Rail bridge system track irregularity signal inversion method, system, device and medium

Abstract

The application provides a vehicle-rail bridge system track irregularity signal inversion method, system, device and medium, and belongs to the technical field of rail transit. The method is characterized in that a sensor array is arranged at a key position of a vehicle-rail bridge system, multi-source time-domain original signals are collected, and dynamic parameters and three-dimensional coordinates of the sensors are obtained. After differential preprocessing, features related to vertical and lateral track irregularities are extracted, and a feature matrix is constructed. By using dynamic graph modeling and dynamic coupling calculation, a signal transmission contribution degree matrix and a dynamic response prediction value set are output, and then the track irregularity time-domain signal is inverted and optimized, and finally the accuracy of the inverted signal is verified by a vehicle-rail bridge dynamic positive model. The vehicle-rail bridge system track irregularity signal inversion method, system, device and medium can realize high-precision inversion of the track irregularity signal, and provide technical support for safe operation and maintenance of rail transit.

Description

Technical Field

[0001] This invention relates to the field of rail transit technology, and in particular to a method, system, equipment, and medium for inverting track irregularity signals in a vehicle-track-bridge system. Background Technology

[0002] Rail transit, as an efficient and high-capacity mode of transportation, occupies an important position in the modern transportation system. However, track irregularities are one of the key factors affecting the safety, comfort, and operational efficiency of rail transit. Track irregularities cause additional vibrations and shocks during train operation, which not only reduces passenger comfort but may also cause fatigue damage to vehicle components, increase maintenance costs, and even threaten operational safety in extreme cases.

[0003] Traditional methods for detecting track irregularities primarily rely on equipment such as track inspection vehicles to directly measure track geometric parameters to obtain irregularity information. However, these methods have some limitations. On the one hand, the inspection frequency of track inspection vehicles is relatively low, making it impossible to monitor track conditions in real time and continuously. On the other hand, there is a certain difference between the directly measured geometric parameters and the irregularities experienced by the vehicle during actual operation, and these methods cannot fully and accurately reflect the true situation under the interaction between the vehicle and the track.

[0004] In recent years, with the development of sensor technology, signal processing technology, and dynamic modeling technology, the method for inverting track irregularity signals based on vehicle-track-bridge coupled systems has gradually become a research hotspot. By deploying sensors at key locations in the vehicle-track-bridge system to collect vibration, displacement, and other signals during vehicle operation, and then using dynamic models and signal processing algorithms to invert track irregularity signals, the track condition can be reflected more timely and accurately, providing strong support for the maintenance and management of rail transit.

[0005] However, existing methods for inverting track irregularity signals in vehicle-rail-bridge systems still have some shortcomings. For example, some methods are not refined enough in the signal preprocessing and feature extraction stages, resulting in low inversion accuracy; some methods fail to fully consider the complex coupling characteristics of the vehicle-rail-bridge system during dynamic coupling calculations, causing some deviation between the inversion results and the actual situation; and some methods lack effective constraints and correction mechanisms in the irregularity signal inversion optimization stage, affecting the reliability and practicality of the inverted signals.

[0006] Therefore, there is an urgent need for a more accurate, efficient and adaptable method for inverting track irregularity signals in vehicle-rail-bridge systems to overcome the shortcomings of existing technologies. Summary of the Invention

[0007] The purpose of this invention is to provide a method, system, equipment, and medium for inverting track irregularity signals in a vehicle-rail-bridge system. Through multi-source signal acquisition, feature extraction, dynamic modeling, and optimized inversion technology, it achieves high-precision reconstruction of vertical and lateral track irregularity signals, thereby providing reliable technical support for the safe operation, fault early warning, and precise maintenance of rail transit, and improving the overall operational efficiency and safety of rail transit.

[0008] To achieve the above objectives, the present invention provides a method for inverting track irregularity signals in a vehicle-rail bridge system, comprising the following steps:

[0009] Step S1: Deploy sensor arrays on the operating vehicles, track structures, and key sections of the bridge in the vehicle-rail-bridge system to collect multi-source time-domain raw signals; synchronously acquire the dynamic parameters of the vehicle-rail-bridge and the three-dimensional physical coordinates of all sensors;

[0010] Step S2: Perform differential preprocessing on the multi-source time-domain raw signals obtained in Step S1 to obtain multi-source time-domain clean signals; extract features related to track vertical irregularities and track lateral irregularities; construct vertical sub-feature matrix, lateral sub-feature matrix and total feature matrix based on the extracted features;

[0011] Step S3: Perform dynamic graph modeling and dynamic coupling calculation, and output the signal transmission contribution matrix and the set of predicted dynamic response values;

[0012] Step S4: Perform irregularity signal inversion and optimization to obtain the final time-domain signals of vertical and lateral track irregularities;

[0013] Step S5: Input the final track vertical irregularity time domain signal and the final track lateral irregularity time domain signal obtained in step S4 into the vehicle-track-bridge dynamics positive model to verify the consistency between the predicted response output by the model and the multi-source time domain clean signal obtained in step S2.

[0014] Preferably, in step S1, the sensor array includes: a vertical vibration sensor, a lateral vibration sensor, a sound pressure sensor, a vertical displacement sensor, and a lateral displacement sensor; the multi-source time-domain raw signals include: the raw vertical vibration signal, the raw lateral vibration signal, the raw sound pressure signal, the raw vertical displacement signal, and the raw lateral displacement signal; the dynamic parameters include: vehicle vertical suspension stiffness, vertical suspension damping, lateral suspension stiffness, lateral suspension damping, wheel-rail vertical contact stiffness, wheel-rail lateral contact stiffness, and bridge bending stiffness.

[0015] Preferably, in step S2, the multi-source time-domain clean signal includes: vertical vibration clean signal, lateral vibration clean signal, sound pressure clean signal, vertical displacement clean signal, and lateral displacement clean signal; the vertical irregularity correlation feature satisfies the absolute value of the Pearson correlation coefficient ≥ 0.7, and the lateral irregularity correlation feature satisfies the absolute value of the Pearson correlation coefficient ≥ 0.7.

[0016] Preferably, in step S2, the vertical irregularity correlation features include: wheel-rail vertical force fluctuation coefficient, peak frequency of axle box vertical acceleration power spectrum, vertical displacement fluctuation of rail, and vertical impact sound pressure level; the lateral irregularity correlation features include: wheel-rail lateral force fluctuation coefficient, rail lateral displacement slope, and axle box vertical-lateral acceleration coupling coefficient.

[0017] Preferably, in step S3, dynamic graph modeling and dynamic coupling calculation are performed to output the signal transmission contribution matrix and the set of predicted dynamic response values, including the following steps:

[0018] Step S31: Construct the initial topology graph of the dual drive system. Using the sensor measurement points deployed in step S1 as nodes, and combining the three-dimensional physical coordinates of each sensor with the vertical and lateral dynamic parameters of the vehicle-rail-bridge system, generate an initial adjacency matrix containing vertical and lateral weights.

[0019] Step S32: Dynamically update the adjacency matrix. Based on the vertical sub-feature matrix and horizontal sub-feature matrix constructed in step S2 and the real-time train operation parameters, the initial adjacency matrix is ​​adjusted through an attention mechanism to obtain a time-varying adjacency matrix.

[0020] Step S33: Spatiotemporal convolution and dynamic coupling calculation. The multi-source temporal clean signal obtained in step S2 is input into the stacked spatiotemporal graph convolution module. During the convolution calculation process, the vehicle-rail-bridge coupled dynamic equation is embedded, and the vertical signal transmission contribution matrix, the lateral signal transmission contribution matrix and the set of dynamic response prediction values ​​are output.

[0021] Preferably, the set of predicted dynamic response values ​​includes predicted values ​​for vertical vibration response, predicted values ​​for lateral vibration response, predicted values ​​for vertical displacement response, and predicted values ​​for lateral displacement response.

[0022] Preferably, in step S4, the irregularity signal is inverted and optimized to obtain the final time-domain signals of vertical and lateral track irregularities, including the following steps:

[0023] Step S41: Locate the dominant vibration source. Based on the vertical signal transmission contribution matrix output in step S3, filter the sensor nodes corresponding to the dominant vertical vibration source of the wheel-rail system. Based on the lateral signal transmission contribution matrix, filter the sensor nodes corresponding to the dominant lateral vibration source of the wheel-rail system.

[0024] Step S42: Invert the vertical irregularity signal of the track. Combine the predicted value of the vertical vibration response with the residuals of the clean vertical vibration signal and the clean vertical displacement signal obtained in step S2. Solve the time domain signal of the vertical irregularity of the track using the weighted least squares iterative method. Introduce wavelength constraints to optimize the solution results and obtain the final time domain signal of the vertical irregularity of the track.

[0025] Step S43: Invert the track lateral irregularity signal, combine the predicted lateral vibration response with the residuals of the lateral vibration clean signal and the lateral displacement clean signal, and solve the track lateral irregularity time domain signal by weighted least squares iterative method. Introduce wavelength constraints and vertical-lateral coupling correction to optimize the solution results and obtain the final track lateral irregularity time domain signal.

[0026] Step S44: Determine the validity of the inversion and verify the amplitude and iterative convergence of the final orbital vertical irregularity time domain signal and the final orbital lateral irregularity time domain signal.

[0027] This invention also provides a track irregularity signal inversion system for a vehicle-track-bridge system, used to perform the track irregularity signal inversion method for a vehicle-track-bridge system as described above, including:

[0028] The sensor array module is used to deploy sensors on the operating vehicles, track structures, and key sections of the bridge in the vehicle-rail-bridge system to collect multi-source time-domain raw signals and obtain the dynamic parameters of the vehicle-rail-bridge and the three-dimensional physical coordinates of all sensors.

[0029] The signal preprocessing and feature extraction module is used to perform differentiated preprocessing on the multi-source time-domain raw signals to obtain multi-source time-domain clean signals, and extract features related to the vertical and lateral irregularities of the track to construct the vertical sub-feature matrix, the lateral sub-feature matrix and the total feature matrix.

[0030] The dynamic graph modeling and dynamic coupling calculation module is used to construct a dual-drive initial topology graph, dynamically update the adjacency matrix, perform spatiotemporal convolution and dynamic coupling calculation, and output a signal transmission contribution matrix and a set of dynamic response prediction values.

[0031] The irregularity signal inversion and optimization module is used to locate the dominant vibration source, invert the time-domain signals of vertical and lateral track irregularities, and introduce wavelength constraints and coupling corrections for optimization to obtain the final track irregularity time-domain signal.

[0032] The verification module is used to input the track irregularity signal obtained by inversion into the vehicle-track-bridge dynamics positive model to verify the consistency between the predicted response and the multi-source time-domain clean signal.

[0033] The present invention also provides a computer device, including: a memory and a processor; the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for inverting track irregularity signals in a vehicle-rail-bridge system.

[0034] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for inverting track irregularity signals in a vehicle-rail-bridge system.

[0035] Therefore, the present invention employs the above-mentioned method, system, equipment, and medium for inverting track irregularity signals in a vehicle-rail bridge system, and the beneficial technical effects are as follows:

[0036] (1) This invention acquires multi-source time-domain raw signals by deploying sensor arrays on the operating vehicles, track structure, and key sections of the bridge in the vehicle-rail-bridge system, and simultaneously obtains the dynamic parameters of the vehicle-rail-bridge and the three-dimensional physical coordinates of all sensors, thus comprehensively and accurately acquiring the dynamic information of the vehicle-rail-bridge system. Compared with traditional methods, this invention can more accurately invert the time-domain signals of vertical and lateral irregularities of the track, improving the inversion accuracy.

[0037] (2) This invention employs dynamic graph modeling and dynamic coupling calculation to construct a dual-drive initial topology graph and dynamically update the adjacency matrix. Combining spatiotemporal convolution and dynamic coupling calculation, it outputs a signal transmission contribution matrix and a set of predicted dynamic response values. This method can fully consider the complex coupling characteristics of the vehicle-track-bridge system. Compared with existing technologies, it can more accurately reflect the interaction between the vehicle and the track, and improve the reliability of the inversion results.

[0038] (3) In the process of inverting the irregularity signal, this invention introduces wavelength constraints and vertical-lateral coupling corrections to optimize the solution results and ensure the smoothness and accuracy of the inverted signal. Furthermore, the inverted track irregularity signal is input into the vehicle-track-bridge dynamics positive model for verification, ensuring the consistency between the model's predicted response and the multi-source time-domain clean signal. This process not only improves the reliability of the inverted signal but also enhances the robustness of the system.

[0039] (4) This invention performs differentiated preprocessing on multi-source time-domain raw signals, extracts features related to vertical and lateral track irregularities, and constructs vertical sub-feature matrices, lateral sub-feature matrices, and a total feature matrix. This multi-source signal fusion and feature extraction method can make full use of different types of sensor data, improve the accuracy and representativeness of features, and compared with existing technologies, can more effectively extract features related to track irregularities, providing more reliable data support for subsequent inversion calculations. Attached Figure Description

[0040] Figure 1This is a flowchart of the method for inverting track irregularity signals in the vehicle-rail bridge system according to the present invention;

[0041] Figure 2 This is a diagram illustrating the architecture of the track irregularity signal inversion method for the vehicle-rail-bridge system of the present invention. Detailed Implementation

[0042] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0043] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0044] Example 1

[0045] like Figure 1 As shown, the method for inverting track irregularity signals in a vehicle-rail bridge system includes the following steps:

[0046] Step S1: Deploy sensor arrays on the operating vehicles, track structures, and key sections of the bridge in the vehicle-rail-bridge system to collect multi-source time-domain raw signals; synchronously acquire the dynamic parameters of the vehicle-rail-bridge and the three-dimensional physical coordinates of all sensors.

[0047] The sensor array includes: vertical vibration sensors, lateral vibration sensors, sound pressure sensors, vertical displacement sensors, and lateral displacement sensors. Specifically, the vertical vibration sensors are deployed on axle boxes, rails, bridges, and the vehicle frame to collect raw vertical vibration signals; the lateral vibration sensors are deployed on axle boxes, rails, bridges, and the vehicle frame to collect raw lateral vibration signals; the vertical displacement sensors are deployed on rails, sleepers, and bridges to collect raw vertical displacement signals; the lateral displacement sensors are deployed on rails, sleepers, and bridges to collect raw lateral displacement signals; and the sound pressure sensors are deployed on both sides of the track and under the bridge spans to collect raw sound pressure signals.

[0048] Dynamic parameters include: vehicle vertical suspension stiffness, vehicle vertical suspension damping, vehicle lateral suspension stiffness, vehicle lateral suspension damping, wheel-rail vertical contact stiffness, wheel-rail lateral contact stiffness, bridge bending stiffness, and bridge vertical damping ratio. and bridge lateral damping ratio The vertical damping ratio of a bridge corresponds to the vertical damping coefficient of the bridge. , Indicates the equivalent mass of a bridge. This indicates the vertical stiffness of the bridge, the lateral damping ratio of the bridge, and the lateral damping coefficient of the bridge. , This indicates the lateral stiffness of the bridge.

[0049] The three-dimensional coordinates of all sensors are as follows: , , Indicates the horizontal coordinate along the track. Represents the longitudinal coordinate along the track. Represents the vertical coordinate. Indicates the total number of sensors. Indicates the number of sensors index.

[0050] Step S2: Perform differential preprocessing on the multi-source time-domain raw signals obtained in step S1 to obtain multi-source time-domain clean signals.

[0051] Original signal of vertical vibration The signal at the axle box is filtered by a 50Hz notch filter (IIR filter, stopband 49.5-50.5Hz, attenuation ≥40dB) + a 3rd-order Butterworth low-pass filter (cutoff 500Hz). The signal at the rail is denoised by a 4-layer thresholding method using a db6 wavelet + least squares detrending term to obtain the vertical vibration clean signal. .

[0052] Original signal of transverse vibration The signal at the axle box is filtered by a 50Hz notch filter and a 3rd-order Butterworth low-pass filter (cutoff 300Hz). The signal at the rail is denoised by a 3-layer threshold denoising using a db4 wavelet and a trend term removal process to obtain the clean lateral vibration signal. .

[0053] Original signal of vertical displacement The vertical displacement cleanliness signal is obtained by linear interpolation completion (missing rate ≤ 5%) + 5th order moving average filtering. .

[0054] Original signal of lateral displacement The lateral displacement cleanliness signal is obtained after linear interpolation completion and third-order moving average filtering. .

[0055] Raw sound pressure signal The clean sound pressure signal is obtained after A-weighted filtering and 3-layer wavelet packet denoising using db4 wavelets. .

[0056] Extract features related to vertical and lateral track irregularities; the vertical irregularity correlation features must satisfy an absolute Pearson correlation coefficient ≥ 0.7, and the lateral irregularity correlation features must satisfy an absolute Pearson correlation coefficient ≥ 0.7.

[0057] The characteristics associated with vertical irregularities include: wheel-rail vertical force fluctuation coefficient, peak frequency of axle box vertical acceleration power spectrum, vertical displacement fluctuation of rail, and vertical impact sound pressure level.

[0058] Wheel-rail vertical force fluctuation coefficient The calculation is as follows:

[0059] ;

[0060] ;

[0061] in, This represents the time-domain signal of the wheel-rail vertical force. Indicates the mass of the axle box. Indicates the vertical cleanliness signal of the axle box, taken from... , Indicates the vertical suspension stiffness of the vehicle. Indicates the vertical suspension damping of the vehicle. This represents the vertical displacement obtained by the second integration of the vertical cleanliness signal of the axle box. This represents the vertical velocity obtained by integrating the vertical cleanliness signal of the axle box once. , , These represent functions for calculating the maximum, minimum, and average values, respectively.

[0062] Peak frequency of vertical acceleration power spectrum of axle box The calculation is as follows:

[0063] ;

[0064] in, Indicates vertical vibration clean signal The power spectral density, Represents a frequency variable.

[0065] Vertical displacement fluctuation of rail The calculation is as follows:

[0066] ;

[0067] in, Indicates the vertical cleanliness signal of the rail, taken from .

[0068] Vertical impact sound pressure level The calculation is as follows:

[0069] ;

[0070] in, This indicates a synchronous sound pressure clean signal, taken from... .

[0071] Lateral irregularity correlation characteristics include: wheel-rail lateral force fluctuation coefficient, rail lateral displacement slope, and axle box vertical-lateral acceleration coupling coefficient.

[0072] Wheel-rail lateral force fluctuation coefficient The calculation is as follows:

[0073] ;

[0074] ;

[0075] in, This represents the time-domain signal of the wheel-rail lateral force. Indicates the axle box lateral cleanliness signal, taken from , This indicates the lateral suspension stiffness of the vehicle. Indicates the lateral suspension damping of the vehicle. This represents the lateral displacement obtained by the second integration of the lateral cleanliness signal of the axle box. This represents the lateral velocity obtained by integrating the lateral cleanliness signal of the axle box once.

[0076] Rail lateral displacement slope The calculation is as follows:

[0077] ;

[0078] in, Indicates the lateral cleanliness signal of the rail, taken from .

[0079] Vertical-lateral acceleration coupling coefficient of axle box The calculation is as follows:

[0080] ;

[0081] in, Describing covariance, Indicates variance.

[0082] Construct a vertical sub-feature matrix based on the extracted features. (dimension) ), horizontal sub-feature matrix (dimension) ) and total characteristic matrix (dimension) ).

[0083] Step S3: Perform dynamic graph modeling and dynamic coupling calculation, and output the signal transmission contribution matrix and the set of predicted dynamic response values.

[0084] Step S31: Construct the initial topology graph of the dual drive system. Using the sensor measurement points deployed in step S1 as nodes, and combining the three-dimensional physical coordinates of each sensor with the vertical and lateral dynamic parameters of the vehicle-rail-bridge system, generate an initial adjacency matrix containing vertical and lateral weights.

[0085] Initial adjacency matrix include:

[0086] Vertical weight matrix:

[0087] ;

[0088] Horizontal weight matrix:

[0089] ;

[0090] in, Represents a node and nodes physical distance, , , These represent the vertical signal matching coefficient, stiffness correlation coefficient, and damping correlation coefficient, respectively. , , These represent the transverse signal matching coefficient, stiffness correlation coefficient, and damping correlation coefficient, respectively.

[0091] Initial matrix fusion obtained :

[0092] ;

[0093] Step S32: Dynamically update the adjacency matrix. Based on the vertical sub-feature matrix and horizontal sub-feature matrix constructed in step S2, and the real-time train operation parameters (the real-time train operation parameters include train speed, train axle load, and time as a variable), the initial adjacency matrix is ​​adjusted through an attention mechanism to obtain a time-varying adjacency matrix that changes with time.

[0094] Attention coefficient calculate:

[0095] ;

[0096] in, The weight matrix representing the attention mechanism. express Time Node eigenvectors, express Time Node eigenvectors, Indicates the train's speed. express Train axle load at all times The bias term representing the attention mechanism. This indicates transpose.

[0097] Time-varying adjacency matrix elements as follows:

[0098] ;

[0099] And the constraints are satisfied:

[0100] .

[0101] in, Represents the initial adjacency matrix. row and number Column elements.

[0102] Step S33: Spatiotemporal convolution and dynamic coupling calculation. The multi-source temporal clean signal obtained in step S2 is input into the stacked spatiotemporal graph convolution module. During the convolution calculation process, the vehicle-rail-bridge coupled dynamic equation is embedded, and the vertical signal transmission contribution matrix, the lateral signal transmission contribution matrix and the set of dynamic response prediction values ​​are output.

[0103] (a) Embedding of core dynamic equations.

[0104] (1) Wheel-rail vertical contact constraint equation:

[0105] ;

[0106] in, express The vertical displacement of the wheel at any moment express The vertical displacement of the rail at all times express At any given moment, the wheel and rail undergo vertical contact deformation. express The vertical irregularity of the orbit in the time domain signal at any given moment. express The vertical displacement of the rail is constantly monitored and cleaned.

[0107] ;

[0108] in, This represents the vertical force between the wheel and the rail. This indicates the vertical contact stiffness between the wheel and the rail.

[0109] Simultaneously, the vertical contact damping force between the wheel and rail Need to be included in the equation:

[0110] ;

[0111] in, This represents the wheel-rail vertical contact damping coefficient. , These represent the vertical speeds of the wheel and the rail, respectively.

[0112] (2) Differential equation of vertical vibration of bridge:

[0113] ;

[0114] ;

[0115] ;

[0116] in, Indicates the equivalent mass of a bridge. express The vertical acceleration of the bridge at all times, express The vertical damping coefficient of the bridge at any given time. Indicates the bridge vibration frequency. This represents the vertical damping coefficient of the foundation. express The vertical velocity of the bridge at all times, Indicates the vertical stiffness of the bridge. Indicates the flexural stiffness of the bridge section. Indicates the bridge span. express The bridge's vertical displacement at all times. express The vertical force exerted by the wheels on the bridge at all times.

[0117] (3) Equation of lateral vibration of vehicle:

[0118] ;

[0119] in, Indicates the sprung mass of a vehicle. express The lateral acceleration of the sprung mass at any given moment. , These represent the vehicle's lateral suspension damping and stiffness, respectively. , They represent The lateral velocity and lateral displacement of the sprung mass at any given time. , They represent Horizontal velocity and lateral displacement of the axle box at all times. This indicates the lateral force between the wheel and rail.

[0120] (II) Spatiotemporal convolution and loss function optimization.

[0121] (1) Signal input and convolution calculation:

[0122] The multi-source time-domain clean signal obtained in step S2 Input stacked spatiotemporal graph convolutional module (containing 3 spatiotemporal convolutional layers, each with a kernel size of [size missing]) The above dynamic equations are embedded during the convolution process to ensure that signal transmission conforms to the coupling law between the vehicle, track, and bridge.

[0123] (2) Construction of the coupling loss function:

[0124] Total loss function Used to balance signal reconstruction accuracy and dynamic consistency:

[0125] ;

[0126] ;

[0127] ;

[0128] in, Indicates the signal reconstruction loss. Indicates kinetic loss, This represents the reconstructed signal value output by the convolution module. This represents the set of measured velocity signals, obtained by integrating the clean signal. This represents the set of velocity signals predicted by the dynamic model.

[0129] (3) Definition of output result:

[0130] 3.1 Signal Transmission Contribution Matrix:

[0131] Vertical signal transmission contribution matrix (element Representation Nodes For nodes The proportion of vertical signal transmission contribution, );

[0132] Lateral signal transmission contribution matrix (element Define the same vertical direction to represent the node. For nodes The proportion of transverse signal transmission contribution is used to locate the dominant source of transverse vibration.

[0133] 3.2 Set of predicted dynamic response values :

[0134] ;

[0135] in, This represents the matrix of predicted vertical vibration responses. This represents the matrix of predicted lateral vibration responses. This represents the matrix of predicted vertical displacement responses. This represents the matrix of predicted lateral displacement responses.

[0136] Step S4: Perform irregularity signal inversion and optimization to obtain the final time-domain signals of vertical and lateral irregularities of the track.

[0137] Step S41: Locate the dominant vibration source. Based on the vertical signal transmission contribution matrix output in step S3, filter the sensor nodes corresponding to the dominant vertical vibration source of the wheel-rail system. Based on the lateral signal transmission contribution matrix, filter the sensor nodes corresponding to the dominant lateral vibration source of the wheel-rail system.

[0138] Location of the dominant source of vertical vibration:

[0139] Contribution matrix to vertical signal transmission For each row, calculate the maximum value of that row. Filter to meet The nodes constitute the vertically dominant source node set. Combined with the sensor deployment location in step S1, The wheel-rail contact area corresponding to the vertical sensor of the center axle box and the vertical sensor of the rail is the main source area of ​​vertical vibration (the signal in this area is most sensitive to vertical irregularities).

[0140] Contribution matrix to transverse signal transmission For each row, calculate the maximum value of that row. Filter to meet The nodes constitute the horizontal dominant source node set. Combined with the sensor deployment location in step S1, The wheel-rail contact area corresponding to the lateral sensor in the center axle box and the lateral sensor in the rail is the main source area of ​​lateral vibration.

[0141] Step S42: Invert the vertical irregularity signal of the track. Combine the predicted vertical vibration response with the residuals of the clean vertical vibration signal and clean vertical displacement signal obtained in step S2. Solve the time domain signal of the vertical irregularity of the track using the weighted least squares iterative method. Introduce wavelength constraints (wavelength range 0.1-100 meters) to optimize the solution results and obtain the final time domain signal of the vertical irregularity of the track.

[0142] The objective function is to minimize the vertical signal residual while constraining smoothness. for:

[0143] ;

[0144] in, express The vertical irregularity of the orbit at any given time (the quantity to be inverted). , They represent The first and second derivatives with respect to time.

[0145] Iterative solution and wavelength optimization:

[0146] Initial value: (Assume the initial trajectory has no vertical irregularities).

[0147] Iterative update (Gauss-Newton method):

[0148] ;

[0149] in, Describes the vertical Jacobian matrix. Represents the vertical weighting matrix. This represents the vertical regularization parameter. Represents the identity matrix. Indicates the first The residual vector of the next iteration; Indicates the first The time-domain signal of vertical track irregularities in the next iteration.

[0150] Convergence condition: .

[0151] Wavelength constraint: Perform a Fourier transform on the converged orbital vertical irregularity time-domain signal and preserve the wavelength [0.1, 100]. The components are used to obtain the final orbital vertical irregularity time-domain signal.

[0152] Step S43: Invert the track lateral irregularity signal. Combine the residuals of the predicted lateral vibration response, the clean lateral vibration signal, and the clean lateral displacement signal. Solve the time-domain signal of the track lateral irregularity using the weighted least squares iterative method. Introduce wavelength constraints (wavelength range 0.1-100 meters) and vertical-lateral coupling corrections to optimize the solution results, obtaining the final time-domain signal of the track lateral irregularity. .

[0153] The inversion process is similar to that in the vertical direction, and the objective function is... for:

[0154] ;

[0155] in, express The time-domain signal of lateral irregularity in the orbit at any given moment (the quantity to be inverted). , They represent The first and second derivatives with respect to time.

[0156] Iterative solution and wavelength optimization:

[0157] Initial value: (Assume the initial track has no lateral irregularities).

[0158] Iterative update (Gauss-Newton method):

[0159] ;

[0160] in, Represents the horizontal Jacobian matrix. Represents a horizontally weighted matrix. This represents the horizontal regularization parameter. Represents the identity matrix. Indicates the first The residual vector of the next iteration; Indicates the first The time-domain signal of the lateral irregularity of the track in the next iteration.

[0161] Convergence condition: .

[0162] Wavelength constraint: Perform a Fourier transform on the converged orbital lateral irregularity time-domain signal and preserve the wavelength [0.1, 100]. The components were used to obtain the wavelength-optimized time-domain signal of the orbital lateral irregularities. .

[0163] Vertical-lateral coupling correction:

[0164] Based on the dynamic coupling characteristics of the vehicle-rail-bridge system in the vertical and lateral directions (vertical irregularities can induce additional lateral forces between the wheel and rail, thus interfering with the inversion results of lateral irregularities), a coupling correction term is introduced to eliminate cross-interference. The correction formula is as follows:

[0165] ;

[0166] in, This represents the time-domain signal indicating the final lateral irregularity of the orbit. , These represent the vertical irregularity coupling coefficient (characterizing the static coupling effect of vertical irregularity on the lateral direction) and the vertical irregularity velocity coupling coefficient (characterizing the dynamic coupling effect of the rate of change of vertical irregularity on the lateral direction), respectively. express The first derivative with respect to time.

[0167] Step S44: Determine the validity of the inversion and verify the amplitude and iterative convergence of the final orbital vertical irregularity time domain signal and the final orbital lateral irregularity time domain signal.

[0168] The validity of the inversion must be determined by simultaneously meeting the following three indicators:

[0169] Amplitude constraints:

[0170] Vertical unevenness amplitude: ;

[0171] Lateral unevenness amplitude: ;

[0172] Convergence constraints:

[0173] The number of vertical iterations and the number of horizontal iterations are both less than 50.

[0174] Dynamic correlation error constraints:

[0175] To calculate the mean absolute error (MAE) of the vertical and lateral velocity responses, the following must be satisfied:

[0176] ;

[0177] ;

[0178] in, , This represents the measured velocity signal obtained by integrating the clean signal in step S2. , This represents the velocity signal predicted by the dynamic model in step S3.

[0179] Step S5: Input the final track vertical irregularity time domain signal and the final track lateral irregularity time domain signal obtained in step S4 into the vehicle-track-bridge dynamics positive model to verify the consistency between the predicted response output by the model and the multi-source time domain clean signal obtained in step S2.

[0180] Construction of a positive model for the dynamics of a vehicle-track bridge.

[0181] (1) Subsystem modeling.

[0182] Vehicle model: A 15-DOF operating vehicle model was established using SIMPACK multibody dynamics software, taking into account axle box mass and vehicle suspension parameters (vertical suspension stiffness, vertical suspension damping, lateral suspension stiffness, and lateral suspension damping). The wheel-rail contact adopted the Hertz contact model (including wheel-rail vertical contact stiffness, wheel-rail lateral contact stiffness, wheel-rail vertical contact damping, and wheel-rail lateral contact damping parameters).

[0183] Track-bridge model: Constructed using ANSYS finite element software; rails are modeled using BEAM188 beam elements (elastic modulus 210 GPa, density 7850 kg / m³). 3 The sleepers are represented by SOLID185 solid elements, and the bridge is represented by SHELL181 shell elements, taking into account the bridge's bending stiffness, vertical damping coefficient, and lateral damping coefficient.

[0184] Co-simulation interface: The vehicle-rail-bridge system dynamics co-simulation is realized through the SIMPACK-ANSYS data interface. The data transmission frequency is ≥1000Hz to ensure that the dynamic parameters of the positive model are completely consistent with those of step S1.

[0185] (2) Verify the input and output definitions.

[0186] Input parameters:

[0187] Uneven signals: The final vertical uneven time domain signal and the final lateral uneven time domain signal of the track obtained in step S4 (the time length is consistent with the sampling time in step S1, usually 60-300s).

[0188] Operating parameters: real-time train speed and axle load (the same parameters used in step S3 dynamic diagram modeling);

[0189] Boundary conditions: Bridge bearing constraints (fixed hinged bearings / rolling bearings, consistent with actual bridges), and track fastener stiffness.

[0190] Output parameters:

[0191] The set of predicted responses output by the positive model contains four types of signals:

[0192] Vertical vibration prediction verification signal: The dimension is consistent with the vertical vibration clean signal in step S2 (number of vertical vibration sensors) (Total number of sampling times)

[0193] Lateral vibration prediction verification signal: The dimension is consistent with the lateral vibration clean signal in step S2 (number of lateral vibration sensors) (Total number of sampling times)

[0194] Vertical displacement prediction verification signal: The dimension is consistent with the vertical displacement clean signal in step S2 (number of vertical displacement sensors) (Total number of sampling times)

[0195] Lateral displacement prediction verification signal: The dimension is consistent with the lateral displacement cleanliness signal in step S2 (number of lateral displacement sensors) (Total number of sampling times).

[0196] (3) Calculation of consistency verification index.

[0197] The consistency between the predicted response and the multi-source time-domain clean signal in step S2 is quantified using two types of indicators: mean absolute error (MAE) and correlation coefficient. The judgment criteria for the four core verification indicators are as follows:

[0198] Vertical vibration consistency:

[0199] Mean Absolute Error (MAE): The average of the absolute values ​​of the differences between the measured clean signal and the predicted verification signal at all sampling times using all vertical vibration sensors, which must be ≤0.05 m / s. 2 ;

[0200] Correlation coefficient: The linear correlation between the measured clean signal and the predicted verification signal of vertical vibration, which must be ≥0.85 (dimensionless, value range [-1,1], the closer to 1, the stronger the consistency).

[0201] Vertical displacement consistency:

[0202] Mean Absolute Error (MAE): The average of the absolute values ​​of the differences between the measured clean signal and the predicted verification signal at all sampling times for all vertical displacement sensors, which must be ≤0.003m;

[0203] Correlation coefficient: The degree of linear correlation between the measured clean signal and the predicted verification signal of vertical displacement, which must be ≥0.9.

[0204] Lateral vibration consistency:

[0205] Mean Absolute Error (MAE): The average of the absolute values ​​of the differences between the measured clean signal and the predicted verification signal at all sampling times using all lateral vibration sensors, which must be ≤0.03 m / s. 2 ;

[0206] Correlation coefficient: The linear correlation between the measured clean signal and the predicted verification signal of transverse vibration, which must be ≥0.85.

[0207] Lateral displacement consistency:

[0208] Mean Absolute Error (MAE): The average of the absolute values ​​of the differences between the measured clean signal and the predicted verification signal at all sampling times for all lateral displacement sensors, which must be ≤0.002m;

[0209] Correlation coefficient: The degree of linear correlation between the measured clean signal and the predicted verification signal of lateral displacement, which must be ≥0.9.

[0210] (4) Output and visualization of verification results.

[0211] Result determination:

[0212] Validity determination: The mean absolute error (MAE) and correlation coefficient of the above four types of signals all meet the threshold requirements, and the inversion result is determined to be valid;

[0213] Invalid handling: If any indicator does not meet the threshold, return to step S4 to adjust the vertical-horizontal coupling coefficient (static coupling coefficient, dynamic coupling coefficient) or regularization parameter (vertical regularization parameter, horizontal regularization parameter), and re-execute the non-smooth signal inversion.

[0214] Example 2

[0215] like Figure 2 As shown, the track irregularity signal inversion system of the vehicle-rail bridge system includes:

[0216] The sensor array module is used to deploy sensors on the operating vehicles, track structures, and key sections of the bridge in the vehicle-rail-bridge system to collect multi-source time-domain raw signals and obtain the dynamic parameters of the vehicle-rail-bridge and the three-dimensional physical coordinates of all sensors.

[0217] The signal preprocessing and feature extraction module is used to perform differentiated preprocessing on the multi-source time-domain raw signals to obtain multi-source time-domain clean signals, and extract features related to the vertical and lateral irregularities of the track to construct the vertical sub-feature matrix, the lateral sub-feature matrix and the total feature matrix.

[0218] The dynamic graph modeling and dynamic coupling calculation module is used to construct a dual-drive initial topology graph, dynamically update the adjacency matrix, perform spatiotemporal convolution and dynamic coupling calculation, and output a signal transmission contribution matrix and a set of dynamic response prediction values.

[0219] The irregularity signal inversion and optimization module is used to locate the dominant vibration source, invert the time-domain signals of vertical and lateral track irregularities, and introduce wavelength constraints and coupling corrections for optimization to obtain the final track irregularity time-domain signal.

[0220] The verification module is used to input the track irregularity signal obtained by inversion into the vehicle-track-bridge dynamics positive model to verify the consistency between the predicted response and the multi-source time-domain clean signal.

[0221] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0222] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0223] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0224] It is worth noting that all contents not described in detail in this invention are existing technologies and are well known to those skilled in the art.

[0225] Therefore, the present invention employs the above-mentioned method, system, equipment and medium for inverting track irregularity signals in a vehicle-rail-bridge system, which can achieve high-precision inversion of track irregularity signals and provide technical support for the safe operation and maintenance of rail transit.

[0226] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for inverting track irregularity signals in a vehicle-rail bridge system, characterized in that, Includes the following steps: Step S1: Deploy sensor arrays on the operating vehicles, track structures, and key sections of the bridge in the vehicle-rail-bridge system to collect multi-source time-domain raw signals; synchronously acquire the dynamic parameters of the vehicle-rail-bridge and the three-dimensional physical coordinates of all sensors; Step S2: Perform differential preprocessing on the multi-source time-domain raw signals obtained in step S1 to obtain multi-source time-domain clean signals; extract features related to track vertical irregularities and track lateral irregularities. Based on the extracted features, construct the vertical sub-feature matrix, the horizontal sub-feature matrix, and the total feature matrix; Step S3: Perform dynamic graph modeling and dynamic coupling calculation, and output the signal transmission contribution matrix and the set of predicted dynamic response values; Step S4: Perform irregularity signal inversion and optimization to obtain the final time-domain signals of vertical and lateral track irregularities; Step S5: Input the final track vertical irregularity time domain signal and the final track lateral irregularity time domain signal obtained in step S4 into the vehicle-track-bridge dynamics positive model to verify the consistency between the predicted response output by the model and the multi-source time domain clean signal obtained in step S2.

2. The method for inverting track irregularity signals in a vehicle-rail bridge system according to claim 1, characterized in that, In step S1, the sensor array includes: a vertical vibration sensor, a lateral vibration sensor, a sound pressure sensor, a vertical displacement sensor, and a lateral displacement sensor; the multi-source time-domain raw signals include: the raw vertical vibration signal, the raw lateral vibration signal, the raw sound pressure signal, the raw vertical displacement signal, and the raw lateral displacement signal; the dynamic parameters include: vehicle vertical suspension stiffness, vertical suspension damping, lateral suspension stiffness, lateral suspension damping, wheel-rail vertical contact stiffness, wheel-rail lateral contact stiffness, and bridge bending stiffness.

3. The method for inverting track irregularity signals in a vehicle-rail bridge system according to claim 1, characterized in that, In step S2, the multi-source time-domain clean signals include: vertical vibration clean signals, lateral vibration clean signals, sound pressure clean signals, vertical displacement clean signals, and lateral displacement clean signals; the vertical irregularity correlation features satisfy the absolute value of the Pearson correlation coefficient ≥ 0.7, and the lateral irregularity correlation features satisfy the absolute value of the Pearson correlation coefficient ≥ 0.

7.

4. The method for inverting track irregularity signals in a vehicle-rail bridge system according to claim 3, characterized in that, In step S2, the vertical irregularity correlation features include: wheel-rail vertical force fluctuation coefficient, peak frequency of axle box vertical acceleration power spectrum, vertical displacement fluctuation of rail, and vertical impact sound pressure level; the lateral irregularity correlation features include: wheel-rail lateral force fluctuation coefficient, rail lateral displacement slope, and axle box vertical-lateral acceleration coupling coefficient.

5. The method for inverting track irregularity signals in a vehicle-rail bridge system according to claim 1, characterized in that, In step S3, dynamic graph modeling and dynamic coupling calculation are performed to output the signal transmission contribution matrix and the set of predicted dynamic response values, including the following steps: Step S31: Construct the initial topology graph of the dual drive system. Using the sensor measurement points deployed in step S1 as nodes, and combining the three-dimensional physical coordinates of each sensor with the vertical and lateral dynamic parameters of the vehicle-rail-bridge system, generate an initial adjacency matrix containing vertical and lateral weights. Step S32: Dynamically update the adjacency matrix. Based on the vertical sub-feature matrix and horizontal sub-feature matrix constructed in step S2 and the real-time train operation parameters, the initial adjacency matrix is ​​adjusted through an attention mechanism to obtain a time-varying adjacency matrix. Step S33: Spatiotemporal convolution and dynamic coupling calculation. The multi-source temporal clean signal obtained in step S2 is input into the stacked spatiotemporal graph convolution module. During the convolution calculation process, the vehicle-rail-bridge coupled dynamic equation is embedded, and the vertical signal transmission contribution matrix, the lateral signal transmission contribution matrix and the set of dynamic response prediction values ​​are output.

6. The method for inverting track irregularity signals in a vehicle-rail bridge system according to claim 5, characterized in that, The set of predicted dynamic response values ​​includes predicted values ​​for vertical vibration response, lateral vibration response, vertical displacement response, and lateral displacement response.

7. The method for inverting track irregularity signals in a vehicle-rail bridge system according to claim 6, characterized in that, In step S4, the irregularity signal is inverted and optimized to obtain the final time-domain signals of vertical and lateral track irregularities, including the following steps: Step S41: Locate the dominant vibration source. Based on the vertical signal transmission contribution matrix output in step S3, filter the sensor nodes corresponding to the dominant vertical vibration source of the wheel-rail system. Based on the lateral signal transmission contribution matrix, filter the sensor nodes corresponding to the dominant lateral vibration source of the wheel-rail system. Step S42: Invert the vertical irregularity signal of the track. Combine the predicted value of the vertical vibration response with the residuals of the clean vertical vibration signal and the clean vertical displacement signal obtained in step S2. Solve the time domain signal of the vertical irregularity of the track using the weighted least squares iterative method. Introduce wavelength constraints to optimize the solution results and obtain the final time domain signal of the vertical irregularity of the track. Step S43: Invert the track lateral irregularity signal, combine the predicted lateral vibration response with the residuals of the lateral vibration clean signal and the lateral displacement clean signal, and solve the track lateral irregularity time domain signal by weighted least squares iterative method. Introduce wavelength constraints and vertical-lateral coupling correction to optimize the solution results and obtain the final track lateral irregularity time domain signal. Step S44: Determine the validity of the inversion and verify the amplitude and iterative convergence of the final orbital vertical irregularity time domain signal and the final orbital lateral irregularity time domain signal.

8. A track irregularity signal inversion system for a vehicle-rail bridge system, characterized in that, The method for inverting track irregularity signals in a vehicle-rail bridge system as described in any one of claims 1-7 includes: The sensor array module is used to deploy sensors on the operating vehicles, track structures, and key sections of the bridge in the vehicle-rail-bridge system to collect multi-source time-domain raw signals and obtain the dynamic parameters of the vehicle-rail-bridge and the three-dimensional physical coordinates of all sensors. The signal preprocessing and feature extraction module is used to perform differentiated preprocessing on the multi-source time-domain raw signals to obtain multi-source time-domain clean signals, and extract features related to the vertical and lateral irregularities of the track to construct the vertical sub-feature matrix, the lateral sub-feature matrix and the total feature matrix. The dynamic graph modeling and dynamic coupling calculation module is used to construct a dual-drive initial topology graph, dynamically update the adjacency matrix, perform spatiotemporal convolution and dynamic coupling calculation, and output a signal transmission contribution matrix and a set of dynamic response prediction values. The irregularity signal inversion and optimization module is used to locate the dominant vibration source, invert the time-domain signals of vertical and lateral track irregularities, and introduce wavelength constraints and coupling corrections for optimization to obtain the final track irregularity time-domain signal. The verification module is used to input the track irregularity signal obtained by inversion into the vehicle-track-bridge dynamics positive model to verify the consistency between the predicted response and the multi-source time-domain clean signal.

9. A computer device, comprising: Memory and processor; The memory stores a computer program, characterized in that when the processor executes the computer program, it implements the steps of the method for inverting track irregularity signals in a vehicle-rail-bridge system as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for inverting track irregularity signals in a vehicle-rail-bridge system as described in any one of claims 1-7.

Images