An intelligent rapid classification method and system for engineering surrounding rock

Abstract

The present application relates to the field of intelligent rapid classification of engineering surrounding rock, and particularly relates to an intelligent rapid classification method and system for engineering surrounding rock. The method comprises the following steps: obtaining three-dimensional characteristic parameters of a surrounding rock structure surface according to stereophotogrammetry; obtaining comprehensive quality indexes of the surrounding rock according to dynamic response testing; obtaining rock mass uniaxial compressive strength and rock mass integrity coefficient based on the three-dimensional characteristic parameters and the comprehensive quality indexes; constructing an evolutionary support vector regression model using a support vector regression algorithm and an evolutionary algorithm; constructing an intelligent classification model based on the evolutionary support vector regression model; and realizing intelligent rapid classification of engineering surrounding rock through the intelligent classification model. The present application establishes a functional relationship model between surrounding rock grades and nondestructive testing indexes, simplifies the field data collection process, reduces the dependence on the experience of operators, realizes rapid and accurate detection of surrounding rock grades, and greatly reduces construction costs and shortens the project cycle.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent rapid classification of engineering surrounding rock, and particularly relates to an intelligent rapid classification method and system for engineering surrounding rock. BACKGROUND

[0002] In engineering geological exploration and underground engineering design, surrounding rock classification is an important basis for evaluating the stability of underground structures and guiding the development of construction plans. The traditional surrounding rock classification method mainly relies on the experience of on-site engineers and geological exploration data, which not only takes a long time, but is also easily affected by the personal experience level of engineers, resulting in a large subjective and uncertain classification result. In recent years, with the rapid development of stereophotogrammetry technology and dynamic response testing technology, new technical approaches have been provided for the intelligentization and rapidization of surrounding rock classification.

[0003] In the prior art, surrounding rock classification mainly relies on manual observation and engineer experience, which has the following disadvantages: first, it is highly subjective, lacks unified standards and objective quantitative indicators, and the classification result is affected by the personal experience of engineers, resulting in uncertainty and inconsistency, affecting accuracy and construction safety management; second, data collection and processing are complicated, requiring on-site investigation, sampling and laboratory testing, which is time-consuming and labor-intensive, increasing construction costs, prolonging the work cycle, and affecting engineering progress and efficiency; third, the technical means is single, focusing on the application of a single technical means, such as geological exploration and mechanical testing, ignoring the comprehensive application of multiple technical means, resulting in incomplete and inaccurate surrounding rock quality evaluation, which is difficult to meet the high requirements of modern underground engineering.

[0004] At present, there is not enough research on intelligent rapid classification of engineering surrounding rock, and there is no specific intelligent rapid classification method for engineering surrounding rock based on stereophotogrammetry and dynamic response testing. SUMMARY

[0005] In view of the defects in the prior art, the present application provides an intelligent rapid classification method and system for engineering surrounding rock.

[0006] In a first aspect, the present application provides an intelligent rapid classification method for engineering surrounding rock, comprising the following steps: obtaining three-dimensional characteristic parameters of a surrounding rock structure surface according to stereophotogrammetry; obtaining a comprehensive quality index of the surrounding rock according to dynamic response testing; obtaining rock mass uniaxial compressive strength and rock mass integrity coefficient based on the three-dimensional characteristic parameters and the comprehensive quality index; constructing an evolutionary support vector regression model using a support vector regression algorithm and an evolutionary algorithm based on the rock mass uniaxial compressive strength and the rock mass integrity coefficient; constructing an intelligent classification model for outputting a surrounding rock grade based on the evolutionary support vector regression model; and realizing intelligent rapid classification of engineering surrounding rock through the intelligent classification model. The present application efficiently and accurately obtains three-dimensional characteristic parameters of a surrounding rock structure surface through stereophotogrammetry, significantly improving the accuracy and efficiency of data acquisition; comprehensively evaluates the quality of the surrounding rock through dynamic response testing, enriching the dimensions of the surrounding rock quality evaluation, and making the evaluation results more comprehensive and reliable; provides key parameters for constructing an intelligent classification model by obtaining rock mass uniaxial compressive strength and rock mass integrity coefficient, and enhances the prediction accuracy and applicability of the model; and realizes rapid and intelligent classification of engineering surrounding rock quality by constructing an intelligent classification model, greatly improving the classification efficiency and accuracy, and providing strong technical support for roadway surrounding rock engineering design and construction.

[0007] Optionally, the obtaining three-dimensional characteristic parameters of a surrounding rock structure surface according to stereophotogrammetry comprises: obtaining real-world coordinates of the surrounding rock using a total station; obtaining rock mass images using a binocular vision camera based on the real-world coordinates; constructing a sparse point cloud model based on a motion recovery structure algorithm using a three-dimensional real scene modeling software based on the rock mass images; constructing a high-density point cloud model using a three-dimensional multi-view stereo vision algorithm according to the sparse point cloud model; constructing a three-dimensional digital surface model using a triangular mesh editing function of the three-dimensional real scene modeling software according to the high-density point cloud model; classifying structure surface point cloud data and obtaining a classification result using a clustering algorithm based on the three-dimensional digital surface model; and calculating structure surface spacing and rock mass volumetric joint number based on the classification result. The present application significantly improves the accuracy and efficiency of data acquisition by obtaining real-world coordinates of the surrounding rock using a total station and capturing rock mass images using a binocular vision camera, ensuring the reliability of the structure surface characteristic parameters; realizes the generation of point cloud data from sparse to dense and from rough to fine by constructing a sparse point cloud model based on a motion recovery structure algorithm and further generating a high-density point cloud model using a three-dimensional multi-view stereo vision algorithm, laying a solid foundation for constructing an accurate three-dimensional digital surface model; and classifies structure surface point cloud data using a triangular mesh editing function of a three-dimensional real scene modeling software and a clustering algorithm, and accurately calculates structure surface spacing and rock mass volumetric joint number, providing key evidence for evaluating the stability of the surrounding rock and developing engineering measures.

[0008] Optionally, the method of obtaining a comprehensive quality index of the surrounding rock based on the dynamic response test includes: obtaining a comprehensive stress wave velocity index of the surrounding rock based on the dynamic response test, and the comprehensive stress wave velocity index is used to characterize the quality of the surrounding rock; obtaining a comprehensive rebound strength index of the surrounding rock based on the dynamic response test, and the comprehensive rebound strength index is used to characterize the strength of the surrounding rock. The present invention obtains the comprehensive stress wave velocity index of the surrounding rock through the dynamic response test, which intuitively and quickly reflects the overall quality and internal defects of the surrounding rock, provides a scientific basis for evaluating the stability of the surrounding rock, and significantly improves the accuracy and efficiency of the surrounding rock quality assessment; by introducing the comprehensive rebound strength index, the strength characteristics of the surrounding rock are effectively characterized, which provides an important reference for judging the bearing capacity and deformation characteristics of the surrounding rock, further enriches the dimensions of the surrounding rock quality assessment, and enhances the reliability and practicality of the assessment results.

[0009] Optionally, the comprehensive stress wave velocity index satisfies the following relationship:

[0010] ,

[0011] in, is the comprehensive stress wave velocity index, is the distance between the center points of the two sensors, is the propagation time difference; the comprehensive rebound strength index satisfies the following relationship:

[0012] ,

[0013] in, is a comprehensive rebound strength index. For the lithologic measurement area Lithologic measurement area Rebound strength at each measuring point, For the Lithologic measurement area Rebound strength at each measuring point, For the The area of ​​lithologic measurement area, For the The present invention provides a theoretical basis for accurately measuring and analyzing the stress wave propagation characteristics of surrounding rocks by correlating wave velocity, sensor spacing, and propagation time difference, which helps to accurately assess the overall quality and internal structure of the surrounding rocks. By combining the rebound strength with the area of ​​different lithologic measurement areas and comprehensively considering the impact of lithologic differences on rebound strength, the rebound strength assessment is made more comprehensive and accurate, providing a scientific basis for judging the strength characteristics of the surrounding rocks.

[0014] Optionally, the obtaining of the uniaxial compressive strength of the rock mass and the rock mass integrity coefficient based on the three-dimensional characteristic parameters and the comprehensive quality index comprises: obtaining the uniaxial compressive strength of the rock mass based on the comprehensive rebound strength index; and obtaining the rock mass integrity coefficient based on the comprehensive stress wave velocity index and the three-dimensional characteristic parameters. The uniaxial compressive strength of the rock mass is obtained, thereby providing strong support for the engineering design and construction of the surrounding rock of the roadway; the rock mass integrity coefficient is obtained, the internal structure and stress wave propagation characteristics of the surrounding rock are comprehensively considered, and the evaluation of the integrity coefficient is more accurate and scientific.

[0015] Optionally, the construction of the evolutionary support vector regression model by using the support vector regression algorithm and the evolutionary algorithm based on the uniaxial compressive strength of the rock mass and the rock mass integrity coefficient comprises: constructing a support vector regression model by using the support vector regression algorithm based on the uniaxial compressive strength of the rock mass and the rock mass integrity coefficient; and constructing an evolutionary support vector regression model by using a particle swarm optimization algorithm according to the support vector regression model. The support vector regression model is constructed, the complex relationship between the rock mass mechanical properties and parameters is accurately captured, the prediction accuracy is improved, and a scientific basis is provided for the rock mass classification and stability evaluation; the particle swarm optimization algorithm is introduced to optimize the support vector regression model, the model parameter selection problem is effectively solved, the generalization ability and robustness of the model are improved, the model can perform well under different geological conditions; the evolutionary support vector regression model not only improves the accuracy of the rock mass classification and stability evaluation, but also provides a more reliable prediction tool for engineering practice, which helps to optimize the engineering scheme, reduce the safety risk, and has significant innovation and practical value.

[0016] Optionally, the construction of the evolutionary support vector regression model by using the particle swarm optimization algorithm according to the support vector regression model comprises: selecting a radial function kernel as a kernel function according to the support vector regression model; setting a hyperparameter based on the kernel function, the hyperparameter comprising a penalty parameter and a width parameter of the radial function kernel; searching for an optimal value of the hyperparameter by using the particle swarm optimization algorithm to obtain an optimal hyperparameter; and constructing the evolutionary support vector regression model by using the optimal hyperparameter. The radial function kernel is selected as the kernel function, the nonlinear relationship between data is flexibly captured, the prediction accuracy of the model is improved, and a more accurate tool is provided for the rock mass classification and stability evaluation under complex geological conditions; the hyperparameter is set, and the optimal value is searched by using the particle swarm optimization algorithm, thereby effectively solving the hyperparameter selection problem, avoiding the complexity and subjectivity of parameter debugging in the traditional method, and improving the automation degree and efficiency of the model; the evolutionary support vector regression model constructed by the optimal hyperparameter not only improves the prediction performance, but also enhances the generalization ability of the model, so that the model can perform well under different geological environments and engineering conditions, and has significant innovation and practical value.

[0017] Optionally, the constructing the intelligent grading model for outputting the surrounding rock grade based on the evolutionary support vector regression model comprises: integrating the obtained three-dimensional characteristic parameters and the comprehensive quality index into first data samples; obtaining second data samples of the basic quality sub-grade grading surrounding rock grade corresponding to the three-dimensional characteristic parameters and the comprehensive quality index; integrating the first data samples and the second data samples into comprehensive data samples; training the evolutionary support vector regression model by using the comprehensive data samples to obtain a training result; establishing a function relationship between the three-dimensional characteristic parameters, the comprehensive quality index and the surrounding rock grade according to the training result; and constructing the intelligent grading model for outputting the surrounding rock grade through the function relationship. The three-dimensional characteristic parameters and the comprehensive quality index are integrated into data samples, various information sources are fully utilized, the comprehensiveness and accuracy of the surrounding rock quality evaluation are improved, and a solid foundation is laid for the construction of the intelligent grading model. The evolutionary support vector regression model is trained by using the comprehensive data samples, the prediction ability of the model is enhanced, and the function relationship between the three-dimensional characteristic parameters, the comprehensive quality index and the surrounding rock grade is established. The surrounding rock grade is automatically output through the intelligent grading model, the evaluation efficiency and accuracy are greatly improved, and a more scientific and convenient surrounding rock quality evaluation tool is provided for engineering practice.

[0018] Optionally, the obtaining of the second data samples of the basic quality sub-grade grading surrounding rock grade corresponding to the three-dimensional characteristic parameters and the comprehensive quality index comprises: substituting the rock mass uniaxial compressive strength and the rock mass integrity coefficient into a surrounding rock basic quality grading formula to obtain a grading index; judging the surrounding rock grade according to the grading index; and integrating the grading index and the surrounding rock grade into the second data samples. The rock mass uniaxial compressive strength and the rock mass integrity coefficient are substituted into the surrounding rock basic quality grading formula, the grading index is accurately calculated, a scientific basis is provided for the division of the surrounding rock grade, and the subjectivity and uncertainty of the grading standard in the traditional method are effectively avoided. The grading index is used to judge the surrounding rock grade, the quantitative evaluation of the surrounding rock quality is realized, and the accuracy and objectivity of the evaluation are improved. The grading index and the surrounding rock grade are integrated into the second data samples, rich data support is provided for the training of the intelligent grading model, the prediction ability and generalization performance of the model are improved, and a solid foundation is laid for the intelligent evaluation of the surrounding rock quality grade.

[0019] In a second aspect, the present invention provides an intelligent and rapid grading system for engineering surrounding rocks, comprising an input device, a processor, an output device, and a memory, wherein the input device, the processor, the output device, and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, the processor is configured to call the program instructions, and the system uses the intelligent and rapid grading method for engineering surrounding rocks. The system integrates binocular vision cameras and total station equipment to achieve rapid and accurate collection of field data; by constructing a high-density point cloud model based on motion recovery structure algorithm and three-dimensional multi-view stereo vision algorithm, it greatly improves the efficiency of data collection and processing, and provides a solid foundation for surrounding rock classification; by adopting a clustering algorithm to classify the structural surface point cloud, and accurately calculate the structural surface spacing and the number of rock volume joints, combined with non-destructive testing indicators such as rebound strength and stress wave velocity, it realizes intelligent and objective judgment of surrounding rock grade, effectively reducing the dependence on the operator's work experience; the application of the system significantly simplifies the surrounding rock classification process, shortens the work cycle, reduces construction costs, and at the same time improves the accuracy and reliability of the classification results, providing a strong guarantee for the safe construction of the project, and demonstrating its great potential and application value in engineering practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a flow chart of a method for intelligent and rapid classification of surrounding rock in an engineering project according to an embodiment of the present invention;

[0021] Figure 2 A flowchart of three-dimensional digital model reconstruction according to an embodiment of the present invention;

[0022] Figure 3 A diagram of a method for arranging measuring points of a rebound hammer according to an embodiment of the present invention;

[0023] Figure 4 This is a diagram of an echo detector device according to an embodiment of the present invention;

[0024] Figure 5 This is a flow chart of constructing an optimal support vector regression model using a particle swarm optimization algorithm according to an embodiment of the present invention;

[0025] Figure 6 This is a schematic structural diagram of an intelligent rapid grading system for engineering surrounding rock according to an embodiment of the present invention;

[0026] Figure 7 This is an operation flow chart of the intelligent rapid grading system for engineering surrounding rock according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] Specific embodiments of the present application will now be described in detail with reference to the drawings, which are provided by way of example and not limitation. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one of ordinary skill in the art that the present application can be practiced without the specific details. In other instances, well-known circuits, software or methods have not been described in detail in order to avoid obscuring the present application. It will be appreciated that the various units shown in the figures can be combined in a system-on-a-chip (SoC) along with other units or they can be provided separately.

[0028] Reference throughout this specification to "an embodiment", "embodiments", "one example", or "examples" means that a particular feature, structure, or characteristic described in connection with the embodiment or example is included in at least one embodiment of the present application. Therefore, appearances of the phrases "in one embodiment", "in embodiments", "one example" or "an example" in various places throughout this specification are not necessarily referring to the same embodiment or example. Furthermore, the particular features, structures, or characteristics can be combined in any suitable

[0029] Reference throughout this specification to "an embodiment", "embodiments", "one example", or "examples" means that a particular feature, structure, or characteristic described in connection with the embodiment or example is included in at least one embodiment of the present application. Therefore, appearances of the phrases "in one embodiment", "in embodiments", "one example" or "an example" in various places throughout this specification are not necessarily referring to the same embodiment or example. Furthermore, the particular features, structures, or characteristics can be combined in any suitable Figure 1 The embodiments of the present application provide an intelligent rapid classification method for surrounding rock, which comprises the following steps:

[0030] S1. Obtain the three-dimensional characteristic parameters of the surrounding rock structure surface according to stereophotogrammetry.

[0031] In one embodiment, the real world coordinates of the surrounding rock are first obtained by using a total station.

[0032] Specifically, a certain number of control points are arranged on the surface of the surrounding rock measurement area; the equipment is checked and the parameters are set, and then the instrument is calibrated; after selecting a suitable measurement method, the total station is installed and aimed at the target point, and the angle, height and slant distance data are recorded; the data is sorted and error analysis is performed to ensure accuracy; and reliable real world coordinates of the surrounding rock are obtained.

[0033] Further, binocular vision cameras are used to shoot images of the roadway face and side rock.

[0034] Specifically, the two cameras are first calibrated to obtain internal and external parameters and homography matrix, and distortion correction is performed; the cameras are fixed, and the same rock region is synchronously shot to obtain two images; image matching is performed according to the polar constraint to establish the corresponding relationship between the feature points; the three-dimensional geometric information of the rock mass is restored by calculating the positional deviation between the corresponding points of the images, combined with the camera parameters, so as to obtain detailed rock mass images.

[0035] Further, based on a three-dimensional real scene modeling software platform, a sparse point cloud model based on a structure from motion (SFM) algorithm is constructed.

[0036] Specifically, the rock mass image is imported into the three-dimensional real scene modeling software, feature point extraction and matching are performed, and the camera pose is estimated; a preliminary sparse point cloud is obtained through triangulation, and the camera parameters and point cloud coordinates are optimized using bundle adjustment (BA) to improve the accuracy; a sparse point cloud model is generated, providing a basis for subsequent three-dimensional reconstruction.

[0037] Further, according to the sparse point cloud model, a three-dimensional multi-view stereo vision (PMVS) algorithm is used to construct a high-density point cloud model.

[0038] Specifically, according to the sparse point cloud model, the sparse point cloud is expanded to generate more candidate points; matching features are searched in multi-view images, and the depth value of each candidate point is calculated through a depth map fusion strategy; the depth map is filtered and optimized to obtain more accurate point cloud data; the optimized point cloud data is integrated to construct a high-density and high-precision three-dimensional point cloud model.

[0039] It should be noted that the PMVS algorithm includes the following detailed steps:

[0040] For each feature point f on the image , an array containing all feature points of the same type on the same epipolar line in other images is associated with it;

[0041] A feature point f' is selected from the array to form a pair of homonymous image points (f, f'), and the three-dimensional coordinates of the object point are calculated through the forward intersection in triangulation;

[0042] The three basic attributes of the patch are initialized: the object point coordinates c(p) as the center point of the patch, the image as the reference image, and the unit normal vector n(p) is constructed according to the direction of the line connecting the object point coordinates and the photographic center;

[0043] The visible set V(p) is determined by the angle between the line connecting the object point and other photographic centers and its unit normal vector, and the updated visible set V*(p) is further determined within the range of plus or minus 60 degrees based on the visible set; the threshold value a is taken as 0.6, i.e., the imaging difference coefficient h(p, I, R(p)) between the reference image should be less than 0.6;

[0044] The goal is to minimize the total imaging difference function g*(p) of the patch; the position and pose of the patch are corrected through the conjugate gradient method;

[0045] After the pose of the patch is changed, the update visible set V*(p) is also changed, at this time, α takes 0.3, if the number of images in the update visible set V*(p) is greater than the threshold γ, it means that the patch appears on at least γ low imaging difference images, the reconstruction of the patch is successful, the patch is stored in the corresponding grid unit, and the related data structure is updated;

[0046] For a new patch p', initialize its c(p'), n(p') and V(p') attributes, n(p') and V(p') can use the corresponding attributes of the patch p as the initial value, and c(p') takes the intersection point of the line connecting the photographic center of the reference image and the center of the grid unit and the plane where the patch p is located, V*(p') is obtained from V(p');

[0047] After the update visible set V*(p') is known, the total imaging difference function g*(p') of the new patch is determined, and the position and pose of the patch are corrected by the conjugate gradient method to minimize g*(p');

[0048] Since the patch p' is different from p, new images can be added to V(p') through the disparity map test, and V*(p') is obtained from V(p') again;

[0049] If the number of images in the update visible set V*(p') is greater than the threshold γ, it means that the new patch is reconstructed successfully, the patch is stored in the corresponding grid unit, and the related data structure is updated.

[0050] Further, according to the high-density point cloud model, a three-dimensional digital surface model is constructed by using a triangular mesh editing function of the three-dimensional real scene modeling software. The reconstruction process of the three-dimensional digital surface model is as shown in Figure 2 .

[0051] Specifically, first, the point cloud data is imported into the three-dimensional real scene modeling software, and a suitable mesh resolution is set; a triangular subdivision algorithm is used to generate an initial triangular mesh, and a smoothing algorithm is used to optimize the mesh to reduce noise and unevenness; the mesh defects are repaired manually or automatically to ensure the integrity of the model; a fine three-dimensional digital surface model is generated to provide a basis for subsequent analysis and application.

[0052] Further, based on the three-dimensional digital surface model, a clustering algorithm based on density and noise (DBSCAN) is used to classify the structural surface point cloud data and obtain a classification result.

[0053] Specifically, first, the point set to be clustered is loaded; the number of points in the neighborhood of each point is calculated, which is described by the following mathematical expression:

[0054] ,

[0055] wherein, Representative Points The set of points in the neighborhood of represents the set of all points, Indicates a point and The distance between is the set distance threshold of the neighboring points of the detection object; the core points are identified based on the calculated number of points in the neighborhood of each point. A point is considered a core point if the number of points in its neighborhood is greater than or equal to a preset minimum number of points; the identified core points are used to refine the structural surface into multiple clusters, that is, starting from a core point, continuously looking for other points that can be reached by its density to form a set of density-connected points, that is, a cluster; then find the next core point that has not been classified into any cluster, and repeat the above process until all core points are classified into a cluster or all points are processed.

[0056] Furthermore, based on the classification results, the structural plane spacing and the number of rock volume joints are calculated.

[0057] Specifically, the normal vector and horizontal direction vector of the structural surface are obtained through principal component analysis, and the classification clusters are projected onto the plane along the horizontal projection direction. The linear regression equation of the scattered points is calculated and fitted into line segments on the projection plane. The spacing of the fitted line segments is calculated on the plane, where the normal distance between each group of structural surfaces is calculated along the normal direction of the structural surface (i.e., the survey line). In addition, the calculation formula for the number of rock volume joints is as follows:

[0058]

[0059] in, is the number of structural planes per cubic meter of rock mass volume, For the The number of structural surfaces per meter along the normal direction of the group structure, It is the number of non-grouped joints per cubic meter of rock mass.

[0060] S2. Obtain comprehensive quality indicators of surrounding rock based on dynamic response testing.

[0061] Among them, S2 includes the following steps:

[0062] S21. Based on the dynamic response test, obtain the comprehensive stress wave velocity index of the surrounding rock.

[0063] In one embodiment, based on the dynamic response test, a comprehensive stress wave velocity index for characterizing the surrounding rock quality is obtained; the comprehensive stress wave velocity index satisfies the following relationship:

[0064]

[0065] wherein, is the comprehensive stress wave velocity index of the surrounding rock, accurate to 1 ms, is the distance (mm) between the centers of the two sensors, generally 300 mm, is the propagation time difference (μs).

[0066] Further, the comprehensive stress wave velocity index of the surrounding rock is calculated by the above expression.

[0067] S22. Based on the dynamic response test, the comprehensive rebound strength index of the surrounding rock is obtained.

[0068] In one embodiment, based on the dynamic response test, the comprehensive rebound strength index for characterizing the strength of the surrounding rock is obtained.

[0069] Specifically, after each cycle of the tunneling construction, the rock debris and floating rock on the newly exposed surrounding rock are cleaned, the rebound measuring points are arranged, and the changes in the lithology of the surrounding rock are observed. Generally, no less than 5 measuring points should be arranged on the surrounding rock of the same lithology, and the distance between adjacent measuring points is within the range of 0.2 m to 2 m.

[0070] Further, the rebound response test of the surrounding rock of the site tunnel is carried out by using the rebound instrument, and the measuring point arrangement is as shown in Figure 3 Each measuring point is measured 10 times, the highest value and the two lowest values are removed, and the arithmetic mean value is taken as the rebound strength of the measuring point.

[0071] Further, based on the rebound strength of the measuring point, the comprehensive rebound strength index of the surrounding rock on the surface of the tunnel of a single cycle of the tunneling construction is obtained, and the comprehensive rebound strength index satisfies the following relationship:

[0072]

[0073] wherein, is the comprehensive rebound strength index, is the rebound strength of the th measuring point in the th lithology measuring area, is the rebound strength of the th measuring point in the th lithology measuring area, is the area of the th lithology measuring area, is the area of the th lithology measuring area.

[0074] S3. Based on the three-dimensional characteristic parameters and the comprehensive quality index, the uniaxial compressive strength of the rock mass and the rock mass integrity coefficient are obtained.

[0075] wherein, S3 further includes the following steps:

[0076] S31. Obtain the uniaxial compressive strength of the rock mass.

[0077] In one embodiment, the uniaxial compressive strength of the rock mass is inferred from the measured rebound value, and the formula is as follows:

[0078]

[0079] wherein, is the uniaxial compressive strength of the rock mass, , is a weight coefficient, is the measured rebound value.

[0080] S32. Obtain the integrity coefficient of the rock mass.

[0081] In one embodiment, on the basis of the rebound measuring points arranged in step S22, the measuring points are arranged again, the stress wave velocity of the newly exposed surrounding rock is detected by using a wireless echo detector, and the stress wave velocity is recorded. The related equipment of the echo detector is shown in Figure 4 .

[0082] Further, the comprehensive data obtained by the wave velocity and the point cloud model are used to characterize the integrity coefficient of the rock mass .

[0083] S4. Based on the uniaxial compressive strength of the rock mass and the integrity coefficient of the rock mass, an evolutionary support vector regression model is constructed.

[0084] In one embodiment, first, based on the uniaxial compressive strength of the rock mass and the integrity coefficient of the rock mass, a target function is constructed by using a support vector regression (SVR) algorithm, and the target function is as follows:

[0085]

[0086] wherein, is a weight vector, is a bias vector, is the number of samples, is a penalty parameter, is a sample serial number, and is a slack variable used to handle the non-separable case. The target function has the minimum model complexity under the condition of meeting the existing insensitive loss function. The target function is also called an SVR model.

[0087] Further, based on the target function, a radial function (RBF) kernel is selected as a kernel function.

[0088] Further, based on the kernel function, a hyperparameter is set, and the hyperparameter includes a penalty parameter and a width parameter of the radial function kernel.

[0089] Furthermore, the particle swarm optimization (PSO) algorithm is used to search for the optimal value of the hyperparameters and obtain the optimal hyperparameters. The PSO algorithm simulates the behavior of a flock of birds foraging and iteratively updates the speed and position of each particle to search for the optimal solution. Each particle represents a potential solution, and its speed and position represent the search direction and the current solution respectively. The algorithm tracks the individual optimal solution of each particle ( pbest ) and the group optimal solution ( gbest ) to guide the particles to move towards the optimal solution.

[0090] Specifically, the group position and velocity are initialized first, and the initial position of the particles is randomly generated in the solution space, and then the initial velocity is generated. The velocity update formula is as follows:

[0091]

[0092] The position update formula is as follows:

[0093]

[0094] in, Indicates the Particles in The speed of time, Indicates the Particles in The location at the moment, Indicates the Particles in The speed of time, Indicates the Particles in Position at the moment; The inertia weight controls the degree to which the particle inherits its previous velocity; and is the acceleration factor, which controls the particle pbest and gbest degree of mobility; and is a random number between [0, 1].

[0095] Furthermore, the PSO algorithm is applied to SVR parameter optimization. The specific process is as follows: Figure 5 shown.

[0096] Specifically, the particle swarm is randomly initialized, each particle representing a set of SVR parameters, the prediction accuracy of the SVR parameters corresponding to each particle is evaluated by using a method such as cross-validation, for example, root mean square error (RMSE) or mean absolute error (MAE), and the prediction accuracy is taken as the fitness value of the particle; the speed and position of each particle are updated according to the updating formula of the PSO algorithm, so as to search for a new set of SVR parameters; the above steps are repeated until a termination condition is met, a maximum number of iterations is reached, or a preset accuracy requirement is met; the SVR parameters corresponding to the particle with the highest fitness value are selected as the optimal parameters, and the optimal parameters are used to construct a final SVR model, that is, an evolutionary support vector regression model.

[0097] S5. Constructing an intelligent classification model based on the evolutionary support vector regression algorithm according to the evolutionary support vector regression model.

[0098] In one embodiment, first, based on step S1, data samples of nondestructive detection indexes such as three-dimensional characteristic parameters of surrounding rock structural planes, comprehensive rebound strength indexes of surrounding rock, and comprehensive stress wave velocity indexes of surrounding rock are obtained, which are named as first data samples.

[0099] Further, the surrounding rock grades after the basic quality (BQ) sub-classification are obtained, and data samples are formed, which are named as second data samples.

[0100] Specifically, the rock mass uniaxial compressive strength and the rock mass integrity coefficient are substituted into the surrounding rock basic quality classification formula to obtain a classification index; the surrounding rock basic quality formula is as follows:

[0101]

[0102] wherein, is a rock mass basic quality index, is a rock mass saturated uniaxial compressive strength (unit: MPa), is a rock mass integrity coefficient, dimensionless;

[0103] Further, the surrounding rock grade is determined according to the classification index; specifically as follows:

[0104] When , it is an excellent grade surrounding rock, with extremely high stability and almost no need for support;

[0105] When , it is a good grade surrounding rock, with relatively high stability and a small amount of support required;

[0106] When , it is a medium grade surrounding rock, with general stability and appropriate support required;

[0107] When When the value of the index is between 0 and 0.2, the surrounding rock is of poor grade, and the stability is poor, and more support is needed.

[0108] When the value of the index is between 0 and 0.2, the surrounding rock is of poor grade, and the stability is poor, and more support is needed. When the value of the index is between 0 and 0.2, the surrounding rock is of poor grade, and the stability is poor, and more support is needed.

[0109] Further, the grading index and the surrounding rock grade are integrated into a data sample, which is named as a second data sample.

[0110] Further, the first data sample and the second data sample are combined to form a comprehensive data sample, and the comprehensive data sample includes a training set, a test set and a validation set.

[0111] Further, the comprehensive data sample is used to train an evolutionary support vector regression model, and a function relationship model of the non-destructive testing index and the surrounding rock grade is established, and the function relationship model is an intelligent grading model based on the evolutionary support vector regression algorithm.

[0112] S6. The intelligent rapid grading of the engineering surrounding rock is realized through the intelligent grading model.

[0113] In one embodiment, the obtained non-destructive testing index is substituted into the intelligent grading model to quickly obtain the surrounding rock grade, so as to realize the intelligent rapid grading of the engineering surrounding rock.

[0114] Please refer to Figure 6 , Figure 6 is a structural schematic diagram of an engineering surrounding rock intelligent rapid grading system in an embodiment of the present application. The system includes an input device, a processor, an output device, a memory, and the input device, the processor, the output device and the memory are connected with each other, wherein the memory is used for storing a computer program, the computer program includes program instructions, the processor is configured to call the program instructions, the system uses the engineering surrounding rock intelligent rapid grading method, and the specific running process is as shown in Figure 7 .

[0115] In the embodiment, the input device includes a wave velocity testing device, an energy rebound device, a binocular vision camera and a total station, and is used for acquiring the characteristic parameters of the strength and quality of the surrounding rock.

[0116] The processor is the core component of the system, and is built-in PSO, SVR and an intelligent grading model, and is responsible for executing various algorithms and programs, and realizes the processing and analysis of data.

[0117] The output device is another important interface for the system to interact with the user, including a display and a printer, the display is used to visually display the results of the surrounding rock classification, including classification grade, related parameters and charts;

[0118] A detailed report is automatically generated according to the classification results, including classification basis and recommended measures, to facilitate subsequent decision-making and management of the user; the classification results and related data are sent to external systems or devices, such as databases and remote servers, through a data interface, to realize data sharing and exchange.

[0119] The storage adopts a high-speed solid-state hard disk, which has the characteristics of fast read-write speed, large capacity and high reliability, and is mainly used for storing data obtained by the input device and result data processed by the processor, and can meet the demand of large data storage.

[0120] In summary, the engineering surrounding rock intelligent rapid classification method provided by the application uses three-dimensional feature parameters and comprehensive quality indexes as data samples, fully utilizes multiple information sources, improves the comprehensiveness and accuracy of surrounding rock quality evaluation, and lays a solid foundation for the construction of an intelligent classification model; the evolutionary support vector regression model is trained using comprehensive data samples, which not only enhances the prediction ability of the model, but also establishes a functional relationship between three-dimensional feature parameters, comprehensive quality indexes and surrounding rock grades; through the intelligent classification model, the surrounding rock grade is automatically output, which greatly improves the evaluation efficiency and accuracy, and provides a more scientific and convenient surrounding rock quality evaluation tool for engineering practice.

[0121] The engineering surrounding rock intelligent rapid classification system provided by the application realizes rapid and accurate collection of field data by integrating binocular vision cameras and total station devices; a high-density point cloud model based on a motion recovery structure algorithm and a three-dimensional multi-view stereo vision algorithm is constructed, which greatly improves the efficiency of data collection and processing, and provides a solid foundation for surrounding rock classification; the clustering algorithm is used to classify the structure surface point cloud, and the structure surface spacing and rock mass volume joint number are accurately calculated, combined with non-destructive testing indexes such as rebound strength and stress wave velocity, to realize intelligent and objective determination of the surrounding rock grade, effectively reducing the dependence on the working experience of the operator; the application of the system significantly simplifies the surrounding rock classification process, shortens the working cycle, reduces the construction cost, and improves the accuracy and reliability of the classification results, providing a strong guarantee for engineering safety construction, and showing its great potential and application value in engineering practice.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and they should be covered in the scope of the claims and the description of the present application.

Claims

1. An intelligent and rapid classification method for engineering surrounding rocks, characterized in that: The method comprises the following steps: Based on stereo photogrammetry, the three-dimensional characteristic parameters of the surrounding rock structure surface are obtained, including: Use a total station to obtain the real-world coordinates of the surrounding rock; Based on the real-world coordinates, using a binocular vision camera, acquiring a rock mass image; Based on the rock mass image, a sparse point cloud model based on a motion recovery structure algorithm is constructed using three-dimensional real scene modeling software; Based on the sparse point cloud model, a high-density point cloud model is constructed using a three-dimensional multi-view stereo vision algorithm; Based on the high-density point cloud model, a three-dimensional digital surface model is constructed using the triangular mesh editing function of the three-dimensional real scene modeling software; Based on the three-dimensional digital surface model, a clustering algorithm is used to classify the structural surface point cloud data and obtain a classification result; Based on the classification results, calculating the spacing between structural planes and the number of rock mass volume joints; Based on the dynamic response test, the comprehensive quality indicators of the surrounding rock are obtained, including: Based on the dynamic response test, a comprehensive stress wave velocity index of the surrounding rock is obtained, and the comprehensive stress wave velocity index is used to characterize the quality of the surrounding rock; The comprehensive stress wave velocity index satisfies the following relationship: ， in, is the comprehensive stress wave velocity index, is the distance between the center points of the two sensors, is the propagation time difference; the comprehensive rebound strength index satisfies the following relationship: ， in, is a comprehensive rebound strength index. For the Lithologic measurement area Rebound strength at each measuring point, For the Lithologic measurement area Rebound strength at each measuring point, For the The area of ​​lithologic measurement area, For the The area of ​​lithologic measurement area; Based on the dynamic response test, a comprehensive rebound strength index of the surrounding rock is obtained, and the comprehensive rebound strength index is used to characterize the strength of the surrounding rock; Obtaining the uniaxial compressive strength and rock integrity coefficient of the rock mass based on the three-dimensional characteristic parameters and the comprehensive quality index; Based on the uniaxial compressive strength of the rock mass and the rock mass integrity coefficient, an evolutionary support vector regression model is constructed using a support vector regression algorithm and an evolutionary algorithm; Based on the evolutionary support vector regression model, an intelligent classification model for outputting surrounding rock grades is constructed; The intelligent grading model is used to achieve intelligent and rapid grading of engineering surrounding rocks.

2. The method for intelligent and rapid classification of surrounding rock in an engineering project according to claim 1, characterized in that: The obtaining of the uniaxial compressive strength and the rock mass integrity coefficient based on the three-dimensional characteristic parameters and the comprehensive quality index includes: Based on the comprehensive rebound strength index, the uniaxial compressive strength of the rock mass is obtained; The rock mass integrity coefficient is obtained based on the comprehensive stress wave velocity index and the three-dimensional characteristic parameters.

3. The method for intelligent and rapid classification of surrounding rock in an engineering project according to claim 1, characterized in that: The constructing of an evolutionary support vector regression model based on the uniaxial compressive strength of the rock mass and the rock mass integrity coefficient by using a support vector regression algorithm and an evolutionary algorithm comprises: Based on the uniaxial compressive strength of the rock mass and the rock mass integrity coefficient, a support vector regression model is constructed using a support vector regression algorithm; According to the support vector regression model, an evolutionary support vector regression model is constructed using a particle swarm optimization algorithm.

4. The method for intelligent and rapid classification of surrounding rock in an engineering project according to claim 3, characterized in that: The method of constructing an evolutionary support vector regression model using a particle swarm optimization algorithm based on the support vector regression model includes: According to the support vector regression model, a radial function kernel is selected as the kernel function; Based on the kernel function, setting hyperparameters, the hyperparameters including a penalty parameter and a width parameter of the radial function kernel; Using a particle swarm optimization algorithm, searching for the optimal value of the hyperparameter to obtain the optimal hyperparameter; An evolutionary support vector regression model is constructed using the optimal hyperparameters.

5. The method for intelligent and rapid classification of surrounding rock in an engineering project according to claim 1, characterized in that: The step of constructing an intelligent classification model for outputting surrounding rock grades based on the evolved support vector regression model includes: Integrating the acquired three-dimensional feature parameters and the comprehensive quality index into a first data sample; Obtain a second data sample of the surrounding rock grade after basic quality sub-grade classification corresponding to the three-dimensional characteristic parameters and the comprehensive quality index; combining the first data sample and the second data sample into a composite data sample; Using the comprehensive data sample, training the evolved support vector regression model to obtain a training result; Establishing a functional relationship between the three-dimensional feature parameters, the comprehensive quality index and the surrounding rock grade according to the training results; Through the functional relationship, an intelligent classification model for outputting surrounding rock grades is constructed.

6. The method for intelligent and rapid classification of surrounding rock in an engineering project according to claim 5, characterized in that: The second data sample of the surrounding rock grade after obtaining the basic quality sub-grade classification corresponding to the three-dimensional characteristic parameters and the comprehensive quality index includes: Substituting the uniaxial compressive strength of the rock mass and the rock mass integrity coefficient into the basic quality classification formula of the surrounding rock to obtain a classification index; Determine the surrounding rock grade according to the classification index; The classification index and the surrounding rock grade are integrated into a second data sample.

7. An intelligent rapid grading system for engineering surrounding rocks, the system using an intelligent rapid grading method for engineering surrounding rocks according to any one of claims 1 to 6, characterized in that: The system includes an input device, a processor, an output device and a memory, wherein the input device, the processor, the output device and the memory are connected to each other, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions.

Images