Tunnel face monitoring system and method

Abstract

The present invention discloses a tunnel face monitoring system and method, which relate to the technical field of rock and soil geological monitoring. The system collects reflected signals of rock strata at the tunnel face through geological radar equipment and performs denoising processing through wavelet transform to generate a noise reduction data set D1, which effectively reduces noise interference in the data and improves the availability of the original data. Secondly, radar signal features are extracted to form a feature vector V, so that the physical characteristics of different geological layers can be accurately expressed. Then, a support vector machine classification algorithm is used for analysis to obtain a preliminary demarcation point set P1, and preliminary identification of geological layer boundaries is quickly completed. Further correction is performed to obtain a corrected demarcation point set P2, eliminating boundary errors that may be introduced by the classification algorithm. The corrected demarcation point set P2 is then compared with the demarcation stability state threshold Tthe to generate a geological stability state identifier S, clarify the geological stability within the tunnel face area, and realize a complete closed loop of demarcation modeling and construction plan generation.

Description

Technical Field

[0001] The present invention relates to the technical field of rock and soil geological monitoring, and in particular to a tunnel face monitoring system and method. Background Art

[0002] Tunnel construction is a key branch of civil engineering, involving the planning, design, and construction of infrastructure such as tunnels, underground spaces, and bridges. The safety and efficiency of the construction process are core concerns in engineering technology. During tunnel construction, the complexity of the surrounding rock geology makes geological layer monitoring a key technology for ensuring construction safety. The tunnel face, as the working surface in direct contact with the surrounding rock, determines the clarity of its geological layer boundaries, which not only determines the adjustment of the construction plan but also directly affects the design and implementation of the support structure.

[0003] For example, at the junction of different geological layers, if the transition area between the weak layer and the hard layer is not accurately identified, it may cause support failure and surrounding rock instability, thus affecting the safety of tunnel construction.

[0004] In current tunnel construction practice, the identification of geological layer boundaries primarily relies on geological radar detection technology, which analyzes rock layer boundaries through radar reflection wave signals. However, this method faces multiple challenges in practical application. First, geological radar data is subject to severe noise interference, resulting in unclear boundary recognition results, especially in areas with multiple layers of geological interlaced. Second, traditional boundary detection algorithms have limited data processing capabilities and cannot fully utilize existing sensor data, resulting in low recognition accuracy. Finally, due to the resolution limitations of current monitoring technology, construction teams often need to combine manual experience with data analysis, which not only increases time costs but also increases the risk of misjudgment. These issues make it difficult for existing geological layer monitoring technology to fully meet the safety requirements of construction under complex geological conditions. Summary of the Invention

[0005] In view of the deficiencies of the prior art, the present invention provides a tunnel face monitoring system and method, which solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A tunnel face monitoring method, comprising the following steps:

[0007] S1, using geological radar equipment to collect the reflected signal of the rock formation at the tunnel face to form the original data set D0, and then using wavelet transform to denoise the data set D1;

[0008] S2. Extract radar signal features from the noise reduction dataset D1 to obtain rock material difference features, signal attenuation features, and main frequency offset features to form a feature vector V;

[0009] S3, using the support vector machine classification algorithm to perform preliminary identification of the address layer boundary on the feature vector V, and obtaining a preliminary demarcation point set P1;

[0010] S4, correcting the obtained preliminary demarcation point set P1 to obtain a corrected demarcation point set P2;

[0011] S5. By using the demarcation point set P2 to establish a three-dimensional model of the geological layer boundary, a three-dimensional model of the geological layer of the tunnel face is obtained, and the model is simultaneously compared with the boundary stability state threshold Tthe to generate a geological stability state identifier S and a construction support plan.

[0012] Preferably, said S1 includes S11 and S12;

[0013] S11. Use geological radar equipment to scan the tunnel face, obtain radar reflection signals, and record them as the original data set D0. Specifically, the geological radar transmits high-frequency electromagnetic waves into the rock layer. Different rock layers have different reflection intensities and propagation times for electromagnetic waves. The returned radar reflection signals contain the dielectric properties of the rock layer and the boundary information between geological layers, including the echo time T, the reflection signal intensity I, and the signal main frequency Fp. The radar echo signal S = {T, I, Fp} of the scanning line i is composed of

[0014] The original data set D0 is specifically D0={Si(t), i=1, 2, 3, ..., n}, where Si(t) represents the radar echo signal of the i-th scan line at time t, and n represents the total number of scan lines.

[0015] Preferably, S12, performing wavelet transform denoising processing on the acquired original data set D0, separating different frequency bands of the signal, removing the noise signal and retaining the high-frequency components of the geological information, to form a denoised data set D1.

[0016] Preferably, said S2 includes S21 and S22;

[0017] S21. Extract radar signal features from the noise reduction data set D1 to obtain the rock material difference feature △Er, the signal attenuation feature A, and the main frequency offset feature △Fp, forming a feature vector V = {△Er, A, △Fp};

[0018] Among them, the difference characteristic △Er of the rock material is obtained by extracting the dielectric constant change rate to reflect the response characteristics of the rock material to electromagnetic waves. Specifically, by comparing the echo time T1 and echo time T2 of different geological layers, the dielectric constant change rate between the two layers is calculated to obtain the difference characteristic △Er of the rock material;

[0019] The signal attenuation feature A is obtained by extracting the attenuation of the signal intensity with the propagation distance to reflect the changes in the density and water content of the geological layer. Specifically, the signal attenuation state of the reflected signal intensity I1 and the reflected signal intensity I2 at different geological interfaces is compared to obtain the signal attenuation feature A;

[0020] The main frequency offset feature △Fp is obtained by analyzing the signal spectrum and extracting the main frequency change characteristics of different geological layer interfaces. Specifically, the main frequency offset value is obtained by calculating the main frequency peak Fp1 and the main frequency peak Fp2 of different geological interfaces to obtain the main frequency offset feature △Fp.

[0021] Preferably, said S3 includes S31 and S32;

[0022] S31, performing data normalization processing on the feature vector V to obtain a normalized standard feature vector Vn, and then using a support vector machine classification algorithm to preliminarily identify the address layer boundary, specifically finding a hyperplane through the support vector machine classification algorithm;

[0023] The hyperplane is obtained by the following target classification function:

[0024]

[0025] Where f(F) represents the output of the target classification function, specifically the category of the Fth standard feature vector in the standard feature vector Vn. The category is judged specifically based on the output f(F). sign represents the sign function, m represents the number of support vectors, Sj represents the jth support vector, wj represents the weight value of the jth support vector, K(F, Sj) represents the kernel function, b represents the bias term of the hyperplane, which is specifically used to adjust the position of the hyperplane, and m represents the number of support vectors.

[0026] Preferably, S32, classifying the standard feature vector Vn according to the found hyperplane, and simultaneously separating the boundary points of different categories of geological layer data to obtain a preliminary boundary point set P1;

[0027] The preliminary demarcation point set P1 is specifically formed by inferring the input standard feature vector Vn using the target classification function, and statistically analyzing the classification results of each demarcation point according to the output f(F) of the target classification function, and then marking the position of the boundary point to form the preliminary demarcation point set P1;

[0028] The preliminary demarcation point set P1 is obtained by the calculation formula P1={z|f(F)=0}, which specifically represents a collection of feature vector points that satisfy the target classification function output of 0, representing the boundary points of the geological layer.

[0029] Preferably, said S4 includes S41;

[0030] S41, revising the obtained preliminary demarcation point set P1, including analyzing the distribution of neighboring points of the demarcation point position in the preliminary demarcation point set P1, including determining that the current demarcation point needs to be revised if the position of the neighboring point deviates by 10% from the average position of the neighboring points, and substituting the revised demarcation point back into the preliminary demarcation point set P1 to form a revised demarcation point set P2;

[0031] The correction includes using a neighboring point interpolation method to perform correction and smooth the position of the dividing point.

[0032] Preferably, said S5 includes S51 and S52;

[0033] S51, establishing a three-dimensional model of geological layer boundaries by using the demarcation point set P2 to obtain a three-dimensional model M of the geological layer of the tunnel face, and generating a continuous surface of the geological layer boundary by performing spatial interpolation fitting on the discrete demarcation points in the demarcation point set P2, including fitting the demarcation values Ms (x, y, z) in the obtained three-dimensional interface models;

[0034] The demarcation value Ms(x, y, z) is obtained by the following calculation formula:

[0035]

[0036] Where Ms(x, y, z) represents the boundary value in the three-dimensional interface model, Wm represents the interpolation weight, which is related to the spatial distribution of the boundary point, Qm(x, y, z) represents the interpolation basis function of the mth boundary point, and M represents the total number of boundary points, specifically the total number of boundary points in the boundary point set P2.

[0037] Preferably, S52 compares several boundary values Ms(x, y, z) in the three-dimensional model M of the tunnel face geological layer with the boundary stability threshold Tthe, and associates them with the tolerance range RC to generate a geological stability status identifier S and a construction support plan. The geological stability status identifier S is obtained by the following comparison method:

[0038]

[0039] When S=1, obtain the stable result of the demarcation value Ms(x, y, z), generate the construction support plan and maintain the original plan for support;

[0040] When S=0, the unstable result of the demarcation value Ms(x, y, z) is obtained, and a construction support plan is generated. While maintaining the original support plan, additional support measures related to anchor rods, concrete and steel mesh are planned for the distribution position of the unstable demarcation value Ms(x, y, z).

[0041] A tunnel face monitoring system includes a signal acquisition module, a signal feature extraction module, a preliminary identification module, a correction module, and a modeling and evaluation module;

[0042] The signal acquisition module collects the reflected signal of the rock formation at the tunnel face through the geological radar equipment to form the original data set D0, and generates the noise-reduced data set D1 through wavelet transform denoising.

[0043] The signal feature extraction module extracts radar signal features from the noise reduction data set D1, obtains rock material difference features, signal attenuation features and main frequency offset features, and forms a feature vector V;

[0044] The preliminary identification module performs preliminary identification of the address layer boundary by using the support vector machine classification algorithm on the feature vector V to obtain a preliminary demarcation point set P1;

[0045] The correction module corrects the obtained preliminary demarcation point set P1 to obtain a corrected demarcation point set P2;

[0046] The modeling and evaluation module establishes a three-dimensional model of the geological layer boundary by using the boundary point set P2, obtains the three-dimensional model M of the geological layer of the tunnel face, compares it with the boundary stability state threshold Tthe, generates a geological stability state mark S, and generates a construction support plan.

[0047] The present invention provides a tunnel face monitoring system and method, which have the following beneficial effects:

[0048] (1) The reflected signal of the face rock formation is collected by geological radar equipment and processed by wavelet transform to generate a noise reduction data set D1, which effectively reduces the noise interference in the data and improves the usability of the original data. Secondly, the radar signal features are extracted to form a feature vector V, so that the physical characteristics of different geological layers can be accurately expressed. Then, the support vector machine classification algorithm is used for analysis to obtain a preliminary demarcation point set P1, which quickly completes the preliminary identification of the geological layer boundary. Further correction is performed to obtain a corrected demarcation point set P2, eliminating the boundary error that may be introduced by the classification algorithm and significantly improving the accuracy of the demarcation point. Finally, by constructing a three-dimensional model M of the geological layer boundary of the face and comparing it with the boundary stability threshold Tthe, a geological stability state identifier S is generated to clarify the geological stability within the face area. At the same time, a scientific and reasonable construction support plan is provided, realizing a complete closed loop of demarcation modeling and construction plan generation, solving the defects of the traditional method in insufficient accuracy of geological layer demarcation point identification, delayed geological stability assessment, and lack of targeted support plan, and significantly improving the safety and construction efficiency during tunnel construction.

[0049] (2) By standardizing the data of the eigenvector V, the standardized standard eigenvector Vn is obtained, and the support vector machine classification algorithm is used to perform preliminary identification of the geological layer boundary and find the optimal classification hyperplane. The hyperplane is determined by the target classification function f(F) and accurately distinguishes the category of the standard eigenvector Vn. The low-dimensional features are simultaneously mapped to the high-dimensional space, effectively dealing with the nonlinear boundary problem under complex geological conditions. By counting and marking the classification boundary points, a preliminary boundary point set P1 is formed. Each eigenvector point that meets the classification boundary is processed in a set, accurately representing the preliminary boundary state of the geological layer. It can efficiently and accurately identify the boundary position of different geological layers on the face, overcoming the problems of long time consumption and strong subjectivity in the traditional manual judgment process, providing accurate initial boundary data for subsequent boundary correction and modeling, and significantly improving the efficiency and reliability of geological layer classification.

[0050] (3) By analyzing and correcting the distribution of neighboring points of the initial demarcation point set P1, a more accurate corrected demarcation point set P2 was constructed, effectively solving the problem of discontinuous distribution of demarcation points caused by the deviation of the initial demarcation point positions. The correction process uses the neighboring point interpolation method to smooth the demarcation point positions, ensuring that the spatial distribution of the demarcation points is more coherent and consistent with the actual boundary characteristics of the geological layer. Subsequently, the three-dimensional model M of the geological layer of the tunnel face is constructed using the demarcation point set P2, realizing spatial interpolation fitting from discrete demarcation points to continuous demarcation surfaces, and generating a complete geological layer boundary model. A geological stability indicator S is generated by comparing several boundary values Ms(x, y, z) in the 3D model M with the boundary stability threshold Tthe, combined with the tolerance range RC. Based on this, the support scheme further distinguishes between stable and unstable areas and proposes additional support measures for unstable areas. This provides a more comprehensive basis for geological boundary demarcation and support scheme optimization for tunnel construction, significantly improving the accuracy of geological boundary modeling at the tunnel face and the safety of construction support. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic diagram of the steps of a tunnel face monitoring method according to the present invention;

[0052] Figure 2 The figure is a schematic block diagram of a tunnel face monitoring system according to the present invention. DETAILED DESCRIPTION

[0053] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0054] Example 1

[0055] The present invention provides a tunnel face monitoring method, please refer to Figure 1 , including the following steps:

[0056] S1, using geological radar equipment to collect the reflected signal of the rock formation at the tunnel face to form the original data set D0, and then using wavelet transform to denoise the data set D1;

[0057] S2. Extract radar signal features from the noise reduction dataset D1 to obtain rock material difference features, signal attenuation features, and main frequency offset features to form a feature vector V;

[0058] S3, using the support vector machine classification algorithm to perform preliminary identification of the address layer boundary on the feature vector V, and obtaining a preliminary demarcation point set P1;

[0059] S4, correcting the obtained preliminary demarcation point set P1 to obtain a corrected demarcation point set P2;

[0060] S5. By using the demarcation point set P2 to establish a three-dimensional model of the geological layer boundary, a three-dimensional model of the geological layer of the tunnel face is obtained, and the model is simultaneously compared with the boundary stability state threshold Tthe to generate a geological stability state identifier S and a construction support plan.

[0061] In this embodiment, rock stratum reflection signals from the tunnel face are collected using geological radar equipment and denoised using wavelet transform to generate a de-noised dataset D1. This effectively reduces noise interference in the data and improves the usability of the original data. Next, radar signal features are extracted to form a feature vector V, enabling accurate representation of the physical properties of different geological strata. This is then analyzed using a support vector machine classification algorithm to obtain a preliminary demarcation point set P1, rapidly identifying the boundaries of geological strata. Further corrections are performed to obtain a corrected demarcation point set P2, eliminating boundary errors that may be introduced by the classification algorithm and significantly improving the accuracy of demarcation points. Finally, a three-dimensional model M of the tunnel face geological stratum boundary is constructed and compared with the boundary stability threshold Tthe to generate a geological stability indicator S, clarifying the geological stability within the tunnel face area. A scientific and reasonable construction support plan is also provided, completing a complete closed-loop process of boundary modeling and construction plan generation. This addresses the shortcomings of traditional methods, such as insufficient accuracy in identifying geological stratum boundary points, delayed geological stability assessment, and lack of targeted support plans, significantly improving safety and efficiency during tunnel construction.

[0062] Example 2

[0063] This embodiment is explained in Example 1, please refer to Figure 1, specifically: the S1 includes S11 and S12;

[0064] S11. Use geological radar equipment to scan the tunnel face, obtain radar reflection signals, and record them as the original data set D0. Specifically, the geological radar transmits high-frequency electromagnetic waves into the rock layer. Different rock layers have different reflection intensities and propagation times for electromagnetic waves. The returned radar reflection signals contain the dielectric properties of the rock layer and the boundary information between geological layers, including the echo time T, the reflection signal intensity I, and the signal main frequency Fp. The radar echo signal S = {T, I, Fp} of the scanning line i is composed of

[0065] The original data set D0 is specifically D0={Si(t), i=1, 2, 3, ..., n}, where Si(t) represents the radar echo signal of the i-th scan line at time t, and n represents the total number of scan lines.

[0066] S12, performing wavelet transform denoising processing on the acquired original data set D0, separating different frequency bands of the signal, removing the noise signal and retaining the high-frequency components of the geological information, to form a denoised data set D1;

[0067] Among them, the wavelet transform denoising process is specifically processed by the following method:

[0068] Signal decomposition: The radar echo signal Si(t) in the original data set D0 is input into the wavelet transform model. Based on the multi-resolution characteristics of the wavelet, the signal is decomposed into high-frequency components and low-frequency components in different frequency bands. The low-frequency components contain the overall trend information of the signal, and the high-frequency components include boundary and geological layer interface information.

[0069] Threshold processing: The high-frequency components obtained by decomposition are filtered by using a preset processing threshold. The noise is concentrated in the high-frequency part, and its amplitude is small. By setting a threshold, the signal below the threshold is set to zero, thereby retaining the main geological information and removing noise interference;

[0070] Signal reconstruction: The high-frequency components that have been thresholded are recombined with the original low-frequency components to restore them to a noise-reduced signal. This process can preserve the boundary characteristics of the geological layer as much as possible while eliminating unnecessary noise components.

[0071] Said S2 includes S21 and S22;

[0072] S21. Extract radar signal features from the noise reduction data set D1 to obtain the rock material difference feature △Er, the signal attenuation feature A, and the main frequency offset feature △Fp, forming a feature vector V = {△Er, A, △Fp};

[0073] Among them, the difference characteristic △Er of the rock material is obtained by extracting the dielectric constant change rate to reflect the response characteristics of the rock material to electromagnetic waves. Specifically, by comparing the echo time T1 and echo time T2 of different geological layers, the dielectric constant change rate between the two layers is calculated to obtain the difference characteristic △Er of the rock material;

[0074] The rock material difference characteristic ΔEr is obtained by the following calculation formula:

[0075]

[0076] Where c represents the propagation speed of electromagnetic waves in a vacuum, and d represents the spacing between ground radar antennas;

[0077] The signal attenuation feature A is obtained by extracting the attenuation of the signal intensity with the propagation distance to reflect the changes in the density and water content of the geological layer. Specifically, the signal attenuation state of the reflected signal intensity I1 and the reflected signal intensity I2 at different geological interfaces is compared to obtain the signal attenuation feature A;

[0078] The signal attenuation characteristic A is obtained by the following calculation formula:

[0079]

[0080] Where, e represents the vertical spatial distance between different geological interfaces, and ln represents the natural logarithm;

[0081] The main frequency offset feature △Fp is obtained by analyzing the signal spectrum and extracting the main frequency variation characteristics of different geological layer interfaces. Specifically, the main frequency offset value is obtained by calculating the main frequency peak value Fp1 and the main frequency peak value Fp2 of different geological interfaces to obtain the main frequency offset feature △Fp;

[0082] The main frequency offset characteristic ΔFp is obtained by the calculation formula ΔFp=|Fp1-Fp2|.

[0083] In this example, a geological radar device is used to scan the tunnel face, acquiring radar echo signals and recording them as a raw data set D0. The acquired signals contain the dielectric properties of the rock strata and information about the boundaries between geological strata. The echo time T, reflected signal intensity I, and signal dominant frequency Fp are recorded in detail, laying a high-resolution foundation for subsequent processing. Secondly, wavelet transform denoising is used to decompose, threshold, and reconstruct the raw data set D0, generating a de-noised data set D1. This accurately preserves the high-frequency components of the geological stratum boundary characteristics, effectively removes environmental noise interference, and ensures the integrity of the geological information in the signal. Furthermore, radar signal feature extraction is performed to obtain the rock stratum material difference characteristic ΔEr, signal attenuation characteristic A, and dominant frequency offset characteristic ΔFp, forming a feature vector V. This accurately quantifies the dielectric constant variation, density and water content differences, and frequency characteristics of the geological strata, enabling a precise description of the physical property differences between geological strata. This effectively addresses the severe noise interference and inaccurate physical property extraction issues in traditional geological radar data, significantly improving the accuracy and reliability of data processing and providing a high-quality input foundation for subsequent geological stratum classification and boundary modeling.

[0084] Example 3

[0085] This embodiment is explained in Example 2, please refer to Figure 1 , specifically: said S3 includes S31 and S32;

[0086] S31, performing data normalization processing on the feature vector V to obtain a normalized standard feature vector Vn, and then using a support vector machine classification algorithm to preliminarily identify the address layer boundary, specifically finding a hyperplane through the support vector machine classification algorithm;

[0087] The hyperplane is obtained by the following target classification function:

[0088]

[0089] Wherein, f(F) represents the output of the target classification function, specifically represents the category of the Fth standard feature vector in the standard feature vector Vn, and the category is judged specifically according to the output f(F). Sign represents the sign function, and specifically returns its sign information according to the positive and negative situation in the sign(*) brackets. When * in sign(*) is greater than 0, the sign function sign returns 1 to indicate a positive number. When * in sign(*) is equal to 0, the sign function sign returns 0 to indicate zero. When * in sign(*) is less than 0, the sign function sign returns -1 to indicate a negative number. When the output f(F) is greater than 0, it indicates that the current standard feature vector Vn belongs to one category. When the output f(F) is less than 0, it indicates that the current standard feature vector Vn belongs to another category. When the output f(F) is equal to 0, it indicates that indicates that the current standard eigenvector Vn belongs to the classification boundary, specifically the preliminary dividing point, m represents the number of support vectors, specifically represents the total number of samples actually participating in the classification in the target classification function, Sj is the jth support vector, specifically represents that the sample dividing point selected during the classification training process is automatically selected through the training process of the support vector machine, the support vector is the training sample point close to the classification boundary, and plays a major role in the shape and position of the classification hyperplane, wj represents the weight value of the jth support vector, K(F, Sj) represents the kernel function, including the Gaussian kernel, which is specifically used to map to the high-dimensional feature space, specifically used to measure the similarity between the Fth standard eigenvector and the Sjth support vector, b represents the bias term of the hyperplane, specifically used to adjust the position of the hyperplane, and m represents the number of support vectors.

[0090] S32, classifying the standard feature vector Vn according to the found hyperplane, and simultaneously separating the boundary points of different categories of geological layer data to obtain a preliminary boundary point set P1;

[0091] The preliminary demarcation point set P1 is specifically formed by inferring the input standard feature vector Vn using the target classification function, and statistically analyzing the classification results of each demarcation point according to the output f(F) of the target classification function, and then marking the position of the boundary point to form the preliminary demarcation point set P1;

[0092] The preliminary demarcation point set P1 is obtained by the calculation formula P1={z|f(F)=0}, which specifically represents a collection of feature vector points that satisfy the target classification function output of 0, representing the boundary points of the geological layer.

[0093] In this embodiment, the data of the eigenvector V is standardized to obtain the standardized standard eigenvector Vn, and the support vector machine classification algorithm is used to perform preliminary identification of the geological layer boundary, and the optimal classification hyperplane is found. The hyperplane is determined by the target classification function f(F), and accurately distinguishes the category of the standard eigenvector Vn; the low-dimensional features are simultaneously mapped to the high-dimensional space, and the nonlinear boundary problem under complex geological conditions is effectively handled. By counting and marking the classification boundary points, a preliminary boundary point set P1 is formed, and each eigenvector point that meets the classification boundary is processed in a set, which accurately represents the preliminary boundary state of the geological layer, and can efficiently and accurately identify the boundary position of different geological layers on the face, overcoming the problems of long time consumption and strong subjectivity in the traditional manual judgment process, providing accurate initial boundary data for subsequent boundary correction and modeling, and significantly improving the efficiency and reliability of geological layer classification.

[0094] Example 4

[0095] This embodiment is explained in Example 3, please refer to Figure 1 Specifically: the S4 includes S41;

[0096] S41, revising the obtained preliminary demarcation point set P1, including analyzing the distribution of neighboring points of the demarcation point position in the preliminary demarcation point set P1, including determining that the current demarcation point needs to be revised if the position of the neighboring point deviates by 10% from the average position of the neighboring points, and substituting the revised demarcation point back into the preliminary demarcation point set P1 to form a revised demarcation point set P2;

[0097] The correction includes using a neighboring point interpolation method to perform correction and smooth the position of the dividing point.

[0098] Said S5 includes S51 and S52;

[0099] S51, establishing a three-dimensional model of geological layer boundaries by using the demarcation point set P2 to obtain a three-dimensional model M of the geological layer of the tunnel face, and generating a continuous surface of the geological layer boundary by performing spatial interpolation fitting on the discrete demarcation points in the demarcation point set P2, including fitting the demarcation values Ms (x, y, z) in the obtained three-dimensional interface models;

[0100] The demarcation value Ms(x, y, z) is obtained by the following calculation formula:

[0101]

[0102] Where Ms(x, y, z) represents the boundary value in the three-dimensional interface model, Wm represents the interpolation weight, which is related to the spatial distribution of the boundary point, Qm(x, y, z) represents the interpolation basis function of the mth boundary point, specifically representing the influence of the mth boundary point on the interpolation point, and M represents the total number of boundary points, specifically representing the total number of boundary points in the boundary point set P2.

[0103] S52. Compare several boundary values Ms(x, y, z) in the three-dimensional model M of the tunnel face geological layer with the boundary stability threshold value Tthe, and associate them with the tolerance range RC to generate a geological stability status identifier S and a construction support plan. The geological stability status identifier S is obtained by the following comparison method:

[0104]

[0105] When S=1, obtain the stable result of the demarcation value Ms(x, y, z), generate the construction support plan and maintain the original plan for support;

[0106] When S=0, the unstable result of the demarcation value Ms(x, y, z) is obtained, and a construction support plan is generated. While maintaining the original support plan, additional support measures related to anchor rods, concrete and steel mesh are planned for the distribution position of the unstable demarcation value Ms(x, y, z).

[0107] In this example, by analyzing and correcting the distribution of neighboring points in the initial demarcation point set P1, a more accurate revised demarcation point set P2 was constructed. This effectively addresses the discontinuous distribution of demarcation points caused by deviations in the initial demarcation point positions. This correction process utilizes neighboring point interpolation to smooth the demarcation point positions, ensuring a more consistent spatial distribution of demarcation points that conforms to the actual boundary characteristics of the geological strata. Subsequently, a three-dimensional model M of the geological strata at the tunnel face was constructed using demarcation point set P2. This spatial interpolation fit from discrete demarcation points to a continuous demarcation surface was achieved, resulting in a complete geological stratum boundary model. Based on the comparison of several boundary values Ms(x, y, z) in the three-dimensional model M with the boundary stability threshold Tthe, and combined with the tolerance range RC, a geological stability state identifier S is generated. Based on the generated geological stability state identifier S, the support scheme further distinguishes between stable and unstable areas and proposes additional support measures for unstable areas. This provides a more comprehensive basis for geological layer boundaries and support scheme optimization for tunnel construction, significantly improving the accuracy of geological layer boundary modeling at the tunnel face and the safety of construction support.

[0108] Example 5

[0109] A tunnel face monitoring system, please refer to Figure 2,Specifically: including signal acquisition module, signal feature extraction module, ,preliminary recognition module, correction module and modeling evaluation module;

[0110] The signal acquisition module collects the reflected signal of the rock formation at the tunnel face through the geological radar equipment to form the original data set D0, and generates the noise-reduced data set D1 through wavelet transform denoising.

[0111] The signal feature extraction module extracts radar signal features from the noise reduction data set D1, obtains rock material difference features, signal attenuation features and main frequency offset features, and forms a feature vector V;

[0112] The preliminary identification module performs preliminary identification of the address layer boundary by using the support vector machine classification algorithm on the feature vector V to obtain a preliminary demarcation point set P1;

[0113] The correction module corrects the obtained preliminary demarcation point set P1 to obtain a corrected demarcation point set P2;

[0114] The modeling and evaluation module establishes a three-dimensional model of the geological layer boundary by using the boundary point set P2, obtains the three-dimensional model M of the geological layer of the tunnel face, compares it with the boundary stability state threshold Tthe, generates a geological stability state mark S, and generates a construction support plan.

[0115] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A tunnel face monitoring method, characterized by: The following steps are involved: S1, using geological radar equipment to collect the reflected signal of the rock formation at the tunnel face to form the original data set D0, and then using wavelet transform to denoise the data set D1; S2. Extract radar signal features from the noise reduction dataset D1 to obtain rock material difference features, signal attenuation features, and main frequency offset features to form a feature vector V; S3, using the support vector machine classification algorithm to perform preliminary identification of the address layer boundary on the feature vector V, and obtaining a preliminary demarcation point set P1; Said S3 includes S31 and S32; S31, performing data normalization processing on the feature vector V to obtain a normalized standard feature vector Vn, and then using a support vector machine classification algorithm to preliminarily identify the address layer boundary, specifically finding a hyperplane through the support vector machine classification algorithm; The hyperplane is obtained by the following target classification function: ； Where f(F) represents the output of the target classification function, specifically the category of the Fth standard feature vector in the standard feature vector Vn. The category is judged specifically based on the output f(F). sign represents the sign function, m represents the number of support vectors, Sj represents the jth support vector, wj represents the weight value of the jth support vector, K(F, Sj) represents the kernel function, b represents the bias term of the hyperplane, which is specifically used to adjust the position of the hyperplane, and m represents the number of support vectors. S32, classifying the standard feature vector Vn according to the found hyperplane, and simultaneously separating the boundary points of different categories of geological layer data to obtain a preliminary boundary point set P1; The preliminary demarcation point set P1 is specifically formed by inferring the input standard feature vector Vn using the target classification function, and statistically analyzing the classification results of each demarcation point according to the output f(F) of the target classification function, and then marking the position of the boundary point to form the preliminary demarcation point set P1; The preliminary demarcation point set P1 is obtained by the calculation formula P1={z|f(F)=0}, which specifically represents a collection of feature vector points that satisfy the target classification function output of 0, representing the boundary points of the geological layer; S4, correcting the obtained preliminary demarcation point set P1 to obtain a corrected demarcation point set P2; Said S4 includes S41; S41, revising the obtained preliminary demarcation point set P1, including analyzing the distribution of neighboring points of the demarcation point position in the preliminary demarcation point set P1, including determining that the current demarcation point needs to be revised if the position of the neighboring point deviates by 10% from the average position of the neighboring points, and re-substituting the revised demarcation point into the preliminary demarcation point set P1 to form a revised demarcation point set P2; The correction includes using a neighboring point interpolation method to correct and smooth the position of the dividing point; S5. Use the demarcation point set P2 to establish a three-dimensional model of the geological layer boundary, obtain a three-dimensional model M of the geological layer of the tunnel face, and compare it with the boundary stability state threshold Tthe to generate a geological stability state identifier S, and generate a construction support plan; Said S5 includes S51 and S52; S51, establishing a three-dimensional model of geological layer boundaries by using the boundary point set P2 to obtain a three-dimensional model M of the geological layer of the tunnel face, and generating a continuous surface of the geological layer boundary by performing spatial interpolation fitting on the discrete boundary points in the boundary point set P2, including fitting the boundary values Ms (x, y, z) in the obtained three-dimensional interface models; The demarcation value Ms (x, y, z) is obtained by the following calculation formula: ； Where Ms(x, y, z) represents the cutoff value in the three-dimensional interface model, Wm represents the interpolation weight, which is related to the spatial distribution of the cutoff point, Qm(x, y, z) represents the interpolation basis function of the mth cutoff point, and M represents the total number of cutoff points, specifically the total number of cutoff points in the cutoff point set P2; S52. Compare several boundary values Ms (x, y, z) in the three-dimensional model M of the tunnel face geological layer with the boundary stability threshold value Tthe, and associate them with the tolerance range RC to generate a geological stability status identifier S and a construction support plan. The geological stability status identifier S is obtained by the following comparison method: ； When S=1, obtain the stable result of the demarcation value Ms (x, y, z), generate the construction support plan and maintain the original plan for support; When S=0, the unstable result of the demarcation value Ms (x, y, z) is obtained, and a construction support plan is generated. While maintaining the original support plan, additional support measures related to anchor rods, concrete and steel mesh are planned for the distribution position of the unstable demarcation value Ms (x, y, z).

2. The tunnel face monitoring method according to claim 1, characterized in that: Said S1 includes S11 and S12; S11. Use geological radar equipment to scan the tunnel face, obtain radar reflection signals, and record them as the original data set D0. Specifically, the geological radar transmits high-frequency electromagnetic waves into the rock layer. Different rock layers have different reflection intensities and propagation times for electromagnetic waves. The returned radar reflection signals contain the dielectric properties of the rock layer and the boundary information between geological layers, including the echo time T, the reflection signal intensity I, and the signal main frequency Fp. The radar echo signal S = {T, I, Fp} of the scanning line i is composed of The original data set D0 is specifically D0={Si(t), i=1, 2, 3, ..., n}, where Si(t) represents the radar echo signal of the i-th scan line at time t, and n represents the total number of scan lines.

3. The tunnel face monitoring method according to claim 2, characterized in that: S12. Perform wavelet transform denoising processing on the original data set D0 to separate different frequency bands of the signal, remove the noise signal and retain the high-frequency components of the geological information to form the denoised data set D1.

4. The tunnel face monitoring method according to claim 3, characterized in that: Said S2 includes S21 and S22; S21. Extract radar signal features from the noise reduction data set D1 to obtain the rock material difference feature △Er, the signal attenuation feature A, and the main frequency offset feature △Fp, and form a feature vector V = {△Er, A, △Fp}; Among them, the difference characteristic △Er of the rock material is obtained by extracting the dielectric constant change rate to reflect the response characteristics of the rock material to electromagnetic waves. Specifically, by comparing the echo time T1 and echo time T2 of different geological layers, the dielectric constant change rate between the two layers is calculated to obtain the difference characteristic △Er of the rock material; The signal attenuation feature A is obtained by extracting the attenuation of the signal intensity with the propagation distance to reflect the changes in the density and water content of the geological layer. Specifically, the signal attenuation state of the reflected signal intensity I1 and the reflected signal intensity I2 at different geological interfaces is compared to obtain the signal attenuation feature A; The main frequency offset feature △Fp is obtained by analyzing the signal spectrum and extracting the main frequency change characteristics of different geological layer interfaces. Specifically, the main frequency offset value is obtained by calculating the main frequency peak Fp1 and the main frequency peak Fp2 of different geological interfaces to obtain the main frequency offset feature △Fp.

5. A tunnel face monitoring system, applied to a tunnel face monitoring method according to any one of claims 1 to 4, characterized in that: It includes signal acquisition module, signal feature extraction module, preliminary recognition module, correction module and modeling evaluation module; The signal acquisition module collects the reflected signal of the rock formation at the tunnel face through the geological radar equipment to form the original data set D0, and generates the noise-reduced data set D1 through wavelet transform denoising. The signal feature extraction module extracts radar signal features from the noise reduction data set D1 to obtain rock material difference features, signal attenuation features, and main frequency offset features. Form the eigenvector V; The preliminary identification module performs preliminary identification of the address layer boundary by using the support vector machine classification algorithm on the feature vector V to obtain a preliminary demarcation point set P1; The correction module corrects the obtained preliminary demarcation point set P1 to obtain a corrected demarcation point set P2; The modeling and evaluation module establishes a three-dimensional model of the geological layer boundary by using the boundary point set P2, obtains the three-dimensional model M of the geological layer of the tunnel face, compares it with the boundary stability state threshold Tthe, generates a geological stability state mark S, and generates a construction support plan.

Images