Tunnel geological exploration-based advanced geological prediction method and system

Abstract

The application relates to the technical field of geological exploration, and discloses an advanced geological prediction method and system based on tunnel geological exploration. The method comprises the following steps: performing data denoising on a plurality of source data sets collected in advance in a tunnel construction area to obtain original data sets; performing preprocessing and multi-level data fusion on the original data sets to obtain fused data sets; performing three-dimensional space modeling and feature extraction on the fused data sets to obtain intelligent geological prediction results; performing multi-factor risk assessment on the intelligent geological prediction results to obtain a geological risk distribution map in front of the tunnel; performing construction scheme optimization on the geological risk distribution map to obtain a construction guidance scheme; and comprehensively analyzing and dynamically updating real-time monitoring data, the construction guidance scheme, the intelligent geological prediction results and the geological risk distribution map to obtain an advanced geological prediction result. The method and system overcome the defects of insufficient prediction accuracy, reliability and accuracy rate of traditional geological prediction methods, and reduce safety risks.

Description

Technical Field

[0001] This invention relates to the field of geological exploration technology, and in particular to an advanced geological prediction method and system based on tunnel geological exploration. Background Technology

[0002] Advanced geological forecasting in tunnel construction is a key technology for ensuring construction safety and efficiency. Traditional advanced geological forecasting methods mainly rely on single detection means, such as ground-penetrating radar and seismic wave methods, combined with experience-based judgment to predict the geological conditions ahead. These methods can detect abnormal geological bodies to a certain extent, but their prediction accuracy and reliability are limited.

[0003] However, information obtained from a single detection method is often incomplete and fails to fully reflect complex geological conditions. Furthermore, traditional methods lack the ability to comprehensively analyze multi-source data, making it difficult to fully utilize the advantages of various detection data. In addition, traditional methods lack dynamic update mechanisms, making it impossible to promptly correct prediction results based on real-time monitoring data during construction. These shortcomings limit the accuracy and timeliness of advanced geological forecasting, increasing safety risks during tunnel construction. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide an advanced geological prediction method and system based on tunnel geological exploration, which solves the defects of traditional geological prediction methods such as insufficient prediction accuracy, reliability and precision, and high safety risks, and significantly improves the comprehensiveness and accuracy of geological prediction.

[0005] This invention provides an advanced geological prediction method based on tunnel geological exploration, comprising: denoising a multi-source dataset pre-collected in the tunnel construction area to obtain an original dataset, wherein the multi-source dataset includes historical geological data, real-time monitoring data, ground-penetrating radar data, seismic wave detection data, advanced drilling data, and infrared detection data; preprocessing and multi-level data fusion of the original dataset to obtain a fused dataset; performing three-dimensional spatial modeling and feature extraction on the fused dataset to obtain intelligent geological prediction results; conducting multi-factor risk assessment on the intelligent geological prediction results to obtain a geological risk distribution map ahead of the tunnel; optimizing the construction plan based on the geological risk distribution map to obtain a construction guidance plan; and comprehensively analyzing and dynamically updating the collected real-time monitoring data, the construction guidance plan, the intelligent geological prediction results, and the geological risk distribution map to obtain advanced geological prediction results.

[0006] This invention also provides an advanced geological prediction system based on tunnel geological exploration, comprising:

[0007] The denoising module is used to denoise the multi-source datasets pre-collected in the tunnel construction area to obtain the original dataset. The multi-source datasets include historical geological data, real-time monitoring data, ground-penetrating radar detection data, seismic wave detection data, advanced drilling data, and infrared detection data.

[0008] The fusion module is used to preprocess the original dataset and perform multi-level data fusion to obtain a fused dataset.

[0009] The extraction module is used to perform three-dimensional spatial modeling and feature extraction on the fused dataset to obtain intelligent geological prediction results;

[0010] The assessment module is used to perform multi-factor risk assessment on the intelligent geological prediction results and obtain a geological risk distribution map in front of the tunnel.

[0011] The optimization module is used to optimize the construction plan based on the geological risk distribution map to obtain a construction guidance plan;

[0012] The update module is used to comprehensively analyze and dynamically update the collected real-time monitoring data, the construction guidance plan, the intelligent geological prediction results, and the geological risk distribution map to obtain advanced geological prediction results.

[0013] The technical solution provided by this invention significantly improves the comprehensiveness and accuracy of geological prediction by comprehensively utilizing various information sources, including historical geological data, real-time monitoring data, and ground-penetrating radar data, through multi-source data fusion technology. Multi-level data processing and analysis workflows, including data cleaning, spatial registration, and scale normalization, effectively improve data quality and comparability, laying a reliable foundation for subsequent analysis. The application of three-dimensional spatial modeling and feature extraction technologies makes the representation of geological structures more intuitive and precise, helping to identify potential risks under complex geological conditions. The probability distribution information contained in the intelligent geological prediction results provides a more scientific basis for risk assessment, making risk management more refined and quantifiable. The multi-factor risk assessment method considers the interaction and weight of geological factors, generating a geological risk distribution map that is closer to the actual situation, which is conducive to developing targeted construction plans. During the construction plan optimization process, steps such as mechanical analysis, support requirement analysis, and numerical simulation ensure the scientific and economical nature of excavation methods and support parameters, improving the safety and efficiency of tunnel construction. A key innovation of this method is the dynamic update mechanism. By comparing and correcting real-time monitoring data with prediction results, geological forecasts can be continuously optimized as construction progresses, improving their timeliness and accuracy. The introduction and weighting of prediction correction indicators make the update process more flexible and precise, enabling rapid responses to changes in actual geological conditions. Overlay analysis and dynamic risk assessment techniques allow for real-time updates of the risk distribution map, providing timely and reliable data for construction decisions. Adaptability analysis and adjustment suggestions for construction plans ensure that the construction process can flexibly respond to changes in geological conditions, enhancing the adaptability and resilience of the construction. Attached Figure Description

[0014] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0015] Figure 1 This is a flowchart of an advanced geological prediction method based on tunnel geological exploration in an embodiment of the present invention;

[0016] Figure 2 This is a schematic diagram of an advanced geological prediction system based on tunnel geological exploration in an embodiment of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0019] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0020] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 , Figure 1 This is a flowchart of an advanced geological prediction method based on tunnel geological exploration according to an embodiment of the present invention, such as... Figure 1 As shown, it includes the following steps:

[0021] S101. Denoise the multi-source datasets pre-collected in the tunnel construction area to obtain the original datasets. The multi-source datasets include historical geological data, real-time monitoring data, ground-penetrating radar data, seismic wave detection data, advanced drilling data, and infrared detection data.

[0022] S102. Preprocess the original dataset and perform multi-level data fusion to obtain a fused dataset;

[0023] S103. Perform three-dimensional spatial modeling and feature extraction on the fused dataset to obtain intelligent geological prediction results;

[0024] S104. Conduct a multi-factor risk assessment on the intelligent geological prediction results to obtain a geological risk distribution map in front of the tunnel;

[0025] S105. Optimize the construction plan based on the geological risk distribution map to obtain a construction guidance plan;

[0026] S106. The collected real-time monitoring data, construction guidance plans, intelligent geological prediction results and geological risk distribution maps are comprehensively analyzed and dynamically updated to obtain advanced geological prediction results.

[0027] Specifically, after pre-collecting multi-source datasets in the tunnel construction area, the data is first denoised. For historical geological data, digitization techniques are used to convert paper documents into digital format and perform hierarchical classification. Real-time monitoring data undergoes filtering algorithms to remove high-frequency noise, such as wavelet transform denoising. Ground-penetrating radar data uses adaptive filtering techniques to enhance effective signals. Seismic wave detection data undergoes spectral analysis to extract characteristic frequencies. Advanced drilling data obtains lithological distribution information through core analysis. Infrared detection data undergoes temperature gradient analysis to identify anomalous areas. Preprocessing the original dataset and multi-level data fusion are key steps. First, spatial registration is performed on data from different sources to ensure the data are in the same coordinate system. Then, scale normalization is performed to unify data at different scales to the same dimension. Outlier detection and interpolation are performed on monitoring data to ensure data continuity. Stratigraphic interface data has features extracted using image segmentation algorithms and is geometrically corrected. Seismic velocity structure data is converted into lithological properties through waveform inversion. Lithological distribution data undergoes spatial statistical analysis to obtain probability distributions. Finally, Bayesian inference was used to fuse the multi-source data to obtain a comprehensive fused dataset.

[0028] The 3D spatial modeling of the fused dataset employs the kriging interpolation method to generate a continuous geological attribute field, followed by meshing to form a discretized 3D geological distribution field. Geological unit classification and boundary refinement are performed on the discretized data to construct a digital twin of the target geological structure. Random fields are generated for geological parameters using Monte Carlo simulation, obtaining multiple sets of geological parameter implementations. Statistical analysis is performed on these implementations to extract statistical characteristics and spatial correlations of the geological parameters. Geostatistical methods such as variogram analysis and kriging interpolation are used to obtain a spatial distribution prediction map of the geological parameters. Finally, cluster analysis is used to identify key risk areas and infer geological conditions, resulting in intelligent geological prediction results containing probability distribution information. When conducting multi-factor risk assessment on the intelligent geological prediction results, geological factors are first decomposed into multiple dimensions, such as lithology, structure, and hydrology. The Analytic Hierarchy Process (AHP) is used to assign weights to each factor, obtaining a factor weight matrix. Principal component analysis is used to extract the combination of controlling factors. Fuzzy comprehensive evaluation is performed on the controlling factors to obtain a risk level score. After normalization of the score, a continuous risk distribution field is generated using kriging interpolation. Finally, the area was divided and the boundaries were optimized to obtain a geological risk distribution map in front of the tunnel.

[0029] When optimizing construction plans based on geological risk distribution maps, the process begins by dividing areas according to risk levels and matching appropriate excavation methods to different areas. Finite element analysis is then performed on the preliminary excavation plan to calculate stress distribution and safety factors. Support requirements are determined based on the analysis results, and support parameters are optimized. Numerical simulations are used to predict the support effect and verify its safety. A detailed construction sequence and process flow are developed, taking into account both excavation and support plans. Finally, construction cycle estimation and resource allocation optimization are performed to obtain a complete construction guidance plan. To achieve dynamic updates, continuous real-time monitoring data collection is necessary. The monitoring data is cleaned and trend analyzed, compared with intelligent geological prediction results, and prediction deviations are calculated. Correction indicators are set based on the deviations, and the original prediction results are weighted and corrected. The updated prediction results are overlaid with the risk distribution map for analysis, resulting in a dynamic risk assessment. Based on the latest risk assessment results, the construction plan is adaptively analyzed and optimized. Finally, the optimized construction plan, updated geological prediction results, and risk distribution are fused to form the latest advanced geological forecast results.

[0030] For example, during tunnel construction, reflected wave data was obtained through ground-penetrating radar. First, the raw data was denoised using wavelet transform to remove high-frequency noise. High-frequency coefficients were then removed through thresholding, yielding the denoised signal. Next, the denoised signal was enhanced using an adaptive filtering algorithm. The enhanced signal was obtained by iteratively optimizing the filter coefficients. Image segmentation was then performed on the enhanced signal to extract stratigraphic interface information. Finally, the processed radar data was fused with other detection data using the Dempster-Shafer evidence theory method. This multi-source data fusion provides a foundation for subsequent geological prediction and risk assessment. In this case, through data processing and fusion, a fault zone was successfully identified 50 meters ahead of the tunnel, approximately 5 meters wide and with a dip angle of 75 degrees. Based on this prediction, the construction plan was adjusted accordingly, employing short-cut excavation and advanced small-diameter pipe support, effectively controlling construction risks.

[0031] By executing the above steps and utilizing multi-source data fusion technology, comprehensively leveraging historical geological data, real-time monitoring data, and ground-penetrating radar data, the comprehensiveness and accuracy of geological prediction are significantly improved. Multi-level data processing and analysis workflows, including data cleaning, spatial registration, and scale normalization, effectively enhance data quality and comparability, laying a reliable foundation for subsequent analysis. The application of 3D spatial modeling and feature extraction technologies makes the representation of geological structures more intuitive and precise, aiding in the identification of potential risks under complex geological conditions. The probability distribution information contained in the intelligent geological prediction results provides a more scientific basis for risk assessment, enabling more refined and quantifiable risk management. The multi-factor risk assessment method considers the interaction and weight of geological factors, generating a geological risk distribution map that more closely reflects actual conditions, facilitating the development of targeted construction plans. During the construction plan optimization process, steps such as mechanical analysis, support requirement analysis, and numerical simulation ensure the scientific and economical nature of excavation methods and support parameters, improving the safety and efficiency of tunnel construction. A key innovation of this method is the dynamic update mechanism. By comparing and correcting real-time monitoring data with prediction results, geological forecasts can be continuously optimized as construction progresses, improving their timeliness and accuracy. The introduction and weighting of prediction correction indicators make the update process more flexible and precise, enabling rapid responses to changes in actual geological conditions. Overlay analysis and dynamic risk assessment techniques allow for real-time updates of the risk distribution map, providing timely and reliable data for construction decisions. Adaptability analysis and adjustment suggestions for construction plans ensure that the construction process can flexibly respond to changes in geological conditions, enhancing the adaptability and resilience of the construction.

[0032] In one specific embodiment, the process of performing step S101 may specifically include the following steps:

[0033] (1) Digitize historical geological data to obtain digitized geological data, and classify the digitized geological data into layers to obtain layered geological information;

[0034] (2) The real-time monitoring data is denoised and filtered to obtain cleaned monitoring data, and the cleaned monitoring data is subjected to time series analysis to obtain monitoring trend data.

[0035] (3) The ground-penetrating radar detection data is signal-enhanced to obtain enhanced radar data, and the enhanced radar data is image-segmented to obtain segmented stratigraphic interface data.

[0036] (4) Perform spectral analysis on the seismic wave detection data to obtain spectral characteristic data, and perform inversion calculation on the spectral characteristic data to obtain the formation velocity structure data;

[0037] (5) Core analysis of the advanced drilling data was performed to obtain lithological distribution data, and spatial interpolation of the lithological distribution data was performed to obtain three-dimensional lithological distribution data;

[0038] (6) Perform temperature gradient analysis on infrared detection data to obtain temperature anomaly area data, and perform cluster analysis on temperature anomaly area data to obtain location information of potential geological anomaly points.

[0039] (7) The layered geological information, monitoring trend data, segmented stratigraphic interface data, stratigraphic velocity structure data, three-dimensional lithological distribution data and potential geological anomaly location information are merged into the original dataset.

[0040] Specifically, historical geological data is digitized, converting paper geological maps and drilling records into digital format. Optical Character Recognition (OCR) technology is used to recognize textual information, and vectorization is employed to extract geological boundaries and structural information. The digitized geological data is then stratified and classified according to lithology, stratigraphic age, and structural characteristics, forming a layered geological information database. Real-time monitoring data, including parameters such as displacement, stress, and water pressure, requires noise reduction and filtering. Wavelet transform is used to remove high-frequency noise, and Kalman filtering is applied to eliminate random errors, resulting in cleaned monitoring data. Time series analysis is performed on the cleaned data, and an Autoregressive Integral Moving Average (ARIMA) model is used to predict the changing trends of the monitoring parameters.

[0041] Ground-penetrating radar (GPR) data is processed to improve the signal-to-noise ratio. Frequency domain filtering and amplitude equalization techniques are used to enhance the radar signal. Then, a region-growing-based image segmentation algorithm is employed to identify stratigraphic interfaces, yielding segmented stratigraphic interface data. Spectral analysis is performed on the seismic wave data, and Fast Fourier Transform (FFT) is used to extract spectral features. Inversion calculations are then performed using the spectral feature data, and the least squares method is used to obtain stratigraphic velocity structure data.

[0042] Advanced drilling data yields lithological distribution information through core analysis. Spatial interpolation of the lithological distribution data is performed using Kriging interpolation to generate three-dimensional lithological distribution data, achieving a spatially continuous representation of lithological information. Infrared detection data identifies anomalous areas through temperature gradient analysis. The spatial gradient of the temperature field is calculated, and a threshold method is used to determine temperature anomaly areas. K-means clustering is applied to the temperature anomaly area data to identify the location information of potential geological anomalies.

[0043] Finally, the layered geological information, monitoring trend data, segmented stratigraphic interface data, stratigraphic velocity structure data, three-dimensional lithological distribution data, and potential geological anomaly location information were integrated into a unified raw dataset. Geographic Information System (GIS) technology was used during the data integration process to establish a unified spatial reference system, ensuring spatial consistency and comparability of various data types.

[0044] Taking a tunnel project as an example, historical geological data included a 1:5000 geological map and data from 20 boreholes. Through digital processing, layered geological information comprising 5 stratigraphic units and 3 faults was generated. Real-time monitoring collected 1000 sets of displacement data, which, after denoising and filtering, yielded 950 valid data points. Ground-penetrating radar (GPR) exploration along the tunnel axis collected 2000m of profile data, identifying 15 major stratigraphic interfaces through signal enhancement and image segmentation. Seismic wave detection data, after spectral analysis and inversion calculations, yielded the P-wave velocity distribution within a depth range of 0-500m, with velocity values ​​varying between 3000-6000m / s. Advanced drilling was conducted in 5 boreholes ahead of the tunnel face; core analysis results, through spatial interpolation, generated a 100m×100m×50m three-dimensional lithological distribution model. Infrared detection covered a 100m radius around the tunnel; temperature gradient analysis identified 3 temperature anomaly areas, and cluster analysis determined 5 potential geological anomalies.

[0045] In one specific embodiment, the process of performing step S102 may specifically include the following steps:

[0046] (1) Spatial registration processing is performed on the layered geological information to obtain the registered geological data, and the scale normalization is performed on the registered geological data to obtain standardized geological data.

[0047] (2) Detect outliers in the monitoring trend data to obtain outlier-marked data, and then perform interpolation on the outlier-marked data to obtain corrected monitoring data;

[0048] (3) Extract the stratigraphic interface from the segmented stratigraphic interface data to obtain interface feature data, and perform geometric correction on the interface feature data to obtain corrected stratigraphic structure data.

[0049] (4) Perform waveform inversion on the formation velocity structure data to obtain velocity field data, and perform geological parameter conversion on the velocity field data to obtain lithological property data;

[0050] (5) Perform spatial statistical analysis on the three-dimensional lithology distribution data to obtain a lithology probability distribution map, and perform threshold segmentation on the lithology probability distribution map to obtain the boundaries of candidate geological bodies;

[0051] (6) Morphological processing is performed on the location information of potential geological anomalies to obtain the outline of the anomaly area, and feature matching is performed on the outline of the anomaly area to obtain the target geological structure information.

[0052] (7) Multi-source data fusion is performed on standardized geological data, corrected monitoring data, corrected stratigraphic structure data, lithological attribute data, candidate geological body boundaries and target geological structure information to obtain a fused dataset.

[0053] Specifically, in advanced geological prediction methods based on tunnel geological exploration, data preprocessing and fusion are key steps. First, spatial registration is performed on layered geological information using affine transformation and control point matching techniques to unify geological data from different sources into the same coordinate system, resulting in registered geological data. Next, the registered geological data undergoes scale normalization using a min-max normalization method to standardize all geological parameters to the [0,1] interval, forming standardized geological data. Then, outlier detection is performed on the monitoring trend data, using the moving window method and the 3σ criterion to identify anomalous data points, resulting in outlier-marked data. Linear interpolation is then used to interpolate the marked outliers, ensuring data continuity and reliability, thus obtaining corrected monitoring data.

[0054] For the segmented stratigraphic interface data, an edge detection algorithm is used to extract the stratigraphic interfaces and obtain interface feature data. Then, a polynomial fitting method is used to geometrically correct the interface feature data, eliminating geometric deformations caused by detection equipment or terrain, and finally obtaining corrected stratigraphic structure data. The stratigraphic velocity structure data undergoes waveform inversion processing. A full waveform inversion algorithm is used to solve the inverse problem of the wave equation through iterative optimization to obtain refined velocity field data. Subsequently, based on a rock physics model, the velocity field data is converted into lithological property data, including parameters such as porosity and permeability.

[0055] Spatial statistical analysis was performed on the three-dimensional lithology distribution data. Variation function analysis and Kriging interpolation were used to generate a lithology probability distribution map. By setting an appropriate probability threshold, the lithology probability distribution map was segmented to obtain the boundaries of candidate geological bodies. Morphological processing techniques, such as dilatation and erosion operations, were used to extract the contours of potential geological anomalies. Feature matching was performed on the extracted contours, and methods such as template matching or feature descriptor comparison were used to identify target geological structural information, such as faults and folds.

[0056] Finally, standardized geological data, corrected monitoring data, corrected stratigraphic structure data, lithological attribute data, candidate geological body boundaries, and target geological structure information are fused from multiple sources. A Bayesian network model is used to consider the uncertainties and correlations of each data source, generating a comprehensive fused dataset.

[0057] Taking a tunnel project as an example, the layered geological information contains five main stratigraphic units. After spatial registration, it achieves precise correspondence with the engineering coordinate system, with the registration error controlled within 0.5m. Scale normalization unifies the lithological strength index to the 0-1 range, facilitating subsequent analysis. Fifteen outlier points were identified in the monitoring trend data; after correction through interpolation, the data continuity was significantly improved. Seven main stratigraphic interfaces were extracted and identified, and geometric correction eliminated the 2-3% deformation caused by topographic undulations. The velocity field data was inverted to obtain the P-wave velocity distribution within a 500m radius around the tunnel, with an accuracy of ±50m / s, and converted into lithological attribute data, including porosity (range 10-30%) and permeability (10^-3-10^-6 m / s). The lithological probability distribution map generated by spatial statistical analysis shows that within a ±50m radius of the tunnel axis, sandstone accounts for 60%, mudstone for 30%, and limestone for 10%. Morphological processing and feature matching identified two main faults and one fold structure. After fusion of multi-source data, a comprehensive dataset containing geological structure, physical properties, and anomaly features was formed, providing comprehensive and reliable data support for subsequent tunnel construction risk assessment and construction scheme optimization.

[0058] In one specific embodiment, the process of performing step S103 may specifically include the following steps:

[0059] (1) Spatial interpolation is performed on the fused dataset to obtain a continuously distributed geological attribute field, and the continuously distributed geological attribute field is divided into grids to obtain a discretized three-dimensional geological distribution field.

[0060] (2) Classify the discretized three-dimensional geological distribution field into geological units to obtain a geological unit partition map, and refine the boundary of the geological unit partition map to obtain the target geological structure twin;

[0061] (3) Assign geological parameters to the target geological structure twin to obtain a parameterized geological twin, and quantify the uncertainty of the parameterized geological twin to obtain the parameter probability density function;

[0062] (4) Generate a random field from the probability density function of the parameters to obtain multiple sets of geological parameter realizations, and perform statistical analysis on the multiple sets of geological parameter realizations to obtain the statistical characteristics of the geological parameters;

[0063] (5) Perform spatial correlation analysis on the statistical characteristics of geological parameters to obtain the variation function of geological parameters, and perform geostatistical interpolation on the variation function of geological parameters to obtain a spatial distribution prediction map of geological parameters.

[0064] (6) Extract abnormal features from the spatial distribution prediction map of geological parameters to obtain potential geological anomaly areas, and calculate the risk probability of potential geological anomaly areas to obtain a geological risk probability distribution map.

[0065] (7) Perform cluster analysis on the geological risk probability distribution map to obtain key risk areas, and infer the geological conditions of key risk areas to obtain intelligent geological prediction results.

[0066] Specifically, Kriging interpolation is used to spatially interpolate the fused dataset to obtain a continuously distributed geological attribute field. This process considers spatial autocorrelation, ensuring the smoothness and accuracy of the interpolation results. Subsequently, an octree algorithm is used to adaptively mesh the continuously distributed geological attribute field, resulting in a discretized three-dimensional geological distribution field. The meshing accuracy is dynamically adjusted according to the gradient of geological attribute changes; the mesh density is higher in complex areas and sparser in uniform areas, ensuring both computational accuracy and efficiency.

[0067] Next, the discretized 3D geological distribution field is classified into geological units using the Support Vector Machine (SVM) algorithm to obtain a geological unit partition map. The SVM algorithm achieves accurate partitioning of different geological units by establishing an optimal separating hyperplane. Then, the level set method is used to refine the boundaries of the geological unit partition map, resulting in a target geological structure twin. The level set method effectively captures complex geological interface morphologies, improving the precision of geological structure representation. Assigning geological parameters to the target geological structure twin is the next crucial step. Here, a Bayesian inference method is used, comprehensively considering prior knowledge and observational data to obtain a parameterized geological twin. To quantify uncertainty, Monte Carlo simulation is used to sample the parameterized geological twin multiple times, generating a parametric probability density function.

[0068] Based on the parametric probability density function, a sequential Gaussian simulation algorithm is used to generate multiple sets of geological parameters. This random field generation method can preserve the spatial structure characteristics of the original data while reflecting the uncertainty of the parameters. Statistical analysis is performed on the multiple sets of geological parameter realizations, calculating statistical characteristics such as mean, variance, skewness, and kurtosis to obtain a statistical description of the geological parameters. Subsequently, spatial correlation analysis is performed on the statistical characteristics of the geological parameters, and the spatial autocorrelation structure of the geological parameters is calculated using the variogram method to obtain the variogram function of the geological parameters. The variogram function describes the spatial variation law of the geological parameters, providing a theoretical basis for subsequent interpolation. Ordinary kriging is used to perform geostatistical interpolation on the variogram function of the geological parameters to obtain a high-precision spatial distribution prediction map of the geological parameters.

[0069] Extracting anomalous features from the spatial distribution prediction map of geological parameters is crucial for identifying potential geological risks. Here, the Local Anomaly Factor (LOF) algorithm is employed to identify potential geological anomaly areas by calculating the local density ratio between a data point and its neighborhood. For each identified potential anomaly area, a logistic regression model is used to calculate the risk probability of the anomaly, resulting in a geological risk probability distribution map. Finally, the DBSCAN density clustering algorithm is applied to this map to identify key risk areas. The DBSCAN algorithm effectively handles irregularly shaped clusters and is suitable for risk area identification under complex geological conditions. For the identified key risk areas, a decision tree algorithm is used to infer geological conditions, comprehensively considering multiple geological factors to obtain the final intelligent geological prediction result.

[0070] For example, in a 5-kilometer-long tunnel project, Kriging interpolation was first performed on the fused dataset to generate a continuous geological attribute field with a resolution of 100m×100m×10m. An octree algorithm was used for adaptive mesh generation, improving the mesh accuracy to 10m×10m×1m in complex areas such as fault zones, while maintaining a resolution of 100m×100m×10m in homogeneous rock areas, ultimately forming a discretized three-dimensional geological distribution field containing approximately 5 million mesh cells. The SVM algorithm classified these mesh cells, identifying five main geological units: sandstone, mudstone, limestone, granite, and metamorphic rock. The level set method refined the cell boundaries, accurately depicting the geometry of two main faults and one complex fold structure. Bayesian inference, combining measured data and regional geological background, assigned parameters such as lithological strength, porosity, and permeability to each geological unit. Monte Carlo simulation generated 1000 sets of parameter samples, constructing a parameter probability density function. Sequential Gaussian simulations, based on these probability density functions, generated 100 sets of geological parameters. Statistical analysis showed that the average uniaxial compressive strength of the sandstone along the tunnel was 50 MPa, with a standard deviation of 5 MPa; the average permeability of the mudstone was 1 × 10⁻⁶ MPa. -6 m / s, which follows a log-normal distribution.

[0071] Variational function analysis showed that the correlation distance of lithological strength was approximately 200m in the horizontal direction and 20m in the vertical direction. Ordinary Kriging interpolation generated a spatial distribution prediction map of geological parameters with a resolution of 10m×10m×1m. The LOF algorithm identified 15 potential geological anomaly areas in the prediction map, of which 3 were located near faults and 2 were located in fold cores. Logistic regression model calculations showed that the geological risk probability of these anomaly areas ranged from 0.1 to 0.8. The DBSCAN algorithm ultimately clustered 4 key risk areas, and decision tree analysis inferred that these areas may have risks of geological hazards such as water inrush, karst collapse, and rock bursts.

[0072] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0073] (1) Decompose the intelligent geological prediction results into geological factors to obtain multidimensional geological factor data, and assign weights to the multidimensional geological factor data to obtain the factor weight matrix.

[0074] (2) Perform hierarchical analysis on the factor weight matrix to obtain the factor importance ranking, and perform threshold screening on the factor importance ranking to obtain the set of key influencing factors;

[0075] (3) Conduct correlation analysis on the set of key influencing factors to obtain the factor correlation coefficient matrix, and extract principal components from the factor correlation coefficient matrix to obtain the combination of main controlling factors;

[0076] (4) Perform fuzzy comprehensive evaluation on the combination of main control factors to obtain risk level scores, and normalize the risk level scores to obtain standardized risk indexes.

[0077] (5) Spatial interpolation of the standardized risk index is performed to obtain a continuous risk distribution field, and the continuous risk distribution field is divided into regions to obtain a risk zoning map;

[0078] (6) Optimize the risk zoning map to obtain a refined risk area, and assign risk level values ​​to the refined risk area to obtain a geological risk distribution map in front of the tunnel.

[0079] Specifically, the intelligent geological prediction results are decomposed into geological factors using Principal Component Analysis (PCA) to break down complex geological conditions into multiple independent geological factors, resulting in multidimensional geological factor data. Subsequently, the Analytic Hierarchy Process (AHP) is used to assign weights to these geological factors, constructing a factor weight matrix. The AHP method ensures the scientific validity and reliability of the weight allocation through expert scoring and consistency checks.

[0080] A hierarchical analysis is performed on the factor weight matrix, and the relative importance of each factor is calculated using the eigenvalue method to obtain a ranking of factor importance. An importance threshold λ is set, and factors with importance greater than λ are selected to form a set of key influencing factors. This process can be expressed by the following formula:

[0081]

[0082] Among them, W i Let a represent the weight of the i-th factor. ijLet represent the importance ratio of the i-th factor to the j-th factor, where m and n are the number of rows and columns, respectively. Correlation analysis is performed on the set of key influencing factors to calculate the Pearson correlation coefficient, resulting in the factor correlation coefficient matrix R. Then, principal component analysis is performed on R to extract principal components with eigenvalues ​​greater than 1, yielding the combination of controlling factors. The formula for principal component analysis is as follows:

[0083]

[0084] Here, F k Let l represent the k-th principal component. ki X is the i-th component of the eigenvector corresponding to the k-th principal component. i 'p' represents the total number of variables, and 'p' represents the original variables. A fuzzy comprehensive evaluation is performed on the combination of controlling factors, establishing membership functions and a fuzzy evaluation matrix. The risk level score is obtained through fuzzy synthesis operations. The score results are then processed using max-min normalization to obtain a standardized risk index.

[0085] Spatial interpolation of the standardized risk index was performed, and the inverse distance weighted method (IDW) was used to generate a continuous risk distribution field. Then, the K-means clustering algorithm was used to divide the continuous risk distribution field into regions, obtaining a preliminary risk zoning map. To optimize the boundaries of the risk zones, an active contour model (Snake algorithm) was applied to refine the boundaries, resulting in refined risk regions. Finally, based on the numerical range of the risk index, risk levels were assigned to the refined risk regions, generating the final geological risk distribution map ahead of the tunnel.

[0086] For example, in a 10-kilometer-long tunnel project, intelligent geological prediction results, through PCA decomposition, yielded eight main geological factors, including lithological strength, groundwater content, fault density, and karst development degree. AHP analysis showed that lithological strength had a weight of 0.3, groundwater content 0.25, fault density 0.2, karst development degree 0.15, and other factors accounted for 0.1. An importance threshold λ = 0.1 was set, and the top five factors were selected as the key influencing factor set. Correlation analysis showed that the correlation coefficient between lithological strength and fault density was -0.6, exhibiting a strong negative correlation. Principal component analysis extracted two principal components, cumulatively explaining 85% of the variance. The first principal component mainly reflected rock mass integrity, while the second principal component mainly reflected hydrogeological conditions.

[0087] Fuzzy comprehensive evaluation established five risk levels (extremely low, low, medium, high, and extremely high), with risk level scores obtained through expert scoring and fuzzy synthesis. After normalization, the risk index ranged from 0 to 1. Spatial interpolation generated a continuous risk distribution field with a resolution of 25m × 25m × 5m. K-means clustering (K = 5) initially divided the tunnel into five risk regions, and the Snake algorithm optimization yielded risk boundaries that better reflected geological realities. The final risk distribution map showed three high-risk regions along the tunnel, located at 2km-3km, 5.5km-6km, and 8km-9km respectively. These regions were mainly affected by fault fracture zones and karst development. Medium-risk regions accounted for 30% of the total length, mainly distributed at 4km-5km and 7km-8km, corresponding to areas rich in groundwater. Low-risk and extremely low-risk regions accounted for 50% of the total length, mainly consisting of intact hard rock sections.

[0088] In one specific embodiment, the process of executing step S105 may specifically include the following steps:

[0089] (1) Divide the geological risk distribution map into regions to obtain risk level zones, and match the excavation methods for the risk level zones to obtain preliminary excavation plans;

[0090] (2) Perform mechanical analysis on the preliminary excavation scheme to obtain stress distribution data, and calculate the safety factor on the stress distribution data to obtain the excavation stability assessment results;

[0091] (3) Analyze the support requirements based on the excavation stability assessment results, obtain support type recommendations, optimize the parameters of the support type recommendations, and obtain a preliminary support scheme;

[0092] (4) Numerical simulation of the preliminary support scheme is performed to obtain support effect prediction data, and the support effect prediction data is verified for safety to obtain optimized support parameters.

[0093] (5) A comprehensive analysis of the optimized support parameters and preliminary excavation scheme was conducted to obtain the construction sequence arrangement, and the construction sequence arrangement was optimized to obtain the preliminary construction procedures.

[0094] (6) Estimate the construction cycle of the preliminary construction procedures to obtain the construction schedule plan, and optimize the resource allocation of the construction schedule plan to obtain the construction guidance plan.

[0095] Specifically, cluster analysis was used to divide the geological risk distribution map into regions, resulting in risk level zones. The K-means clustering algorithm was used to group regions with similar risk indices together, forming different risk level zones. Subsequently, based on the risk level zones and geological conditions, an expert system method was used to match excavation methods for each zone, resulting in preliminary excavation schemes. The next important step is to perform mechanical analysis on the preliminary excavation schemes. The finite element method was used to numerically simulate the excavation process, obtaining stress distribution data. The formula for calculating stress distribution is:

[0096]

[0097] Where, σ ij Represents the stress tensor, u i and u j Let x represent the displacement components in the i and j directions, respectively. j and x i Using spatial coordinates. Based on stress distribution data, the safety factor of the excavation face is calculated. The safety factor is calculated using the Mohr-Coulomb strength criterion:

[0098]

[0099] Here, FS is the safety factor, c is the rock mass cohesion, and σ n Let φ be the normal stress, φ be the internal friction angle, and τ be the shear stress. The excavation stability assessment result is obtained by comparing the calculated safety factor with the set critical value. Support requirements are analyzed based on the excavation stability assessment result. A fuzzy inference system is used, considering factors such as rock mass quality, groundwater conditions, and excavation cross-sectional dimensions, to derive a recommended support type. Then, a particle swarm optimization algorithm is used to optimize the support parameters, resulting in a preliminary support scheme. The objective function of particle swarm optimization can be expressed as:

[0100] f(x) = w1·C(x) + w2·S(x)

[0101] Where f(x) represents the support parameter vector, C(x) is the support cost function, S(x) is the support safety function, and w1 and w2 are weighting coefficients. Numerical simulation of the preliminary support scheme is performed using the three-dimensional finite difference method to simulate the interaction between the support structure and the surrounding rock, obtaining predicted support effect data. The predicted data includes the stress and deformation of the support structure and the displacement of the surrounding rock. The predicted data are then used for safety verification using the limit state method to calculate the bearing capacity of the support structure and the stability of the surrounding rock, obtaining optimized support parameters. The optimized support parameters are then comprehensively analyzed with the preliminary excavation scheme, and the critical path method (CPM) is used to determine the construction sequence. The CPM method considers the duration and interdependencies of each process to calculate the optimal construction sequence. Finally, the construction sequence is optimized using value engineering methods, balancing construction efficiency, safety, and economy to obtain the preliminary construction procedures.

[0102] Finally, the construction period for the preliminary construction procedures is estimated using the PERT (Project Review and Evaluation Technique) method. This method considers the optimal, most likely, and worst-case times for each procedure to calculate the expected duration and standard deviation. The PERT estimation formula is:

[0103]

[0104] Among them, T e For the expected construction period, T o For optimal time, T m For the most likely time, T pThe worst-case scenario is considered. Based on the estimation results, a construction schedule is developed, and linear programming is used to optimize resource allocation, ultimately yielding a detailed construction guidance plan. For example, in a 5-kilometer-long tunnel project, the geological risk distribution map is divided into 5 risk level zones using K-means clustering (K=5). High-risk zones (1.5km-2km section) are excavated using the bench method, medium-risk zones (0-1.5km and 2km-3.5km sections) are excavated using full-face excavation, and low-risk zones (3.5km-5km section) are excavated using smooth blasting. Mechanical analysis shows that the maximum principal stress in the high-risk zone reaches 25MPa, with a safety factor of 1.2, lower than the set critical value of 1.5, requiring reinforced support. The fuzzy inference system recommends using steel arch frames combined with anchor bolts in the high-risk zone, with a preliminary design of H200 type steel arch frames spaced 0.8m apart, coupled with system anchor bolts of L=4m and spaced 1m×1m. After optimization using the particle swarm optimization algorithm, the spacing between the steel arch frames was adjusted to 0.6m, and the anchor bolt length was increased to 4.5m. Numerical simulation results show that the optimized support scheme reduced the maximum deformation of the surrounding rock from 15cm to 8cm, meeting safety requirements. CPM analysis determined the key construction steps of "advanced geological prediction - tunnel face excavation - initial support - secondary lining". The total tunnel construction period estimated by the PERT method is 550 days, with a standard deviation of 30 days. After resource optimization, a scheme of simultaneous advancement of three construction faces was determined, which is expected to shorten the construction period to 500 days.

[0105] In one specific embodiment, the process of executing step S106 may specifically include the following steps:

[0106] (1) Clean the real-time monitoring data to obtain effective monitoring data, and perform trend analysis on the effective monitoring data to obtain the trend of monitoring parameter changes;

[0107] (2) Compare and analyze the changing trends of monitoring parameters with the results of intelligent geological prediction to obtain prediction deviation data, and make threshold judgments on the prediction deviation data to obtain prediction correction indicators.

[0108] (3) The prediction correction index is weighted to obtain the correction weight matrix, and the correction weight matrix is ​​weighted and fused with the intelligent geological prediction results to obtain the updated geological prediction results.

[0109] (4) Overlay analysis of the updated geological prediction results and geological risk distribution map to obtain dynamic risk assessment results, and classify the dynamic risk assessment results into levels to obtain the updated risk distribution map.

[0110] (5) Conduct an adaptability analysis on the updated risk distribution map and construction guidance plan, obtain suggestions for plan adjustment, conduct a feasibility assessment on the suggestions for plan adjustment, and obtain an optimized construction plan.

[0111] (6) The optimized construction plan, the updated geological prediction results and the updated risk distribution map are fused to obtain advanced geological prediction results.

[0112] Specifically, real-time monitoring data undergoes data cleaning. Outlier detection algorithms, such as the Local Anomaly Factor (LOF) method, are used to identify and remove outlier data points. Simultaneously, moving average filtering is used to remove high-frequency noise, resulting in valid monitoring data. Then, time series analysis methods, such as the Autoregressive Integrated Moving Average (ARIMA) model, are applied to the valid monitoring data to analyze the trends of the monitoring parameters and obtain their changing trends. These trends are then compared with the intelligent geological prediction results. The Dynamic Time Warping (DTW) algorithm is used to calculate the similarity between the two, and the prediction deviation data is obtained through differencing. A threshold is set for the prediction deviation data, such as using the 3σ criterion. Deviations exceeding the threshold are marked as items requiring correction, thus obtaining prediction correction indicators. The next crucial step is weighting the prediction correction indicators. Using the Analytic Hierarchy Process (AHP), considering the importance and reliability of each indicator, a judgment matrix is ​​constructed, and eigenvectors are calculated to obtain the correction weight matrix. The correction weight matrix is ​​then weighted and fused with the intelligent geological prediction results using a Bayesian update method, comprehensively considering prior prediction results and new observation data, to obtain updated geological prediction results.

[0113] Subsequently, the updated geological prediction results were overlaid and analyzed with the original geological risk distribution map. Spatial analysis functions of a Geographic Information System (GIS) were used to perform raster calculations to obtain dynamic risk assessment results. Cluster analysis methods, such as the Fuzzy C-means (FCM) clustering algorithm, were applied to the dynamic risk assessment results to classify risk values ​​into different levels, generating an updated risk distribution map. An adaptability analysis was performed on the updated risk distribution map and the existing construction guidance plan. A decision tree algorithm was used to evaluate the applicability of the current construction plan based on the new risk distribution, and suggestions for plan adjustment were derived. Then, multi-criteria decision analysis (MCDA) methods, such as the TOPSIS method, were used to conduct a feasibility assessment of the plan adjustment suggestions. Taking into account factors such as technical feasibility, economy, and safety, an optimized construction plan was obtained.

[0114] Finally, the optimized construction plan, updated geological prediction results, and updated risk distribution map are fused together. Using the Dempster-Shafer evidence theory, considering the reliability and uncertainty of each data source, and through evidence synthesis rules, the final advanced geological prediction results are obtained.

[0115] For example, in an 8-kilometer-long tunnel project, the real-time monitoring system collects data hourly, including parameters such as surrounding rock displacement, groundwater level, and borehole stress. During data cleaning, the LOF algorithm identified 15 outliers, and moving average filtering reduced the noise level by 60%. ARIMA model analysis showed that in the 3km-4km section, surrounding rock displacement exhibited an accelerating growth trend, with the growth rate increasing from 0.5mm / day to 1.2mm / day. The DTW algorithm's prediction results showed a similarity of 0.85 with the actual monitoring data, and difference analysis revealed a significant deviation at 3.5km, exceeding the set 2σ threshold. AHP analysis showed that the weights of surrounding rock displacement, groundwater level, and borehole stress were 0.5, 0.3, and 0.2, respectively. After a Bayesian update, the geological prediction result at 3.5km was adjusted from "medium risk" to "high risk." GIS spatial analysis results indicated that the area of ​​the high-risk zone increased by 15%. FCM clustering reclassified the risk level into 5 levels, with the 3.4km-3.6km section being classified as the highest risk level.

[0116] Decision tree analysis recommended using advanced small-diameter pipe reinforcement in the 3.4km-3.6km section. TOPSIS assessment showed that this proposed adjustment had a feasibility score of 0.82, higher than other options. After integrating various data using Dempster-Shafer evidence theory, a fault fracture zone was ultimately identified in the 3.4km-3.6km section, with a predicted inflow of 300-500 m³ / h. 3 / h, it is recommended to use curtain grouting and double-layer support.

[0117] This invention also provides an advanced geological prediction system based on tunnel geological exploration, such as... Figure 2 As shown, this advanced geological prediction system based on tunnel geological exploration specifically includes:

[0118] The denoising module 201 is used to denoise the multi-source dataset pre-collected in the tunnel construction area to obtain the original dataset. The multi-source dataset includes historical geological data, real-time monitoring data, ground-penetrating radar detection data, seismic wave detection data, advanced drilling data, and infrared detection data.

[0119] The fusion module 202 is used to preprocess the original dataset and perform multi-level data fusion to obtain a fused dataset;

[0120] Extraction module 203 is used to perform three-dimensional spatial modeling and feature extraction on the fused dataset to obtain intelligent geological prediction results;

[0121] The evaluation module 204 is used to perform a multi-factor risk assessment on the intelligent geological prediction results to obtain a geological risk distribution map in front of the tunnel.

[0122] Optimization module 205 is used to optimize the construction plan based on the geological risk distribution map to obtain a construction guidance plan;

[0123] The update module 206 is used to comprehensively analyze and dynamically update the collected real-time monitoring data, the construction guidance plan, the intelligent geological prediction results, and the geological risk distribution map to obtain advanced geological prediction results.

[0124] Through the collaborative work of the aforementioned modules and the use of multi-source data fusion technology, the comprehensiveness and accuracy of geological predictions are significantly improved by integrating historical geological data, real-time monitoring data, and ground-penetrating radar data. Multi-level data processing and analysis workflows, including data cleaning, spatial registration, and scale normalization, effectively enhance data quality and comparability, laying a reliable foundation for subsequent analysis. The application of 3D spatial modeling and feature extraction technologies makes the representation of geological structures more intuitive and precise, aiding in the identification of potential risks under complex geological conditions. The probability distribution information contained in the intelligent geological prediction results provides a more scientific basis for risk assessment, enabling more refined and quantifiable risk management. The multi-factor risk assessment method considers the interaction and weight of geological factors, generating a geological risk distribution map that more closely reflects actual conditions, facilitating the development of targeted construction plans. During the construction plan optimization process, steps such as mechanical analysis, support requirement analysis, and numerical simulation ensure the scientific and economical nature of excavation methods and support parameters, improving the safety and efficiency of tunnel construction. A key innovation of this method is the dynamic update mechanism. By comparing and correcting real-time monitoring data with prediction results, geological forecasts can be continuously optimized as construction progresses, improving their timeliness and accuracy. The introduction and weighting of prediction correction indicators make the update process more flexible and precise, enabling rapid responses to changes in actual geological conditions. Overlay analysis and dynamic risk assessment techniques allow for real-time updates of the risk distribution map, providing timely and reliable data for construction decisions. Adaptability analysis and adjustment suggestions for construction plans ensure that the construction process can flexibly respond to changes in geological conditions, enhancing the adaptability and resilience of the construction.

[0125] The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method for advanced geological prediction based on tunnel geological exploration, characterized in that, include: The multi-source datasets pre-collected in the tunnel construction area are denoised to obtain the original datasets. The multi-source datasets include historical geological data, real-time monitoring data, ground-penetrating radar detection data, seismic wave detection data, advanced drilling data, and infrared detection data. The original dataset is preprocessed and multi-level data fusion is performed to obtain a fused dataset; The fused dataset is used for 3D spatial modeling and feature extraction to obtain intelligent geological prediction results, including: spatial interpolation processing of the fused dataset to obtain a continuously distributed geological attribute field, and meshing the continuously distributed geological attribute field to obtain a discretized 3D geological distribution field; geological unit classification of the discretized 3D geological distribution field to obtain a geological unit partition map, and boundary refinement of the geological unit partition map to obtain a target geological structure twin; geological parameter assignment of the target geological structure twin to obtain a parameterized geological twin, and uncertainty quantification of the parameterized geological twin to obtain a parameter probability density function; and further processing of the parameter probability density function. A random field is generated to obtain multiple sets of geological parameters. Statistical analysis is performed on these multiple sets of geological parameters to obtain statistical characteristics of the geological parameters. Spatial correlation analysis is then performed on the statistical characteristics of the geological parameters to obtain the variogram of the geological parameters. Geostatistical interpolation is then performed on the variogram of the geological parameters to obtain a spatial distribution prediction map of the geological parameters. Anomaly features are extracted from the spatial distribution prediction map of the geological parameters to obtain potential geological anomaly areas. The risk probability of these potential geological anomaly areas is then calculated to obtain a geological risk probability distribution map. Cluster analysis is then performed on the geological risk probability distribution map to obtain key risk areas. Geological conditions are then inferred from these key risk areas to obtain the intelligent geological prediction result. A multi-factor risk assessment is performed on the intelligent geological prediction results to obtain a geological risk distribution map ahead of the tunnel. This includes: decomposing the intelligent geological prediction results into geological factors to obtain multi-dimensional geological factor data; assigning weights to the multi-dimensional geological factor data to obtain a factor weight matrix; performing hierarchical analysis on the factor weight matrix to obtain a ranking of factor importance; filtering the ranking of factor importance using a threshold to obtain a set of key influencing factors; performing correlation analysis on the set of key influencing factors to obtain a factor correlation coefficient matrix; extracting principal components from the factor correlation coefficient matrix to obtain a combination of controlling factors; performing fuzzy comprehensive evaluation on the combination of controlling factors to obtain a risk level score; normalizing the risk level score to obtain a standardized risk index; performing spatial interpolation on the standardized risk index to obtain a continuous risk distribution field; dividing the continuous risk distribution field into regions to obtain a risk zoning map; optimizing the boundary of the risk zoning map to obtain refined risk regions; and assigning risk level values ​​to the refined risk regions to obtain a geological risk distribution map ahead of the tunnel. The construction plan was optimized based on the geological risk distribution map to obtain a construction guidance plan; The collected real-time monitoring data, the construction guidance plan, the intelligent geological prediction results, and the geological risk distribution map are comprehensively analyzed and dynamically updated to obtain advanced geological prediction results.

2. The advanced geological prediction method based on tunnel geological exploration according to claim 1, characterized in that, The step of denoising the multi-source dataset pre-collected in the tunnel construction area to obtain the original dataset, wherein the multi-source dataset includes historical geological data, real-time monitoring data, ground-penetrating radar detection data, seismic wave detection data, advanced drilling data, and infrared detection data, includes: Historical geological data are digitized to obtain digitized geological data, and then the digitized geological data is classified into layers to obtain layered geological information. The real-time monitoring data is denoised and filtered to obtain cleaned monitoring data, and time series analysis is performed on the cleaned monitoring data to obtain monitoring trend data. Signal enhancement processing is performed on the ground-penetrating radar detection data to obtain enhanced radar data, and image segmentation is performed on the enhanced radar data to obtain segmented stratigraphic interface data. Spectral analysis was performed on the seismic wave detection data to obtain spectral characteristic data, and inversion calculations were performed on the spectral characteristic data to obtain the formation velocity structure data; Core analysis was performed on the advanced drilling data to obtain lithological distribution data, and spatial interpolation was performed on the lithological distribution data to obtain three-dimensional lithological distribution data. Temperature gradient analysis was performed on infrared detection data to obtain temperature anomaly area data, and cluster analysis was performed on temperature anomaly area data to obtain location information of potential geological anomalies. The layered geological information, the monitoring trend data, the segmented stratigraphic interface data, the stratigraphic velocity structure data, the three-dimensional lithological distribution data, and the location information of potential geological anomalies are merged into the original dataset.

3. The advanced geological prediction method based on tunnel geological exploration according to claim 2, characterized in that, The steps of preprocessing and multi-level data fusion of the original dataset to obtain the fused dataset include: Spatial registration processing is performed on the layered geological information to obtain registered geological data, and scale normalization is performed on the registered geological data to obtain standardized geological data. Anomaly detection is performed on the monitoring trend data to obtain anomaly-marked data, and the anomaly-marked data is interpolated to obtain corrected monitoring data; The segmented stratigraphic interface data is extracted to obtain interface feature data, and the interface feature data is geometrically corrected to obtain corrected stratigraphic structure data. Waveform inversion is performed on the aforementioned formation velocity structure data to obtain velocity field data, and geological parameter conversion is performed on the velocity field data to obtain lithological property data; Spatial statistical analysis is performed on the three-dimensional lithology distribution data to obtain a lithology probability distribution map, and threshold segmentation is performed on the lithology probability distribution map to obtain the boundaries of candidate geological bodies; Morphological processing is performed on the location information of the potential geological anomalies to obtain the outline of the anomaly region, and feature matching is performed on the outline of the anomaly region to obtain the target geological structure information. The standardized geological data, the corrected monitoring data, the corrected stratigraphic structure data, the lithological attribute data, the candidate geological body boundaries, and the target geological structure information are fused from multiple sources to obtain the fused dataset.

4. The advanced geological prediction method based on tunnel geological exploration according to claim 1, characterized in that, The steps for optimizing the construction plan based on the geological risk distribution map to obtain the construction guidance plan include: The geological risk distribution map is divided into regions to obtain risk level zones, and excavation methods are matched for the risk level zones to obtain preliminary excavation plans; A mechanical analysis was performed on the preliminary excavation scheme to obtain stress distribution data, and a safety factor was calculated on the stress distribution data to obtain the excavation stability assessment results. The support requirements are analyzed based on the excavation stability assessment results to obtain support type recommendations. The parameters of the support type recommendations are then optimized to obtain a preliminary support scheme. Numerical simulation was performed on the preliminary support scheme to obtain support effect prediction data, and the support effect prediction data was then subjected to safety verification to obtain optimized support parameters. A comprehensive analysis of the optimized support parameters and the preliminary excavation scheme is conducted to obtain the construction sequence arrangement, and the construction sequence arrangement is further optimized to obtain the preliminary construction procedures. The construction period of the preliminary construction procedures is estimated to obtain a construction schedule plan, and the resource allocation of the construction schedule plan is optimized to obtain the construction guidance scheme.

5. The advanced geological prediction method based on tunnel geological exploration according to claim 4, characterized in that, The step of comprehensively analyzing and dynamically updating the collected real-time monitoring data, the construction guidance plan, the intelligent geological prediction results, and the geological risk distribution map to obtain advanced geological prediction results includes: The real-time monitoring data is cleaned to obtain effective monitoring data, and trend analysis is performed on the effective monitoring data to obtain the changing trends of the monitoring parameters. The trend of the monitoring parameters is compared and analyzed with the intelligent geological prediction results to obtain prediction deviation data, and the prediction deviation data is judged by a threshold to obtain prediction correction index. The prediction correction index is weighted to obtain a correction weight matrix, and the correction weight matrix is ​​weighted and fused with the intelligent geological prediction result to obtain the updated geological prediction result. The updated geological prediction results are overlaid with the geological risk distribution map to obtain dynamic risk assessment results, and the dynamic risk assessment results are classified into levels to obtain an updated risk distribution map. An adaptability analysis is performed on the updated risk distribution map and the construction guidance plan to obtain suggestions for plan adjustment. A feasibility assessment is then conducted on the suggestions for plan adjustment to obtain an optimized construction plan. The optimized construction plan, the updated geological prediction results, and the updated risk distribution map are fused to obtain the advanced geological prediction results.

6. A tunnel geological exploration-based advanced geological prediction system, used to execute the tunnel geological exploration-based advanced geological prediction method as described in any one of claims 1 to 5, characterized in that, include: The denoising module is used to denoise the multi-source datasets pre-collected in the tunnel construction area to obtain the original dataset. The multi-source datasets include historical geological data, real-time monitoring data, ground-penetrating radar detection data, seismic wave detection data, advanced drilling data, and infrared detection data. The fusion module is used to preprocess the original dataset and perform multi-level data fusion to obtain a fused dataset; An extraction module is used to perform three-dimensional spatial modeling and feature extraction on the fused dataset to obtain intelligent geological prediction results. This includes: performing spatial interpolation on the fused dataset to obtain a continuously distributed geological attribute field; dividing the continuously distributed geological attribute field into a grid to obtain a discretized three-dimensional geological distribution field; classifying the discretized three-dimensional geological distribution field into geological units to obtain a geological unit partition map; refining the boundaries of the geological unit partition map to obtain a target geological structure twin; assigning geological parameters to the target geological structure twin to obtain a parameterized geological twin; quantifying the uncertainty of the parameterized geological twin to obtain a parameter probability density function; and then... The system generates multiple sets of geological parameters using a random field, performs statistical analysis on these parameters to obtain statistical characteristics, analyzes spatial correlation to obtain variability functions, interpolates these variability functions using geostatistics to obtain a spatial distribution prediction map of the geological parameters, extracts anomaly features from the spatial distribution prediction map to identify potential geological anomaly areas, calculates the risk probability of these areas to obtain a geological risk probability distribution map, performs cluster analysis on the geological risk probability distribution map to identify key risk areas, and infers the geological conditions of these key risk areas to obtain the intelligent geological prediction result. The evaluation module is used to perform multi-factor risk assessment on the intelligent geological prediction results to obtain a geological risk distribution map ahead of the tunnel. This includes: decomposing the intelligent geological prediction results into geological factors to obtain multi-dimensional geological factor data; assigning weights to the multi-dimensional geological factor data to obtain a factor weight matrix; performing hierarchical analysis on the factor weight matrix to obtain a ranking of factor importance; performing threshold filtering on the ranking of factor importance to obtain a set of key influencing factors; performing correlation analysis on the set of key influencing factors to obtain a factor correlation coefficient matrix; extracting principal components from the factor correlation coefficient matrix to obtain a combination of controlling factors; performing fuzzy comprehensive evaluation on the combination of controlling factors to obtain a risk level score; normalizing the risk level score to obtain a standardized risk index; performing spatial interpolation on the standardized risk index to obtain a continuous risk distribution field; dividing the continuous risk distribution field into regions to obtain a risk zoning map; optimizing the boundary of the risk zoning map to obtain refined risk regions; and assigning risk level values ​​to the refined risk regions to obtain a geological risk distribution map ahead of the tunnel. The optimization module is used to optimize the construction plan based on the geological risk distribution map to obtain a construction guidance plan; The update module is used to comprehensively analyze and dynamically update the collected real-time monitoring data, the construction guidance plan, the intelligent geological prediction results, and the geological risk distribution map to obtain advanced geological prediction results.

Images