corresponding peak point into a cross-path abnormality time offset set.

[0092] The cross-path anomaly time offset set is a three-dimensional database storing anomaly propagation events. It contains velocity, path identifier, and amplitude fields, quantifying the spatial distribution and intensity of anomalies. Its storage structure supports multidimensional retrieval by time, location, and type. The corresponding path identifier is the name of the adjacent sensor path combination where the abnormal current propagation velocity was detected. The corresponding peak amplitude is the instantaneous amplitude of the current peak point that triggered the abnormal velocity determination.

[0093] In an embodiment of the present application, if Δt exceeds the range of the longitudinal current propagation speed of the rail [L / v_max, L / v_min], then mark this speed as an abnormal candidate value. In other words, compare the calculated actual current propagation speed v with the preset range. If v < v_min or v > v_max, then store v, the corresponding path identifier, and the amplitude of the corresponding peak point in the cross-path abnormal time offset set, where v_min = L / v_max and v_max = L / v_min. When storing the amplitude in the cross-path abnormal time offset set, it can be stored in a key-value manner, where the key is the path identifier and the value is the structured data of the abnormal speed, timestamp, and amplitude.

[0094] Step 135: Generate a dynamic influence weight corresponding to each path according to the abnormal frequency of each path in the cross-path abnormal time offset set and the amplitude of the corresponding peak point.

[0095] Among them, the dynamic influence weight refers to the weight coefficient (such as in the range of 0 - 1) synthesized by normalizing the abnormal frequency and amplitude, which is used to quantify the contribution difference of the abnormalities of different paths to the system synchronization. The higher the weight, the more likely this path is to be the fault source.

[0096] In an embodiment of the present application, according to the abnormal frequency of each path in the cross-path abnormal time offset set and the amplitude of the corresponding peak point, calculate the dynamic influence weight through a weighting formula: weight W = (abnormal frequency × normalized amplitude) / total abnormal frequency. Among them, the normalized amplitude is the ratio of the current amplitude to the historical maximum amplitude. The weight result is used to characterize the influence degree of this path on the system synchronization.

[0097] Step 136: Perform collaborative analysis according to the dynamic influence weight corresponding to each path to generate a synchronization deviation feature.

[0098] Among them, collaborative analysis refers to a statistical analysis method that fuses multi-path data based on weights, extracts cross-path common features through a weighted covariance matrix, and suppresses single-sensor noise interference.

[0099] In an embodiment of the present application, step 136 can perform collaborative analysis on the multi-path current waveforms based on the generated dynamic influence weight. Specifically: apply weights to the peak time series of each path, calculate the weighted average time offset; evaluate the synchronization deviation degree between multi-paths through standard deviation analysis; finally generate a synchronization deviation feature, and the feature data is stored in matrix form for triggering an alarm or adjusting the monitoring strategy.

[0100] The following is a specific example: During monitoring of the K325+200 to K325+600 section of a high-speed rail line, the system first divides the time-aligned multipath current waveform data into multiple analysis windows of 1-second duration. Within each window, the peak points and corresponding timestamps of the current waveforms for paths P1-P3 are extracted: 1000.5μs and 1500.2μs for P1, and 1002.3μs and 1503.1μs for P2. Bidirectional timing matching is then performed: a forward and reverse comparison of the peak timestamps of adjacent paths P1-P2 is performed. The forward matching time difference is 1.8μs, and the reverse matching time difference is -1.8μs. Because the absolute values ​​of the differences are equal and their signs are opposite, the valid timestamp difference is determined to be 1.8μs. Based on the difference, combined with the longitudinal rail spacing Δx = 200m and the preset current propagation speed range of 200-300m / μs, the actual propagation speed v = 200m / 1.8μs ≈ 111m / μs is calculated. Because the speed of 111m / μs is lower than the lower threshold of 200m / μs, the system stores the path combination P1-P2, the abnormal speed of 111m / μs, and the corresponding peak amplitude of 85A in the cross-path abnormal time offset set. Further statistics for 10 consecutive analysis windows show that the P1-P2 combination has an abnormal frequency of 7 times, with an average peak amplitude of 90A. Calculated according to the dynamic impact weight formula:

[0101] Weight = 0.7 × (abnormal frequency ratio) + 0.3 × (amplitude ratio) = 0.7 × 0.7 + 0.3 × (90A / 200A) = 0.64. Finally, through multi-path collaborative analysis, a synchronization deviation characteristic ΔΦ = 12°, an amplitude fluctuation variance σA = 20A², and an abnormality ratio of 70% were generated. Combined with the dynamic model, it was determined that an impedance anomaly at K325+450 was caused by aging of the rail insulation pad. On-site inspection confirmed that the carbonization rate of the insulation pad at this location was 60%, consistent with the system's positioning results, verifying the reliability of the entire process from data segmentation, time series matching, to dynamic weight collaborative analysis.

[0102] By executing steps 131-136, this embodiment of the present application utilizes timing matching and dynamic weight assignment to achieve anomaly detection sensitivity as small as 0.1μs, significantly improving sensitivity compared to traditional threshold methods. Bidirectional timing matching eliminates noise mismatches, while the dynamic weighting mechanism reduces false positives such as ground interference. Synchronous deviation features can distinguish complex operating conditions such as harmonic distortion and transient shocks, improving classification accuracy.

[0103] In a possible embodiment, step 136, performing collaborative analysis based on the dynamic impact weights corresponding to each path to generate a synchronization deviation feature, includes:

[0104] Step a1: perform weighted superposition on multi-path current waveform data using the dynamic impact weights corresponding to all paths to generate cross-path synergy reference waveform data.

[0105] Among them, the cross-path collaborative reference waveform data refers to the reference signal generated by weighted fusion of multi-path current waveforms, which reflects the common propagation characteristics of rail current and is used to compare single-path abnormal deviations.

[0106] In the embodiment of the present application, step a1 uses the dynamic influence weights corresponding to all paths to perform weighted superposition on the multi-path current waveform data after time axis alignment to generate cross-path synergy reference waveform data. The specific implementation is: for each time point t, calculate the reference waveform amplitude ,in is the reference waveform amplitude, N is the total number of paths, is the current amplitude of the i-th path at time t. For example, if the amplitudes of the three paths at t = 1000 μs are 120 A, 80 A, and 100 A, and the weights are 0.8, 0.5, and 0.6, then the reference amplitude Aref = 0.8 × 120 + 0.5 × 80 + 0.6 × 100 = 196 A. This reference waveform characterizes the common characteristics of the multipath currents and suppresses noise interference from low-weight paths.

[0107] Step a2: Based on the cross-path synergy reference waveform data and the time-axis aligned multi-path current waveform data, the phase difference and amplitude deviation between the multi-path current waveform and the cross-path synergy reference waveform are calculated point by point to form a synchronization deviation measurement sequence.

[0108] Phase and amplitude deviations are quantified by the phase difference, which is the angular offset between a single-path current and a reference waveform, and the amplitude deviation, which is the percentage difference in relative amplitude. The synchronization deviation metric sequence is a chronologically ordered dataset of phase and amplitude deviations used for statistical analysis and threshold determination.

[0109] In the embodiment of the present application, step a2 calculates the phase difference and amplitude deviation point by point based on the cross-path cooperative reference waveform data and the multi-path current waveform data after time axis alignment: for the complex current of the i-th path at time t in, is the complex current of the i-th path at time t, is the current amplitude of the ith path at time t, is the current phase angle of the i-th path at time t. Calculate its difference with the reference waveform ,in, is the complex current of the cross-path cooperativity reference waveform, is the weighted superposition amplitude of the reference waveform, is the weighted average phase angle of the reference waveform. Calculate the phase difference. The formula for the phase difference is: ,in, is the phase difference of the i-th path at time t, and the result is limited to [-180°, 180°]. Calculate the relative deviation rate: ,in, The relative deviation rate of the amplitude of the i-th path at time t. and Arrange them in chronological order to form a synchronization deviation measurement sequence.

[0110] Step a3: Slide the analysis window along the time axis and accumulate and sum the synchronization deviation measurement sequence. When the accumulated value exceeds the distortion threshold preset in the dynamic association model, a synchronization deviation feature is generated. The synchronization deviation feature includes the path identifier with the largest amplitude deviation.

[0111] Cumulative summation accumulates deviation metrics within a sliding time window, amplifying persistent abnormal signals and suppressing transient noise interference. The distortion threshold is a preset cumulative deviation threshold based on historical normal data statistics or electromagnetic propagation models. Any deviation exceeding this threshold is considered an abnormal current event.

[0112] In the embodiment of the present application, the synchronization deviation measurement sequence is accumulated and summed along the time axis using a sliding window, and the cumulative phase deviation is calculated as the sum of all paths in the window. The summation formula is: ,in, is the total accumulated phase deviation. The accumulated amplitude deviation is calculated as the total of all paths in the window. The summation formula is: ,in, is the total cumulative amplitude deviation. If the distortion threshold preset in the dynamic correlation model is exceeded, it is determined to be an abnormal current event, and the current in the window is extracted. The largest path identifier generates the synchronization deviation feature as the maximum deviation path = P2, and the cumulative deviation value = 520°.

[0113] The following is a specific example: In the high-speed rail line monitoring scenario, based on the path reliability weights of sensors No. 3, No. 4, and No. 5 in the dynamic correlation model, the path reliability weights are set to 0.4, 0.3, and 0.3 respectively. The multi-path current waveform data are weighted and superimposed to generate cross-path collaborative reference waveform data; the multi-path waveform data after time axis alignment, sensor No. 3 compensated 0.6μs and No. 5 compensated 1.2μs and compared point by point with the reference waveform, and the phase difference and amplitude deviation are calculated to form a synchronization deviation measurement sequence; a 200ms sliding window is used to accumulate and sum the deviation sequence. When the accumulated value in the window exceeds the distortion threshold, the synchronization deviation feature is triggered, and sensor No. 3 with the largest amplitude deviation is marked as an abnormal path. Combined with the wave speed of 250m / μs and the time delay difference of 0.6μs, the abnormal point is located at K325+450, which is consistent with the location of the damaged insulating gasket on site.

[0114] By executing steps a1 to a3, the embodiment of the present application generates a reference waveform through dynamic weighted fusion, quantifies the coordinated deviation of multi-path currents point by point, and combines the sliding window accumulation to detect anomalies. It can accurately identify the distortion characteristics caused by loss of synchronization or overload between paths in the power system in real time, improve the fault detection sensitivity and positioning accuracy of the multi-path system, and reduce the misjudgment rate.

[0115] In a possible embodiment, S12, performing time-frequency domain joint noise reduction processing on the multi-path current waveform data, aligning the processed multi-path current waveform data with the time axis, and obtaining time-axis aligned multi-path current waveform data, includes:

[0116] Step 121 : Decompose the multi-path current waveform data into frequency band component signals of multiple levels using a multi-scale decomposition method.

[0117] Among them, the multi-scale decomposition method refers to decomposing the signal into sub-signals of different frequency ranges through time-frequency analysis technology (such as wavelet packet decomposition hierarchy) to separate noise and effective components. Its hierarchical division is based on the typical frequency bands of rail current, such as high-frequency transients, medium-frequency harmonics, and low-frequency fundamental waves.

[0118] In this embodiment, a multi-scale decomposition method is used to decompose multipath current waveform data into multiple levels of frequency band component signals. The specific process involves performing a three-layer wavelet packet decomposition on the current waveform of each path, generating sub-signals of different scales, such as high-frequency, intermediate-frequency, and low-frequency. Each level of frequency band component signal contains waveform characteristics within a specific frequency range. For example, the original signal of path 1 is decomposed into high-frequency components D1, D2, and D3, intermediate-frequency component A3, and low-frequency component A2, forming a multi-level signal set.

[0119] Step 122 : For the frequency band component signals of each layer, calculate the dynamic noise threshold of the current layer based on the statistical distribution difference of the frequency band component signals of adjacent layers.

[0120] The statistical distribution difference refers to the difference measure of the probability density function of the signal amplitudes of adjacent levels, which is quantified by the Kullback-Leibler Divergence (KLD). The larger the difference, the greater the noise interference.

[0121] In an embodiment of the present application, for the frequency band component signals of each level, the dynamic noise threshold of the current level is calculated by the Kulbeck-Leibler divergence based on the statistical distribution differences of adjacent levels. Optionally, the specific formula is: dynamic noise threshold = α × KLD (current level || adjacent level) + β × current level standard deviation, α = 0.6, β = 0.4. For example, the KLD value of high frequency D1 is 0.8 and the standard deviation is 5A, then the dynamic noise threshold = 0.6 × 0.8 + 0.4 × 5 = 2.48A, and the signal segment below this value is regarded as noise.

[0122] Step 123 : Filter the frequency band component signals whose amplitudes are lower than the dynamic noise threshold of the corresponding level, and perform cross-level fusion on the filtered frequency band component signals of all levels to generate processed multi-path current waveform data.

[0123] The dynamic noise threshold is a noise threshold adaptively calculated based on signal statistics. It is used to dynamically filter out invalid components with amplitudes below the threshold. Its calculation formula incorporates the distribution differences between layers and the standard deviation of the layer. Cross-layer fusion reconstructs a complete waveform by weighting the effective signal components of multiple frequency bands. This method preserves key features while suppressing high-frequency noise and low-frequency drift.

[0124] In this embodiment, frequency band component signals with amplitudes below the dynamic noise threshold are screened at each level. The retained valid signals are then fused across levels using an inverse wavelet packet transform to reconstruct processed multipath current waveform data. For example, the high-frequency component of path 1 retains the pulse segment with D1 > 2.48A, while all mid- and low-frequency components are retained. This fusion improves the waveform's signal-to-noise ratio.

[0125] Step 124 : extract the phase principal component sequence of each path from the processed multi-path current waveform data, and calculate the relative time offset of the current waveform data of each path relative to the time axis reference time according to the starting phase mutation point of the phase principal component sequence of each path.

[0126] The phase principal component sequence refers to the time series data extracted through principal component analysis that reflects the dominant pattern of current phase variation and is used to characterize the global phase evolution of the current waveform. The starting phase mutation point is the time point at which the first phase jump occurs in the phase principal component sequence and is used to identify the starting propagation moment of the current waveform. The relative time offset is the time difference between the starting phase mutation point and a unified time reference, reflecting the time delay difference in the current wave propagating to different sensors.

[0127] In this embodiment, a Hilbert transform is first performed on the current signal of each path to obtain the instantaneous phase. Then, principal component analysis is performed to extract the first principal component sequence that dominates the phase change. Based on the starting phase mutation point of each path principal component sequence, its relative time offset Δt relative to the time axis reference time is calculated as: mutation point timestamp - reference time.

[0128] Step 125 : Based on the relative time offset, perform reverse time shift correction on the current waveform data of each path so that the starting point of the phase principal component sequence of the corrected current waveform data of each path is aligned with the time axis reference time, thereby obtaining the time axis aligned multi-path current waveform data.

[0129] Among them, reverse time shift correction refers to eliminating the difference in sensor layout delay through time domain shift operation. The shift amount is determined by the offset between the phase mutation point and the reference time, ensuring the uniformity of the time reference of multi-path data.

[0130] In this embodiment, based on the relative time offset Δt, the reverse time shift correction is performed on the current waveform data of each path. This involves shifting the entire path data left or right by Δt sampling points, so that the starting point of each path's phase principal component sequence is strictly aligned with the time axis reference moment. For example, the starting phase mutation point of the original waveform of path 1 is shifted from t = 10 μs to t = 0, ultimately generating time-aligned multi-path current waveform data for subsequent cross-path collaborative analysis.

[0131] The following is a specific example: In the high-speed rail line monitoring scenario, the multi-path current waveform data of sensors No. 3, No. 4, and No. 5 are decomposed into five levels of frequency band component signals through the wavelet multi-scale decomposition method, of which the 1150Hz target frequency band is distributed in the second level; for the frequency band components of each level, the dynamic noise threshold is calculated based on the divergence difference of the energy distribution of adjacent levels and the unbiased risk estimation threshold method; the components with amplitudes lower than the threshold at each level are screened, and the noise-reduced waveform data is generated through cross-level fusion through inverse wavelet transform; the phase principal component sequence of the 1150Hz frequency band is extracted from the processed waveform, and the relative time offset of each path relative to the reference time is calculated; the waveform data of No. 3 and No. 5 are respectively subjected to reverse time shift correction to align the starting points of the phase principal components of the three sensors to the reference time, and finally the time-axis aligned multi-path current waveform data is generated, which supports the subsequent dynamic correlation model to accurately extract the phase difference of 15 degrees and the abnormal speed ratio of 85%, and locate the damaged insulating gasket at K325+450.

[0132] By executing steps 121 to 125, the embodiment of the present application effectively removes noise interference in the multi-path current waveform through multi-scale decomposition and dynamic noise suppression; through phase principal component extraction and time offset correction, high-precision time synchronization of multi-path signals is achieved, solving the waveform distortion problem caused by transmission delay or sampling asynchrony, improving the accuracy and reliability of current data analysis, and is suitable for scenarios such as power system fault location and harmonic analysis.

[0133] In a possible embodiment, step 124, calculating the relative time offset of the current waveform data of each path relative to the time axis reference time based on the starting phase mutation point of the phase principal component sequence of each path, includes:

[0134] Step b1: performing differential processing on the phase principal component sequence of each path to generate a corresponding phase difference sequence representing the phase change.

[0135] The phase difference sequence refers to the first-order difference result of the phase principal component sequence, which is used to amplify the phase mutation characteristics. The calculation formula is the phase difference value of the latter term minus the former term.

[0136] Step b2: In the phase difference sequence, detect phase mutation intervals that exceed the dynamic mutation threshold for N consecutive times. The dynamic mutation threshold is calculated by the statistical variance of the phase principal component sequence in the non-phase mutation interval, and N is greater than or equal to 2.

[0137] The dynamic mutation threshold is a phase mutation threshold adaptively calculated based on the statistical characteristics of the non-mutation interval (such as variance). It is used to distinguish normal fluctuations from abnormal jumps and can be set to three times the standard deviation. The non-mutation interval and the phase mutation interval are different intervals.

[0138] Step b3: Mark the moment when the phase mutation threshold is exceeded for the first time within the phase mutation interval as the starting phase mutation point.

[0139] The phase mutation interval refers to the segment in the phase difference sequence that continuously exceeds the threshold value. Its length is controlled by the parameter N (N≥2) and is used to eliminate the interference of isolated noise points.

[0140] Step b4: Taking the time axis reference time as the reference zero point, calculate the relative time offset between the starting phase mutation point corresponding to each path current waveform data and the reference zero point.

[0141] The starting phase mutation point refers to the moment when the phase mutation interval exceeds the dynamic mutation threshold for the first time, and is used to identify the starting propagation time of the abnormal current event.

[0142] The following is a specific example: The phase principal component sequence of path P1, 10°, 12°, 15°, 25°, and 40°, was differentially processed to generate the phase difference sequence of 2°, 3°, 10°, and 15°. A dynamic mutation threshold of T = 4.5° was calculated within the non-mutation interval. Three consecutive differential values ​​of 5°, 6°, and 5.5° exceeding the threshold were detected within the window of t = 6-8 μs and marked as a mutation interval. t = 6 μs was backtracked to determine the starting phase mutation point. Its offset, Δt = 6 μs, relative to the reference time t = 0, was calculated. By achieving time alignment by shifting the P1 waveform left by 6 μs, the carbonization fault of the insulating gasket was ultimately located at K12+300, with a maintenance verification error of 0.15 meters.

[0143] By executing steps b1 to b4, the embodiment of the present application accurately identifies phase jumps caused by faults or interference in the power system and eliminates noise misjudgments through differential amplification of phase mutation characteristics, adaptive threshold judgment and continuous over-limit detection; through starting mutation point calibration and relative time offset calculation, a high-precision basis is provided for time synchronization correction of multi-path current waveforms, thereby improving the accuracy of power system status monitoring and fault location.

[0144] In a possible embodiment, step 122, calculating the dynamic noise threshold of the current level based on the statistical distribution difference of the frequency band component signals of adjacent levels, includes:

[0145] Step c1: constructing a probability density distribution function of the current layer and a probability density distribution function of the adjacent layer for the frequency band component signal of the current layer and the frequency band component signal of the adjacent layer respectively.

[0146] Among them, the probability density distribution function refers to a continuous function that describes the probability of signal amplitude distribution generated by kernel density estimation. It reflects the possibility density of different amplitudes and is used to quantify the statistical characteristics of the signal.

[0147] Step c2: Calculate the cumulative probability difference between the probability density distribution function of the current level and the probability density distribution function of the adjacent level within a preset amplitude range, and use the cumulative probability difference as an indicator of the statistical distribution difference between the levels.

[0148] The cumulative probability difference refers to the absolute difference in the probability cumulative values ​​of two levels within the preset amplitude range, which is used to measure the similarity of statistical distribution between levels. The larger the difference, the greater the noise interference.

[0149] Step c3: construct a dynamic noise threshold calculation function based on the statistical distribution difference index and the energy proportion of the frequency band component signal at the current level.

[0150] The energy percentage refers to the ratio of the current level's signal energy to the total energy of the multi-scale decomposition, and is used to assess the importance of the signal at that level. The dynamic noise threshold calculation function is a linear function that integrates statistical distribution differences with energy weights and is used to adaptively generate the noise threshold. This embodiment does not specifically limit its expression.

[0151] Step c4: input the amplitude median of the frequency band component signal of the current level into the dynamic noise threshold calculation function, and output the dynamic noise threshold of the current level.

[0152] The median amplitude refers to the middle value of the signal amplitude at the current level, which reflects the central trend of the amplitude distribution and is used for normalized threshold calculation.

[0153] Step c5: Perform cross-level smoothing constraints on the dynamic noise threshold according to the amplitude extreme difference of the frequency band component signals of adjacent levels to generate the dynamic noise threshold of the current level.

[0154] The amplitude range, defined as the difference between the maximum and minimum signal amplitudes, characterizes the signal's dynamic range and is used for cross-level threshold smoothing. Cross-level smoothing constraints utilize the signal characteristics of adjacent levels to modify the threshold of the current level, avoiding threshold distortion caused by single-level anomalies.

[0155] The following is a specific example: After the traction current signal of a high-speed railway line is decomposed into high-frequency D1, medium-frequency A3, and low-frequency A2 sub-bands through multi-scale decomposition, a probability density distribution function is constructed for the D1 sub-band, and its cumulative probability in the [-2A, 2A] interval is calculated to be 0.85, and the adjacent A3 sub-band is 0.45, with a statistical distribution difference index ΔP=0.4; combined with the D1 sub-band energy proportion of 0.15, through the dynamic noise threshold function T=0.6×0.4+0.4×0.15=0.3; the median amplitude of the input D1 is The value of 0.8A is calculated, and T_final = 0.3×0.8 = 0.24A is obtained. Based on the amplitude range of 6A in the A3 sub-band, the D1 range of 4A is smoothed and constrained to generate the final threshold T_smooth = 0.24×(4 / 6) = 0.16A. After filtering out the noise segment with amplitude <0.16A in D1, the waveform signal-to-noise ratio is improved, and the carbonization fault of the insulating gasket at K325+450 is synchronously located. The actual maintenance verification shows that the error between the carbonization area and the detection result is small, which verifies the effectiveness of this scheme.

[0156] By executing steps c1 to a5, the embodiment of the present application dynamically generates noise thresholds that adapt to different frequency band characteristics through inter-level distribution difference analysis, energy proportion weighting and cross-level smoothing constraints. It can accurately distinguish between valid signals and noise, and is particularly suitable for multi-scale denoising of non-stationary signals. It improves the adaptability and accuracy of noise suppression, while avoiding the over-smoothing or under-filtering problems caused by traditional fixed threshold methods.

[0157] In a possible embodiment, S14, determining an abnormal current event caused by harmonic distortion or transient impact and the longitudinal location information of the abnormal current event in the rail by matching and verifying the synchronization deviation characteristics with a preset dynamic change interval of the traction load, includes:

[0158] Step 141: Calculate the equivalent time delay between the current propagation direction corresponding to the path with the maximum amplitude deviation in the synchronization deviation feature and the adjacent path by combining the rail longitudinal current propagation velocity range and the rail longitudinal spacing parameter preset in the dynamic correlation model, where N is greater than or equal to 2.

[0159] Among them, the equivalent time delay refers to the theoretical propagation delay of adjacent paths calculated based on the preset current propagation speed, which is used to compare with the actual delay to verify anomalies.

[0160] Step 142: Match the amplitude deviation of the path with the maximum amplitude deviation with the upper and lower bounds of the dynamic change range of the traction load point by point.

[0161] The dynamic change range of the traction load refers to the allowable range threshold of the current amplitude deviation under normal operating conditions, the upper limit is the maximum allowable distortion rate, and the lower limit is the minimum transient sensitivity threshold.

[0162] Step 143: If the amplitude deviation exceeds the upper bound N times in a row and the phase difference fluctuates periodically, a harmonic distortion determination index is generated. If the amplitude deviation is lower than the lower bound and the equivalent time delay exceeds the preset transient propagation delay threshold, a transient impulse determination index is generated.

[0163] Among them, the harmonic distortion judgment index is a quantitative parameter reflecting the degree of harmonic component exceeding the limit, including the total harmonic distortion rate and characteristic harmonic order. The transient impact judgment index is a parameter that characterizes the intensity of transient events, such as the current change rate and peak duration.

[0164] Step 144 : Based on the harmonic distortion determination index or the transient impulse determination index, the equivalent propagation path length of the abnormal current event between adjacent paths is calculated, and an initial positioning interval is generated according to the longitudinal coordinate of the path identifier.

[0165] Among them, the equivalent propagation path length refers to the equivalent distance that the abnormal current propagates along the rail, which is calculated by the product of time delay and velocity.

[0166] Step 145: Perform weighted correction on the initial positioning interval to generate an optimized positioning interval in the longitudinal direction of the rail. The boundary value of the optimized positioning interval is dynamically adjusted by multiplying the equivalent propagation path length by the dynamic influence weight.

[0167] Among them, the initial positioning interval is the possible distribution range of the anomaly calculated based on the equivalent path length and path coordinates.

[0168] Step 146: Calculate the abnormality confidence of the optimized positioning interval, calibrate the reference coordinates of the abnormal current event in the longitudinal direction of the rail based on the path identifier of the path with the maximum amplitude deviation, and obtain positioning information including the abnormality type, optimized positioning interval, abnormality confidence and reference coordinates.

[0169] The optimized positioning interval is the probability distribution range of the abnormal location after weight adjustment, which improves accuracy compared to the initial interval. Anomaly types mainly include harmonic distortion and transient shock. Anomaly confidence refers to the reliability of the positioning result and is calculated based on data consistency. The reference coordinates are the reference starting position of the abnormal current event in the longitudinal direction of the rail.

[0170] The following is a specific example: The equivalent time delay Δt = 0.8 μs was calculated for the path P2-P3. The amplitude deviation ΔA was detected to have exceeded the upper bound five times in a row, indicating harmonic distortion. The equivalent path L = 0.8 μs × 250 m / μs = 200 m was calculated, generating an initial interval K325+300-500. This interval was then corrected to K325+220-580 using a weight of 0.8. With a calculated confidence level of 67%, the reference coordinate K325+300 was calibrated. During actual maintenance, carbonization of the insulation gasket was discovered at K325+450, which fell within the optimized interval, verifying the effectiveness of the proposed solution.

[0171] By executing steps 141 to 146, the embodiment of the present application achieves high-precision positioning and classification of abnormal current events of high-speed rail current through equivalent delay verification, abnormality type determination and dynamic weight correction, effectively distinguishes harmonic distortion from transient shocks, optimizes positioning interval accuracy to the meter level, and quantitatively evaluates abnormality confidence to improve the reliability of operation and maintenance decisions, and adapts to the needs of multi-source interference suppression under complex working conditions.

[0172] Figure 2 A schematic diagram of the structure of a railway traction return current monitoring and analysis system based on multi-path current fusion provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system includes:

[0173] The generation module 21 is used to detect the phase change of the optical signal through the distributed optical fiber sensing devices arranged at intervals along the longitudinal direction of the rail, and generate multi-path current waveform data based on the mapping relationship between the phase change and the traction return current intensity.

[0174] The processing module 22 is configured to perform time-frequency domain joint noise reduction processing on the multi-path current waveform data, align the processed multi-path current waveform data with the time axis, and obtain time-axis aligned multi-path current waveform data.

[0175] The extraction module 23 performs cross-path synergy analysis on the time-axis aligned multi-path current waveform data through a dynamic correlation model of the multi-path current, and extracts the synchronization deviation characteristics of the multi-path current. The dynamic correlation model is constructed based on the spatial gradient constraint conditions of the longitudinal current propagation of the rail.

[0176] The verification module 24 is used to verify the matching between the synchronization deviation characteristics and the preset dynamic change range of the traction load, and determine the abnormal current event caused by the harmonic distortion or transient impact and the positioning information of the abnormal current event in the longitudinal direction of the rail.

[0177] Figure 2 The railway traction return current monitoring and analysis system based on multi-path current fusion can perform Figure 1 The implementation principles and technical effects of the railway traction return current monitoring and analysis method based on multi-path current fusion described in the illustrated embodiment are not further elaborated. The specific manner in which the various modules and units perform operations in the railway traction return current monitoring and analysis system based on multi-path current fusion in the above embodiment have been described in detail in the relevant embodiments of the method and will not be elaborated on here.

[0178] In one possible design, Figure 2 The railway traction return current monitoring and analysis system based on multi-path current fusion of the embodiment shown can be implemented as a computing device, such as Figure 3As shown, the computing device may include a storage component 31 and a processing component 32 .

[0179] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0180] The processing component 32 is used to: detect the phase change of the optical signal through distributed optical fiber sensing devices arranged at intervals along the longitudinal direction of the rail, and generate multi-path current waveform data based on the mapping relationship between the phase change and the traction return current intensity; perform time-frequency domain joint noise reduction processing on the multi-path current waveform data, align the processed multi-path current waveform data with the time axis, and obtain the multi-path current waveform data after time axis alignment; perform cross-path synergy analysis on the multi-path current waveform data after time axis alignment through a dynamic correlation model of the multi-path current, and extract the synchronization deviation characteristics of the multi-path current. The dynamic correlation model is constructed based on the spatial gradient constraint conditions of the longitudinal current propagation of the rail; and determine the abnormal current events caused by harmonic distortion or transient impact and the positioning information of the abnormal current events in the longitudinal direction of the rail by matching and verifying the synchronization deviation characteristics with the preset dynamic change interval of the traction load.

[0181] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0182] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as random access memory (RAM), static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0183] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0184] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0185] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0186] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0187] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment provides a railway traction return current monitoring and analysis method based on multi-path current fusion.

[0188] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0189] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0190] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0191] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A railway traction return current monitoring and analysis method based on multi-path current fusion, characterized in that: include: The distributed optical fiber sensing devices arranged at intervals along the longitudinal direction of the rail are used to detect the phase change of the optical signal, and multi-path current waveform data is generated based on the mapping relationship between the phase change and the traction return current intensity; Performing a time-frequency domain joint noise reduction process on the multi-path current waveform data, aligning the processed multi-path current waveform data with a time axis, and obtaining time-axis aligned multi-path current waveform data; Performing cross-path synergy analysis on the time-axis aligned multi-path current waveform data using a dynamic correlation model of multi-path current to extract synchronization deviation characteristics of the multi-path current, wherein the dynamic correlation model is constructed based on spatial gradient constraints on rail longitudinal current propagation; By matching and verifying the synchronization deviation characteristics with a preset dynamic change interval of the traction load, the abnormal current event caused by the harmonic distortion or transient impact and the positioning information of the abnormal current event in the longitudinal direction of the rail are determined.

2. The method according to claim 1, characterized in that The method of performing cross-path synergy analysis on the time-axis aligned multi-path current waveform data using a dynamic correlation model of the multi-path current to extract synchronization deviation features of the multi-path current includes: The multi-path current waveform data after time axis alignment is divided into multiple analysis windows according to a fixed time length, and the peak point and corresponding time stamp of each path current waveform are extracted in each analysis window; Within each analysis window, based on the peak points and corresponding timestamps of the current waveforms of each path, bidirectional timing matching is performed on the peak point timestamps of adjacent paths. If the forward time difference and the reverse time difference in the bidirectional timing matching result are equal, the forward time difference is used as the timestamp difference of the peak points between the adjacent paths. Based on the timestamp difference, combined with the rail longitudinal current propagation velocity range and rail longitudinal spacing parameters preset in the dynamic association model, an actual current propagation velocity is generated; The actual current propagation speed exceeding the rail longitudinal current propagation speed range, the corresponding path identifier and the amplitude of the corresponding peak point are stored in a cross-path abnormal time offset set; Generating a dynamic impact weight corresponding to each path according to the abnormal frequency of each path in the cross-path abnormal time offset set and the amplitude of the corresponding peak point; A collaborative analysis is performed based on the dynamic impact weights corresponding to the paths to generate synchronization deviation features.

3. The method according to claim 2, characterized in that The collaborative analysis based on the dynamic impact weights corresponding to the paths to generate synchronization deviation features includes: Performing weighted superposition on the multi-path current waveform data using the dynamic impact weights corresponding to all the paths to generate cross-path synergy reference waveform data; Calculating the phase difference and amplitude deviation between the multipath current waveform and the cross-path cooperativity reference waveform point by point based on the cross-path cooperativity reference waveform data and the time-axis aligned multipath current waveform data to form a synchronization deviation measurement sequence; The analysis window is slid along the time axis, and the synchronization deviation measurement sequence is accumulated and summed. When the accumulated value exceeds the distortion threshold preset in the dynamic association model, a synchronization deviation feature is generated, and the synchronization deviation feature includes a path identifier with the largest amplitude deviation.

4. The method according to claim 1, wherein The performing time-frequency domain joint noise reduction processing on the multi-path current waveform data, aligning the processed multi-path current waveform data with the time axis, and obtaining time-axis aligned multi-path current waveform data, includes: Decomposing the multi-path current waveform data into frequency band component signals at multiple levels by a multi-scale decomposition method; For the frequency band component signals at each level, based on the statistical distribution difference of the frequency band component signals at adjacent levels, calculating a dynamic noise threshold of the current level; Filtering frequency band component signals whose amplitudes are lower than the dynamic noise threshold of the corresponding level, and cross-level fusing the filtered frequency band component signals of all levels to generate processed multi-path current waveform data; Extracting a phase principal component sequence of each path from the processed multi-path current waveform data, and calculating a relative time offset of the current waveform data of each path relative to a time axis reference time based on a starting phase mutation point of the phase principal component sequence of each path; Based on the relative time offset, the current waveform data of each path is subjected to reverse time shift correction so that the starting point of the phase principal component sequence of the corrected current waveform data of each path is aligned with the time axis reference time, thereby obtaining the multi-path current waveform data after time axis alignment.

5. The method according to claim 4, characterized in that The step of calculating the relative time offset of the current waveform data of each path relative to the time axis reference time according to the starting phase mutation point of the phase principal component sequence of each path includes: performing differential processing on the phase principal component sequences of each path to generate corresponding phase difference sequences representing phase changes; In the phase difference sequence, detecting a phase mutation interval that exceeds a dynamic mutation threshold for N consecutive times, where the dynamic mutation threshold is calculated by calculating the statistical variance of the phase principal component sequence in a non-phase mutation interval, and N is greater than or equal to 2; Marking the moment when the phase mutation threshold is exceeded for the first time within the phase mutation interval as the starting phase mutation point; Taking the time axis reference time as the reference zero point, the relative time offset between the starting phase mutation point corresponding to each path current waveform data and the reference zero point is calculated.

6. The method according to claim 4, characterized in that The calculating of the dynamic noise threshold of the current level based on the statistical distribution difference of the frequency band component signals of adjacent levels includes: Constructing a probability density distribution function of the current layer and a probability density distribution function of the adjacent layer for the frequency band component signal of the current layer and the frequency band component signal of the adjacent layer respectively; Calculating the cumulative probability difference between the probability density distribution function of the current level and the probability density distribution function of the adjacent level within a preset amplitude range, and using the cumulative probability difference as an indicator of the statistical distribution difference between the levels; Constructing a dynamic noise threshold calculation function according to the statistical distribution difference index and the energy proportion of the frequency band component signal at the current level; Inputting the amplitude median of the frequency band component signal of the current layer into the dynamic noise threshold calculation function, and outputting the dynamic noise threshold of the current layer; According to the amplitude extreme difference of the frequency band component signals of adjacent levels, a cross-level smoothing constraint is performed on the dynamic noise threshold to generate a dynamic noise threshold of the current level.

7. The method according to claim 1, characterized in that The matching verification between the synchronization deviation feature and the preset dynamic change interval of the traction load is performed to determine the abnormal current event caused by the harmonic distortion or transient impact and the positioning information of the abnormal current event in the longitudinal direction of the rail, including: In combination with the preset rail longitudinal current propagation velocity range and rail longitudinal spacing parameters in the dynamic correlation model, the equivalent time delay between the current propagation direction corresponding to the path with the maximum amplitude deviation in the synchronization deviation feature and the adjacent path is calculated; Matching the amplitude deviation of the maximum amplitude deviation path with the upper and lower bounds of the dynamic change interval of the traction load point by point; If the amplitude deviation exceeds the upper bound N times in a row and the phase difference fluctuates periodically, a harmonic distortion determination index is generated, where N is greater than or equal to 2; if the amplitude deviation is lower than the lower bound and the equivalent time delay exceeds a preset transient propagation delay threshold, a transient impact determination index is generated; Based on the harmonic distortion determination index or the transient impulse determination index, the equivalent propagation path length of the abnormal current event between adjacent paths is calculated, and an initial positioning interval is generated according to the longitudinal coordinate of the path identifier; Performing weighted correction on the initial positioning interval to generate an optimized positioning interval in the longitudinal direction of the rail, wherein the boundary value of the optimized positioning interval is dynamically adjusted by multiplying the equivalent propagation path length by a dynamic influence weight; The abnormality confidence of the optimized positioning interval is calculated, and the reference coordinates of the abnormal current event in the longitudinal direction of the rail are calibrated based on the path identifier of the path with the maximum amplitude deviation, so as to obtain positioning information including the abnormality type, optimized positioning interval, abnormality confidence and reference coordinates.

8. A railway traction return current monitoring and analysis system based on multi-path current fusion, characterized in that: include: a generation module configured to detect a phase change of an optical signal through distributed optical fiber sensing devices spaced apart along the longitudinal direction of the rail, and to generate multipath current waveform data based on a mapping relationship between the phase change and the traction return current intensity; a processing module, configured to perform time-frequency domain joint noise reduction processing on the multi-path current waveform data, align the processed multi-path current waveform data with a time axis, and obtain time-axis aligned multi-path current waveform data; an extraction module that performs cross-path synergy analysis on the time-axis aligned multi-path current waveform data using a dynamic correlation model of multi-path current, and extracts synchronization deviation characteristics of the multi-path current, wherein the dynamic correlation model is constructed based on spatial gradient constraints on the longitudinal current propagation of the rail; The verification module is used to match and verify the synchronization deviation characteristics with the preset dynamic change interval of the traction load to determine the abnormal current event caused by harmonic distortion or transient impact and the positioning information of the abnormal current event in the longitudinal direction of the rail.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the railway traction return current monitoring and analysis method based on multi-path current fusion as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the railway traction return current monitoring and analysis method based on multi-path current fusion as described in any one of claims 1 to 7 is implemented.

Images