Geological radar water content measurement method and system suitable for ballast bed

Abstract

The application discloses a geological radar water content measurement method and system suitable for ballast bed, and the method comprises the following steps: generating a radar data matrix with depth as rows and detection distance as columns by using a geological radar; obtaining purified data through preprocessing; automatically picking up a ballast top interface based on envelope and gradient characteristics, and locking a target depth window under a sleeper; performing gain normalization or data rejection based on incident surface peak energy ratio, peak width and low frequency proportion characteristics; obtaining a comprehensive speed in a sliding time window, and calculating an effective dielectric constant; substituting the effective dielectric constant into a three-phase dielectric mixed model to obtain volume water content through inversion; and performing error propagation evaluation in the inversion process, and outputting a continuous volume water content curve along the line. The application significantly improves the inversion stability and precision under the conditions of rain and complex coupling environment while realizing non-destructive and continuous measurement, and provides a directly applicable water correction quantity for ballast dirt evaluation.

Description

Technical Field

[0001] This invention relates to the field of railway track inspection and maintenance technology, and more specifically to a ground-penetrating radar method and system for measuring the moisture content of crushed stone track beds. Background Technology

[0002] As a crucial component of the track structure, the condition of the railway ballast bed directly affects the smoothness and stability of the line. The moisture content of the ballast bed is a key parameter for evaluating its drainage performance, degree of soiling, and load-bearing capacity. After rainfall, the moisture content of the ballast bed changes rapidly. Accurately understanding its spatial distribution and temporal evolution is of great significance for post-rain safety assessment, soiling identification, and maintenance decisions.

[0003] Currently, the methods for obtaining the moisture content of track bed have the following limitations:

[0004] Sampling-drying method: This method has clear indicators, but it requires destructive sampling, has sparse sampling points, long cycle, high cost, is difficult to cover long sections, and cannot reflect rapid changes after rain.

[0005] Buried sensor method (such as TDR, capacitive sensor): It is widely used in fine-grained soil, but in coarse aggregate medium such as crushed stone roadbed, there are natural difficulties in electrode contact and volumetric characterization. At the same time, it requires long-term burial and maintenance, making it difficult to achieve network coverage and temporary inspection and retesting.

[0006] In the field of ground-penetrating radar (GPR)-based non-destructive testing, existing research has proposed directly assessing the dirt and grime status of track beds by analyzing the time-domain characteristics of radar signals (such as the number of axis intersections and time-domain integrals). This method avoids complex signal preprocessing and transformation, reducing reliance on human experience. However, such methods are mainly applicable to dry and stable conditions. After rain or in humid environments, strong wave impedance interfaces form inside the track bed due to dirt stratification and humidity differences, generating strong reflection signals. This intensity of reflection energy can suppress or even obscure crucial target information, significantly reducing the accuracy of analysis methods that primarily rely on time-domain amplitude characteristics.

[0007] Furthermore, existing GPR-based moisture content inversion methods mostly focus on relatively homogeneous scenarios such as soil and railway subgrade, often using the Topp empirical formula or similar semi-empirical models to map the effective dielectric constant to volumetric moisture content. These models, when directly applied to crushed stone track beds with a three-phase coexistence of aggregate, air, and water, will produce significant deviations because the porosity and frequency sensitivity of the track bed make its effective dielectric behavior far more complex than that of homogeneous soil. Especially after rain, surface water accumulation drastically alters near-surface reflection and system coupling states, making the inversion extremely unstable.

[0008] Therefore, there is an urgent need for a non-destructive measurement method and system that can adapt to the characteristics of crushed stone track bed media and can still stably, continuously, and accurately invert the volumetric moisture content after rain and under complex working conditions. Summary of the Invention

[0009] To address the shortcomings of existing technologies, the present invention aims to provide a ground-penetrating radar method and system for measuring moisture content in crushed stone track beds. This invention can acquire the electromagnetic response of the target window under the sleeper in real time without altering the track structure, stably retrieve the volumetric moisture content, and output a mileage sequence. This enables continuous mapping and early warning of abnormal sections after rain and under complex operating conditions, improving the comparability and engineering usability of results from different equipment and different tracks.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] In a first aspect, embodiments of the present invention provide a ground-penetrating radar method for measuring the moisture content of crushed stone roadbeds, comprising the following steps:

[0012] Data acquisition and synchronization steps: The ground-penetrating radar acquisition unit moves along the railway track bed, emits multi-frequency electromagnetic waves and receives echo signals, and simultaneously records the mileage location information through the encoder to generate a radar data matrix with depth as the row and detection distance as the column.

[0013] Preprocessing steps: The radar echo signal is preprocessed, including DC offset removal, coupling interference and standing wave suppression, and noise removal, to obtain cleaned data;

[0014] Target window and incident surface identification steps: Based on the envelope and gradient features of the purification data, the interface at the top of the track bed is automatically picked up, and the target depth window under the pillow is locked;

[0015] Surface water accumulation identification and compensation steps: Based on the peak energy ratio, peak width and low frequency ratio of the incident surface of the target depth window, identify surface water accumulation or strong coupling conditions, and perform gain normalization or data removal.

[0016] Multi-frequency feature extraction and group velocity estimation steps: Within the sliding time window, perform time-frequency analysis on the radar signal to extract local spectral features, calculate the phase velocity using the cross-correlation time delay method, calculate the group velocity using the phase gradient method, and obtain the comprehensive velocity by using variance reciprocal weighted fusion, and calculate the effective dielectric constant.

[0017] Dielectric mixing inversion and calibration steps: Substitute the effective dielectric constant into the three-phase dielectric mixing model to obtain the volumetric water content, and introduce the fine particle volume fraction for correction;

[0018] Results generation steps: Evaluate the propagation of errors in the inversion process and output a continuous volumetric water content curve along the line.

[0019] In one embodiment, the DC offset removal in the preprocessing step employs in-channel mean subtraction to perform zero-point correction on each scan channel;

[0020] The coupling interference and standing wave suppression are achieved by calculating the total energy of each channel within a local window, setting a threshold to exclude high-energy channels, and subtracting the average value of the background channels from the original data.

[0021] The noise removal includes energy criterion identification of outliers, matched filtering, and two-dimensional wavelet denoising.

[0022] In one embodiment, the energy criterion is achieved by calculating the ratio of air layer energy to total channel energy, and data exceeding a preset threshold is marked as noisy. The matched filtering uses Ricker wavelet for correlation matching. The two-dimensional wavelet denoising includes threshold shrinking and / or removal of high-frequency coefficients after wavelet domain decomposition.

[0023] In one embodiment, the multi-frequency feature extraction and group velocity estimation steps include:

[0024] Within the sliding time window, a short-time Fourier transform is performed on the radar signal to obtain the local spectrum of the signal as it changes with depth.

[0025] In the time domain, the time delay is calculated by normalizing the cross-correlation of signals from adjacent depth layers, and then the phase velocity is calculated.

[0026] In the frequency domain, the group delay is obtained by calculating the derivative of the phase with respect to the frequency, and then the group velocity is calculated.

[0027] The variances of the estimated phase velocity and group velocity are calculated separately, and the two are weighted and averaged using the reciprocal of the variance as the weight to obtain the combined velocity; and the effective dielectric constant is obtained.

[0028] In one embodiment, in the time domain, the time delay is calculated by performing normalized cross-correlation on signals from adjacent depth layers, and then the phase velocity is calculated; the formula is as follows:

[0029] (1)

[0030] (2)

[0031] In the formula: This represents the radar echo signal within a sliding time window at depth or sample point z. This represents the depth interval between two adjacent analytical layers used for velocity estimation. Indicates depth The echo sequence at that point is shifted along the time axis. The resulting sequence; Represents the time shift, from the sampling point With sampling interval Conversion, ; Indicates phase velocity; This represents the estimated time delay when the normalized cross-correlation reaches its maximum value.

[0032] In one embodiment, in the frequency domain, the group delay is obtained by calculating the derivative of the phase with respect to frequency, and then the group velocity is calculated; the calculation formula is as follows:

[0033] (3)

[0034] (4)

[0035] In the formula: The group delay, which is the negative derivative of the phase with respect to frequency, is the propagation time of the energy envelope. f Indicates frequency; Indicates frequency differential interval; Indicates at the center frequency Nearby, phase Follow The amount of change in radian; Indicates group velocity.

[0036] In one embodiment, the variances of the phase velocity and group velocity estimates are calculated separately, and the two are weighted and averaged using the reciprocal of the variance as the weight to obtain the combined propagation velocity and the effective dielectric constant; the calculation formula is as follows:

[0037] The overall speed is obtained by weighting and merging the inverse variances. :

[0038] (5)

[0039] In the formula: , They represent and The variance of the estimated value is obtained; the effective dielectric constant is then obtained.

[0040] (6)

[0041] In the formula: The effective relative permittivity at depth z is represented by c; the speed of light in vacuum is represented by c. .

[0042] In one embodiment, the three-phase dielectric hybrid model is a CRIM extended model, expressed as:

[0043]

[0044] In the formula: Indicates volumetric moisture content; Indicates the porosity of the track bed; , , These represent the dielectric constants of water, aggregate, and air, respectively.

[0045] Secondly, embodiments of the present invention provide a geological radar moisture content measurement system suitable for crushed stone roadbeds, the system comprising:

[0046] The ground-penetrating radar acquisition unit includes a multi-frequency antenna and a synchronization triggering device;

[0047] Encoder used for mileage synchronization;

[0048] Power and charging management module;

[0049] A data processing unit is configured to perform the steps of the method as described in any of the first aspects.

[0050] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following:

[0051] This invention provides a ground-penetrating radar (GPR) method and system for measuring the moisture content of crushed stone roadbeds. Within the target window acquired by GPR, the invention obtains effective dielectric constant characteristics through a combination of short-time spectrum and group velocity estimation. It establishes a CRIM extended expression suitable for the aggregate-air-water three-phase system and explicitly incorporates the influence of fine aggregate volume fraction. Combining strong reflection characteristics of the incident surface with an energy ratio detection mechanism, it identifies and eliminates surface water accumulation and strong coupling effects, outputting a volumetric moisture content curve along the track, enabling comparability of results across different tracks and equipment. Compared to existing methods based on sampling and sensor installation, this invention significantly improves the stability and accuracy of inversion under post-rain and complex coupling environments while providing non-destructive, continuous measurement, and offers directly applicable moisture correction for roadbed contamination assessment. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0053] Figure 1 This is a flowchart of a ground-penetrating radar method for measuring the moisture content of crushed stone roadbeds, provided in an embodiment of the present invention.

[0054] Figure 2This is a schematic diagram of the waveforms before and after DC offset removal adjustment provided in an embodiment of the present invention;

[0055] Figure 3 This is a schematic diagram of the data waveforms before and after removing coupling interference provided in an embodiment of the present invention;

[0056] Figure 4 This is a schematic diagram of the data waveforms before and after background noise removal provided in an embodiment of the present invention;

[0057] Figure 5 This is a schematic diagram of a field test provided in an embodiment of the present invention;

[0058] Figure 6 This is a comparison chart of the original data and the purified data provided in this embodiment of the invention;

[0059] Figure 7 This is a schematic diagram of automatic pickup of the top surface of the track bed provided in an embodiment of the present invention;

[0060] Figure 8 This is a schematic diagram of the data before and after gain adjustment provided in an embodiment of the present invention;

[0061] Figure 9 This is an inversion volumetric water content curve provided in an embodiment of the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] The technical problem this invention aims to solve is to obtain continuous and quantifiable volumetric moisture content along railway ballast track without disrupting traffic flow, providing a reliable basis for rapid post-rain assessment and contamination identification. The ballast track is a coarse-grained, porous, and heterogeneous medium; rainfall and temperature changes cause significant spatiotemporal fluctuations in moisture content. Industry experimental studies have shown that changes in moisture content significantly alter electromagnetic wave propagation characteristics: under saturated humidity conditions, the relative permittivity of the ballast track can be more than twice that of a dry track, leading to a significant reduction in electromagnetic wave propagation speed and causing signal energy attenuation, waveform distortion, and bandwidth compression. These effects, existing at the level of physical principles, are difficult to overcome effectively by simply improving equipment hardware or conventional signal processing methods.

[0064] Reference Figure 1As shown in the figure, this invention discloses a ground-penetrating radar (GPR) method for measuring the moisture content of crushed stone track beds. The GPR used comprises a GPR acquisition unit (including a multi-frequency antenna and a synchronous triggering device), an encoder, and a power supply and charging management module. This GPR can be installed on a hand-push platform or maintenance vehicle, meeting the requirements for rapid scanning during train operation or short-term maintenance windows.

[0065] The working principle is as follows: electromagnetic pulses are emitted and received within the target depth window under the pillow. The electromagnetic waves propagate in the three-phase medium of "aggregate-air-water". The effective dielectric properties determine the group velocity, phase delay, and spectral morphology. Increased moisture content increases the effective dielectric constant and changes the dispersion and attenuation characteristics. To avoid the impact of single-band and near-surface coupling on the inversion stability, this invention uses multi-frequency characteristics and group velocity estimation as the core, combined with short-time spectrum and phase gradient methods to robustly estimate the effective dielectric constant. At the same time, the surface water accumulation / strong coupling condition is determined and adaptively compensated by the strong reflection energy ratio of the incident surface, the envelope peak width, and the low-frequency energy ratio. The effective dielectric constant is mapped to volumetric moisture content through the three-phase dielectric hybrid model, generating a continuous moisture content curve and anomaly segment location along the line, which are used as the moisture correction amount for subsequent fouling assessment. Specifically, it includes the following steps S1~S7:

[0066] S1. Data Acquisition and Synchronization Steps: The ground-penetrating radar acquisition unit moves along the railway track bed, emits multi-frequency electromagnetic waves and receives echo signals, and simultaneously records mileage location information through an encoder to generate a radar data matrix with depth as the row and detection distance as the column.

[0067] In this step, the radar wave mileage position can be corrected based on the encoder and ledger data, and the original radar data can be read to form a matrix with depth as the row and detection distance as the column.

[0068] S2. Preprocessing steps: Preprocess the radar echo signal, including DC offset removal, coupling interference and standing wave suppression, and noise removal, to obtain cleaned data.

[0069] To achieve accurate moisture content measurement, targeted preprocessing of the raw radar signal is a prerequisite for overcoming humidity interference. Unlike existing technologies that aim to directly assess dirt status without complex preprocessing, the preprocessing link of this invention is specifically designed to separate and suppress specific interference signals caused by moisture, ensuring the reliability of subsequent feature extraction.

[0070] ① DC offset removal: Zero-point correction is performed on each scan channel using in-channel mean subtraction to ensure signal symmetry about the zero axis, establishing a benchmark for subsequent accurate energy and phase analysis. For example... Figure 2As shown, the horizontal axis represents time in nanoseconds (ns), and the vertical axis represents amplitude in mV. The red line represents the curve before adjustment, and the blue line represents the curve after adjustment. This eliminates fixed biases introduced by hardware and the system, providing an accurate and reliable signal reference for all subsequent fine-grained feature extraction and inversion calculations.

[0071] ② Coupling Interference and Standing Wave Suppression: By calculating the total energy of each track within a local window and setting a threshold, high-energy tracks caused by strong reflectors inside the track bed (such as sleepers) are intelligently excluded. Only the low-energy "background" tracks are averaged and subtracted from the original data. This method effectively overcomes the defect of traditional global averaging methods that produce artificial artifacts at strong reflection interfaces, and is specifically used to suppress strong standing wave interference caused by system coupling. Figure 3 As shown, the upper part (a) is the data with coupling interference, and the lower part (b) is the data with coupling interference removed.

[0072] ③ Noise Removal: A hybrid denoising strategy combining energy criteria and waveform characteristics is employed. First, abnormal channels severely affected by electromagnetic interference are identified by calculating the ratio of "air layer energy" to "total channel energy," with a threshold of 0.1; data exceeding this threshold is marked as noisy. Then, the noisy data undergoes matched filtering with the radar transmitted wavelet (e.g., Ricker wavelet) to enhance the signal-to-noise ratio of the useful echo. Finally, two-dimensional wavelet denoising is performed, treating the channel as a grayscale image with approximate Gaussian noise contamination. After wavelet domain decomposition, the diagonal and vertical coefficients of high frequencies are thresholded / removed, preserving effective low-frequency and horizontal information. Through this four-step process of "standing wave subtraction—energy ratio denoising—matched filtering—wavelet denoising," significantly denoised detection data is obtained. Figure 4 As shown, the horizontal axis represents the amplitude in mV; the vertical axis represents time in ns; the left half is the effect before background denoising, and the right half is the effect after background denoising.

[0073] S3. Target window and incident surface identification steps: Based on the envelope and gradient features of the purification data, automatically pick the interface at the top of the bed and lock the target depth window under the pillow;

[0074] S4. Surface water accumulation identification and compensation step: Based on the peak energy ratio, peak width and low frequency ratio characteristics of the incident surface of the target depth window, surface water accumulation or strong coupling conditions are identified, and gain normalization or data removal is performed; this step can ensure the stability of subsequent inversion.

[0075] S5. Multi-frequency feature extraction and group velocity estimation steps: Within the sliding time window, perform time-frequency analysis on the radar signal to extract local spectral features, calculate the phase velocity using the cross-correlation time delay method, calculate the group velocity using the phase gradient method, and obtain the comprehensive velocity by using the inverse variance weighted fusion, and calculate the effective dielectric constant.

[0076] In this step, a sliding time window is used to perform a short-time Fourier transform (STFT) on the radar signal to obtain the local frequency domain characteristics of the signal as a function of depth (time). This avoids the frequency information ambiguity caused by performing a global Fourier transform on the entire signal in traditional methods, and can capture the dispersion characteristics of media at different depths more precisely. At the same time, the cross-correlation time delay method (estimates wave propagation delay in the time domain) and the phase gradient method (directly calculates group velocity in the frequency domain) are combined to calculate the variance of multiple velocity estimates obtained by different methods, and a weighted average is performed using the reciprocal of the variance as the weight. This method can automatically assign higher weights to more reliable estimates, significantly improve the anti-interference ability and robustness of the final group velocity results, and effectively solve the problem of large fluctuations in single-method estimates caused by the uneven electrical properties of the track bed after rain.

[0077] Perform normalized cross-correlation on adjacent depth layers and calculate the time delay. and phase velocity :

[0078] (1)

[0079] (2)

[0080] In the formula: This represents the radar echo signal at depth (or sample point) z within a sliding time window; This represents the depth interval (m) between two adjacent analytical layers used for velocity estimation. Indicates depth The echo sequence at that point is shifted along the time axis. The resulting sequence; The time shift (s) is represented by the sampling point. With sampling interval Conversion, ; Represents phase velocity (m / s); This represents the estimated time delay when the normalized cross-correlation reaches its maximum value.

[0081] (3)

[0082] (4)

[0083] In the formula: The group delay, which is the negative derivative of the phase with respect to frequency, is the propagation time of the energy envelope. f Indicates frequency; Indicates the frequency differential interval (Hz); Indicates at the center frequency Nearby, phase Follow The change in radians; This represents the group velocity (m / s).

[0084] The overall speed is obtained by weighting and merging the inverse variances. :

[0085] (5)

[0086] In the formula: , express and The variance (uncertainty) of the estimated value, (m / s) 2 ).

[0087] The effective dielectric constant is obtained as follows:

[0088] (6)

[0089] In the formula: The effective relative permittivity at depth z is represented by c; the speed of light in vacuum is represented by c.

[0090] S6. Dielectric mixing inversion and calibration steps: Substitute the effective dielectric constant into the three-phase dielectric mixing model to obtain the volumetric water content, and introduce the fine material volume fraction for correction.

[0091] Traditional empirical models of soil moisture content (such as the Topp formula) or simple dielectric mixing models cannot accurately describe the complex electromagnetic behavior of the three-phase medium of aggregate-air-water in crushed stone roadbeds. This invention presents a physical improvement based on the complex refractive index model (CRIM), the core expression of which is:

[0092] (7)

[0093] In the formula: Indicates volumetric moisture content; Indicates the porosity of the track bed; , , These represent the dielectric constants of water, aggregate, and air, respectively.

[0094] S7. Result Generation Steps: The errors in the inversion process are propagated and evaluated, outputting a continuous volumetric water content curve along the line. For example, errors such as travel time noise, spectral fitting residuals, and calibration errors are propagated to output mileage-based volumetric water content.

[0095] Example:

[0096] This embodiment uses a three-channel mid-to-high frequency antenna combination, which can simultaneously collect complete cross-sectional data of the track center and both sides of the ballast shoulders in a single pass. The test time is 2 hours after rain, and the latter quarter of the test line is the tunnel location, used to visually distinguish different water contents of the track. Figure 5 The image shown is a schematic diagram of the on-site test.

[0097] (1) Preprocessing results

[0098] like Figure 6 As can be seen from the raw data in section (a), there are strip-shaped standing wave interference waveforms in the horizontal direction, which are strongest on the surface of the track bed and gradually weaken with depth. This standing wave interference has a strong suppressive effect on the effective signal. Figure 6 Part (b) is the processed image, in which standing wave interference and direct waves have been removed, allowing for a clearer observation of the details of the signals inside the track bed.

[0099] (2) Automatic picking of the top surface of the track bed

[0100] like Figure 7 As shown, the horizontal axis Distance represents distance in meters (m), and the vertical axis Time represents time in millimeters (ns). The horizontally oriented white undulating curve in the figure represents the location of the track bed top surface automatically picked up by this method based on envelope and gradient features. The figure shows that this automatic picking of the track bed top surface continuously and stably tracked the undulations of the track bed surface throughout the entire detection distance. The identification results perfectly matched the strong reflection interfaces in the radar image, maintaining stable tracking performance even in sections with changes in the geometric shape of the track bed surface (normal lines and tunnels, approximately 150 meters). This reliable interface identification provides an accurate benchmark for subsequently determining the "target depth window under the pillow," ensuring that the electromagnetic features used for volumetric water content inversion all originate from within the track bed itself, effectively avoiding interference from the air layer and track bed interface on the analysis results.

[0101] (3) Gain adjustment

[0102] During GPR (Ground Ground Pulse Radar) detection of railway track beds, the signal amplitude gradually attenuates with increasing propagation depth. This attenuation may be due to the scattering and absorption of electromagnetic waves in the medium. Without proper gain adjustment, the reflected signal from deep targets may be masked by noise, leading to inaccurate detection results. Gain adjustment aims to compensate for the amplitude changes caused by energy attenuation during signal propagation, thereby improving signal detectability and interpretability. Figure 8 In the diagram, (a) shows the schematic before gain adjustment; as can be seen, after gain adjustment, in part (b), the information at the bottom of the track bed is displayed more accurately, improving the resolution and clarity of the target object.

[0103] (4) Inverted volumetric water content

[0104] An inversion algorithm based on multi-frequency characteristics and a three-phase dielectric hybrid model was used, and the volumetric moisture content inversion results for the entire test line showed good agreement with the actual situation. Figure 9 As shown, the volumetric moisture content-mileage curves of the test route and the corresponding comparison of dry and wet sections are plotted; the horizontal axis represents mileage in meters (m), and the vertical axis represents moisture content (dimensionless). The data curves in the figure show that in the open section from mileage 0 to 150m, the volumetric moisture content fluctuates drastically and is relatively high, with many peaks exceeding 0.15, and even reaching 0.24 in some areas (such as at 80m), reflecting the high moisture content of the roadbed after rain. However, in the tunnel section after mileage 150m (corresponding to the last quarter of the test route), the curve drops significantly and tends to stabilize, with the moisture content value basically remaining below 0.05. The results indicate that the volumetric moisture content obtained from electromagnetic characteristic inversion is consistent with the actual dry and wet state of the route in terms of spatial distribution, numerical range, and variation trend. The difference in moisture content between the open section and the tunnel section is accurately reflected, indicating that this inversion method is suitable for engineering applications.

[0105] Compared with existing technologies, the ground-penetrating radar moisture content measurement method for crushed stone roadbeds provided by this invention has the following advantages:

[0106] 1) Continuous, non-destructive, and highly efficient: GPR enables mileage measurement along the route without the need for sampling and drying or long-term sensor installation, making it suitable for rapid inspection and significantly reducing the workload of testing and maintenance.

[0107] 2) More robust under rain and coupled conditions: An adaptive identification and compensation link for surface water accumulation is proposed, combined with standing wave suppression and gain unification, to avoid the misleading effect of strong near-surface reflection and system gain drift after rain on the results, and improve the usability under complex conditions.

[0108] 3) Improved inversion accuracy and stability: The dual-channel estimation of "multi-frequency + group / phase velocity" is adopted and fused with inverse variance weighting to reduce the sensitivity of single frequency band and single feature; combined with a three-phase dielectric hybrid model and the introduction of fine material (dirt) correction, the moisture content inversion is more in line with the real electromagnetic characteristics of the gravel-air-water coarse particle medium, and the influence of moisture content is separated from the dirt index, making dirt assessment and drainage anomaly identification more reliable;

[0109] 4) Flexible deployment and strong adaptability: The device can be mounted on a hand-pushed platform or maintenance vehicle, and the algorithm parameters (window length, threshold, frequency band) can be tuned according to the antenna frequency and line noise to adapt to different lines and environments.

[0110] Based on the same inventive concept, this invention also provides a ground-penetrating radar moisture content measurement system suitable for crushed stone roadbeds. Since the principle of the problem solved by this system is similar to the aforementioned ground-penetrating radar moisture content measurement method suitable for crushed stone roadbeds, the implementation of this system can refer to the implementation of the aforementioned method, and the repeated parts will not be described again.

[0111] The system includes:

[0112] The ground-penetrating radar acquisition unit includes a multi-frequency antenna and a synchronization triggering device;

[0113] Encoder used for mileage synchronization;

[0114] Power and charging management module;

[0115] The data processing unit is used to execute steps S1 to S7 of the ground-penetrating radar moisture content measurement method for crushed stone roadbed as described in the above embodiments.

[0116] This system enables rapid, continuous, and quantitative measurement of the volumetric moisture content of crushed stone ballast along railway lines without causing any damage to the track structure. Its core function is to improve the stability and accuracy of inversion under complex conditions such as post-rain dampness and variable surface coupling states, providing a reliable basis for moisture correction for ballast soil contamination assessment and drainage performance diagnosis, and ensuring the comparability of measurement results between different equipment and different lines.

[0117] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0118] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A ground-penetrating radar method for measuring the moisture content of crushed stone roadbeds, characterized in that, Includes the following steps: Data acquisition and synchronization steps: The ground-penetrating radar acquisition unit moves along the railway track bed, emits multi-frequency electromagnetic waves and receives echo signals, and simultaneously records the mileage location information through the encoder to generate a radar data matrix with depth as the row and detection distance as the column. Preprocessing steps: The radar echo signal is preprocessed, including DC offset removal, coupling interference and standing wave suppression, and noise removal, to obtain cleaned data; Target window and incident surface identification steps: Based on the envelope and gradient features of the purification data, the interface at the top of the track bed is automatically picked up, and the target depth window under the pillow is locked; Surface water accumulation identification and compensation steps: Based on the peak energy ratio, peak width and low frequency ratio of the incident surface of the target depth window, identify surface water accumulation or strong coupling conditions, and perform gain normalization or data removal. Multi-frequency feature extraction and group velocity estimation steps: Within the sliding time window, perform time-frequency analysis on the radar signal to extract local spectral features, calculate the phase velocity using the cross-correlation time delay method, calculate the group velocity using the phase gradient method, and obtain the comprehensive velocity by using variance reciprocal weighted fusion, and calculate the effective dielectric constant. Dielectric mixing inversion and calibration steps: Substitute the effective dielectric constant into the three-phase dielectric mixing model to obtain the volumetric water content, and introduce the fine particle volume fraction for correction; Results generation steps: Evaluate the propagation of errors in the inversion process and output a continuous volumetric water content curve along the line.

2. The method according to claim 1, characterized in that, The DC offset removal in the preprocessing step uses in-channel mean subtraction to perform zero-position correction on each scan channel. The coupling interference and standing wave suppression are achieved by calculating the total energy of each channel within a local window, setting a threshold to exclude high-energy channels, and subtracting the average value of the background channels from the original data. The noise removal includes energy criterion identification of outliers, matched filtering, and two-dimensional wavelet denoising.

3. The method according to claim 2, characterized in that, The energy criterion is achieved by calculating the ratio of air layer energy to total channel energy. Data with a ratio exceeding a preset threshold is marked as noisy. The matched filtering uses Ricker wavelet for correlation matching. The two-dimensional wavelet denoising includes threshold shrinking and / or removal of high-frequency coefficients after wavelet domain decomposition.

4. The method according to claim 1, characterized in that, The multi-frequency feature extraction and group velocity estimation steps include: Within the sliding time window, a short-time Fourier transform is performed on the radar signal to obtain the local spectrum of the signal as it changes with depth. In the time domain, the time delay is calculated by normalizing the cross-correlation of signals from adjacent depth layers, and then the phase velocity is calculated. In the frequency domain, the group delay is obtained by calculating the derivative of the phase with respect to the frequency, and then the group velocity is calculated. The variances of the estimated phase velocity and group velocity are calculated separately, and the two are weighted and averaged using the reciprocal of the variance as the weight to obtain the combined velocity; and the effective dielectric constant is obtained.

5. The method according to claim 4, characterized in that, In the time domain, the time delay is calculated by normalizing and cross-correlating the signals from adjacent depth layers, and then the phase velocity is calculated; the formula is as follows: （1） （2） In the formula: This represents the radar echo signal within a sliding time window at depth or sample point z. This represents the depth interval between two adjacent analytical layers used for velocity estimation. Indicates depth The echo sequence at that point is shifted along the time axis. The resulting sequence; Represents the time shift, from the sampling point With sampling interval Conversion, ; Indicates phase velocity; This represents the estimated time delay when the normalized cross-correlation reaches its maximum value.

6. The method according to claim 5, characterized in that, In the frequency domain, the group delay is obtained by calculating the derivative of the phase with respect to frequency, and then the group velocity is calculated; the calculation formula is as follows: （3） （4） In the formula: The group delay, which is the negative derivative of the phase with respect to frequency, is the propagation time of the energy envelope. f Indicates frequency; Indicates frequency differential interval; Indicates at the center frequency Nearby, phase Follow The amount of change in radian; Indicates group velocity.

7. The method according to claim 6, characterized in that, The variances of the estimated phase velocity and group velocity are calculated separately. Using the reciprocal of the variance as the weight, a weighted average is taken to obtain the combined velocity and the effective dielectric constant. The calculation formula is as follows: The overall speed is obtained by weighting and merging the inverse variances. : （5） In the formula: , They represent and The variance of the estimated value is obtained; the effective dielectric constant is then obtained. （6） In the formula: The effective relative permittivity at depth z is represented by c; the speed of light in vacuum is represented by c. .

8. The method according to claim 7, characterized in that, The three-phase dielectric hybrid model is an extended CRIM model, and its expression is: （7） In the formula: Indicates volumetric moisture content; Indicates the porosity of the track bed; , , These represent the dielectric constants of water, aggregate, and air, respectively.

9. A geological radar moisture content measurement system suitable for crushed stone track beds, characterized in that, The system includes: The ground-penetrating radar acquisition unit includes a multi-frequency antenna and a synchronization triggering device; Encoder used for mileage synchronization; Power and charging management module; A data processing unit is configured to perform the steps of the method as described in any one of claims 1-8.

Images