Method for inverting coal seam thickness by using love-type channel wave dispersion

Abstract

This invention belongs to the field of underground geophysical exploration technology in coal mines, specifically relating to a method for inverting coal seam thickness using a Love-type channel wave dispersion method. This invention introduces regularization constraints and a sparsity factor to optimize the coal thickness inversion objective function, mathematically enhancing the algorithm's noise resistance and boundary identification capabilities. Through a parameter pre-acquisition dimensionality reduction strategy, the complex multi-parameter joint inversion involving velocity, density, and thickness is simplified to a single-parameter (coal thickness only) inversion, fundamentally eliminating multiple solutions. Simultaneously, based on the reduction of unknown parameters and the optimization of the objective function's shape, the conjugate gradient method is used to quickly converge to the optimal solution, significantly improving computational efficiency.

Description

Technical Field

[0001] This invention belongs to the field of underground geophysical exploration technology in coal mines, specifically relating to a method for inverting coal seam thickness using Love-type channel wave dispersion. Background Technology

[0002] Downhole channel wave seismic detection technology has advantages such as long detection distance, high accuracy, and strong anti-interference ability, making it an effective means of detecting geological structures such as faults, collapse columns, and coal thickness variations within the working face. Traditional channel wave coal thickness detection methods mainly rely on single-frequency velocity tomography (velocity CT), which requires acquiring channel wave data, retrieving the channel wave velocity distribution in different areas of the working face through tomography, and then indirectly inferring the coal thickness using the empirical relationship between velocity and coal seam thickness. However, this indirect path of "velocity-empirical formula-coal thickness" introduces large error propagation, and relying on roadway exposure points to fit the "velocity-coal thickness" relationship can lead to low inversion accuracy and poor reliability due to insufficient boreholes and insufficient fitting samples.

[0003] To overcome the shortcomings of traditional methods, existing technologies attempt to directly invert coal seam thickness based on theoretical dispersion equations and using full-band or multi-order channel wave information. In his research paper "Study on Three-Dimensional Dispersion Inversion Imaging of Channel Waves in Coal Seam Working Faces" (Journal of Coal Science and Technology, 2025, Vol. 50 (Issue 12)), Ji Guangzhong proposed using the generalized S-transform (GST) to extract the dispersion curve of Love-type channel waves. Then, the working face is divided into blocks, and the entire dispersion curve of each block is inverted, finally integrating them to obtain the coal thickness distribution of the entire working face. Regarding the inversion algorithm, this research mainly explores nonlinear global optimization algorithms such as genetic algorithms and particle swarm optimization. Chinese patent CN120315027A discloses a working face coal thickness inversion method based on full-band multi-order Love-type channel wave dispersion curves, which directly inverts coal seam thickness using the dispersion curves of Love-type channel waves. Based on the time-frequency domain decomposition of the channel wave, the attenuation coefficient b(v,f) of the channel wave with different frequencies and velocities at each imaging point on the working face is obtained through an inversion algorithm. Then, by measuring the consistency between the theoretical dispersion curve of different coal thicknesses and b(v,f) at each point, the coal thickness at that point is estimated, thus achieving accurate detection of the coal thickness on the working face. The above-mentioned existing technical solution has, to some extent, solved the error propagation caused by the indirect path of "velocity-empirical formula-coal thickness" in traditional channel wave coal thickness detection methods, as well as the problem of poor reliability caused by insufficient boreholes and insufficient fitting samples. However, in practical applications, the following three shortcomings still exist:

[0004] (1) Poor noise resistance and blurred boundaries: Actual seismic data often contains noise, and the existing technical solutions mentioned above lack effective mathematical constraints in the inversion process, resulting in poor noise resistance and easy to cause false oscillations in the inversion results; at the same time, least squares fitting or smooth inversion tends to produce smooth solutions, resulting in unclear identification of the coal seam roof and floor boundaries and low resolution.

[0005] (2) Multiple solutions due to multi-parameter coupling: The dispersion curve is affected by multiple parameters such as coal seam thickness, shear wave velocity of surrounding rock and coal seam, and density. The above-mentioned existing technical solutions involve multi-parameter joint inversion of velocity, density and thickness, which is prone to falling into the "thickness-velocity" coupling effect (that is, thinning of thickness and increase of velocity may produce similar dispersion curves), resulting in non-unique inversion results and poor stability of solutions.

[0006] (3) Low computational efficiency: Although the global optimization methods such as genetic algorithms used in the above-mentioned existing technical solutions can theoretically find the global optimal solution, they require traversing multi-dimensional space, resulting in high computational costs and slow convergence speed, which makes it difficult to meet the efficiency requirements of engineering sites. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the technical problem to be solved by this invention is to provide a method for inverting coal seam thickness using a Love-type channel wave dispersion method, which enhances the algorithm's noise resistance and boundary identification capabilities, optimizes complex multi-parameter joint inversion to eliminate multiple solutions, and significantly improves computational efficiency.

[0008] The core idea of ​​this invention is as follows: by introducing regularization constraints and sparsity factors to optimize the objective function, the algorithm's noise resistance and boundary recognition capabilities are enhanced from a mathematical perspective; by using a parameter pre-acquisition dimensionality reduction strategy, the complex multi-parameter joint inversion is simplified to a single-parameter (coal thickness only) inversion, fundamentally eliminating multiple solutions; at the same time, based on the reduction of unknown parameters and the optimization of the objective function shape (smoother, fewer local extrema, reducing the risk of getting trapped in local optima), the conjugate gradient method is used to quickly converge to the optimal solution, significantly improving computational efficiency.

[0009] The method of the present invention specifically includes the following steps:

[0010] Step S1: A seismic channel wave exploration and observation system is set up at the coal mining face using the transmission method. Seismic sources and geophones are set up in the roadways on both sides of the working face, and seismic channel wave exploration data is obtained using dual-component geophones.

[0011] Step S2 involves filtering and denoising the raw seismic channel wave data acquired in step S1, preprocessing the amplitude gain, removing mine noise interference signals, and improving the signal-to-noise ratio of the seismic signal.

[0012] Step S3: Through wavefield separation, the effective Love-type trough wave signal is separated from the seismic trough wave data preprocessed in step S2, ensuring that subsequent analysis is carried out only for the target trough wave signal;

[0013] Preferably, in step S3, the wavefield separation is performed using a time-frequency domain adaptive covariance matrix polarization filtering method, which includes the following steps:

[0014] The time-frequency transformation of the multi-component seismic data time domain signal after preprocessing in step S2 is performed. At each time-frequency point, a covariance matrix is ​​constructed using the complex spectrum values ​​of different components to quantitatively describe the correlation, amplitude ratio and phase difference between the component signals.

[0015] By performing eigenvalue decomposition on the covariance matrix, the dominant polarization direction and energy characteristics of the signal are extracted: the Love-type slot wave belongs to the SH wave, and its particle vibration direction is perpendicular to the wave propagation direction and is mainly distributed on the horizontal plane; the longitudinal wave P wave in the interference wave has the particle vibration direction consistent with the propagation direction, and the Rayleigh-type slot wave particles are elliptically polarized P-SV waves.

[0016] Based on the differences in polarization characteristics mentioned above, an adaptive filter is designed to identify and retain signal components that conform to the Love wave characteristics at each point in the time-frequency domain, while suppressing or eliminating interference components that conform to the longitudinal wave or Rayleigh wave characteristics. Finally, the filtered data is restored to the time domain signal through inverse transformation to obtain a pure and effective Love wave slot signal.

[0017] Step S4: Perform dispersion analysis on the effective Love-type slot wave signal separated in step S3 to generate a dispersion map. Pick and connect the signals from the dispersion map to form the actual dispersion curve.

[0018] Preferably, in step S4, the dispersion analysis method is an improved S-transform method, which includes the following steps:

[0019] An adjustment coefficient is introduced on the basis of the standard S-transform to flexibly control the width and shape of the transform window in the time-frequency plane, thereby improving the fixed scaling rule used in the standard S-transform where the window function width is usually inversely proportional to the frequency.

[0020] Apply the S-transform with the introduced adjustment coefficient to the effective Love-type slot wave signal separated in step 3. By adjusting the adjustment coefficient, the window function is made to match the dispersion characteristics of the slot wave signal.

[0021] Calculate the magnitude of the transformed complex matrix to obtain the dispersion map of the time-frequency energy distribution.

[0022] Because of the introduction of the adjustment coefficient, the signal energy no longer diverges in the time-frequency plane, but is highly focused on the curve representing the propagation law of the slot wave. The shades of color in the figure directly correspond to the strength of the signal energy (the red area represents the high-energy region, i.e., the dominant frequency dispersion energy cluster of the slot wave).

[0023] Step S5: Based on the three-layer symmetrical geological model of "surrounding rock-coal seam-surrounding rock", standard theoretical dispersion templates are formed under different coal thickness conditions;

[0024] Preferably, in step S5, the phase velocity dispersion equation of the Love-type channel wave is used, and the phase velocity dispersion formula is derived by combining the correlation between the circular frequency, coal seam thickness, channel wave phase velocity, surrounding rock and coal seam shear modulus, shear wave velocity and density; then, by using the implicit function differentiation rule and combining the derivative relationship between group velocity and phase velocity, the group velocity dispersion curve of the Love-type channel wave is calculated, forming a standard theoretical dispersion template under different coal thickness conditions.

[0025] Step S6: With the goal of minimizing the residual between the theoretical dispersion curve obtained in step S5 and the actual dispersion curve obtained in step S4, construct the initial inversion objective function.

[0026] Preferably, in step S6, a regularization constraint term and a sparsity factor are introduced to optimize and form a nonlinear initial inversion objective function with regularization constraints as follows:

[0027] (1),

[0028] In the formula, For coal seam thickness, , The shear wave velocity of the surrounding rock and coal seam. , These are the densities of the surrounding rock and the coal seam, respectively. For the phase velocity of the Love-type slot wave, For the slot wave frequency, For noise variance, As a regularization factor, It is the sparse factor under regularization constraints.

[0029] Regularization constraints can constrain the smoothness or range of variation of inversion results, effectively suppress interference caused by data noise, and prevent drastic fluctuations in inversion results that do not conform to geological laws. The sparsity factor guides the inversion results to converge toward the sparse solution, which helps to highlight the main variation characteristics of coal seam thickness, filter out minor anomalies caused by non-geological factors, and make the results more geologically interpretable.

[0030] Preferably, in step S6, the shear wave velocity and density parameters of the surrounding rock and coal seam are obtained by first arrival wave picking, P-wave and S-wave velocity conversion and Gardner equation, and are used as known constants, with only the coal seam thickness set as the only unknown inversion parameter.

[0031] The dispersion energy of Love-type channel waves is mainly concentrated in the coal seam, but their propagation speed is affected by the surrounding rock. By picking up the first arrival wave and using the travel time of the direct P-wave or S-wave, the average velocity of the roof and floor surrounding rock can be quickly and accurately inverted. Through P-wave and S-wave conversion and the Gardner equation, the S-wave velocity and density can be estimated from the P-wave velocity using empirical rock physics formulas. By using these parameters as known constant inputs, only the coal seam thickness remains as an unknown in the inversion equation. The inversion problem is simplified from a complex "multi-parameter joint inversion" to an efficient "single-parameter (coal thickness) inversion", reducing the degree of freedom of the problem and making the inversion results more stable and unique. More importantly, because the simplified objective function is smoother and has fewer local extrema, the risk of getting trapped in local optima is reduced, making it possible to apply the more computationally efficient conjugate gradient method.

[0032] Step S7: Iteratively solve the initial inversion objective function constructed in step S6. By iterating and correcting the model parameters multiple times, achieve the optimal matching between the theoretical dispersion curve and the actual dispersion curve, and directly invert to obtain the coal seam thickness value on the ray path from the source to the detector.

[0033] Preferably, in step S7, the conjugate gradient method is used to iteratively solve the initial inversion objective function optimized in step S6. The conjugate direction is constructed using the first derivative, and the minimum value of the objective function is searched along this direction. The model parameters are corrected through multiple iterations to achieve the optimal matching between the theoretical dispersion curve and the actual dispersion curve, and the coal seam thickness value on the ray path from the source to the detector is directly obtained through inversion.

[0034] Step S8: The inverted coal thickness data of all effective rays within the working face area are divided into grids, the average coal thickness of each ray within a single grid is calculated, and a contour map of the coal thickness distribution in the working face is generated by spatial interpolation to visually display the coal thickness variation pattern across the entire working face.

[0035] The beneficial effects obtained by adopting the above technical solution are as follows:

[0036] (1) In constructing the inversion objective function, this invention introduces a regularization constraint term and a sparsity factor. The regularization term is used to suppress the violent oscillation of model parameters and smooth noise interference, while the sparsity factor utilizes the sparsity characteristics of the coal seam geological structure to guide the inversion results to converge towards a direction with clear boundaries and uniform background. This effectively solves the problem of false oscillations in the inversion results caused by noise interference from actual data, and overcomes the boundary ambiguity defect caused by smooth inversion, thereby achieving a clear characterization of the coal seam roof and floor boundaries and improving the stability and resolution of the inversion results.

[0037] (2) This invention utilizes first arrival wave picking, P-wave and S-wave velocity conversion and Gardner equation to obtain the physical properties of the surrounding rock and coal seam such as shear wave velocity and density in advance, and inputs them as known constants into the inversion system, setting only the coal seam thickness as the only unknown inversion parameter; it eliminates the problem of multiple solutions caused by the "thickness-velocity" coupling effect, transforms the complex multi-parameter nonlinear inversion into a single-parameter solution problem, and ensures the uniqueness and accuracy of the inversion results.

[0038] (3) Based on the optimized single-parameter simplified model, the present invention uses the conjugate gradient method for iterative solution, uses the first derivative to construct the conjugate direction, and quickly searches for the minimum value of the objective function along the direction, which can quickly converge to the optimal solution, maintain high accuracy, greatly reduce the computational cost, improve the inversion efficiency, and adapt to large-scale engineering applications. Attached Figure Description

[0039] Figure 1 This is a technical roadmap for the Love-type channel wave dispersion inversion of coal seam thickness according to the present invention.

[0040] Figure 2 This is a diagram of the trench wave transmission method exploration and observation system according to an embodiment of the present invention.

[0041] Figure 3 This is a dispersion characteristic diagram of the slot wave of the sixth detector in the 10th shot of the present invention.

[0042] Figure 4 This is a dispersion inversion coal thickness distribution map of working face 11061 in this embodiment of the invention.

[0043] Figure 5 This is a measured coal thickness distribution diagram of the 11061 working face in an embodiment of the present invention. Detailed Implementation

[0044] The technical solution of the present invention will now be described more clearly and completely with reference to the accompanying drawings.

[0045] This embodiment uses the 11061 working face of Yi'an Coal Mine as the application scenario. This working face mines the No. 21 coal seam of the Lower Permian Shanxi Formation. The geological structure is relatively simple, but the coal thickness varies drastically (from 1m to 8m). It is necessary to achieve accurate detection of the coal thickness over the entire working face. The method of this invention is used to complete the coal thickness inversion. The specific implementation steps are as follows:

[0046] Step 1: Start the SUMMIT II Ex channel wave seismograph, excite the seismic sources one by one, and use dual-component geophones to record the seismic records after each source excitation, storing them in standard seismic data format; strictly control the source excitation intensity during the acquisition process to avoid interference from environmental noise on the effective signal.

[0047] Step 2: Perform preprocessing on the raw acquired seismic data, including filtering, noise reduction, amplitude gain adjustment, and trend correction, to obtain seismic data that can be used for wavefield separation.

[0048] Step 3: Through wavefield separation, the effective Love-type trough wave signal is separated from the seismic trough wave data preprocessed in Step 2, ensuring that subsequent analysis is carried out only for the target trough wave signal;

[0049] Preferably, in step 3, the wavefield separation is performed using a time-frequency domain adaptive covariance matrix polarization filtering method, which includes the following steps:

[0050] Time-frequency transformation is performed on the time-domain signal of the multi-component seismic data after step 2 preprocessing. At each time-frequency point, a covariance matrix is ​​constructed using the complex spectrum values ​​of different components to quantitatively describe the correlation, amplitude ratio and phase difference between the component signals.

[0051] By performing eigenvalue decomposition on the covariance matrix, the dominant polarization direction and energy characteristics of the signal are extracted: the Love-type slot wave belongs to the SH wave, and its particle vibration direction is perpendicular to the wave propagation direction and is mainly distributed on the horizontal plane; the longitudinal wave P wave in the interference wave has the particle vibration direction consistent with the propagation direction, and the Rayleigh-type slot wave particles are elliptically polarized P-SV waves.

[0052] Based on the differences in polarization characteristics mentioned above, an adaptive filter is designed to identify and retain signal components that conform to the Love wave characteristics at each point in the time-frequency domain, while suppressing or eliminating interference components that conform to the longitudinal wave or Rayleigh wave characteristics. Finally, the filtered data is restored to the time domain signal through inverse transformation to obtain a pure and effective Love wave slot signal.

[0053] Step 4: Perform dispersion analysis on the effective Love-type slot wave signal separated in Step 3 to generate a dispersion map. Pick and connect the signals from the dispersion map to form the actual dispersion curve.

[0054] Preferably, in step 4, the dispersion analysis method is an improved S-transform method, which includes the following steps:

[0055] An adjustment coefficient is introduced on the basis of the standard S-transform to flexibly control the width and shape of the transform window in the time-frequency plane;

[0056] Apply this S-transform with an adjustment coefficient to the effective Love-type slot wave signal separated in step 3. By adjusting the adjustment coefficient, the window function is made to match the dispersion characteristics of the slot wave signal.

[0057] Calculate the magnitude of the transformed complex matrix to obtain the dispersion map of the time-frequency energy distribution.

[0058] Because of the introduction of the adjustment coefficient, the signal energy no longer diverges in the time-frequency plane, but is highly focused on the curve representing the propagation law of the slot wave. The shades of color in the figure directly correspond to the strength of the signal energy (the red area represents the high-energy region, i.e., the dominant frequency dispersion energy cluster of the slot wave).

[0059] By combining manual picking with automatic identification, the dispersion points with concentrated energy and obvious characteristics in the dispersion map are picked out and connected to form the actual dispersion curve. The low-frequency dispersion segment with a frequency of less than 50Hz (affected by the cutoff frequency and with poor signal quality) and the invalid segment with a frequency of more than 400Hz are removed. Only the effective frequency segment data of 50Hz to 400Hz are selected for subsequent inversion calculation.

[0060] In this embodiment, dispersion curves were picked up one by one for the seismic traces corresponding to 36 detectors and 22 seismic sources. Data from detectors with poor signal quality were removed. Finally, the distribution range of the seismic traces with the picked dispersion curves basically covered the entire working surface and the distribution was relatively uniform.

[0061] Step 5: Based on the three-layer symmetrical geological model of the working face surrounding rock-coal seam-surrounding rock (with consistent lithological parameters of the roof and floor surrounding rocks), the phase velocity dispersion formula is derived by applying the Love-type channel wave phase velocity dispersion equation and combining the correlation between circular frequency, coal seam thickness, channel wave phase velocity, shear modulus of surrounding rock and coal seam, and shear wave velocity.

[0062] (2)

[0063] In the formula For the slot wave frequency, For coal seam thickness, For the phase velocity of the Love-type slot wave, , These are the shear wave velocities of the surrounding rock and the coal seam, respectively. , These are the densities of the surrounding rock and the coal seam, respectively. The dispersion order (taken in this embodiment) (i.e., fundamental mode dispersion).

[0064] By using the implicit function differentiation rule and combining it with group velocity With phase velocity Substituting the derivative relationship into the above phase velocity dispersion formula, the velocity dispersion curve of the Love-type channel wave group is calculated. Combined with the basic parameters of the surrounding rock and coal seam collected in the early stage, a series of theoretical dispersion curves for coal thickness of 1m to 8m (covering the coal thickness range of the working face) are constructed to form a standard theoretical dispersion template.

[0065] Step 6: With minimizing the residual between the theoretical dispersion curve and the actual dispersion curve as the core, construct the initial inversion objective function; in view of the problems of noise interference in actual data, easy inversion getting trapped in local optima, and non-unique results, introduce regularization constraint terms and sparsity factors to optimize and form a nonlinear objective function with regularization constraints (Equation (1)).

[0066] In step 6, when obtaining the shear wave velocity of the surrounding rock, the travel time of the first-arrival refracted longitudinal wave received by the pickup detector is used to calculate the longitudinal wave velocity of the surrounding rock through least squares fitting.

[0067] (3)

[0068] In the formula, Corresponding epicenter to the first The distance between the detectors From the epicenter to The average distance of each detector. For the first The initial arrival P-wave travel time recorded by each detector for Each detector records the average value of the initial P-wave travel time. After obtaining the P-wave velocity in the surrounding rock, the P-wave ratio is calculated using the empirical formula (P / S ratio). The surrounding rock shear wave velocity is obtained by conversion.

[0069] In step 6, the shear wave velocity of the coal seam is determined by the groove wave velocity corresponding to the high-frequency band (350Hz~400Hz) of the dispersion diagram. In this embodiment, the shear wave velocity of the coal seam is determined to be 1050m / s.

[0070] In step 6, based on the Gardner equation and the calculated P-wave velocities of the surrounding rock and coal seam, the densities of the surrounding rock and coal seam are calculated respectively. The objective function only considers the coal seam thickness. By setting it as a unique unknown inversion parameter, the computational complexity of the inversion is simplified and the stability of the solution is improved.

[0071] Step 7: Use the conjugate gradient method to iteratively solve the optimized inversion objective function to obtain coal thickness data from the source to the detector path.

[0072] Step 8: Process the coal thickness data obtained from the inversion using the Kriging interpolation method. Divide the working face of Yi'an Coal Mine 11061 into a grid, calculate the average value of all valid ray coal thickness data in each grid, and use it as the representative value of the coal thickness for that grid. Then, generate a contour map of the coal thickness distribution in the working face using spatial interpolation. Figure 4 This clearly shows the trend of coal thickness distribution in the working face.

[0073] According to the coal thickness distribution map of working face 11061 obtained by dispersion inversion, the thinnest coal seam is located near 150m on the X-axis and 75m on the Y-axis, with a thickness of about 2m. The coal thickness gradually increases from west to east, with the thickest area being on the east side of the working face, at approximately 7m. This is consistent with the measured coal thickness distribution during the actual mining of this working face. Figure 5 (Comparative verification shows that the blank area on the right side of the figure represents the retained coal pillar, for which no test data was obtained.) It can be seen that the coal thickness distribution obtained through inversion is consistent with the overall trend of the actual coal thickness distribution in the working face. The thinnest coal area in the working face is basically consistent with the distribution of the inversion results, with a coal thickness error of about 10%.

Claims

1. A method for inverting coal seam thickness using Love-type channel wave dispersion, characterized in that, Includes the following steps: Step S1: A seismic channel wave exploration and observation system is set up at the coal mining face using the transmission method. Seismic sources and geophones are set up in the roadways on both sides of the working face, and seismic channel wave exploration data is obtained using dual-component geophones. Step S2 involves filtering and denoising the raw seismic channel wave data acquired in step S1, preprocessing the amplitude gain, removing mine noise interference signals, and improving the signal-to-noise ratio of the seismic signal. Step S3: Through wavefield separation, the effective Love-type trough wave signal is separated from the seismic trough wave data preprocessed in step S2, ensuring that subsequent analysis is carried out only for the target trough wave signal; Step S4: Perform dispersion analysis on the effective Love-type slot wave signal separated in step S3 to generate a dispersion map. Pick and connect the signals from the dispersion map to form the actual dispersion curve. Step S5: Based on the three-layer symmetrical geological model of "surrounding rock-coal seam-surrounding rock", standard theoretical dispersion templates are formed under different coal thickness conditions; Step S6: With the goal of minimizing the residual between the theoretical dispersion curve obtained in step S5 and the actual dispersion curve obtained in step S4, construct the initial inversion objective function. Step S7: Iteratively solve the initial inversion objective function constructed in step S6. Correct the model parameters through multiple iterations and directly invert to obtain the coal seam thickness value on the ray path from the source to the detector. Step S8: The inverted coal thickness data of all effective rays within the working face area are divided into grids, the average coal thickness of each ray within a single grid is calculated, and a contour map of the coal thickness distribution in the working face is generated by spatial interpolation to visually display the coal thickness variation pattern across the entire working face.

2. The method for inverting coal seam thickness according to claim 1, characterized in that, In step S3, the wavefield separation is performed using a time-frequency domain adaptive covariance matrix polarization filtering method, including the following steps: The time-frequency transformation of the multi-component seismic data time domain signal after preprocessing in step S2 is performed. At each time-frequency point, a covariance matrix is ​​constructed using the complex spectrum values ​​of different components to quantitatively describe the correlation, amplitude ratio and phase difference between the component signals. By performing eigenvalue decomposition on the covariance matrix, the dominant polarization direction and energy characteristics of the signal are extracted: where... The SH wave is a type of groove wave, in which the particle vibration direction is perpendicular to the wave propagation direction and is mainly distributed on the horizontal plane; the longitudinal wave P wave in the interference wave has the particle vibration direction in the same direction as the propagation direction, and the Rayleigh groove wave has elliptically polarized P-SV waves. Based on the differences in polarization characteristics mentioned above, an adaptive filter is designed to identify and retain signal components that conform to the Love wave characteristics at each point in the time-frequency domain, while suppressing or eliminating interference components that conform to the longitudinal wave or Rayleigh wave characteristics. Finally, the filtered data is restored to the time domain signal through inverse transformation to obtain a pure and effective Love wave slot signal.

3. The method for inverting coal seam thickness according to claim 1, characterized in that, In step S4, the dispersion analysis method is an improved S-transform method, which includes the following steps: An adjustment coefficient is introduced on the basis of the standard S-transform to flexibly control the width and shape of the transform window in the time-frequency plane, thereby improving the fixed scaling rule used in the standard S-transform where the window function width is usually inversely proportional to the frequency. Apply the S-transform with the introduced adjustment coefficient to the effective Love-type slot wave signal separated in step S3. By adjusting the adjustment coefficient, the window function is made to match the dispersion characteristics of the slot wave signal. Calculate the magnitude of the transformed complex matrix to obtain the dispersion map of the time-frequency energy distribution.

4. The method for inverting coal seam thickness according to claim 1, characterized in that, In step S5, the phase velocity dispersion equation of the Love-type channel wave is used, and the phase velocity dispersion formula is derived by combining the correlation between the circular frequency, coal seam thickness, channel wave phase velocity, surrounding rock and coal seam shear modulus, shear wave velocity and density. Then, by using the implicit function differentiation rule and combining the derivative relationship between group velocity and phase velocity, the group velocity dispersion curve of the Love-type channel wave is calculated, forming a standard theoretical dispersion template under different coal thickness conditions.

5. The method for inverting coal seam thickness according to any one of claims 1-4, characterized in that, In step S6, a regularization constraint term and a sparsity factor are introduced to optimize and form a nonlinear initial inversion objective function with regularization constraints as follows: ， In the formula, For coal seam thickness, , The shear wave velocity of the surrounding rock and coal seam. , These are the densities of the surrounding rock and the coal seam, respectively. For the phase velocity of the Love-type slot wave, For the slot wave frequency, For noise variance, As a regularization factor, It is the sparse factor under regularization constraints.

6. The method for inverting coal seam thickness according to claim 5, characterized in that, In step S6, the shear wave velocity of the surrounding rock and the coal seam, and the density of the surrounding rock and the coal seam are obtained by first arrival wave picking, P-wave and S-wave velocity conversion and Gardner equation, and are used as known constants. Only the coal seam thickness is set as the only unknown inversion parameter.

7. The method for inverting coal seam thickness according to claim 6, characterized in that, In step S7, the conjugate gradient method is used to iteratively solve the initial inversion objective function optimized in step S6. The first derivative is used to construct the conjugate direction, and the minimum value of the objective function is searched along the direction. The model parameters are corrected through multiple iterations, and the coal seam thickness value on the ray path from the source to the detector is directly obtained by inversion.

Images