A distributed fiber grating track deformation on-line monitoring method and system

Abstract

The application relates to the technical field of track state online monitoring, in particular to a distributed fiber grating track deformation online monitoring method and system. The method comprises the following steps: collecting center wavelength offset data through a pre-set sensing fiber group in a to-be-monitored steel rail; determining a strain initial value sequence corresponding to each measurement section according to the center wavelength offset data, extracting a time sequence strain curve based on a pre-set cache window, and determining a low-frequency background strain baseline; subtracting the low-frequency background strain baseline from the time sequence strain curve to obtain a high-frequency pulse response parameter, extracting a strain increment corresponding to a target peak value of the high-frequency pulse response parameter, and determining a curvature change amount and a torsion angle change amount corresponding to each measurement section based on the strain increment and a triangular space position of the sensing fiber grating, so as to determine a track deformation monitoring result of the to-be-monitored steel rail. Through the above technical scheme, the online monitoring of track deformation can be more accurately realized.

Description

Technical Field

[0001] This application relates to the field of online track condition monitoring technology, and in particular to a method and system for online monitoring of track deformation using a distributed fiber Bragg grating. Background Technology

[0002] As the core basic component for train load-bearing, guidance, and travel, rail tracks are used in transportation systems such as high-speed railways, conventional railways, urban subways, light rail, and trams. They are also extensively laid in industrial transportation scenarios such as heavy-haul transportation in mines, dedicated freight lines in ports, and heavy locomotive transfer in factories. During train operation, all loads are continuously transferred to the track structure through the wheels. Rail tracks are the infrastructure that ensures the smooth, safe, and continuous passage of vehicles.

[0003] With the continuous expansion of the rail transit network, the increasing density of vehicle operation, and the normalization of heavy-load and high-speed operation, the track needs to withstand reciprocating impact loads and alternating pressures for a long time, and be exposed to complex external environments such as temperature changes and rain erosion around the clock. The stability of the track structure affects the operational efficiency and traffic safety of the line.

[0004] Track deformation increases friction and impact between train wheels and rails, exacerbates wear on rails, wheels, and fasteners, and increases the frequency of track maintenance and repair and operation and maintenance costs. Track deformation also causes trains to ride bumpy and the train body to sway, reducing passenger comfort and, in severe cases, causing abnormal noises from the train wheels and rails and longer braking distances.

[0005] Track deformation can also trigger train derailment, wheel-rail breakage, and track collapse, causing equipment damage or casualties. Local track deformation can force temporary speed limits and line closures for maintenance, reducing the capacity of rail transit. Therefore, it is necessary to monitor track deformation online. Summary of the Invention

[0006] To achieve more accurate online monitoring of track deformation, this application provides a method and system for online monitoring of track deformation using a distributed fiber Bragg grating.

[0007] According to a first aspect of the embodiments of this application, a method for online monitoring of track deformation using distributed fiber optic gratings is provided, comprising: acquiring center wavelength offset data through a pre-set sensing fiber optic group in the rail to be monitored; the sensing fiber optic group includes three sensing fiber optic gratings arranged in a triangle within the cross-section of the rail to be monitored, and the rail to be monitored is divided into multiple measurement sections along the longitudinal direction; determining the strain initial value sequence corresponding to each measurement section based on the center wavelength offset data, and extracting the time-series strain curve based on a preset buffer window, performing extreme value optimization processing on the time-series strain curve to determine a low-frequency background strain baseline characterizing longitudinal shear interference; subtracting the low-frequency background strain baseline from the time-series strain curve to obtain a high-frequency pulse response parameter characterizing the vertical rolling effect, extracting the strain increment corresponding to the high-frequency pulse response parameter at the target peak time, and determining the curvature change and torsion angle change corresponding to each measurement section based on the strain increment and the triangular spatial position of the sensing fiber optic gratings; performing an integral stitching operation on the curvature change according to the distribution order of each measurement section, and determining the three-dimensional deformation data in combination with the torsion angle change to determine the track deformation monitoring result of the rail to be monitored.

[0008] This allows for more accurate online monitoring of track deformation.

[0009] Optionally, the strain initial value sequence corresponding to each measurement section is determined based on the center wavelength offset data, and the time-series strain curve is extracted based on the preset buffer window. This includes: determining the center wavelength offset of each measurement section at the current time based on the calibration initial reference wavelength, mapping the center wavelength offset to the strain initial value at the corresponding spatial location to generate the strain initial value sequence; pushing the strain initial value sequence into the buffer queue associated with the preset buffer window according to the acquisition order, and monitoring the current fluctuation amplitude corresponding to the latest strain initial value stored in the buffer queue; and determining the time series range corresponding to the preset buffer window as the effective data segment in response to the current fluctuation amplitude being greater than the preset static environmental noise floor threshold to generate the time-series strain curve.

[0010] In this way, by calculating the center wavelength offset and monitoring the current fluctuation amplitude, the massive amount of invalid background noise data when no vehicles are passing is eliminated, reducing the computational overhead of subsequent full-line deformation data processing.

[0011] Optionally, extreme value optimization processing is performed on the time-series strain curves to determine the low-frequency background strain baseline characterizing longitudinal shear interference. This includes: extracting the time-series strain curves corresponding to each measurement section, dividing the time-series strain curves into multiple local evaluation intervals on the time axis; extracting the strain minimum points reflecting the wheel rolling gap state in each local evaluation interval, and performing connection and numerical interpolation operations on the time axis for the multiple continuously extracted strain minimum points to obtain the low-frequency background strain baseline.

[0012] Optionally, the high-frequency pulse response parameters characterizing the vertical rolling effect are obtained by subtracting the low-frequency background strain baseline from the time-series strain curve, including: extracting the original strain values ​​corresponding to each time node on the time-series strain curve, and extracting the low-frequency background strain baseline corresponding to each time node on the low-frequency background strain baseline; subtracting the corresponding low-frequency background strain baseline from the original strain values ​​to obtain the high-frequency response difference; and performing time-series recombination of the high-frequency response differences corresponding to consecutive time nodes to obtain the high-frequency pulse response parameters.

[0013] Optionally, the strain increment corresponding to the high-frequency pulse response parameter at the target peak time is extracted, including: traversing the strain change amplitude of the high-frequency pulse response parameter on the time axis, extracting the time node when the high-frequency pulse response parameter reaches the highest amplitude state as the target peak time; synchronously acquiring the strain values ​​corresponding to the three sensing fiber gratings on the same measurement section at the target peak time, and merging the three strain values ​​corresponding to the three sensing fiber gratings at the target peak time to confirm the multi-dimensional strain increment.

[0014] Optionally, based on the strain increment and the triangular spatial position of the sensing fiber grating, the curvature change and torsion angle change corresponding to each measurement section are determined, including: substituting the strain values ​​corresponding to the three sensing fiber gratings into the spatial structure conversion mapping model constructed based on the preset triangular spatial position; and using the spatial structure conversion mapping model to convert the strain values ​​into the vertical curvature change, the transverse curvature change, and the torsion angle change around the axis of the current measurement section.

[0015] In this way, by using the spatial structure conversion mapping model to process multi-dimensional strain values, the mixed linear strain is transformed into vertical curvature change, lateral curvature change and torsional angle change, thereby improving the fidelity of orbit attitude reconstruction under complex loads.

[0016] Optionally, the curvature change is integrated and stitched according to the distribution order of each measurement section, and the three-dimensional deformation data is determined by combining the torsional angle change. This includes: extracting the straight-line spacing parameters of each measurement section along the layout route, using a stable measurement section located at the far end and not subject to external force disturbance as a spatial reference point; using the straight-line spacing parameters, performing a step-by-step accumulation operation on the vertical curvature change and transverse curvature change corresponding to each measurement section distributed along the longitudinal direction to obtain a continuous deflection fluctuation curve; summarizing the torsional angle changes of all measurement sections to generate global three-dimensional attitude data, and combining the continuous deflection fluctuation curve and the global three-dimensional attitude data to determine the three-dimensional deformation data.

[0017] Optionally, the method further includes: when the rail to be monitored is in a light-load operation state, recording the reference values ​​of the baseline subsidence characteristics corresponding to all measurement sections along the entire line; when the rail to be monitored is in a heavy-load operation state, obtaining the actual subsidence test value generated by the current measurement section under heavy-load stress; subtracting the reference subsidence characteristic value from the actual subsidence test value to obtain the subsidence surge difference value; in response to the subsidence surge difference value being greater than the preset settlement surge threshold and deviating from the normal deformation range of adjacent sections, positioning the current measurement section as a weak support point of the track bed; and when performing integral splicing, using a preset elastic inhibition coefficient to limit the output value of the curvature change corresponding to the weak support point of the track bed.

[0018] In this way, by comparing the difference in sudden subsidence, weak support points of the track bed are identified, and the curvature output value of abnormal points is limited by the elastic inhibition coefficient, the spatial integral divergence effect caused by excessive local settlement is suppressed, thus ensuring the calculation stability of the overall continuous deflection undulation curve.

[0019] Optionally, the method further includes: extracting the initial strain values ​​corresponding to two adjacent time nodes arranged in chronological order from the initial strain value sequence; subtracting the initial strain values ​​corresponding to the two adjacent time nodes to obtain the strain difference value; and determining the strain increment rate of change based on the time interval between the strain difference value and the two adjacent time nodes; extracting the preset upper limit reference value of the elastic deformation rate corresponding to the rail to be monitored; and determining that a transient high-energy impact occurs at the time node corresponding to the strain increment rate of change when the strain increment rate of change is greater than the preset upper limit reference value of the elastic deformation rate; and marking the initial strain value collected at the corresponding time node as distorted saturated acquisition data.

[0020] In this way, by dynamically aligning the rate of change of strain increment at adjacent time points with the preset upper limit reference value of elastic deformation rate, abnormal strain jumps caused by wheel-rail contact surface damage can be captured quickly and sensitively.

[0021] Optionally, after marking the initial strain values ​​collected at the corresponding time nodes as distorted saturated acquisition data, the process includes: removing the distorted saturated acquisition data within a preset buffer window; extracting a first stable historical data sequence within a preset first acquisition period span before the occurrence of the transient high-energy impact by backtracking forward in the time series; extracting a second stable recovery data sequence within a preset second acquisition period span after the dissipation of the transient high-energy impact by delaying backward in the time series; performing a smooth interpolation bridging operation on the time series based on the first stable historical data sequence and the second stable recovery data sequence to obtain transition smoothing test parameters, and writing the transition smoothing test parameters into a preset buffer window to replace the distorted saturated acquisition data.

[0022] In this way, by extracting the first stationary historical data sequence and the second stationary recovery data sequence before and after the anomaly and performing smooth interpolation bridging operation, information gaps are filled in while destructive data is removed, and the continuity and integrity of the time series strain curve are maintained.

[0023] Optionally, the track deformation monitoring results of the rail to be monitored are determined, including: extracting the continuous deflection fluctuation curve and global three-dimensional attitude data from the three-dimensional deformation data, and obtaining the pre-set standard alignment change envelope and angle safety limit; using the continuous deflection fluctuation curve and the standard alignment change envelope, determining the over-limit area from the continuous deflection fluctuation curve; using the global three-dimensional attitude data and angle safety limit to determine the torsion angle deviation value; performing trend fitting on the continuous deflection fluctuation curve to obtain the overall track alignment data; mapping the distance coordinates of the over-limit area, the maximum subsidence of the over-limit area, the torsion angle deviation value, and the overall track alignment data to a multi-dimensional feature matrix to construct the track deformation monitoring results.

[0024] According to a second aspect of the present application, an online monitoring system for track deformation of a distributed fiber Bragg grating is provided, comprising: a processor and a memory, wherein the memory stores computer program instructions, and the computer program instructions, when executed by the processor, implement the steps of the online monitoring method for track deformation of a distributed fiber Bragg grating provided in the first aspect of the present application.

[0025] The technical solutions provided by the embodiments of this application may include the following beneficial effects: by synchronously acquiring center wavelength offset data through a triangularly arranged sensing fiber optic grating, a correlation basis between multidimensional strain and spatial deformation is established; by performing extreme value optimization processing on the time-series strain curve to extract the low-frequency background strain baseline, environmental interference in the high-frequency pulse response parameters is effectively separated; by combining spatial position to decouple curvature and torsion angle and performing longitudinal integral stitching, the restoration of the three-dimensional deformation data of the track is realized, improving the accuracy and reliability of online assessment of track deformation state.

[0026] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating an online monitoring method for track deformation of a distributed fiber Bragg grating according to an exemplary embodiment;

[0028] Figure 2 This is a schematic diagram of the spatial distribution of the sensing fiber optic gratings within the cross-section of the rail to be monitored. Figure 3 This is a schematic diagram illustrating the structure of an online monitoring system for track deformation of a distributed fiber Bragg grating, according to an exemplary embodiment. Detailed Implementation

[0029] To achieve more accurate online monitoring of track deformation, embodiments of this application provide a method and system for online monitoring of track deformation using a distributed fiber Bragg grating. Figure 1 This is a flowchart illustrating an online monitoring method for track deformation of a distributed fiber Bragg grating according to an exemplary embodiment, such as... Figure 1 As shown, the method includes the following steps.

[0030] In step S101, the center wavelength offset data of the rail to be monitored is obtained.

[0031] Center wavelength offset data is collected by a pre-installed group of sensing optical fibers in the rail to be monitored.

[0032] Figure 2 This is a schematic diagram of the spatial distribution of the sensing fiber optic gratings within the cross-section of the rail to be monitored, as shown below. Figure 2 As shown, three sensing fiber Bragg gratings are arranged in an isosceles or equilateral triangle within the cross-section of the rail to be monitored. The first sensing fiber Bragg grating is located on the upper surface of the rail bottom to the left of the centerline of the rail waist; the second sensing fiber Bragg grating is located on the upper surface of the rail bottom to the right of the centerline of the rail waist; and the third sensing fiber Bragg grating is located at the lower jaw of the rail head at the centerline of the rail waist. The triangular spatial topology ensures that each sensing fiber Bragg grating maintains a different distance and orientation angle from the geometric center of the rail cross-section.

[0033] The sensing fiber optic assembly includes three sensing fiber gratings arranged in a triangle within the cross-section of the rail to be monitored. The rail to be monitored is divided into multiple measurement sections along its longitudinal direction. During the dynamic operation of the train, the rail to be monitored bears a complex combination of spatial loads, including the vertical downward pressure from the wheels, the lateral thrust caused by the swaying motion, and the torsional force caused by the irregular swaying of the vehicle body. By arranging the sensing fiber optic assembly in the rail to be monitored, it is possible to perceive the stress distribution changes of the material matrix in real time deep inside the rail structure.

[0034] In order to comprehensively capture the stress state in different directions, the sensing fiber array is specifically configured as an array structure containing three sensing fiber gratings. These three sensing fiber gratings are arranged in a triangle within the cross-section of the rail to be monitored. The spatial topology of the triangle ensures that each sensing fiber grating maintains a different distance and orientation angle from the geometric center of the rail cross-section.

[0035] Three sensing fiber optic gratings are arranged in an isosceles or equilateral triangle within the cross-section of the rail to be monitored. For example, the first grating is located on the upper surface of the rail bottom to the left of the center line of the rail web, the second grating is located on the upper surface of the rail bottom to the right of the center line of the rail web, and the third grating is located at the lower jaw of the rail head at the center line of the rail web. The side length of the triangle is approximately 40-80mm, which can be determined according to the cross-sectional dimensions of the rail model.

[0036] When the rail under monitoring undergoes vertical bending, lateral bending, or torsional deformation as a train passes, the three sensing fiber optic gratings located at different spatial coordinates will generate tensile or compressive strains of varying values. By injecting optical signals into the sensing fiber optic group using a broadband swept-frequency light source, the selective reflection characteristics of the grating structure cause light waves of a specific wavelength to be reflected along the original path.

[0037] Changes in environmental strain can cause mechanical deformation of the grating period and elasto-optical changes in the refractive index of the fiber core material, resulting in a continuous drift of the center wavelength of the reflection spectrum. Collecting center wavelength shift data can quantitatively reflect the magnitude of the axial strain borne by the fiber grating at that location.

[0038] To achieve long-distance continuous monitoring, the rail to be monitored is divided into multiple measurement sections at equal intervals or as needed along the longitudinal direction of the line. Each measurement section is equipped with three sensing fiber optic gratings arranged in a triangular pattern, forming a distributed sensing network covering the entire line.

[0039] The longitudinal spacing of the measurement sections is usually 0.5m-2m. For high-speed railways or heavy-haul lines, a smaller value, such as 0.5m-1m, is preferable to capture short-wave irregularities. For conventional speed lines, a larger value, such as 1m-2m, can be used. The spacing should be increased in turnout areas, bridge joints, or historical defect points.

[0040] Multiple measurement sections distributed along the train's direction of travel are simultaneously used to collect data, converging discrete center wavelength offset data into the demodulation device. The densely distributed sampling points provide a data foundation for subsequent splicing and reconstruction of a complete and coherent rail space curve.

[0041] The center wavelength of the sensing fiber optic grating is usually selected in the C-band (1525nm-1565nm) or L-band, the grating length is 5mm-10mm, the reflectivity is greater than 80%, and the 3dB bandwidth is less than 0.3nm. The demodulation equipment adopts a wavelength scanning type or edge filtering type demodulator, and the sampling frequency is not less than 10 data points per axle at the train's maximum passing speed. For example, when the train speed is 350km / h and the bogie wheelbase is 2.5m, the sampling frequency should not be less than 388Hz.

[0042] In step S102, the low-frequency background strain baseline is determined using the center wavelength offset data.

[0043] Based on the center wavelength offset data, the strain initial value sequence corresponding to each measurement section is determined, and the time-series strain curve is extracted based on the preset buffer window. The time-series strain curve is subjected to extreme value optimization processing to determine the low-frequency background strain baseline characterizing longitudinal shear interference.

[0044] The wavelength change of the optical signal cannot be directly used to calculate the mechanical state of the track. However, the information in the wavelength dimension can be mapped to the numerical value of the engineering strain dimension. Through linear calibration coefficients and ambient temperature compensation parameters, the discrete center wavelength offset data is converted point by point into the initial strain value reflecting the tensile or compressive degree of the material. The values ​​are then arranged and combined in chronological order to form a coherent strain initial value sequence.

[0045] Temperature reference fiber optic gratings are placed near the rail, such as on the sleepers, to monitor the wavelength drift caused by ambient temperature in real time. The measured wavelength shift of the sensing fiber optic grating is subtracted from the wavelength shift of the temperature reference fiber optic grating to obtain the net wavelength shift caused only by mechanical strain. The strain sensitivity constant is then substituted to calculate the initial strain value. Alternatively, a dual-wavelength fiber optic grating can be used to achieve temperature self-compensation.

[0046] Extreme value optimization processing of time-series strain curves can be achieved through local minimum value extraction and baseline interpolation reconstruction. Within each local evaluation interval, a sliding window minimum value search algorithm is adopted: the window length is set, the strain minimum value point is located in each window, and pseudo-minimum values ​​caused by noise are eliminated. Pseudo-minimum values ​​can be filtered by difference thresholding with the minimum values ​​of adjacent windows. If the difference exceeds the preset threshold, it is determined to be a pseudo point and eliminated. Cubic spline interpolation or Akima interpolation is performed on the retained effective minimum value points to obtain a continuous low-frequency background strain baseline.

[0047] The dynamic effects of a passing train are characterized by transient occurrence and rapid decay. Therefore, a preset buffer window is used to extract effective data segments on the time axis that closely match the passing train event, integrating the discrete strain initial value sequence into a time-series strain curve with time-related attributes.

[0048] In addition to the elastic deformation caused by the vertical downward pressure of the wheel, the time-series strain curve may also include the longitudinal shear force exerted on the rail by the train traction or braking process, the creep relaxation of the rail itself, and the long-period slow deformation caused by the temperature gradient. These slow change components caused by indirect rolling usually show a low-frequency drift trend on the time axis.

[0049] The process of extreme value optimization on the time-series strain curve aims to extract the potential slowly varying baseline from the complex curve with high-frequency fluctuations, and to search for extreme characteristic points that can represent the gap without direct wheel action within the local interval of the time-series strain curve.

[0050] By continuously extracting multiple such extreme feature points and connecting them using curve fitting or interpolation, the low-frequency background strain baseline can be reconstructed. The low-frequency background strain baseline characterizes the slow drift component caused by the combined effects of longitudinal shear disturbance and environmental factors.

[0051] In one embodiment, the center wavelength offset of each measurement section at the current moment is determined based on the initial reference wavelength of the calibration, and the center wavelength offset is mapped to the strain initial value of the corresponding spatial location to generate a strain initial value sequence; the strain initial value sequence is pushed into the cache queue associated with the preset cache window according to the acquisition order, and the current fluctuation amplitude corresponding to the latest strain initial value stored in the cache queue is monitored; in response to the current fluctuation amplitude being greater than the preset static environmental noise floor threshold, the time series range corresponding to the preset cache window is determined as the effective data segment to generate a time-series strain curve.

[0052] The initial reference wavelength was calibrated during the late night when the rails were completely unloaded, the temperature was uniform, and there was no external seismic interference. The steady-state wavelength parameter was obtained by taking the average of multiple measurements of the sensing fiber optic gratings at each measurement section.

[0053] By subtracting the corresponding initial reference wavelength from the optical wavelength value collected in real time during train operation, the center wavelength offset that eliminates the influence of manufacturing tolerances and installation residual stress can be obtained.

[0054] By utilizing the strain sensitivity constant of the fiber optic grating, the center wavelength offset is divided by the strain sensitivity constant to complete the numerical conversion from the optical domain to the mechanical domain, generating initial strain values ​​that reflect the stress state at the corresponding spatial location. As the sampling frequency continues to advance, the initial strain values ​​are spliced ​​together according to the time sequence to form a continuously extending sequence of initial strain values.

[0055] To balance memory usage and data real-time performance, the capacity of the preset buffer window can be set to accommodate, for example, all sampling point arrays from 2 seconds before a train passes through to 5 seconds after it passes through. The strain initial value sequence is pushed into the buffer queue for temporary storage according to the first-in-first-out management rule.

[0056] To avoid redundant calculations of invalid background noise when no trains are passing, the program can continuously calculate the deviation between the latest initial strain value stored in the buffer queue and the historical average value, and use this as the current fluctuation amplitude. The amplitude reflects the transient severity of the rail under external disturbances.

[0057] The preset static environmental noise threshold is a boundary reference value set based on the maximum envelope of environmental noise under long-term no-vehicle conditions. It is usually taken as 3 to 4 times the standard deviation of environmental noise. When the current fluctuation amplitude is greater than the preset static environmental noise threshold, it can be determined that the train wheelset has entered the mechanical sensing range of the current measurement section.

[0058] The time series range corresponding to the preset cache window is determined, and continuous data within this range is extracted as valid data segments. These data are compact and contain all strain details when the train passes over them. They are then used to assemble and generate time-series strain curves for subsequent in-depth analysis, reducing the computational power consumption of invalid time periods and improving the sensitivity of sudden event capture.

[0059] In one embodiment, the time-series strain curves corresponding to each measurement section are extracted, and the time-series strain curves are divided into multiple local evaluation intervals on the time axis. Within each local evaluation interval, the strain minimum point reflecting the wheel crushing gap state is extracted, and the multiple continuously extracted strain minimum points are connected and numerically interpolated on the time axis to obtain the low-frequency background strain baseline.

[0060] The time-series strain curve contains all strain fluctuations during the period when the train passes through a measurement section. The train is composed of multiple carriages and bogies, and the rolling action of the wheels exhibits obvious periodic pulse characteristics on the time axis.

[0061] To accurately capture the troughs between these periodic pulses, the time-series strain curve can be divided into multiple local evaluation intervals at equal intervals along the time axis or based on the peak intervals.

[0062] The length of the local evaluation interval is adaptively adjusted based on the train speed and axle spacing to ensure that each local evaluation interval contains at least one complete wheel-to-wheel gap time period. Within each local evaluation interval, the local differential zero-finding rule or minimum value comparison algorithm is used to search for the moment when the strain value drops to the local minimum and the corresponding strain minimum point.

[0063] These minimum strain points correspond to an intermediate equilibrium state where neither the front nor rear wheels are directly pressing on the current measurement section, and the rail is recovering from inertia and material elasticity but has not yet fully recovered. Therefore, these minimum strain points reflect the macroscopic, slowly varying load effect added to the rail.

[0064] Since the minimum strain points are discrete coordinates that are discontinuous on the time axis, in order to construct a continuous benchmark that can cover the entire train passage period, cubic spline interpolation function or polynomial smoothing fitting method can be used to perform connection and numerical interpolation operations on multiple minimum strain points. The resulting low-frequency background strain baseline is a curve that smoothly passes through each trough point, showing the bottom drift trend caused by longitudinal shear and temperature drift effects.

[0065] In step S103, the amount of curvature change and the amount of torsion angle change are determined.

[0066] By subtracting the low-frequency background strain baseline from the time-series strain curve, the high-frequency pulse response parameter characterizing the vertical crushing effect is obtained. The strain increment corresponding to the high-frequency pulse response parameter at the target peak time is extracted. Based on the strain increment and the triangular spatial position of the sensing fiber grating, the curvature change and torsion angle change corresponding to each measurement section are determined.

[0067] After obtaining the low-frequency background strain baseline characterizing the slowly varying interference, this interference can be separated from the original mixed signal. The time-series strain curve containing all strain information is subtracted from the corresponding low-frequency background strain baseline one by one according to the time node, so as to achieve high- and low-frequency separation of the signal.

[0068] The remaining part after removing the low-frequency components is the high-frequency pulse response parameter. The high-frequency pulse response parameter eliminates the long-period trend deviation and only retains the high-frequency, large-amplitude strain response caused by the direct downward pressure of the wheel itself under heavy load. The high-frequency pulse response parameter can characterize the rail's ability to withstand the vertical rolling effect.

[0069] At the instant the train wheelset passes the measurement section, the vertical crushing effect reaches a local extreme, and the deformation of the track is the most intense and representative. By traversing the high-frequency pulse response parameters, the target peak moment with the largest amplitude is located, and the strain increment under this specific transient is extracted as the input reference for subsequent spatial calculation.

[0070] Since the strain increment of a single optical fiber only represents the local stretching and shortening at its laying location, it cannot directly describe the overall spatial bending and torsion state of the rail. It can be comprehensively analyzed by combining the geometric layout of multiple sensors. The three sensing fiber gratings present a unique triangular spatial position in the cross-section, so that each of the three sensing fiber gratings feels the lever tension and pressure at different neutral axis distances.

[0071] By substituting the strain increments at three different locations into the mechanical equation matrix, the degree of bending of the cross section around the horizontal axis, bending around the vertical axis, and rotation around the geometric center can be calculated. This allows for the accurate determination of the curvature change and torsional angle change corresponding to each measured cross section, thus converting one-dimensional linear strain into three-dimensional cross-sectional attitude characteristics.

[0072] Let the centroid of the rail section be the origin, and establish a local coordinate system with the x-axis along the transverse direction of the rail, the y-axis along the perpendicular direction of the rail, and the z-axis along the longitudinal direction of the rail; the coordinates of the three sensing fiber optic gratings are respectively ( x 1, y 1) ( x 2, y 2) and ( x 3, y 3).

[0073] According to the plane section assumption in mechanics of materials, at any point on the section ( x ,y The axial strain of ) can be expressed as: e ( x , y )= e 0+ kx y yes x + i oh ( x , y ),in e 0 represents axial strain. kx Let be the vertical curvature about the x-axis. yes Let be the transverse curvature about the y-axis. i Let be the angle of twist about the z-axis. oh ( x , y ) is the warping function.

[0074] For solid steel rails, it can be approximated as follows: oh ( x , y )= x y Alternatively, the Saint-Venant warping solution can be used, substituting the positions of the three optical fibers to obtain a 3×3 or 3×4 linear equation system. The deformation components can then be decoupled and obtained by matrix inversion or least squares solution.

[0075] In one embodiment, the original strain values ​​corresponding to each time node on the time-series strain curve are extracted, and the low-frequency background strain baseline corresponding to each time node on the low-frequency background strain baseline is extracted; the original strain values ​​are subtracted from their respective low-frequency background strain baselines to obtain high-frequency response differences; the high-frequency response differences corresponding to consecutive time nodes are time-series reassembled to obtain high-frequency pulse response parameters.

[0076] To ensure the alignment of the signal subtraction operation on the time axis, the time sequence can be traversed at a set sampling frequency to extract the original strain values ​​of each time node on the time-series strain curve without filtering modification; under the same time coordinate, the low-frequency background strain baseline at the corresponding position on the fitted low-frequency background strain baseline can be extracted.

[0077] In the array computing unit of the microprocessor, the original strain values ​​at the same time point can be subtracted one by one from the low-frequency background strain base. The difference result eliminates the influence of the background trend and obtains the high-frequency response difference excited by the instantaneous dynamic load.

[0078] Discrete high-frequency response differences cannot describe the complete stress change waveform on their own. Therefore, the high-frequency response differences can be spliced ​​together one by one according to the original time progression order to form a time-series reassembly set. The reassembled sequence removes the original baseline tilt phenomenon and presents multiple pulse clusters that are symmetrical or regularly distributed with the horizontal axis as the reference, which constitute a complete high-frequency pulse response parameter and improves the anti-interference margin when extracting feature points in the later stage.

[0079] In one embodiment, the strain change amplitude of the high-frequency pulse response parameter on the time axis is traversed, and the time node when the high-frequency pulse response parameter reaches the highest amplitude state is extracted as the target peak time; the strain values ​​corresponding to the three sensing fiber gratings on the same measurement section at the target peak time are obtained simultaneously, and the three strain values ​​corresponding to the three sensing fiber gratings at the target peak time are merged to confirm the multi-dimensional strain increment.

[0080] The high-frequency pulse response parameter exhibits a dense pattern of alternating peaks and valleys. Each peak corresponds to the mechanical response limit point of a train wheel rolling over the measured cross section. By comparing the amplitude data of all strain changes of the high-frequency pulse response parameter on the time axis, the instant when the amplitude reaches its peak can be determined.

[0081] The critical point at which the high-frequency pulse response parameter changes from an ascending state to a descending state can be extracted. The time node at which the highest amplitude is reached can be marked as the target peak time. To ensure the synchronization of multi-dimensional decoupling operations in the time dimension and to avoid the divergence of the spatial geometric matrix due to microsecond-level phase differences, data extraction can be performed under the unified scheduling of the main control program.

[0082] Simultaneously acquire the independent strain values ​​recorded by three sensing fiber optic gratings at multiple locations, including the bottom left end, bottom right end, and above the rail waist, at the target peak time; these three strain values ​​carry the tensile and compressive deformation information of different regions of the cross section under ultimate load.

[0083] The three strain values ​​corresponding to the target peak time of these three sensing fiber gratings are packaged and spliced ​​according to a fixed channel sequence, and merged to confirm a multi-dimensional strain increment vector containing multiple independent components.

[0084] In one embodiment, the strain values ​​corresponding to the three sensing fiber optic gratings are substituted into a spatial structure conversion mapping model constructed based on the preset triangular spatial position; the spatial structure conversion mapping model is used to convert the strain values ​​into the vertical curvature change, the transverse curvature change, and the torsional angle change of the current measurement section around the horizontal axis.

[0085] By using the inverse matrix operation of the coefficient matrix in the spatial structure conversion mapping model, the mixed strain vector can be decoupled into independent deformation components. Each row of the inverse matrix corresponds to an extraction filter for a deformation component, and the linear separation of each component is achieved through matrix multiplication.

[0086] For open, thin-walled sections like rails, torsional strain calculations must consider the warping constraint effect. In this embodiment, since the measurement section spacing is small (usually less than the warping wavelength), and the focus is primarily on the change in torsional angle rather than the warping normal stress, the Saint-Venant free torsion theory is used as an approximation: the shear strain caused by torsion is linearly distributed along the wall thickness, and the corresponding axial warping strain is antisymmetrically distributed within the section; the three fiber gratings are arranged in a triangle, and the torsional component can be separated using their asymmetric tensile-compressive response difference; for higher accuracy requirements, the Vlasov constrained torsion theory can be introduced to correct the warping function.

[0087] As a complex beam-like solid structure, the bending and torsion of rails conform to the spatial deformation theory of mechanics of materials. The axial strain at any point on the cross-section is a comprehensive result of the superposition of vertical bending moment, lateral bending moment and axial torque.

[0088] The pre-defined spatial position of the triangle is based on the centroid or shear center of the rail as the origin of the coordinate system, reflecting the lateral and vertical offset coordinates of the three sensing fiber optic gratings. Based on the generalized Hooke's law and the plane section assumption, a transformation inverse matrix composed of the sensor coordinate matrix and the bending and torsional section modulus parameters of the material can be constructed. The transformation inverse matrix is ​​the spatial structure conversion mapping model.

[0089] The strain numerical vector containing three components, extracted in the previous stage through three sensing fiber optic gratings, is used as the input parameter and substituted into the spatial structure conversion mapping model for matrix multiplication. Through the cross-separation of the coefficient matrix in the spatial structure conversion mapping model, the mixed stress can be decomposed, and the output end independently generates the vertical curvature change that characterizes the degree of rail head-down or head-up deformation, and the lateral curvature change that characterizes the degree of rail leftward or rightward bending.

[0090] In addition to the two bending modes, the spatial structure conversion mapping model also uses the asymmetric tension-compression response difference of the three vertices of the triangle to derive and calculate the change in torsional angle that characterizes the degree of rolling rotation of the rail around its own central axis.

[0091] In step S104, the track deformation monitoring results of the rail to be monitored are determined.

[0092] The curvature change is integrated and spliced ​​according to the distribution order of each measurement section, and the three-dimensional deformation data is determined by combining the torsion angle change, thus determining the track deformation monitoring result of the rail to be monitored.

[0093] The changes in vertical and lateral curvature on a single measurement section only reflect the degree of bending at that local location and cannot intuitively show the macroscopic deflection curve fluctuations of a rail hundreds of meters long. Since curvature is the second spatial derivative of the deflection curve with respect to distance, the displacement curve must be reconstructed by performing two mathematical integrals along the longitudinal extension of the rail.

[0094] According to the geographical distribution order of each measurement section on the track, the discrete curvature changes are arranged into a spatial sequence. After determining the reference boundary conditions with no relative displacement, a spatial integration splicing operation is performed along the longitudinal mileage coordinate. By accumulating the small bending angles and deflection increments between adjacent sections, the continuous displacement trend of the rail in the vertical and horizontal planes is reconstructed.

[0095] The displacement curve data obtained by longitudinal integration is combined with the torsional angle change calculated independently on each measurement section by vectorization, and the spatial attitude attribute of the displacement curve is given, thus forming complete and realistic three-dimensional deformation data.

[0096] The three-dimensional deformation data comprehensively reproduces the complex micro-morphology of the rails under the pressure of heavy-load trains, such as serpentine meandering, undulating waves, and local overturning. Combined with the allowable deviation threshold in the track maintenance specifications, the extreme points and abrupt change points in the three-dimensional deformation data are quantitatively benchmarked and over-limit judged to determine the track deformation monitoring results containing fault type and fault location information, thus realizing the visual rating of the track status.

[0097] In one embodiment, the linear spacing parameters of each measurement section along the layout route are extracted, and a stable measurement section located at the far end and not subject to external force disturbance is used as a reference point for spatial calculation. Using the linear spacing parameters, the vertical curvature change and the lateral curvature change corresponding to each measurement section distributed along the longitudinal direction are accumulated step by step to obtain a continuous deflection fluctuation curve. The torsional angle change of all measurement sections is summarized to generate global three-dimensional attitude data, and the three-dimensional deformation data is determined by combining the continuous deflection fluctuation curve and the global three-dimensional attitude data.

[0098] The objectively existing straight-line spacing parameter between two adjacent measurement sections can be extracted along the actual layout route. The spacing parameter is used as the independent variable step length in the difference integral algorithm.

[0099] Select a measurement section that is stable when subjected to load, located above the fixed support of the bridge pier or on the hard rock wall of the tunnel. Set the absolute displacement and rotation angle to zero as the reference benchmark for the entire spatial displacement calculation system, thus eliminating the drift uncertainty caused by the unknown constant term in the integration process.

[0100] By using the linear spacing parameter and combining it with the trapezoidal integral formula or Simpson's integral rule, the calculation can be steadily advanced from the reference point along the longitudinal distribution direction. The vertical curvature change corresponding to each measurement section is accumulated twice in the distance domain to derive the vertical displacement. Referring to the method of obtaining the vertical displacement, the lateral curvature change can be accumulated step by step to derive the lateral displacement.

[0101] By combining lateral and longitudinal displacement data with the mileage coordinates of the measurement section, a smooth and intuitive continuous deflection fluctuation curve reflecting the macroscopic curvature of the track is plotted. The torsional angle changes scattered at each node are summarized and arranged according to the mileage coordinates to generate global three-dimensional attitude data reflecting the tilting and overturning of the track at various points.

[0102] By using an affine transformation matrix, the three-dimensional spatial coordinate vectors of each node on the continuous deflection undulation curve are orthogonally fused with the normal rotation tensor in the global three-dimensional attitude data to determine the three-dimensional deformation data, which includes the spatial centerline displacement coordinates and the rotation attitude of the normal section, thus restoring the true three-dimensional distortion of the track under the heavy pressure of the train.

[0103] In one embodiment, when the rail under monitoring is in a light-load operation state, the reference value of the baseline subsidence characteristic corresponding to all measurement sections along the entire line can be recorded; when the rail under monitoring is in a heavy-load operation state, the actual subsidence test value generated by the current measurement section under heavy load force can be obtained; the actual subsidence test value is subtracted from the reference value of the baseline subsidence characteristic to obtain the subsidence surge difference value; in response to the subsidence surge difference value being greater than the preset settlement surge threshold and deviating from the normal deformation range of the adjacent sections, the current measurement section is positioned as a weak support point of the track bed; when performing integral splicing, the output value of the curvature change corresponding to the weak support point of the track bed is limited by the preset elastic inhibition coefficient.

[0104] Uneven settlement of the track bed base can cause local rails to lose effective support and undergo abnormal and severe deflection under train load. In order to capture this hidden danger, when the rail to be monitored is in the state of passing light railcars without passengers or single machine light load operation, the settlement caused by the elastic deformation of the foundation itself is collected and recorded as the reference value of the baseline subsidence characteristics.

[0105] When a heavy-load freight train or a fully loaded passenger car passes at high speed, and the rail to be monitored is in a heavy-load operation state, the more significant actual subsidence test value generated by the current measurement section under the action of axle load is obtained; the actual subsidence test value containing extreme deformation is subtracted from the benchmark subsidence characteristic reference value.

[0106] The differential value is used to separate the common elastic settlement of the track foundation and obtain the differential value of the sudden increase in settlement caused by the surge in load. The preset settlement change threshold is a tolerance limit set by combining the stiffness of the track pad and the normal compaction of the track bed.

[0107] When the sudden increase in subsidence exceeds the preset settlement abruptness threshold, and when it is compared with the deformation of the adjacent sections before and after it is found to be significantly beyond the normal deformation range, exhibiting isolated settlement spike characteristics, it can be determined that the sleeper or ballast support has failed and is suspended, thus accurately locating the current measurement section as the weak support point of the track bed.

[0108] Because the strain at the weak support points of the track bed is extremely large, directly incorporating it into the spatial integration algorithm may cause severe trapezoidal integral divergence and artifacts, resulting in huge distortion of the deformation curve along the entire line. Therefore, when performing integral splicing, a preset elastic suppression coefficient of less than 1 can be introduced and multiplied by the curvature of the outlier point to limit and attenuate the output value of the curvature change corresponding to the weak support point of the track bed, thus maintaining the rigid smoothness and numerical stability of the entire continuous deflection undulation curve.

[0109] In one embodiment, the initial strain values ​​corresponding to two adjacent time nodes arranged in chronological order can be extracted from the initial strain value sequence. The difference between the initial strain values ​​corresponding to the two adjacent time nodes is obtained by subtracting the difference. The strain increment rate is determined based on the time interval between the strain difference and the two adjacent time nodes. The preset upper limit reference value of the elastic deformation rate corresponding to the rail to be monitored is extracted. In response to the strain increment rate being greater than the preset upper limit reference value of the elastic deformation rate, the time node corresponding to the strain increment rate is determined to be the time node where a transient high-energy impact occurs. The initial strain value collected at the corresponding time node is marked as distorted saturated acquisition data.

[0110] Severe working conditions such as wheel tread abrasion or track wavy wear can exert a high-energy instantaneous hammering effect on the rail, causing nonlinear distortion of the sensor signal that exceeds the range. In the data preprocessing stage, the strain initial values ​​corresponding to two adjacent time nodes arranged in a close time sequence are extracted cyclically from the continuous strain initial value sequence.

[0111] The difference between the value at the next time node and the value at the previous time node is used to obtain the strain difference value, which represents the local fluctuation amplitude. The strain difference value is divided by the fixed time interval between two adjacent time nodes to obtain the first derivative with respect to time, which is the strain increment rate of change.

[0112] The elastic deformation rate of rail material is constrained by the stress wave propagation speed and Young's modulus, and there is an objective physical upper limit. The upper limit of the upper limit can be converted into a preset reference value for the upper limit of elastic deformation rate. If the calculated rate of change of strain increment is greater than the preset reference value for the upper limit of elastic deformation rate, it can be determined that a violent rigid impact has occurred at the wheel-rail contact surface.

[0113] The transient high-energy impact phenomenon occurs at the instantaneous time node corresponding to the strain increment rate of change, and the optical wavelength demodulation result is affected by violent oscillation. In order to avoid the destructive impact of such step noise on the subsequent integration and extreme value optimization steps, the initial strain values ​​that exhibit abnormal jump characteristics collected at the corresponding time node can be labeled with abnormal attributes in the database and marked as distorted saturated acquisition data, so as to provide a reference for subsequent targeted isolation and repair mechanisms.

[0114] In one embodiment, after marking the initial strain values ​​collected at the corresponding time nodes as distorted saturated acquisition data, the distorted saturated acquisition data in the preset buffer window can be removed; the first stable historical data sequence within the preset first acquisition period span before the transient high-energy impact is extracted by backtracking in the time series; the second stable recovery data sequence within the preset second acquisition period span after the dissipation of the transient high-energy impact is extracted by delaying in the time series; a smooth interpolation bridging operation is performed on the time series based on the first stable historical data sequence and the second stable recovery data sequence to obtain the transition smoothing test parameters, and the transition smoothing test parameters are written into the preset buffer window to replace the distorted saturated acquisition data.

[0115] After confirming that there is local distortion in the signal, a pointer positioning instruction can be executed to zero out or remove the distorted saturated acquisition data with abnormal attribute labels in the preset buffer window, preventing it from flowing into the main control parsing matrix.

[0116] However, directly discarding data may cause a break in the time-series strain curve on the time axis, which may lead to step size matching errors and frequency domain leakage in the integration algorithm. To address this, the time node where the break occurs can be used as the reference point, and the time series can be traced back in reverse order to extract the first stable historical data sequence that is not affected by transient shocks and is within the preset first acquisition period span.

[0117] By delaying the time sequence and waiting for the oscillation packet to completely decay, the second stable recovery data sequence, which is within the preset second acquisition period span, is extracted and returned to a stable stress state. These two uncontaminated data sequences truly reflect the mechanical transition trend of the rail before and after the impact event.

[0118] The first acquisition cycle is preset to span 0.5-2 seconds before the transient impact (approximately 50-200 data points, depending on the sampling frequency), and the second acquisition cycle is preset to span 0.5-2 seconds after the transient impact; it includes at least one complete wheel-rail contact cycle and ensures that the data is in a steady-state stress state. The steady-state criterion can be used by verifying that the strain standard deviation within the test period is less than twice the standard deviation before the impact.

[0119] By calling the high-order polynomial fitting library function that includes spline curve approximation, based on the first stationary historical data sequence and the second stationary recovered data sequence, smooth interpolation bridging operation can be performed across the fracture interval, taking into account the continuity of the first and second derivatives of the endpoints of the data sequences on both sides of the fracture.

[0120] The transition smoothing test parameters with natural trends and no jumps are derived and reconstructed. The transition smoothing test parameters are written into the corresponding time-series blank area in the preset cache window to replace the original distorted saturated acquisition data, ensuring the continuity and fidelity of the dynamic deformation waveform.

[0121] In one embodiment, determining the track deformation monitoring results of the rail to be monitored includes: extracting the continuous deflection fluctuation curve and global three-dimensional attitude data from the three-dimensional deformation data, and obtaining the pre-set standard alignment change envelope and angle safety limit; using the continuous deflection fluctuation curve and the standard alignment change envelope, determining the over-limit area from the continuous deflection fluctuation curve; using the global three-dimensional attitude data and angle safety limit to determine the torsion angle deviation value; performing trend fitting on the continuous deflection fluctuation curve to obtain the overall track alignment data; mapping the distance coordinates of the over-limit area, the maximum subsidence of the over-limit area, the torsion angle deviation value, and the overall track alignment data to a multi-dimensional feature matrix to construct the track deformation monitoring results.

[0122] In addition to using isolated data fragments, assessing the safety status of a track requires systematic benchmarking and integration. This can be achieved by separating and extracting continuous deflection fluctuation curves reflecting displacement and global three-dimensional attitude data reflecting tilt and roll from a complete three-dimensional deformation data structure.

[0123] The standard alignment variable envelope, which is pre-set according to the train operating speed level, and the angle safety limits for curved and straight sections are retrieved from the track maintenance specification database. In the spatial rectangular coordinate system, the intersection and boundary checks of the coverage area of ​​the continuously deflected undulation curve calculated in real time and the standard alignment variable envelope are determined.

[0124] By scanning clusters of nodes whose differences exceed the safety threshold, we can locate and identify over-limit areas with settlement collapse or uplift anomalies from the continuous deflection fluctuation curve. We can then extract the normal flip angle from the global three-dimensional attitude data and use the normal flip angle to subtract the angle safety limit to determine the torsional angle deviation caused by gauge widening or unilateral settlement.

[0125] For the entire continuous deflection undulation curve, the least squares method or curve smoothing filtering algorithm can be applied to eliminate the trend fitting of local jumps and obtain the overall track alignment data that reflects the long-wave irregularity.

[0126] To make the output warning and assessment information more machine-readable and intuitively instructive, the exact distance coordinates of the over-limit area on the track line, the extreme maximum subsidence at the center of the over-limit area, the torsional angle deviation value of the cross-sectional instability characteristics, and the overall track alignment data describing the macroscopic geometry can be jointly compiled.

[0127] The aforementioned core indicators characterizing fault features are mapped to a multidimensional feature matrix according to dimensional order and structural definition. The multidimensional feature matrix comprehensively displays the synergistic correlation between microscopic defects and macroscopic deformation, and constructs track deformation monitoring results that include diagnostic ratings and maintenance targets, thereby reducing operation and maintenance costs and improving the overall operational safety factor of the track network.

[0128] Figure 3 This is a schematic diagram illustrating the structure of an online track deformation monitoring system 1000 based on a distributed fiber Bragg grating, according to an exemplary embodiment. (Refer to...) Figure 3 The distributed fiber Bragg grating track deformation online monitoring system 1000 includes a processor 1100 and a memory 1200. The memory 1200 stores computer program instructions, which, when executed by the processor 1100, implement all or part of the steps of the distributed fiber Bragg grating track deformation online monitoring method in this application.

[0129] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only.

[0130] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for online monitoring of track deformation of a distributed fiber Bragg grating, characterized in that, include: Center wavelength offset data is collected by a pre-set group of sensing optical fibers in the rail to be monitored. The sensing fiber optic assembly includes three sensing fiber gratings arranged in a triangle within the cross-section of the rail to be monitored. The rail to be monitored is divided into multiple measurement sections along the longitudinal direction. Based on the center wavelength offset data, the strain initial value sequence corresponding to each measurement section is determined, and the time-series strain curve is extracted based on the preset buffer window. The time-series strain curve is subjected to extreme value optimization processing to determine the low-frequency background strain baseline characterizing longitudinal shear interference. Subtract the low-frequency background strain baseline from the time-series strain curve to obtain the high-frequency pulse response parameter characterizing the vertical rolling effect. Extract the strain increment corresponding to the high-frequency pulse response parameter at the target peak time. Based on the strain increment and the triangular spatial position of the sensing fiber grating, determine the curvature change and torsion angle change corresponding to each measurement section. The curvature change is integrated and spliced ​​according to the distribution order of each measurement section, and the three-dimensional deformation data is determined by combining the torsion angle change, thus determining the track deformation monitoring result of the rail to be monitored.

2. The method for online monitoring of track deformation of a distributed fiber Bragg grating according to claim 1, characterized in that, Based on the center wavelength offset data, the initial strain value sequence corresponding to each measurement section is determined, and the time-series strain curve is extracted based on a preset buffer window, including: The center wavelength offset of each measurement section at the current moment is determined based on the initial reference wavelength of the calibration, and the center wavelength offset is mapped to the strain initial value of the corresponding spatial position to generate a strain initial value sequence. The strain initial value sequence is pushed into the cache queue associated with the preset cache window according to the acquisition order, and the current fluctuation amplitude corresponding to the latest strain initial value stored in the cache queue is monitored. In response to the current fluctuation amplitude being greater than the preset static environment noise threshold, the time series range corresponding to the preset cache window is determined as the effective data segment to generate the time series strain curve.

3. The method for online monitoring of track deformation of a distributed fiber Bragg grating according to claim 1, characterized in that, The time-series strain curves are subjected to extreme value optimization to determine the low-frequency background strain baseline characterizing longitudinal shear disturbance, including: Extract the time-series strain curves corresponding to each measurement section, and divide the time-series strain curves into multiple local evaluation intervals on the time axis; Within each local evaluation interval, strain minimum points reflecting the wheel crushing gap state are extracted. Multiple continuously extracted strain minimum points are connected on the time axis and numerical interpolation is performed to obtain the low-frequency background strain baseline.

4. The method for online monitoring of track deformation of a distributed fiber Bragg grating according to claim 1, characterized in that, Subtracting the low-frequency background strain baseline from the time-series strain curve yields high-frequency impulse response parameters characterizing the vertical rolling effect, including: Extract the original strain values ​​corresponding to each time point on the time-series strain curve, and extract the low-frequency background strain baseline corresponding to each time point on the low-frequency background strain baseline. The high-frequency response difference is obtained by subtracting the corresponding low-frequency background strain base from the original strain values. The high-frequency response difference corresponding to the continuous time nodes is then time-series reassembled to obtain the high-frequency impulse response parameter.

5. The method for online monitoring of track deformation of a distributed fiber optic grating according to claim 1, characterized in that, Extract the strain increment corresponding to the high-frequency impulse response parameters at the target peak time, including: Traverse the strain change amplitude of the high-frequency pulse response parameter on the time axis, and extract the time node when the high-frequency pulse response parameter reaches the highest amplitude state as the target peak time; The strain values ​​of three sensing fiber optic gratings at the target peak time are acquired simultaneously on the same measurement section. The three strain values ​​of the three sensing fiber optic gratings at the target peak time are combined and confirmed as a multi-dimensional strain increment.

6. The method for online monitoring of track deformation of a distributed fiber Bragg grating according to claim 1, characterized in that, Based on the strain increment and the triangular spatial position of the sensing fiber grating, the curvature change and torsion angle change corresponding to each measurement section are determined, including: The strain values ​​corresponding to the three sensing fiber optic gratings are substituted into the spatial structure conversion mapping model constructed based on the preset triangular spatial positions; The spatial structure conversion mapping model is used to convert the strain values ​​into the vertical curvature change, the transverse curvature change, and the torsional angle change of the current measurement section around the horizontal axis.

7. The method for online monitoring of track deformation of a distributed fiber Bragg grating according to claim 1, characterized in that, The curvature changes are integrated and stitched according to the distribution order of each measurement section, and the three-dimensional deformation data is determined by combining the torsional angle changes, including: Extract the straight-line spacing parameters of each measurement section along the layout route, and use the stable measurement section located at the far end and not subject to external force disturbance as the reference benchmark point for spatial calculation. By using the linear spacing parameter, the vertical curvature change and the transverse curvature change corresponding to each measurement section distributed along the longitudinal direction are accumulated step by step to obtain a continuous deflection fluctuation curve. The torsional angle changes of all measured sections are aggregated to generate global three-dimensional attitude data. The three-dimensional deformation data are determined by combining the continuous deflection fluctuation curve with the global three-dimensional attitude data.

8. The method for online monitoring of track deformation of a distributed fiber Bragg grating according to claim 1, characterized in that, The method further includes: When the rail under monitoring is in a light-load operation state, record the reference value of the benchmark subsidence characteristic corresponding to all measurement sections along the entire line; when the rail under monitoring is in a heavy-load operation state, obtain the actual subsidence test value generated by the current measurement section under heavy load. The difference in subsidence is obtained by subtracting the benchmark subsidence characteristic reference value from the actual subsidence test value. If the difference in subsidence exceeds the preset settlement change threshold and deviates from the normal deformation range of the adjacent cross sections, the current measurement cross section is located as a weak support point of the track bed. When performing integral splicing, the output value of the curvature change corresponding to the weak support point of the track bed is limited by the preset elastic inhibition coefficient.

9. The method for online monitoring of track deformation of a distributed fiber Bragg grating according to claim 1, characterized in that, The method further includes: Extract the initial strain values ​​corresponding to two adjacent time nodes in the strain initial value sequence arranged in time sequence, subtract the initial strain values ​​corresponding to the two adjacent time nodes to obtain the strain difference value, and determine the strain increment change rate based on the strain difference value and the time interval between the two adjacent time nodes. Extract the preset upper limit reference value of the elastic deformation rate corresponding to the rail to be monitored. In response to the strain increment change rate being greater than the preset upper limit reference value of the elastic deformation rate, determine the time node corresponding to the strain increment change rate to determine the transient high-energy impact, and mark the initial strain value collected at the corresponding time node as distorted saturated acquisition data.

10. The method for online monitoring of track deformation of a distributed fiber Bragg grating according to claim 1, characterized in that, Determine the track deformation monitoring results of the rail to be monitored, including: Extract the continuous deflection fluctuation curve and global three-dimensional attitude data from the three-dimensional deformation data, and obtain the pre-set standard linear deformation envelope and angle safety limit. Use the continuous deflection fluctuation curve and the standard linear deformation envelope to determine the over-limit area from the continuous deflection fluctuation curve. The deviation of the torsion angle is determined by using global three-dimensional attitude data and angle safety limits, and the overall track alignment data is obtained by trend fitting of the continuous deflection fluctuation curve. The distance coordinates of the out-of-limit area, the maximum subsidence of the out-of-limit area, the deviation of the torsion angle, and the overall track alignment data are mapped to a multi-dimensional feature matrix to construct the track deformation monitoring results.

Images