An intelligent identification processing method, system and platform for tunnel adverse geological body based on improved blood-sucking leech optimization algorithm combined with near-infrared spectroscopy

Abstract

This invention discloses a method, system, platform, and storage medium for intelligent identification and processing of adverse geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy. The invention deeply integrates the improved leech optimization algorithm with near-infrared spectroscopy detection technology to construct a complete intelligent identification system for adverse geological bodies in tunnels. The optimization algorithm is used to globally and collaboratively optimize key model parameters throughout the spectral data processing process, improving the dimensionality reduction effect of Multidimensional Scaling Analysis (MDS), the outlier removal capability of Fuzzy C-means Clustering (FCM), and the classification accuracy of Least Squares Support Vector Machine (LSSVM). Ultimately, in complex tunnel construction environments, it achieves rapid and accurate identification and real-time early warning of four types of adverse geological bodies: cavities, faults, fracture zones, and water-rich areas, improving the reliability of geological prediction and the safety of tunnel construction operations.

Description

Technical Field

[0001] This invention belongs to the field of tunnel and underground engineering construction safety technology, specifically involving a method, system, platform, and storage medium for intelligent identification and processing of adverse geological bodies in tunnels based on an improved leech optimization algorithm and near-infrared spectroscopy. Background Technology

[0002] During tunnel drilling and blasting construction, unfavorable geological formations ahead of the tunnel face (such as cavities, faults, fracture zones, and water-rich areas) pose a serious threat to construction safety and efficiency. Traditional geophysical exploration methods, such as resistivity methods and mechanical vibration signal analysis, suffer from insufficient inversion accuracy and limited detection depth. Although ground-penetrating radar technology is widely used, electromagnetic wave attenuation is severe in high-conductivity strata, significantly shortening the effective detection range. High-frequency electromagnetic wave technologies such as X-rays pose ionizing radiation risks, limiting their widespread application in tunnel construction.

[0003] Near-infrared spectroscopy, as a non-ionizing, non-contact rapid detection method, can obtain information such as the mineral composition, moisture content, and structural characteristics of surrounding rocks by analyzing the spectral reflectance signals of rock masses. However, near-infrared spectral data is typically characterized by high dimensionality, redundancy, and low signal-to-noise ratio. Directly using raw spectral data for the identification of adverse geological bodies presents challenges such as high computational cost and difficulty in feature extraction. Traditional machine learning models, such as Support Vector Machines (SVM) and Backpropagation Neural Networks, struggle to fully exploit deep nonlinear features when dealing with spectral data under complex geological conditions, and their performance is highly dependent on the optimization of key parameters. Traditional optimization methods, such as Particle Swarm Optimization (PSO) and Genetic Algorithms, suffer from low optimization efficiency and a tendency to get trapped in local optima, which limits the performance improvement of spectral recognition models, resulting in insufficient model recognition accuracy and robustness, and hindering the achievement of accurate and real-time geological early warning.

[0004] Therefore, in view of the above-mentioned technical problems and defects, there is an urgent need to design and develop an intelligent identification and processing method, system, platform and storage medium for tunnel adverse geological bodies based on an improved leech optimization algorithm and near-infrared spectroscopy. Summary of the Invention

[0005] To overcome the shortcomings and difficulties of the existing technology, the purpose of this invention is to provide a method, system, platform and storage medium for intelligent identification and processing of unfavorable geological bodies in tunnels based on an improved leech optimization algorithm and near-infrared spectroscopy, so as to efficiently process near-infrared spectral data and have strong optimization capabilities for intelligent identification of unfavorable geological bodies.

[0006] The first objective of this invention is to provide an intelligent identification and processing method for unfavorable geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy; the second objective of this invention is to provide an intelligent identification and processing system for unfavorable geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy; the third objective of this invention is to provide an intelligent identification and processing platform for unfavorable geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy; the fourth objective of this invention is to provide a computer-readable storage medium; and the fifth objective of this invention is to provide a near-infrared spectroscopy drilling detection device.

[0007] The first objective of this invention is achieved as follows: the method includes: constructing a first model corresponding to the unfavorable geological body of the tunnel, optimizing the first model based on an improved leech optimization algorithm, and creating first data corresponding to the unfavorable geological body of the tunnel; wherein, the first model is a multidimensional scaling (MDS) model; and the first data is data with the optimal combination of dimensionality reduction parameters.

[0008] A second set of data corresponding to the tunnel borehole wall is created, generated, and acquired. Based on the first data, the second data is dimensionality-reduced, and a third set of data corresponding to the adverse geological body of the tunnel is generated. The second data is the original near-infrared spectral data, and the third data is the spectral principal component vector set data characterizing the features of the adverse geological body.

[0009] The third data is processed by cluster analysis using the fuzzy C-means clustering (FCM) method, and a fourth data corresponding to the unfavorable geological bodies in the tunnel is constructed; wherein, the fourth data is geological anomaly sample data;

[0010] Based on the fourth data, a corresponding second model is constructed and generated. Based on the second model and combined with real-time near-infrared spectral data, a corresponding fifth data is identified and generated. The second model is a least squares support vector machine (LSSVM) classification model. The fifth data is data on the types of adverse geological bodies in tunnels.

[0011] Furthermore, the construction of a first model corresponding to the unfavorable geological features of the tunnel, the optimization of the first model based on an improved leech optimization algorithm, and the creation of first data corresponding to the unfavorable geological features of the tunnel further include:

[0012] A population corresponding to leeches and consisting of multiple individuals is constructed, the population is initialized, and a corresponding first set is created. At the same time, an initial amount of resources is allocated to each host. The first set is an initial host set composed of individuals with the highest fitness ranking in the population.

[0013] A sixth data point corresponding to the population fitness is calculated and generated. Based on the amount of host resources, a corresponding seventh data point is created by combining the sixth data point. The sixth data point is the selection probability. The seventh data point is the operation control data, including the execution of host utilization operation control or random exploration operation control.

[0014] Furthermore, the calculation and generation of the sixth data corresponding to the population fitness, based on the amount of host resources, and in conjunction with the sixth data, to create and generate the corresponding seventh data, also includes:

[0015] When performing host exploitation operations, the location data of the processed individual is updated based on the positional relationship between the individual leech and the selected host, as well as random perturbations, and the amount of resources of the exploited host is reduced accordingly.

[0016] When a host's resources fall below a preset threshold, it is removed from the host set, and the individual with the best fitness from the current population is selected for replacement.

[0017] Furthermore, the calculation and generation of the sixth data corresponding to the population fitness, based on the amount of host resources, and in conjunction with the sixth data, to create and generate the corresponding seventh data, also includes:

[0018] Calculate the selection probability corresponding to the host; the calculation formula is as follows:

[0019] (2)

[0020] In the formula, To adjust the parameters.

[0021] Furthermore, the step of combining the fuzzy C-means clustering (FCM) method to perform cluster analysis on the third data and constructing the fourth data corresponding to the adverse geological features of the tunnel also includes:

[0022] Abnormal spectral sample data is identified and removed, and valid sample data is classified into predefined categories of unfavorable geological bodies; wherein, the predefined categories include cavities, faults, fracture zones and water-rich areas;

[0023] The process of removing abnormal spectral sample data includes: calculating and generating an eighth set of data corresponding to each sample and relative to each cluster center; identifying samples whose maximum membership data in the third set of data is lower than a preset threshold as abnormal samples and removing them; wherein, the eighth set of data is membership data.

[0024] The second objective of this invention is achieved as follows: a data construction and generation unit is used to construct a first model corresponding to the unfavorable geological body of the tunnel, optimize the first model based on an improved leech optimization algorithm, and create first data corresponding to the unfavorable geological body of the tunnel; wherein, the first model is a multidimensional scaling analysis (MDS) model; and the first data is optimal dimensionality reduction parameter combination data;

[0025] A data creation and generation unit is used to create, generate, and acquire second data corresponding to the tunnel borehole wall, and based on the first data, perform dimensionality reduction processing on the second data to generate third data corresponding to the adverse geological body of the tunnel; wherein, the second data is raw near-infrared spectral data; and the third data is spectral principal component vector set data characterizing the features of the adverse geological body;

[0026] The data processing and generation unit is used to combine the fuzzy C-means clustering (FCM) method to perform cluster analysis on the third data and construct and generate fourth data corresponding to the unfavorable geological bodies of the tunnel; wherein, the fourth data is geological anomaly sample data;

[0027] The data recognition and generation unit is used to construct and generate a corresponding second model based on the fourth data, and to identify and generate corresponding fifth data based on the second model and combined with real-time near-infrared spectral data; wherein the second model is a least squares support vector machine (LSSVM) classification model; and the fifth data is tunnel adverse geological body type data.

[0028] Furthermore, the data construction and generation unit further includes:

[0029] The first generation module is used to construct a population corresponding to leeches and consisting of multiple individuals, initialize the population, create a corresponding first set, and allocate initial resources to each host; wherein, the first set is an initial host set composed of individuals with high fitness ranking in the population.

[0030] The second generation module is used to calculate and generate sixth data corresponding to the population fitness, and based on the host resource quantity, and in combination with the sixth data to create and generate corresponding seventh data; the sixth data is the selection probability; the seventh data is operation control data, including execution host utilization operation control or random exploration operation control;

[0031] And / or, the second generation module further includes:

[0032] The first processing module is used to update the location data of the leech individual based on the positional relationship between the leech individual and the selected host and random perturbations when performing host utilization operations, and the amount of resources of the utilized host is reduced accordingly.

[0033] The second processing module is used to remove a host from the host set when the host's resource quantity is lower than a preset threshold, and to select the individual with the best fitness from the current population for replacement processing.

[0034] The generation module is used to calculate the selection probability corresponding to the host; the calculation formula is as follows:

[0035] (2)

[0036] In the formula, To adjust the parameters; For the first The probability that a host is selected by an individual leech; Describes the size of the host set, that is, the top hosts selected from the current population ranked by fitness. Individuals form the initial host set; The current resource quantity of the j-th host; Indicates the first Each host corresponds to a candidate solution individual; The fitness function represents the host The value of is used to measure the quality of the candidate solution; for minimization problems such as MDS hyperparameter optimization in this invention, The smaller the value, the better the solution.

[0037] And / or, the data processing and generation unit further includes:

[0038] The type classification module is used to identify and remove abnormal spectral sample data, and classify the valid sample data into predefined categories of unfavorable geological bodies; wherein, the predefined categories include cavities, faults, fracture zones, and water-rich areas; the removal of abnormal spectral sample data includes: calculating and generating an eighth set of data corresponding to each sample and relative to each cluster center, and determining samples whose maximum membership data in the third set of data is lower than a preset threshold as abnormal samples and removing them; wherein, the eighth set of data is membership data.

[0039] The third objective of this invention is achieved as follows: It includes a processor, a memory, and a control program for an intelligent identification and processing platform for unfavorable geological features in tunnels, based on an improved leech-based optimization algorithm and fusing near-infrared spectroscopy. The processor executes the control program, which is stored in the memory. This control program implements the intelligent identification and processing method for unfavorable geological features in tunnels based on an improved leech-based optimization algorithm and fusing near-infrared spectroscopy.

[0040] The fourth objective of this invention is achieved as follows: the computer-readable storage medium stores a control program for an intelligent identification and processing platform for unfavorable geological features in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy. This control program implements the intelligent identification and processing method for unfavorable geological features in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy.

[0041] The fifth objective of the present invention is achieved as follows: the device includes a probe section configured as a cylindrical structure for threaded connection to the tail of a drill bit; and a light source module, a beam splitter module, and at least one optical probe respectively disposed inside the probe section; wherein the light source module is used to generate near-infrared light; the beam splitter module is used to decompose the near-infrared light into spectral signals of different frequencies; the optical probe is arranged circumferentially along the probe section and is used to emit the spectral signals to the borehole wall and receive the reflected spectral signals; the optical probe is an arc-shaped probe and its end face is flush with or slightly protruding from the outer wall of the probe section;

[0042] The device further includes a spectral driver disposed inside the probe section for controlling spectral acquisition and processing the received spectral signals; a probe brush ring configured to clean the end face of the optical probe when the probe section rotates; and a circumferential elastic damping belt disposed on the probe section for attenuating vibrations during drilling.

[0043] This invention constructs a first model corresponding to adverse geological features in a tunnel using a method, optimizes the first model based on an improved leech optimization algorithm, and generates first data corresponding to the adverse geological features in the tunnel. The first model is a multidimensional scaling (MDS) model; the first data is data with the optimal combination of dimensionality reduction parameters. Second data corresponding to the tunnel borehole wall is created, generated, and acquired. Based on the first data, the second data is dimensionality-reduced, and third data corresponding to the adverse geological features in the tunnel is generated. The second data is raw near-infrared spectral data; the third data is a set of spectral principal component vectors characterizing the features of the adverse geological features. Fuzzy C-means clustering (FC) is then used. The method (M) involves cluster analysis to process the third data and construct a fourth data corresponding to adverse geological bodies in the tunnel; wherein the fourth data is geological anomaly sample data; based on the fourth data, a corresponding second model is constructed and generated, and based on the second model and combined with real-time near-infrared spectral data, a corresponding fifth data is identified and generated; wherein the second model is a least squares support vector machine (LSSVM) classification model; the fifth data is data on the types of adverse geological bodies in the tunnel, as well as the system, platform and storage medium corresponding to the method, and the near-infrared spectral drilling detection device used for the method, which can realize rapid, accurate and real-time identification of adverse geological bodies in complex tunnel environments.

[0044] In other words, this invention deeply integrates an improved leech optimization algorithm with near-infrared spectroscopy detection technology to construct a complete intelligent identification system for adverse geological bodies in tunnels. By utilizing the optimization algorithm to globally and collaboratively optimize key model parameters throughout the entire spectral data processing workflow, the dimensionality reduction effect of Multidimensional Scaling Analysis (MDS), the outlier removal capability of Fuzzy C-means Clustering (FCM), and the classification accuracy of Least Squares Support Vector Machine (LSSVM) are improved. Ultimately, this enables rapid and accurate identification and real-time early warning of four types of adverse geological bodies—cavities, faults, fracture zones, and water-rich areas—in complex tunnel construction environments, thereby enhancing the reliability of geological predictions and the safety of tunnel construction operations. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of the process steps of an intelligent identification and processing method for unfavorable geological bodies in tunnels based on an improved leech optimization algorithm and near-infrared spectroscopy according to the present invention.

[0047] Figure 2 This is a schematic diagram of the process for identifying undesirable geological features in tunnels, based on an improved leech optimization algorithm and near-infrared spectroscopy, according to an embodiment of the present invention.

[0048] Figure 3 This is a schematic diagram of the intelligent identification and processing system architecture for tunnel adverse geological bodies based on an improved leech optimization algorithm and near-infrared spectroscopy according to the present invention.

[0049] Figure 4 This is a schematic diagram of the functional module architecture of an embodiment of the intelligent identification and processing system for tunnel adverse geological bodies based on an improved leech optimization algorithm and near-infrared spectroscopy according to the present invention.

[0050] Figure 5 This is a schematic diagram of the intelligent identification and processing platform architecture for tunnel adverse geological bodies based on an improved leech optimization algorithm and near-infrared spectroscopy according to the present invention.

[0051] Figure 6 This is a schematic diagram of a computer-readable storage medium architecture in one embodiment of the present invention;

[0052] Figure 7 This is a schematic diagram of one structure of a near-infrared spectroscopy drilling detection device according to the present invention;

[0053] Figure 8 This is a second schematic diagram of the structure of a near-infrared spectroscopy drilling detection device according to the present invention;

[0054] In the diagram, 1-detection section; 2-circumferential elastic damping strip; 3-spectral driver; 4-dynamic data acquisition system; 5-sleeve; 12-infrared laser; 13-interferometer; 14-infrared optical probe; 15-fiber bundle; 151-input fiber bundle; 152-output fiber bundle; 153-sampling fiber bundle; 16-probe brush ring; 18-fiber duct; 181-main duct; 182-branch duct; 183-connector branch. Detailed Implementation

[0055] To facilitate a clearer understanding of the objectives, technical solutions, and advantages of this invention, the invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Those skilled in the art can easily understand other advantages and effects of this invention from the content disclosed in this specification.

[0056] This invention can also be implemented or applied through other different specific examples, and various details in this specification can also be modified and changed based on different viewpoints and applications without departing from the spirit of this invention.

[0057] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0058] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Secondly, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0059] Preferably, the intelligent identification and processing method for tunnel adverse geological bodies based on an improved leech optimization algorithm and near-infrared spectroscopy is applied in one or more terminals or servers. The terminal is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions. Its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0060] The terminal can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal can interact with the customer via a keyboard, mouse, remote control, touchpad, or voice control device.

[0061] This invention provides a method, system, platform, and storage medium for intelligent identification and processing of adverse geological features in tunnels based on an improved leech optimization algorithm and near-infrared spectroscopy.

[0062] like Figure 1 The diagram shown is a flowchart of an intelligent identification and processing method for tunnel adverse geological bodies based on an improved leech optimization algorithm fused with near-infrared spectroscopy, provided by an embodiment of the present invention.

[0063] In this embodiment, the intelligent identification and processing method for tunnel adverse geological bodies based on the improved leech optimization algorithm and near-infrared spectroscopy can be applied to terminals or fixed terminals with display functions. The terminals are not limited to personal computers, smartphones, tablets, desktop computers or all-in-one computers with cameras, etc.

[0064] The intelligent identification and processing method for tunnel adverse geological bodies based on the improved leech-based optimization algorithm and near-infrared spectroscopy can also be applied to a hardware environment consisting of a terminal and a server connected to the terminal via a network. The network includes, but is not limited to, wide area networks (WANs), metropolitan area networks (MANs), or local area networks (LANs). The intelligent identification and processing method for tunnel adverse geological bodies based on the improved leech-based optimization algorithm and near-infrared spectroscopy in this embodiment of the invention can be executed by the server, by the terminal, or by both the server and the terminal.

[0065] For example, for a terminal requiring intelligent identification and processing of tunnel unfavorable geological bodies based on an improved leech-based optimization algorithm fused with near-infrared spectroscopy, the intelligent identification and processing function of tunnel unfavorable geological bodies based on an improved leech-based optimization algorithm fused with near-infrared spectroscopy, provided by the method of this invention, can be directly integrated onto the terminal, or a client for implementing the method of this invention can be installed. Alternatively, the method provided by this invention can also run on servers or other devices in the form of a Software Development Kit (SDK). The SDK provides an interface for the intelligent identification and processing function of tunnel unfavorable geological bodies based on an improved leech-based optimization algorithm fused with near-infrared spectroscopy. Terminals or other devices can then implement this intelligent identification and processing function through the provided interface. The invention will be further described below with reference to the accompanying drawings.

[0066] like Figures 1-2 As shown, this invention provides an intelligent identification and processing method for unfavorable geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy. The method includes the following steps:

[0067] S01. Construct a first model corresponding to the unfavorable geological body of the tunnel, and optimize the first model based on the improved leech optimization algorithm to create and generate first data corresponding to the unfavorable geological body of the tunnel; wherein, the first model is a multidimensional scaling analysis (MDS) model; the first data is the optimal combination of dimensionality reduction parameters.

[0068] S02. Create, generate, and acquire second data corresponding to the tunnel borehole wall, and based on the first data, perform dimensionality reduction processing on the second data to generate third data corresponding to the unfavorable geological body of the tunnel; wherein, the second data is the original near-infrared spectral data; and the third data is the spectral principal component vector set data characterizing the features of the unfavorable geological body;

[0069] S03. Using the fuzzy C-means clustering (FCM) method, the third data is processed by cluster analysis, and a fourth data corresponding to the unfavorable geological bodies of the tunnel is constructed; wherein, the fourth data is geological anomaly sample data;

[0070] S04. Based on the fourth data, a corresponding second model is constructed and generated, and based on the second model and combined with real-time near-infrared spectral data, a corresponding fifth data is identified and generated; wherein, the second model is a least squares support vector machine (LSSVM) classification model; and the fifth data is tunnel adverse geological body type data.

[0071] The process of constructing a first model corresponding to the unfavorable geological features of the tunnel, optimizing the first model based on an improved leech optimization algorithm, and creating first data corresponding to the unfavorable geological features of the tunnel further includes:

[0072] S011. Construct a population corresponding to leeches and consisting of multiple individuals, initialize the population, and create a corresponding first set, while allocating initial resources to each host; wherein, the first set is an initial host set composed of individuals with the highest fitness ranking in the population.

[0073] S012. Calculate and generate a sixth data corresponding to the population fitness, and based on the host resource quantity, create and generate a corresponding seventh data in combination with the sixth data; the sixth data is the selection probability; the seventh data is operation control data, including execution host utilization operation control or random exploration operation control.

[0074] The calculation and generation of the sixth data corresponding to the population fitness, based on the host resource quantity, and combined with the sixth data to create the corresponding seventh data, also includes:

[0075] S0121. When performing host utilization operations, the location data of the processed individual is updated according to the positional relationship between the leech individual and the selected host and random perturbations, and the amount of resources of the utilized host is reduced accordingly.

[0076] S0122. When the host's resource quantity is lower than the preset threshold, remove it from the host set and select the individual with the best fitness from the current population for replacement.

[0077] The calculation and generation of the sixth data corresponding to the population fitness, based on the host resource quantity, and combined with the sixth data to create the corresponding seventh data, also includes:

[0078] S0123. Calculate the selection probability corresponding to the host; the calculation formula is as follows:

[0079] (2)

[0080] In the formula, To adjust the parameters; For the first The probability that a host is selected by an individual leech; Describes the size of the host set, that is, the top hosts selected from the current population ranked by fitness. Individuals form the initial host set; The current resource quantity of the j-th host; Indicates the first Each host corresponds to a candidate solution individual; The fitness function represents the host The value of is used to measure the quality of the candidate solution; for minimization problems such as MDS hyperparameter optimization in this invention, The smaller the value, the better the solution.

[0081] The method of combining fuzzy C-means clustering (FCM) to perform cluster analysis on the third data and constructing fourth data corresponding to the adverse geological features of the tunnel also includes:

[0082] S031. Identify and remove abnormal spectral sample data, and classify the valid sample data into predefined categories of unfavorable geological bodies; wherein, the predefined categories include cavities, faults, fracture zones and water-rich areas;

[0083] The process of removing abnormal spectral sample data includes: calculating and generating an eighth set of data corresponding to each sample and relative to each cluster center; identifying samples whose maximum membership data in the third set of data is lower than a preset threshold as abnormal samples and removing them; wherein, the eighth set of data is membership data.

[0084] Specifically, in this embodiment of the invention, a method for intelligent identification and processing of adverse geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy is provided. The method includes the following steps:

[0085] S1. The improved leech optimization algorithm is used to globally optimize the hyperparameters of the multidimensional scaling (MDS) algorithm to obtain the optimal dimensionality reduction parameters suitable for the extraction of spectral features of adverse geological bodies.

[0086] S2. Under the guidance of the optimal parameters determined in step S1, perform MDS dimensionality reduction on the raw spectral data collected by the near-infrared spectral detection device during tunnel excavation to extract the principal component vector set containing the spectral features of the adverse geological body.

[0087] S3. Use the fuzzy C-means clustering (FCM) method to divide the principal components of the dimensionality-reduced spectral data into four preset types of geological anomaly samples, corresponding to cavities, faults, fracture zones and water-rich areas, respectively.

[0088] S4. Use Least Squares Support Vector Machine (LSSVM) to train and model various types of samples, and predict the type of adverse geological bodies based on the new spectral data received in real time during drilling.

[0089] Furthermore, an improved leech optimization algorithm is proposed for hyperparameter optimization in methods for identifying adverse geological bodies, including MDS, fuzzy C-means clustering (FCM), and LSSVM. The specific optimization process is as follows: Let the population size be N, the problem dimension be D, and the population representation be... Each individual This is a candidate solution in the search space. The fitness function f(X) is used to determine the optimal solution. i Assess individual quality and select the top K individuals by fitness from the population as the initial host set. And allocate an initial amount of resources R0 to each host.

[0090] In each iteration t, the number of leech individuals X i The host is selected based on probability P, either by utilizing the host or by conducting random exploration. Host selection probability P j Based on the host's resource quantity R j and fitness f(H) j )calculate:

[0091] (2)

[0092] In the formula, To adjust the parameters; For the first The probability that a host is selected by an individual leech; Describes the size of the host set, that is, the top hosts selected from the current population ranked by fitness. Individuals form the initial host set; The current resource quantity of the j-th host; Indicates the first Each host corresponds to a candidate solution individual; The fitness function represents the host The value of is used to measure the quality of the candidate solution; for minimization problems such as MDS hyperparameter optimization in this invention, The smaller the value, the better the solution.

[0093] Selected host H j Then, the formula for updating the position of individual leeches is:

[0094] (3)

[0095] In the formula, Indicates the first The current position of an individual in the t-th iteration. Indicates that it is in the first The update position in the next iteration Indicates the first The host in the 1st The current position in the next iteration. This represents the step size coefficient along the host direction. The amplitude coefficient represents the random disturbance. This indicates that the value range is [-1, 1]. 3D random vector, Indicates the dimension of the search space. Indicates the iteration round, This indicates the individual index that has been updated. The above symbols collectively represent the host index being utilized. They are used to characterize the update mechanism of directional convergence of individuals around the host and the superposition of random perturbations, in order to advance the search of the hyperparameter space.

[0096] After utilizing the host, its resource quantity decreases by a fixed amount ΔR:

[0097] (4)

[0098] In the formula, Indicates the first The host in the 1st Resource amount at the next iteration. Indicates that it is in the first Resource amount at the next iteration.

[0099] like <0, then another =0.

[0100] When an individual leech does not choose to utilize a host, it randomly generates a new location within the search space:

[0101] (5)

[0102] In the formula, and These are the lower and upper bounds of the search space, respectively. Let the vector be a D-dimensional random vector. If the new position... If the host with better fitness than the worst host is replaced, it is given an initial resource amount of R0.

[0103] After each iteration, check the resource quantity of all hosts. Set a resource quantity threshold of θ. If the resource quantity R of a certain host... j If the value is less than θ, then remove it and select a new individual with the best fitness from the population to supplement the host set, and assign it an initial resource amount of R0.

[0104] When the termination condition is met, the solution of the individual with the best fitness value in the population is output as the optimal solution.

[0105] In other words, this invention provides a method for intelligent identification and processing of unfavorable geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy. The specific steps are as follows:

[0106] S1. The raw spectral data of the tunnel face is collected using the near-infrared spectral drilling detection device for identifying adverse geological bodies proposed in this invention. The data acquisition frequency is once per minute, and a total of 1000 spectral samples are collected. Each sample contains 500 wavelength points, forming a high-dimensional dataset.

[0107] Then, the improved leech optimization algorithm proposed in this invention is used to globally optimize the hyperparameters of the multidimensional scaling (MDS) algorithm to obtain the optimal dimensionality reduction parameters suitable for extracting spectral features of adverse geological bodies. The specific implementation process is as follows:

[0108] The key hyperparameters of the MDS algorithm are determined to be the target dimensionality reduction dimension k and the shrinkage coefficient. Therefore, this problem is modeled as a two-dimensional search space, i.e., each candidate solution... This represents a set of possible parameter configurations.

[0109] Set population size Maximum number of iterations .

[0110] The search range for the dimension reduction dimension k is set to [1, 10], and the shrinkage coefficient is... The search range is [0.1, 1.0].

[0111] Within the aforementioned search space, 50 initial candidate solutions are randomly generated to form the initial population. .

[0112] Through the fitness function f(X) i Assess individual quality.

[0113] (1)

[0114] In the formula, This represents the distance between samples i and j in the original high-dimensional spectral data. This represents the distance between embedding points i and j in the target low-dimensional space.

[0115] This represents the distance between the original high-dimensional spectral samples under the current dimensionality reduction hyperparameter settings. Distance between the target and the corresponding point in the low-dimensional embedding space The cumulative bias, i.e., the distance difference over all sample pairs (i,j). The sum of squares is used to measure the degree of distortion in the reconstruction of geometric relationships. The smaller the value, the better the low-dimensional embedding can preserve the relative distance relationship between the original spectral samples.

[0116] The normalization scale representing the cumulative amount of the above bias is usually expressed as the high-dimensional original distance. The scale is normalized to make the target values ​​under different hyperparameter combinations comparable, eliminating the influence of the overall sample size or the number of sample pairs on the target value.

[0117] fitness in the population Top Individuals were selected as initial hosts to form a host set. and allocate an initial amount of resources to each host. .

[0118] In each iteration middle( ), perform the following operations:

[0119] For each non-host individual X in the population i Calculate each host according to the formula Probability of being selected This probability is determined by the host's resource availability. and fitness If determined jointly, then

[0120] (2)

[0121] In the formula, To adjust the parameters; For the first The probability that a host is selected by an individual leech; Describes the size of the host set, that is, the top hosts selected from the current population ranked by fitness. Individuals form the initial host set; The current resource quantity of the j-th host; Indicates the first Each host corresponds to a candidate solution individual; The fitness function represents the host The value of is used to measure the quality of the candidate solution; for minimization problems such as MDS hyperparameter optimization in this invention, The smaller the value, the better the solution.

[0122] Adjust parameters This is used to balance the weights of resource quantity and fitness. Probability The highest-ranking host is most likely to become X i The "parasitic" target.

[0123] In this embodiment, probability Decide whether to utilize the host. If utilization is chosen, then based on... Randomly select a host

[0124] (3)

[0125] In the formula, Indicates the first The current position of an individual in the t-th iteration. Indicates that it is in the first The update position in the next iteration Indicates the first The host in the 1st The current position in the next iteration. This represents the step size coefficient along the host direction. The amplitude coefficient represents the random disturbance. This indicates that the value range is [-1, 1]. 3D random vector, Indicates the dimension of the search space. Indicates the iteration round, This indicates the individual index that has been updated. The above symbols collectively represent the host index being utilized. They are used to characterize the update mechanism of directional convergence of individuals around the host and the superposition of random perturbations, in order to advance the search of the hyperparameter space.

[0126] In this embodiment, the step size parameter for approaching the host Parameters that control the amplitude of random disturbances , It is a 2-dimensional random vector whose elements are in the interval [0, 1]. The internal components are uniformly and randomly generated.

[0127] like Successfully exploited the host Its resource quantity decreases by a fixed amount ΔR:

[0128] (4)

[0129] In the formula, Indicates the first The host in the 1st Resource amount at the next iteration. Indicates that it is in the first Resource amount at the next iteration.

[0130] In this embodiment, ΔR = 10, if Then let .

[0131] If the host does not become X i The "parasitic" target then needs to be randomly explored again within the entire search space to generate a new location:

[0132] (5)

[0133] In the formula, and These are the lower and upper bounds of the search space, respectively. for A dimensional random vector. If the new position If the host with better fitness than the worst host is replaced, it is given an initial resource amount of R0.

[0134] After each iteration, check the resource quantity of all hosts. Set a resource quantity threshold. If a host of If the host is not found to be fit, it is removed from the host set. Then, a host with fitness is selected from the current population. The highest-ranking non-host individual is added to the host set and given an initial amount of resources. .

[0135] When the number of iterations reaches the preset maximum value When the time is reached, the algorithm terminates. Finally, fitness scores are selected from the entire population. The individual with the highest value The combination of hyperparameters carried by this individual. This is the global optimal solution of the MDS algorithm.

[0136] This embodiment successfully automates the manual parameter adjustment process, ensuring the optimization of MDS dimensionality reduction effect and laying a solid and reliable data foundation for subsequent anomaly removal and geological body classification. It fully demonstrates the advanced nature and practicality of this invention in the field of intelligent monitoring of tunnel engineering.

[0137] S2: The optimal hyperparameters obtained in step S1 Under the guidance of [unspecified authority], multidimensional scaling (MDS) analysis was performed on the raw near-infrared spectral data of the tunnel face to reduce dimensionality and extract principal components containing spectral characteristics of adverse geological bodies. The specific implementation process is as follows:

[0138] First, for each pair of spectral samples x i ,x j Calculate the contracted Mahalanobis distance:

[0139] (6)

[0140] In the formula, in the formula, and Representing samples respectively With sample High-dimensional spectral vectors, parameters This represents the shrinkage coefficient obtained from the global optimization in step S1, which is used to adjust the tradeoff between the original covariance and the identity matrix in the covariance estimation. Indicates the contraction parameter Sample pairs under action ( , The contracted Mahalanobis distance, This represents the covariance matrix obtained from the sample set. Here, I is the sample covariance matrix, and III is the identity matrix. Indicates by With I press The resulting shrinkage covariance matrix ( ).

[0141] Will Squaring each element in the matrix yields the squared distance matrix:

[0142] (7)

[0143] In the formula, Represents the pairwise distance of the samples The distance matrix formed This represents the corresponding squared distance matrix. Represents the original high-dimensional spectral data sample and The distance between them.

[0144] Construct a centralized matrix And for D (2) Implement dual centralization:

[0145] (8)

[0146] In the formula, The centered matrix is ​​defined as follows: ; Represents the number of samples A matching unit matrix; Indicates the number of samples.

[0147] Perform eigenvalue decomposition on matrix B:

[0148] (9)

[0149] In the formula, and

[0150] Indicates by The matrix composed of eigenvectors, Represents the diagonal eigenvalue matrix. Indicates the first The eigenvalues ​​are sorted from largest to smallest to select the first few principal components. for The transpose of .

[0151] The optimal dimension obtained in step S1 Under the guidance of the previous Given the largest eigenvalues ​​and their corresponding eigenvectors, construct:

[0152] (10)

[0153] In the formula, Indicates the preceding A matrix composed of eigenvectors This is the corresponding diagonal eigenvalue matrix.

[0154] The final matrix That is, the principal components of each spectral sample in the low-dimensional embedding space.

[0155] This principal component representation preserves the original geometric relationships of the spectrum while eliminating redundant dimensions and noise features, thereby obtaining spectral principal components that can effectively characterize the features of adverse geological bodies, providing a reliable basis for subsequent anomaly sample removal and classification modeling.

[0156] S3: In this embodiment, the two-dimensional spectral principal component dataset obtained in step S2 contains 1000 spectral samples. To identify unfavorable geological bodies and remove abnormal samples, the specific process is as follows:

[0157] Will The spectral samples are divided into C=4 fuzzy classes, corresponding to cavities, faults, fracture zones, and water-rich areas, respectively. The Fuzzy C-means Clustering (FCM) algorithm achieves clustering by minimizing the following objective function.

[0158] (11)

[0159] In the formula, Indicates the number of samples; This represents the number of cluster categories, which is taken as the number of clusters in the application scenario of this invention. ; This represents the fuzzy weighted index, which is generally taken as... This is used to adjust the strength of the influence of membership degree on weighted distance; Indicates the first Vector representation of each spectral sample in a low-dimensional embedding space; Indicates the first The cluster center corresponding to the class; Indicates sample Cluster centers The membership degree of has a value range of [0,1]. This represents the Euclidean distance between a sample and its cluster center.

[0160] In each iteration, the cluster centers for each category are updated according to the following formula:

[0161] (12)

[0162] In the formula, Indicates the first Cluster center vector of the class, Indicates sample For the Membership degree of a class This represents a fuzzy weighted index. Indicates the first Vector representation of each sample in the low-dimensional embedding space Represents the total number of samples. It is a category index with values ​​ranging from 1 to .

[0163] This update process ensures that the cluster centers continuously approach the centroid of the current sample distribution.

[0164] After updating the cluster centers, recalculate the membership degree of each sample to each cluster center using the following formula:

[0165] (13)

[0166] In the formula, Indicates the first The sample relative to the first Membership degree of cluster centers; Indicates the first Vector representation of each sample in a low-dimensional embedding space; Indicates the first Cluster center vector of the class; Indicates sample With cluster center The Euclidean distance; Indicates the total number of categories; Indicates the total number of samples; This represents the fuzzy weighted index; k is the category index of the summation in the denominator, used to iterate from 1 to... .

[0167] In this embodiment, the closer a sample is to a cluster center, the higher its membership value. The "cluster center update - membership degree update" process is continuously alternated until the objective function J is achieved. m The clustering process converges to a preset threshold or reaches the maximum number of iterations. After clustering is complete, the maximum membership value of each sample in the four categories is calculated:

[0168] (14)

[0169] In the formula, Indicates the first The maximum membership degree of a sample across the four categories; Indicates the first The sample relative to the first Membership degree of cluster centers; This represents the set of category indices for four types of unfavorable geological bodies (cavities, faults, fracture zones, and water-rich areas); a threshold of 0.6 is used to distinguish between samples with clear and unclear classifications.

[0170] In this embodiment, if If so, the sample can be reliably classified into a certain type of adverse geological body;

[0171] like If so, it is determined to be an abnormal spectral sample.

[0172] In this embodiment, 57 spectral samples with a maximum membership degree of less than 0.6 were removed as outliers.

[0173] After removing abnormal samples, the remaining 943 samples were divided into four categories: Category 1: cavity spectral samples; Category 2: fault spectral samples; Category 3: fracture zone spectral samples; and Category 4: water-rich area spectral samples.

[0174] Clustering results show that the four classes of samples form a compact distribution in the two-dimensional embedding space, with clear inter-class boundaries. Most outliers are distributed in the inter-class transition region or far from the main cluster.

[0175] S4: In the initial steps S1 and S2, the dimensionality-reduced principal components of the spectrum were obtained. In step S3, fuzzy C-means clustering (FCM) was used to remove abnormal spectral samples, ultimately yielding 943 valid samples, of which:

[0176] There were 230 cavity samples, 215 fault samples, 248 fracture zone samples, and 250 water-rich area samples.

[0177] To achieve the identification of adverse geological body types, the above 943 samples were input into the least squares support vector machine (LSSVM) model for training.

[0178] LSSVM, based on regularization theory and the cost function of least squares method, constructs the following optimization objective function:

[0179] (15)

[0180] In the formula, c represents the regularization parameter, used to balance model complexity and fitting error; γ i Let be the i-th error term; ω be the weight vector in the high-dimensional space; φ be the nonlinear mapping function from low-dimensional to high-dimensional space; and b be the bias term.

[0181] By introducing Lagrange multipliers and deriving the optimization problem, the regression form of LSSVM is obtained:

[0182] (16)

[0183] In the formula, For Lagrange multipliers, and Let represent two sample vectors in the low-dimensional embedding space. For kernel function, This is a bias term.

[0184] To better handle the nonlinear characteristics of spectral data from adverse geological bodies, the embodiment uses a radial basis function (RBF) as the kernel function.

[0185] (17)

[0186] In the formula, This represents the kernel function bandwidth parameter, used to control the similarity decay scale, and its optimal value is determined by cross-validation. and These represent two sample vectors in the low-dimensional embedding space; This represents the Euclidean distance between the two.

[0187] After the model training is complete, the final discriminant function takes the form of:

[0188] (18)

[0189] In the formula, This represents the kernel function bandwidth parameter, used to control the similarity decay scale, and its optimal value is determined by cross-validation. Representation and Sample The corresponding weights; and These represent two sample vectors in the low-dimensional embedding space.

[0190] This function is used to classify and predict the input spectral samples and output their corresponding adverse geological body categories.

[0191] The four types of spectral principal component samples output by S3 are used as the training set and input into the LSSVM model above.

[0192] By optimizing parameter c and To obtain the optimal model;

[0193] After training, the LSSVM model can establish a mapping relationship between spectral features and geological categories (cavities, faults, fracture zones, and water-rich areas).

[0194] In the actual construction of urban rail transit tunnels, the target identification zone is defined as an area of ​​approximately 5-8 meters in front of the tunnel face. The near-infrared spectral drilling detection device proposed in this invention is installed at the tail of the drill bit via a threaded connection. The device has a cylindrical structure, internally containing a light source module, a spectrometer module, and a spectral driver. An arc-shaped optical probe is arranged circumferentially, and externally, a probe brush ring and a circumferential elastic damping band are installed. During the drill-and-blast method, raw near-infrared spectral data corresponding to the tunnel wall is acquired at a fixed sampling frequency of once per minute as secondary data. 1000 samples (500 wavelength points per sample) are continuously obtained and transmitted back to the processing terminal in real time via the main data link. A multidimensional scaling analysis (MDS) model is constructed as the first model. An improved leech optimization algorithm is used to globally optimize the target dimensionality reduction dimension and shrinkage coefficient, creating the first data set with the optimal combination of dimensionality reduction parameters. This algorithm constructs a population of multiple leech individuals and an initial host set, allocates host resources, determines the host selection probability based on the host resource quantity and fitness, updates the position using a step size parameter that tends towards the host and random perturbation, and dynamically removes and replenishes hosts when their resources fall below a threshold, thereby avoiding getting trapped in local optima and improving optimization efficiency. The first data guides the multidimensional scaling analysis (MDS) dimensionality reduction of the second data, obtaining the spectral principal component vector set characterizing the features of the adverse geological body as the third data set.

[0195] Within this interval, based on the third data, cluster analysis was performed using the fuzzy C-means clustering (FCM) method to calculate the membership data corresponding to each cluster center. Anomalies were identified and removed using a maximum membership threshold of 0.6, generating geological anomaly sample data as the fourth data. Statistical analysis showed that the removed samples accounted for 9.7% of the total, and the overlap between the removed samples and the "high-disturbance periods" (rod replacement, borehole cleaning, high mud and water content) of the construction process was 83.5%, indicating that the anomalies mainly originated from unstable spectra. Regarding the dimensionality reduction geometry, the mean square distance from each category sample to its own cluster center was used as the intra-class dispersion index, averaging 0.42 for the four categories: cavities, faults, fracture zones, and water-rich areas. The minimum pairwise distance between the four cluster centers was used as the inter-class separation index, with a statistical value of 1.85. These two indicators demonstrate that the third data obtained under the guidance of the first data possesses both high intra-class compactness and strong inter-class separability within the low-dimensional embedding space, providing a stable geometric foundation for subsequent classification modeling.

[0196] In the classification and identification stage, a least squares support vector machine (LSSVM) classification model was constructed based on the fourth set of data as the second model. A radial basis function kernel was used, and the regularization parameter and kernel function bandwidth were optimized collaboratively. A layered time window rolling verification method was employed to identify continuously arriving new samples online, achieving an overall accuracy of 95.1% for the four classes. For easily confused class pairs, the recall rate was 94.2% for faults, 92.7% for fractured zones, and 95.4% for water-rich areas. Online early warnings were triggered based on consistent judgments over three consecutive frames, with a false alarm rate of 2.1% and a missed alarm rate of 2.8%. The average early warning distance relative to the revealed or geologically logged confirmed location was 6.2m. In the section from K12+380 to K12+395, the system continuously identified a combination of water-rich area and fractured zone, triggering audible and visual early warnings. The construction unit accordingly reduced drilling speed and increased the density of advanced support. Subsequent revealing and geological logging confirmed the presence of water-rich fractured structures in this section, verifying the real-time performance and accuracy of the proposed solution under complex working conditions.

[0197] To make the technical effects of the present invention more intuitive and clear, in conjunction with the method and parameter settings of this patent, a comparative statistical evaluation was carried out on the same batch of original near-infrared spectral second data (1000 samples, 500 wavelength points per sample) under the same urban rail transit tunnel conditions.

[0198] Comparative Example 1 was set up, retaining only the same hardware conditions and the same second set of data. The improved leech optimization algorithm was not used; instead, fixed hyperparameters were applied to Multidimensional Scaling Analysis (MDS) and Least Squares Support Vector Machine (LSSVM), maintaining the same threshold process for Fuzzy C-means Clustering (FCM). In this case, the average intra-class dispersion of the third data in the low-dimensional embedding space after dimensionality reduction was 0.68, and the minimum inter-class central distance was 1.21. During the clustering stage, the outlier removal rate was 15.4%, and the overlap rate with high-disturbance periods was 61.2%. The overall accuracy for the four classes was 79.3%, with significant confusion between faults and fracture zones. The recall rate was 78.0% for faults, 76.5% for fracture zones, and 80.1% for water-rich areas. The false alarm rate for online early warning was 5.6%, the missed alarm rate was 6.9%, and the average early warning distance was 4.1m. The above results indicate that, due to the lack of global collaborative optimization of the hyperparameters of multidimensional scaling analysis (MDS), the third data exhibits high intra-class dispersion and low inter-class separability in the low-dimensional embedding space, which directly leads to an increase in the complexity of the subsequent two-stage discrimination task.

[0199] Comparative Example 2 maintains the dimensionality reduction using Multidimensional Scaling Analysis (MDS) guided by the first data, but removes outlier removal from Fuzzy C-means Clustering (FCM), directly training and applying Least Squares Support Vector Machine (LSSVM) online using all the third data. In this case, the overall accuracy for the four classes is 66.7%, with increased false alarms near faults, fracture zones, and water-rich areas in the medium transition zone; the false alarm rate for online early warning is 7.8%, the missed alarm rate is 8.5%, and the average early warning distance is 1.7m. The results of Comparative Example 2 demonstrate that in the drilling application scenario of this patent, outlier removal using Fuzzy C-means Clustering (FCM) based on the maximum membership threshold is crucial. The absence of this step significantly reduces the purity of the "fourth data," making it difficult for LSSVM to form a stable decision boundary in the embedding space, directly leading to a systematic degradation of recognition accuracy, early warning stability, and early warning lead time.

[0200] As can be seen from the comparison of the examples and the two sets of comparative examples, the optimal combination of dimensionality reduction parameters generated by the improved leech optimization algorithm enables the third data to achieve significantly lower intra-class dispersion and higher inter-class centroid spacing in the low-dimensional embedding space. Based on this, the membership threshold elimination mechanism of fuzzy C-means clustering (FCM) effectively focuses on the unstable spectrum caused by construction disturbances, thereby improving the purity of the fourth data. Finally, the least squares support vector machine (LSSVM) forms a stable nonlinear discrimination boundary in the online identification of the fifth data, resulting in improved accuracy, reduced false alarms and missed alarms, and increased early warning lead time. This result intuitively demonstrates the technical effects of "improving the dimensionality reduction effect of multi-dimensional scale analysis, enhancing the outlier removal capability of fuzzy C-means clustering (FCM), and improving the classification accuracy of least squares support vector machine (LSSVM)." Combined with the near-infrared spectroscopy drilling detection device of this patent, it meets the engineering requirements for rapid and accurate identification and real-time early warning of cavities, faults, fracture zones, and water-rich areas under conditions of continuous drilling at the face.

[0201]

[0202] Table 1. Comparison of key indicators between the examples and the two comparative sets.

[0203] To achieve the above objectives, this invention also provides an intelligent identification and processing system for tunnel adverse geological bodies based on an improved leech-based optimization algorithm fused with near-infrared spectroscopy, such as... Figures 3-4 As shown, the system is applied to the intelligent identification and processing method for tunnel adverse geological bodies based on an improved leech optimization algorithm fused with near-infrared spectroscopy. The system includes:

[0204] The data construction and generation unit is used to construct a first model corresponding to the unfavorable geological body of the tunnel, optimize the first model based on the improved leech optimization algorithm, and create first data corresponding to the unfavorable geological body of the tunnel; wherein, the first model is a multidimensional scaling analysis (MDS) model; the first data is data with the optimal combination of dimensionality reduction parameters;

[0205] A data creation and generation unit is used to create, generate, and acquire second data corresponding to the tunnel borehole wall, and based on the first data, perform dimensionality reduction processing on the second data to generate third data corresponding to the adverse geological body of the tunnel; wherein, the second data is raw near-infrared spectral data; and the third data is spectral principal component vector set data characterizing the features of the adverse geological body;

[0206] The data processing and generation unit is used to combine the fuzzy C-means clustering method to perform cluster analysis on the third data and construct and generate fourth data corresponding to the unfavorable geological bodies in the tunnel; wherein, the fourth data is geological anomaly sample data;

[0207] The data recognition and generation unit is used to construct and generate a corresponding second model based on the fourth data, and to identify and generate corresponding fifth data based on the second model and combined with real-time near-infrared spectral data; wherein the second model is a least squares support vector machine (LSSVM) classification model; and the fifth data is tunnel adverse geological body type data.

[0208] The data construction and generation unit further includes:

[0209] The first generation module is used to construct a population corresponding to leeches and consisting of multiple individuals, initialize the population, create a corresponding first set, and allocate initial resources to each host; wherein, the first set is an initial host set composed of individuals with high fitness ranking in the population.

[0210] The second generation module is used to calculate and generate sixth data corresponding to the population fitness, and based on the host resource quantity, and in combination with the sixth data to create and generate corresponding seventh data; the sixth data is the selection probability; the seventh data is operation control data, including execution host utilization operation control or random exploration operation control;

[0211] And / or, the second generation module further includes:

[0212] The first processing module is used to update the location data of the leech individual based on the positional relationship between the leech individual and the selected host and random perturbations when performing host utilization operations, and the amount of resources of the utilized host is reduced accordingly.

[0213] The second processing module is used to remove a host from the host set when the host's resource quantity is lower than a preset threshold, and to select the individual with the best fitness from the current population for replacement processing.

[0214] The generation module is used to calculate the selection probability corresponding to the host; the calculation formula is as follows:

[0215] (2)

[0216] In the formula, To adjust the parameters; For the first The probability that a host is selected by an individual leech; Describes the size of the host set, that is, the top hosts selected from the current population ranked by fitness. Individuals form the initial host set; The current resource quantity of the j-th host; Indicates the first Each host corresponds to a candidate solution individual; The fitness function represents the host The value of is used to measure the quality of the candidate solution; for minimization problems such as MDS hyperparameter optimization in this invention, The smaller the value, the better the solution.

[0217] And / or, the data processing and generation unit further includes:

[0218] The type classification module is used to identify and remove abnormal spectral sample data, and classify the valid sample data into predefined categories of unfavorable geological bodies; wherein, the predefined categories include cavities, faults, fracture zones, and water-rich areas; the removal of abnormal spectral sample data includes: calculating and generating an eighth set of data corresponding to each sample and relative to each cluster center, and determining samples whose maximum membership data in the third set of data is lower than a preset threshold as abnormal samples and removing them; wherein, the eighth set of data is membership data.

[0219] Specifically, in the embodiments of the present invention, such as Figure 4 As shown, a tunnel adverse geological body intelligent identification system based on an improved leech optimization algorithm fused with near-infrared spectroscopy is provided, comprising:

[0220] The data acquisition module uses the near-infrared spectroscopy drilling detection device for identifying unfavorable geological bodies proposed in this invention to acquire near-infrared spectral data of unfavorable geological bodies on the borehole wall;

[0221] The data processing module executes the intelligent identification method for adverse geological bodies proposed in this invention. It uses an improved leech optimization algorithm to globally optimize the hyperparameters of the Multidimensional Scaling Analysis (MDS), Fuzzy C-means Clustering (FCM), and Least Squares Support Vector Machine (LSSVM) algorithms. Then, it performs dimensionality reduction, anomaly removal, and sample classification on the original spectral data. Finally, it establishes a LSSVM classification model based on historical spectral samples to identify different categories of adverse geological bodies.

[0222] The data perception module transmits the training results of the data processing module to the data early warning module.

[0223] The data early warning module receives data collected by the data acquisition module from the data sensing module. After identifying data anomalies, the data early warning module automatically triggers an early warning signal and dynamically identifies whether the measurement area is an unfavorable geological body.

[0224] The data storage module is used to archive all data collection data, data processing results, data sensing results, and data early warning records.

[0225] In the system embodiment of the present invention, the specific details of the method steps involved in the intelligent identification and processing of tunnel adverse geological bodies based on an improved leech optimization algorithm and near-infrared spectroscopy have been described above. That is to say, the functional modules in the system are used to implement the steps or sub-steps in the above method embodiment, and will not be repeated here.

[0226] To achieve the above objectives, this invention also provides an intelligent identification and processing platform for tunnel adverse geological bodies based on an improved leech-based optimization algorithm fused with near-infrared spectroscopy, such as... Figure 5 As shown, the system includes a processor, a memory, and a control program for an intelligent identification and processing platform for tunnel unfavorable geological bodies based on an improved leech-based optimization algorithm fused with near-infrared spectroscopy. The processor executes the control program, which is stored in the memory. This control program implements the steps of the intelligent identification and processing method for tunnel unfavorable geological bodies based on an improved leech-based optimization algorithm fused with near-infrared spectroscopy.

[0227] The specific details of the steps have been explained above and will not be repeated here.

[0228] In this embodiment of the invention, the built-in processor of the intelligent identification and processing platform for tunnel adverse geological bodies based on the improved leech-based optimization algorithm and near-infrared spectroscopy can be composed of integrated circuits. For example, it can be composed of a single packaged integrated circuit, or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor connects to various components using various interfaces and lines, and executes programs or units stored in memory, as well as calls data stored in memory, to perform various functions and process data for intelligent identification and processing of tunnel adverse geological bodies based on the improved leech-based optimization algorithm and near-infrared spectroscopy.

[0229] The memory, used to store program code and various data, is installed in the intelligent identification and processing platform for tunnel adverse geological bodies based on an improved leech-based optimization algorithm and near-infrared spectroscopy. During operation, it enables high-speed and automatic access to programs or data. The memory includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0230] To achieve the above objectives, the present invention also provides a computer-readable storage medium, such as... Figure 6 As shown, the computer-readable storage medium stores a control program for an intelligent identification and processing platform for unfavorable geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy. This control program implements the steps of the intelligent identification and processing method for unfavorable geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy. Specific details of these steps have been described above and will not be repeated here.

[0231] In the description of embodiments of the present invention, it should be noted that any process or method description in the flowcharts or otherwise described herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order according to the functions involved, as should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0232] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processing module, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, a “computer-readable medium” can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, the computer-readable medium can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0233] To achieve the above objectives, the present invention also provides a near-infrared spectral drilling detection device, the device comprising a detection sub configured as a cylindrical structure for threaded connection to the tail of the drill bit; and a light source module, a beam splitter module, and at least one optical probe respectively disposed inside the detection sub; wherein, the light source module is used to generate near-infrared light; the beam splitter module is used to decompose the near-infrared light into spectral signals of different frequencies; the optical probe is arranged circumferentially along the detection sub and is used to emit the spectral signals to the borehole wall and receive the reflected spectral signals; the optical probe is an arc-shaped probe and its end face is flush with or slightly protruding from the outer wall of the detection sub;

[0234] The device further includes a spectral driver disposed inside the probe section for controlling spectral acquisition and processing the received spectral signals; a probe brush ring configured to clean the end face of the optical probe when the probe section rotates; and a circumferential elastic damping belt disposed on the probe section for attenuating vibrations during drilling.

[0235] Specifically, in this embodiment of the invention, a near-infrared spectral drilling detection device for identifying adverse geological bodies is provided, comprising: a detection sub, which is a cylindrical structure that provides support and protection for the entire spectral detection system and is threaded to the tail of the drill bit, with its outer diameter being the same as that of the drill bit; multiple infrared lasers, which are used to generate broadband or multi-band near-infrared light sources; and multiple interferometers, which are located adjacent to the infrared lasers and are used to decompose the white light or broadband near-infrared light emitted by the infrared lasers into the original spectrum according to frequency.

[0236] Multiple arc-shaped near-infrared optical probes are distributed along the annular intervals of the detection section, with the probe surface flush with or slightly protruding from the outer wall of the detection section. Each group of probes includes a transmitting optical fiber and a receiving optical fiber, used to illuminate the tunnel borehole wall with the spectrum and receive the spectral signal reflected back from the geological body by reflection.

[0237] Multiple fiber bundles, the fiber bundles being used to transmit spectral signals;

[0238] A spectral driver, fixed inside the detector short circuit, is used to receive transmitted spectral signals and perform synchronous signal acquisition, digital processing, data encoding, and internal storage.

[0239] The probe brush ring is used to simultaneously wipe the probe end face and window surface as the probe section rotates, removing mud and rock powder.

[0240] The interferometer is connected to the infrared laser via an input fiber optic bundle, and to the arc-shaped near-infrared optical probe via an output fiber optic bundle. The spectral driver is directly connected to the interferometer via a sampling fiber. There are two arc-shaped near-infrared optical probes, with the probe end face flush with the outer surface of the short section. The optical path between the probe and the interferometer is realized through a multi-channel fiber optic branch module. The infrared laser is a broadband near-infrared laser.

[0241] The device also includes a circumferential elastic damping belt, which is made of damping rubber pads and is used to attenuate mechanical vibrations caused during drilling.

[0242] Furthermore, such as Figures 7-8 As shown, the near-infrared spectroscopy drilling detection device has a cylindrical structure and is threaded to the drill bit tail. Its outer diameter is the same as that of the drill bit to ensure a seamless connection. The near-infrared spectroscopy drilling detection device includes a detection sub 1, a circumferential elastic damping belt 2, a spectral driver 3, a dynamic data acquisition system 4, and an annular sleeve 5.

[0243] The detection section 1 includes an infrared laser 12, an interferometer 13, an infrared optical probe 14, an optical fiber bundle 15, and a probe brush ring 16. The optical fiber bundle includes an input optical fiber bundle 151, an output optical fiber bundle 152, and a sampling optical fiber bundle 153.

[0244] Infrared laser 12 generates broadband near-infrared light, which is transmitted to interferometer 13 via input fiber bundle 151. Interferometer 13 decomposes the light into raw spectra of different frequencies, which are then transmitted to arc-shaped near-infrared optical probe group 14 via output fiber bundle 152. Spectrum driver 3 is fixed inside probe section 1 and connected to dynamic data acquisition system 3 via sampling fiber 153, responsible for synchronous signal acquisition, digital processing, data encoding, and storage. Probe brush ring 16 rotates with probe section 1, wiping the end face of the optical probe to remove mud and rock powder splashed during drilling, ensuring a clean optical path. Dynamic data acquisition system 4 is connected to spectrum driver 3 to acquire spectral data and identify unfavorable geological bodies (such as soft strata and fault zones) at the tunnel face through analysis algorithms. Circumferential elastic damping strip 2, composed of damping pads, surrounds the outer wall of probe section 1 to attenuate drilling vibrations and ensure measurement stability. This device achieves efficient and accurate identification of unfavorable geological bodies through non-stop drilling and real-time spectral analysis.

[0245] Specifically, in this embodiment, the probe sub 1, the circumferential elastic damping band 2, the spectral actuator 3, and the dynamic data acquisition system 4 are all installed inside the annular sleeve 5, with the sleeve 5 serving as a unified load-bearing and protective outer shell. The probe sub 1 is installed laterally along the inner cavity of the sleeve, with its axis strictly perpendicular to the axis of the sleeve 5 at 90°. The outer side of the probe sub 1 is rigidly fixed to the corresponding lateral mounting position on the inner wall of the sleeve via a positioning shoulder. The circumferential elastic damping band 2 is coaxially fitted around the outer circumference of the probe sub 1, effectively suppressing the micro-vibration of the sub 1 under drilling loads through radial pre-tightening and end limiting. The spectral actuator 3 and the dynamic data acquisition system 4 are respectively arranged along the axial direction of the annular sleeve 5 on its inner wall electronic mounting position and fixed to the annular sleeve 5 with positioning pins. There is no direct rigid connection between the two and the probe sub 1 and the damping band 2, thereby avoiding the transmission of the lateral vibration of the sub 1 to the spectral actuator 3 and the dynamic data acquisition system 4.

[0246] The spectral driver 3 and the dynamic data acquisition system 4 form a single power and data main link: a shielded wire harness and cable clamps are pre-arranged inside the sleeve 5. The dynamic data acquisition system 4 provides power to the spectral driver 3 and issues acquisition and control commands. The driver 3 transmits the synchronously acquired spectral data back to the dynamic data acquisition system 4 in real time via this wire harness for processing, storage, and early warning. The probe section 1 and the spectral driver 3 are connected by a functional flexible connection. The probe section 1 is connected to the spectral driver 3 via an optical fiber bundle. This connection is only an optical-cable coupling and does not form a rigid mechanical fixation. There is no direct electrical connection between the probe section 1 and the dynamic data acquisition system 4; all signals related to the section are first collected by the spectral driver 3 before being sent to the dynamic data acquisition system 4.

[0247] In this embodiment of the invention, there are two optical probes 14. One end of each optical probe 14 contacts the aperture wall, and they are distributed at an annular interval along the detection segment. Each group of probes includes a transmitting optical fiber and a receiving optical fiber. They are responsible for illuminating the decomposed spectrum.

[0248] The other end is connected to an optical fiber conduit 18, which protects the optical fiber bundle 15 from instrument impacts. The optical fiber conduit is divided into a main conduit 181 and a branch conduit 182. A connector branch 183 for flexible conduit segmentation is set in the middle of the optical fiber conduit. The included angle between the main conduit 181 and the branch conduit 182 is 90°. The two transition smoothly at the connector branch 183, and the bending radius of the branch channel must be greater than or equal to the minimum bending radius of the optical fiber bundle 15 to avoid optical fiber signal attenuation or breakage.

[0249] Preferably, the number of optical probes 14 is determined according to monitoring needs and site conditions. In conventional monitoring scenarios, such as... Figure 8 As shown, to reduce costs and simplify installation, this embodiment of the invention uses two optical probes, namely a first optical probe and a second optical probe. These are arranged at 180° intervals via infrared lasers 12. Symmetrically distributed around the outer periphery of the detection device, the end face of each optical probe 14 is flush with or slightly protruding from the outer surface of the detection device to conform to the curvature of the borehole wall and reduce light signal scattering. In complex geological environments or under high-precision monitoring requirements, the number and spacing of probes can be flexibly adjusted according to the tunnel diameter, geological complexity, and construction requirements. For example, in tunnels with larger diameters, it is recommended to increase the number of optical probes 14 to ensure full coverage, or to densify the probes in fault zones to improve identification accuracy. The number of optical probes 14 can be increased to 6 or 8 to ensure 360° coverage of the tunnel borehole wall, thereby improving the coverage and resolution of data acquisition. The surface of the optical probe 14 adopts an arc-shaped design, with a radius of curvature matching the tunnel borehole wall to ensure accurate acquisition of spectral signals. The optical probe 14 is made of high-hardness sapphire or quartz glass, which is wear-resistant and has high light transmittance, suitable for harsh construction environments.

[0250] This invention constructs a first model corresponding to adverse geological features in a tunnel using a method, optimizes the first model based on an improved leech optimization algorithm, and generates first data corresponding to the adverse geological features in the tunnel. The first model is a multidimensional scaling (MDS) model; the first data is data with the optimal combination of dimensionality reduction parameters. Second data corresponding to the tunnel borehole wall is created, generated, and acquired. Based on the first data, the second data is dimensionality-reduced, and third data corresponding to the adverse geological features in the tunnel is generated. The second data is raw near-infrared spectral data; the third data is a set of spectral principal component vectors characterizing the features of the adverse geological features. Fuzzy C-means clustering (FC) is then used. The method (M) involves cluster analysis to process the third data and construct a fourth data corresponding to adverse geological bodies in the tunnel; wherein the fourth data is geological anomaly sample data; based on the fourth data, a corresponding second model is constructed and generated, and based on the second model and combined with real-time near-infrared spectral data, a corresponding fifth data is identified and generated; wherein the second model is a least squares support vector machine (LSSVM) classification model; the fifth data is data on the types of adverse geological bodies in the tunnel, as well as the system, platform and storage medium corresponding to the method, and the near-infrared spectral drilling detection device used for the method, which can realize rapid, accurate and real-time identification of adverse geological bodies in complex tunnel environments.

[0251] In other words, this invention deeply integrates an improved leech optimization algorithm with near-infrared spectroscopy detection technology to construct a complete intelligent identification system for adverse geological bodies in tunnels. By utilizing the optimization algorithm to globally and collaboratively optimize key model parameters throughout the spectral data processing workflow, the dimensionality reduction effect of Multidimensional Scaling Analysis (MDS), the outlier removal capability of fuzzy C-means clustering, and the classification accuracy of Least Squares Support Vector Machine (LSSVM) are improved. Ultimately, this enables rapid and accurate identification and real-time early warning of four types of adverse geological bodies—cavities, faults, fracture zones, and water-rich areas—in complex tunnel construction environments, thereby enhancing the reliability of geological predictions and the safety of tunnel construction operations.

[0252] In other words, the proposed solution first uses a near-infrared spectroscopy detection device to collect spectral data from the tunnel face, then uses an improved leech optimization algorithm to globally optimize the hyperparameters of the multidimensional scaling (MDS) dimensionality reduction model to obtain the optimal solution. Guided by the optimal parameters, dimensionality reduction of the spectral data is performed to extract the principal spectral components containing characteristics of adverse geological bodies. Then, fuzzy C-means clustering (FCM) is used to classify the data, and abnormal spectral samples that significantly deviate from the existing categories are removed based on membership thresholds, classifying the effective samples into four categories of adverse geological bodies: cavities, faults, fracture zones, and water-rich areas. Finally, a classification model based on least squares support vector machine (LSSVM) is constructed to predict the geological body category of new spectral data received in real time during drilling. This invention enables rapid and accurate identification of adverse geological bodies in complex tunnel environments, effectively improving the reliability and real-time performance of geological prediction.

[0253] The near-infrared spectral drilling detection device for identifying adverse geological bodies described in this invention has a detection section that is integrally connected to the drill bit tail via a threaded connection. Inside the section, an infrared laser, an interferometer, two annular near-infrared optical probes, a multi-channel fiber optic bundle, a spectral driver, a probe brush ring, a circumferential elastic damping band, and multiple slip rings and quick-release end caps are arranged sequentially. A dynamic data acquisition system is also included. During drilling, the laser emits broadband near-infrared light, which is decomposed by the interferometer and then irradiated at 320° by the optical probes, which simultaneously receive the reflected signals. The spectral driver drives, acquires, processes, and stores the reflected spectrum in real time. The probe brush ring automatically cleans the probe and window as it rotates. The elastic damping module suppresses drilling vibrations. The dynamic data acquisition system summarizes the various time-series spectral curves and identifies adverse geological bodies such as fractured zones, high-water-bearing layers, and weak surrounding rock online, achieving real-time, accurate geological early warning and positioning at the tunnel face without stopping the drilling rig.

[0254] Furthermore, the improved leech optimization algorithm proposed in this invention, based on the standard leech optimization algorithm, significantly enhances the algorithm's global search capability and robustness in high-dimensional parameter optimization problems by introducing a dynamic host resource management mechanism, a multi-probability selection strategy, and an adaptive replacement rule. Specifically, the algorithm first evaluates the quality of individuals in the population using a fitness function and selects the top K individuals with excellent fitness as the initial host set. Simultaneously, it allocates an initial resource amount R0 to each host to simulate the biological resource consumption process. During iteration, leech individuals choose to utilize a host or conduct random exploration with probability P. The host selection probability Pj is calculated based on a weighted average of the host resource amount Rj and the fitness f(Hj), ensuring that hosts with high resources and high fitness are more likely to be selected, thereby optimizing resource allocation efficiency. Position updates employ a formula combining a step size parameter α that tends towards the host with a random perturbation β, achieving a balance between local utilization and global exploration. Simultaneously, host resources decrease by a fixed amount ΔR and are dynamically removed and replenished when falling below a threshold θ, ensuring continuous optimization of the host set and avoiding the algorithm from falling into local optima traps. Compared with existing methods such as particle swarm optimization (PSO) or genetic algorithm (GA), the algorithm of this invention performs well in the identification of adverse geological bodies in tunnels. It can efficiently optimize the hyperparameters of multidimensional scaling analysis (MDS), fuzzy C-means clustering (FCM), and least squares support vector machine (LSSVM), thereby handling the noise interference and high-dimensional complexity of geological spectral data, improving the convergence speed by 15%-30%, and enhancing the adaptability to uncertain geological environments, ultimately outputting a more stable optimization solution.

[0255] Furthermore, this invention proposes an intelligent identification and processing method for adverse geological bodies in tunnels based on an improved leech optimization algorithm combined with near-infrared spectroscopy. By combining the improved leech optimization algorithm with near-infrared spectroscopy, it achieves accurate real-time identification and classification of adverse geological bodies at the tunnel face. The improved leech optimization algorithm globally optimizes the hyperparameters of the Multidimensional Scaling Analysis (MDS) algorithm to obtain the optimal dimensionality reduction parameters for the near-infrared spectral characteristics of adverse geological bodies. MDS dimensionality reduction processing guided by the optimal parameters can extract high-information principal component vector sets from the original near-infrared spectral data, significantly reducing data dimensionality while retaining key geological features. Then, the fuzzy C-means clustering (FCM) method is introduced to effectively identify and remove abnormal spectral samples, enhancing the purity of the dataset and the generalization ability of the model, and avoiding misjudgment problems caused by noise interference in existing methods. The optimized spectral principal components are divided into four types of geological anomaly samples: fractures, discontinuous faults, fracture zones, and water-rich areas. Finally, least squares support vector machine (LSSVM) is used for training, modeling, and real-time prediction, further improving the recognition accuracy. Compared with existing methods based on traditional spectral analysis or single machine learning, the method of this invention integrates the global optimization capability of optimization algorithms, realizes non-contact, real-time monitoring, reduces safety risks in tunnel construction, improves recognition accuracy and robustness, and reduces the need for manual intervention, thus having higher engineering practicality and economic benefits.

[0256] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for intelligent identification and processing of adverse geological bodies in tunnels based on an improved leech-based optimization algorithm fused with near-infrared spectroscopy, characterized in that, The method includes: A first model corresponding to the unfavorable geological body of the tunnel is constructed, and the first model is optimized based on an improved leech optimization algorithm to generate first data corresponding to the unfavorable geological body of the tunnel; wherein, the first model is a multi-dimensional scale analysis model; the first data is data with the optimal combination of dimensionality reduction parameters; A second set of data corresponding to the tunnel borehole wall is created, generated, and acquired. Based on the first data, the second data is dimensionality-reduced, and a third set of data corresponding to the adverse geological body of the tunnel is generated. The second data is the original near-infrared spectral data, and the third data is the spectral principal component vector set data characterizing the features of the adverse geological body. The third data is processed by cluster analysis using the fuzzy C-means clustering method, and a fourth data corresponding to the unfavorable geological bodies in the tunnel is constructed; wherein, the fourth data is geological anomaly sample data. Based on the fourth data, a corresponding second model is constructed and generated. Based on the second model and combined with real-time near-infrared spectral data, corresponding fifth data is identified and generated. The second model is a least squares support vector machine classification model. The fifth data is data on the types of adverse geological bodies in tunnels.

2. The intelligent identification and processing method for tunnel adverse geological bodies based on an improved leech optimization algorithm fused with near-infrared spectroscopy, as described in claim 1, is characterized in that... The process of constructing a first model corresponding to the unfavorable geological features of the tunnel, optimizing the first model based on an improved leech optimization algorithm, and creating first data corresponding to the unfavorable geological features of the tunnel further includes: A population corresponding to leeches and consisting of multiple individuals is constructed, the population is initialized, and a corresponding first set is created. At the same time, an initial amount of resources is allocated to each host. The first set is an initial host set composed of individuals with the highest fitness ranking in the population. A sixth data point corresponding to the population fitness is calculated and generated. Based on the amount of host resources, a corresponding seventh data point is created by combining the sixth data point. The sixth data point is the selection probability. The seventh data point is the operation control data, including the execution of host utilization operation control or random exploration operation control.

3. The intelligent identification and processing method for tunnel adverse geological bodies based on an improved leech optimization algorithm fused with near-infrared spectroscopy, as described in claim 2, is characterized in that... The calculation and generation of the sixth data corresponding to the population fitness, based on the host resource quantity, and combined with the sixth data to create the corresponding seventh data, also includes: When performing host exploitation operations, the location data of the processed individual is updated based on the positional relationship between the individual leech and the selected host, as well as random perturbations, and the amount of resources of the exploited host is reduced accordingly. When a host's resources fall below a preset threshold, it is removed from the host set, and the individual with the best fitness from the current population is selected for replacement.

4. A method for intelligent identification and processing of tunnel adverse geological bodies based on an improved leech optimization algorithm fused with near-infrared spectroscopy, as described in claim 2 or 3, characterized in that... The calculation and generation of the sixth data corresponding to the population fitness, based on the host resource quantity, and combined with the sixth data to create the corresponding seventh data, also includes: Calculate the selection probability corresponding to the host; the calculation formula is as follows: （2） In the formula, To adjust the parameters; For the first The probability that a host is selected by an individual leech; Describes the size of the host set, that is, the top hosts selected from the current population ranked by fitness. Individuals form the initial host set; The current resource quantity of the j-th host; Indicates the first Each host corresponds to a candidate solution individual; The fitness function represents the host The value that can be taken on.

5. The intelligent identification and processing method for adverse geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy, as described in claim 1, is characterized in that... The method of combining fuzzy C-means clustering to perform cluster analysis on the third data and constructing fourth data corresponding to the adverse geological features of the tunnel also includes: Abnormal spectral sample data is identified and removed, and valid sample data is classified into predefined categories of unfavorable geological bodies; wherein, the predefined categories include cavities, faults, fracture zones and water-rich areas; The process of removing abnormal spectral sample data includes: calculating and generating an eighth set of data corresponding to each sample and relative to each cluster center; identifying samples whose maximum membership data in the third set of data is lower than a preset threshold as abnormal samples and removing them; wherein, the eighth set of data is membership data.

6. A smart identification and processing system for tunnel adverse geological bodies based on an improved leech-based optimization algorithm and near-infrared spectroscopy, characterized in that, The system is applied to the intelligent identification and processing method for tunnel adverse geological bodies based on an improved leech optimization algorithm fused with near-infrared spectroscopy as described in any one of claims 1 to 5, the system comprising: A data construction and generation unit is used to construct a first model corresponding to the unfavorable geological body of the tunnel, optimize the first model based on an improved leech optimization algorithm, and create first data corresponding to the unfavorable geological body of the tunnel; wherein, the first model is a multi-dimensional scale analysis model; the first data is data with the optimal combination of dimensionality reduction parameters; A data creation and generation unit is used to create, generate, and acquire second data corresponding to the tunnel borehole wall, and based on the first data, perform dimensionality reduction processing on the second data to generate third data corresponding to the adverse geological body of the tunnel; wherein, the second data is raw near-infrared spectral data; and the third data is spectral principal component vector set data characterizing the features of the adverse geological body; The data processing and generation unit is used to combine the fuzzy C-means clustering method to perform cluster analysis on the third data and construct and generate fourth data corresponding to the unfavorable geological bodies in the tunnel; wherein, the fourth data is geological anomaly sample data; The data recognition and generation unit is used to construct and generate a corresponding second model based on the fourth data, and to identify and generate corresponding fifth data based on the second model and combined with real-time near-infrared spectral data; wherein the second model is a least squares support vector machine classification model; and the fifth data is tunnel adverse geological body type data.

7. The intelligent identification and processing system for tunnel adverse geological bodies based on an improved leech optimization algorithm and near-infrared spectroscopy as described in claim 6, characterized in that, The data construction and generation unit further includes: The first generation module is used to construct a population corresponding to leeches and consisting of multiple individuals, initialize the population, create a corresponding first set, and allocate initial resources to each host; wherein, the first set is an initial host set composed of individuals with high fitness ranking in the population. The second generation module is used to calculate and generate sixth data corresponding to the population fitness, and based on the host resource quantity, and in combination with the sixth data to create and generate corresponding seventh data; the sixth data is the selection probability; the seventh data is operation control data, including execution host utilization operation control or random exploration operation control; And / or, the second generation module further includes: The first processing module is used to update the location data of the leech individual based on the positional relationship between the leech individual and the selected host and random perturbations when performing host utilization operations, and the amount of resources of the utilized host is reduced accordingly. The second processing module is used to remove a host from the host set when the host's resource quantity is lower than a preset threshold, and to select the individual with the best fitness from the current population for replacement processing. The generation module is used to calculate the selection probability corresponding to the host; the calculation formula is as follows: （2） In the formula, To adjust the parameters; For the first The probability that a host is selected by an individual leech; Describes the size of the host set, that is, the top hosts selected from the current population ranked by fitness. Individuals form the initial host set; The current resource quantity of the j-th host; Indicates the first Each host corresponds to a candidate solution individual; The fitness function represents the host The value of ; And / or, the data processing and generation unit further includes: The type classification module is used to identify and remove abnormal spectral sample data, and classify the valid sample data into predefined categories of unfavorable geological bodies; wherein, the predefined categories include cavities, faults, fracture zones, and water-rich areas; the removal of abnormal spectral sample data includes: calculating and generating an eighth set of data corresponding to each sample and relative to each cluster center, and determining samples whose maximum membership data in the third set of data is lower than a preset threshold as abnormal samples and removing them; wherein, the eighth set of data is membership data.

8. A smart identification and processing platform for tunnel adverse geological bodies based on an improved leech-based optimization algorithm fused with near-infrared spectroscopy, characterized in that, The system includes a processor, a memory, and a control program for an intelligent identification and processing platform for unfavorable geological features in tunnels, based on an improved leech-based optimization algorithm and fusing near-infrared spectroscopy. The processor executes the control program, which is stored in the memory. This control program implements the intelligent identification and processing method for unfavorable geological features in tunnels based on an improved leech-based optimization algorithm and fusing near-infrared spectroscopy, as described in any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a control program for an intelligent identification and processing platform for unfavorable geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy. This control program implements the intelligent identification and processing method for unfavorable geological bodies in tunnels based on an improved leech optimization algorithm fused with near-infrared spectroscopy, as described in any one of claims 1 to 6.

10. A near-infrared spectroscopy drilling detection device for implementing the method of any one of claims 1 to 5, characterized in that, The device includes a probe section configured as a cylinder for threaded connection to the tail of a drill bit; and a light source module, a beam splitter module, and at least one optical probe respectively disposed inside the probe section; wherein, the light source module is used to generate near-infrared light; the beam splitter module is used to decompose the near-infrared light into spectral signals of different frequencies; the optical probe is arranged circumferentially along the probe section and is used to emit the spectral signals to the borehole wall and receive the reflected spectral signals; the optical probe is an arc-shaped probe and its end face is flush with or slightly protruding from the outer wall of the probe section; The device further includes a spectral driver disposed inside the probe section for controlling spectral acquisition and processing the received spectral signals; a probe brush ring configured to clean the end face of the optical probe when the probe section rotates; and a circumferential elastic damping belt disposed on the probe section for attenuating vibrations during drilling.

Images