Hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling

Abstract

The application relates to a hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling, and belongs to the technical field of hydrogeological analysis. The method comprises the following steps: obtaining multi-source data of a hydrogeological structure and performing preprocessing to obtain standardized multi-source data of the hydrogeological structure; integrating the standardized multi-source data of the hydrogeological structure to obtain key multi-source data of the hydrogeological structure; the key multi-source data of the hydrogeological structure comprises special geological body spatial distribution data, underground facility-stratum spatial relationship data and stratum-structure three-dimensional framework data; an initial three-dimensional geological model is constructed based on the key multi-source data of the hydrogeological structure, and the initial three-dimensional geological model is dynamically updated to obtain a dynamic three-dimensional geological model; a representation method of the dynamic three-dimensional geological model is optimized to obtain a high-precision three-dimensional model; and mechanism analysis and correlation mining are carried out based on the high-precision three-dimensional model. The application can improve the accuracy of hydrogeological structure analysis.

Description

Technical Field

[0001] This invention belongs to the field of hydrogeological analysis technology, specifically relating to a hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling. Background Technology

[0002] In hydrogeological research and engineering applications, accurate understanding of geological structures is crucial. However, existing technologies have many shortcomings. Traditional methods often lack key geological information in their 3D models, such as incomplete coverage of faults, fracture zones, underground structures, and adverse geological formations. These elements are critical for hydraulic connections, construction risks, and seepage control; their absence prevents the model from accurately reflecting the complexity of the geological structure. Furthermore, it is difficult to view stratigraphic profiles after model cross-sectioning, failing to demonstrate the relationship between strata, geological structures, and underground structures. This lack of detail limits its application in engineering analysis.

[0003] Existing technologies are mostly static modeling technologies with weak dynamic updating capabilities. They are unable to adjust model parameters in real time according to project progress or geological events, and therefore cannot reflect the real-time state of the study area. Especially in complex scenarios such as key areas, it is difficult to simulate the dynamic evolution of groundwater and ignores the disturbance of engineering construction on the groundwater recharge-drainage relationship and the feedback effect with geological deformation.

[0004] In areas with complex hydrogeological conditions, such as key engineering hubs, the heterogeneous anisotropy of the permeability characteristics of rock masses and structural surfaces is difficult to accurately characterize, leading to large deviations in seepage simulations. Furthermore, existing modeling techniques cannot effectively integrate multi-source data for refined modeling of complex geological bodies, making it difficult to support construction risk early warning and seepage control analysis. In addition, existing methods have a vague understanding of the relationship between geological deformation and groundwater dynamics, focusing only on changes in effective stress in the rock mass and ignoring their mutual feedback relationship. This results in an unclear understanding of the deformation mechanism, making it impossible to support accurate prediction and the formulation of engineering safety countermeasures. Summary of the Invention

[0005] To address the shortcomings of existing technologies, such as insufficient data coverage and refinement, inadequate dynamism and timeliness, insufficient adaptability to complex geological conditions, and insufficient mechanism analysis and correlation mining, this invention provides a hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling, comprising the following steps: S1. Obtain multi-source hydrogeological structure data; the multi-source hydrogeological structure data includes basic geological data, geophysical exploration data, topographic and geomorphological data, underground facility data and groundwater dynamic monitoring data; Specifically, the basic geological data includes borehole data and geological mapping data; the borehole data includes stratum depth and physical and mechanical parameters of core samples; the geological mapping data includes surface lithological outcrops, geological boundaries, and stratum attitude; the geophysical exploration data includes ground-penetrating radar data and transient electromagnetic method data; the topographic data includes digital elevation model data and remote sensing image data; the underground facility data includes design drawings, construction logs, and as-built data of underground facilities; and the groundwater dynamic monitoring data includes water level monitoring data and water quality testing data.

[0006] S2. Preprocess the multi-source hydrogeological structure data to obtain standardized multi-source hydrogeological structure data; Furthermore, an adaptive wavelet threshold denoising algorithm is used to denoise the ground-penetrating radar data while retaining the effective reflection signal.

[0007] S3. The geological information contained in multi-source data is scattered and redundant. It is necessary to extract key information and effectively integrate it to accurately reflect the characteristics of hydrogeological structure. This invention integrates standardized hydrogeological structure multi-source data to obtain key hydrogeological structure multi-source data. Key hydrogeological structure multi-source data includes: spatial distribution data of special geological bodies, spatial relationship data of underground facilities-strata, and three-dimensional framework data of strata-structure.

[0008] Furthermore, special geological bodies, such as faults, fracture zones, karst caves, and aquifers, are key factors affecting groundwater seepage and the safety of underground engineering. Their spatial location and properties must be accurately determined to avoid "ignoring critical hidden dangers" in modeling. Spatial distribution data of special geological bodies is used as a hard constraint in step S4 of the model to prevent misclassification of this area as normal rock mass during subsequent mesh assignment. The spatial distribution data of special geological bodies includes: the spatial coordinates, strike, and dip angle of faults; the center coordinates and volume of karst caves; and the spatial distribution range and permeability coefficient of aquifers. The specific process for obtaining the spatial distribution data of special geological bodies is as follows: For transient electromagnetic data in standardized hydrogeological structure multi-source data, resistivity tomography is used to invert the underground resistivity structure and obtain resistivity profile map; a resistivity difference threshold is set to identify resistivity abrupt change zones, and the fault location is determined by cross-validation with stratigraphic abrupt change points in geological mapping data and core fracture sections in borehole data, and the spatial coordinates, strike, dip, and dip angle of the fault are output. A convolutional neural network model was trained using ground-penetrating radar (GPR) data. The model's inputs were the amplitude, frequency, and travel time of the reflected wave, labeled as "cave" or "non-cave." GPR data from standardized hydrogeological structure multi-source data was then input into the trained model to achieve automatic identification and classification of caves. The model was then validated using transient electromagnetic method data to ultimately obtain the center coordinates and volume of the caves. Based on the low resistivity anomaly zone derived from transient electromagnetic method data inversion in standardized hydrogeological structure multi-source data, combined with the groundwater level depth in borehole data and the water level fluctuation segment in groundwater dynamic monitoring data, the top and bottom plate depths of the aquifer are determined, and the spatial distribution range and permeability coefficient of the aquifer are obtained.

[0009] Existing models are prone to misalignment between underground facilities and geological strata, such as mismatches between the tunnel model and the actual geological interface, abnormal topological relationships, or facilities penetrating strata without being identified. These models fail to reflect the interaction between the actual engineering project and the geological structure, leading to distorted engineering risk assessments. The specific process by which this invention generates underground facility-stratum spatial relationship data is as follows: Based on the design drawings and actual excavation contour data from the construction logs of the underground facilities, the design coordinates of the underground facilities were converted into the Gauss-Kruger coordinate system using the seven-parameter method to obtain the three-dimensional contour of the underground facilities. Based on the basic geological data, the stratigraphic surface was constructed using borehole data and geological mapping data. The vertical distance from each point on the three-dimensional contour of the underground facilities to the interface of the adjacent stratigraphic layers was calculated to determine whether the underground facilities were located inside the stratigraphic layers. The presence of abnormal topological relationships such as penetration or overlap between the underground facilities and the stratigraphic layers was checked. Finally, the spatial relationship data between the underground facilities and the stratigraphic layers was obtained, including the name of the underground facilities, the name of the stratigraphic layer, the distance between the underground facilities and the stratigraphic interface, and the topological relationship status.

[0010] The specific process of generating stratigraphic-structural 3D framework data is as follows: Based on standardized basic geological data, using stratigraphic depth information from borehole data and stratigraphic attitude information from geological mapping data, a layer modeling technique in 3D geological modeling software is used to construct an initial stratigraphic framework. Fault spatial coordinates, strike, dip, and dip angle data are used as structural constraints, and a fault cutting algorithm is employed to correct the initial stratigraphic framework. For faults identified by abrupt changes in stratigraphic attitude and fractured sections in core samples, Boolean operations are used to spatially cut the fault model and stratigraphic model, causing corresponding displacements or faultings of the strata on both sides of the fault, forming a complex stratigraphic structure. Digital elevation model data from topographic data is used as surface boundary conditions to trim the top of the generated stratigraphic-structural framework, ultimately obtaining the 3D stratigraphic-structural framework data.

[0011] S4. Construct an initial three-dimensional geological model based on multi-source data of key hydrogeological structures, and dynamically update the initial three-dimensional geological model to obtain a dynamic three-dimensional geological model. Furthermore, S41, constructing an initial three-dimensional geological model: Based on the stratigraphic-structural three-dimensional framework data, the study area is spatially discretized using a grid partitioning algorithm to generate a three-dimensional grid model containing stratigraphic units and fault structures; the fault, karst cave, and aquifer attribute parameters in the spatial distribution data of special geological bodies are assigned to the corresponding grid units through a grid assignment algorithm; combined with the spatial relationship data between underground facilities and stratigraphy, a three-dimensional solid model of underground facilities is embedded in the three-dimensional grid model, and its spatial distance and topological relationship with the surrounding strata are labeled to form an initial three-dimensional geological model; S42. Dynamically update the initial 3D geological model: Establish a correlation mechanism between groundwater dynamic monitoring data and model parameters, compare the water level monitoring data with the water level calculated by the model, and use a particle swarm optimization hybrid algorithm to invert and adjust the aquifer permeability coefficient and porosity; when new borehole data or geophysical exploration data are acquired, the consistency between the new data and the existing model is judged by root mean square error analysis. If the error exceeds the threshold, the local grid reconstruction and parameter reassignment process of the model is triggered; combined with the long-term dynamic information of vegetation cover change and water system distribution migration in remote sensing image data, the potential impact of surface environment changes on groundwater recharge is identified by machine learning classification algorithm, and the model boundary conditions are corrected to realize the dynamic update of the initial 3D geological model, and finally obtain a dynamic 3D geological model.

[0012] S5. Optimize the characterization method of the dynamic three-dimensional geological model to obtain a high-precision three-dimensional model.

[0013] Furthermore, an improved moving least squares surface fitting algorithm is introduced to optimize the dynamic three-dimensional geological model; the calculation formula of the improved moving least squares surface fitting algorithm is expressed as follows: , , in, Represents the fitted stratigraphic interface; Represents global basis functions; Represents local basis functions; n and e are the number of global basis functions and the number of local basis functions, respectively; , Let them represent the first coefficient to be determined and the second coefficient to be determined, respectively. Represents the region weighting function; Indicator function representing the fault influence zone; , These represent the first weight gain coefficient and the second weight gain coefficient, respectively. By employing a multi-scale grid subdivision strategy, fine grids are used for key areas such as aquifers, faults, and fracture zones, while coarse grids are used for homogeneous rock mass areas. A visualization and hierarchical mechanism for special geological bodies is established, assigning different display priorities and attribute labeling rules based on the volume of karst caves and the scale of faults, resulting in a high-precision 3D model.

[0014] S6. Mechanism analysis and correlation mining based on high-precision three-dimensional models.

[0015] Furthermore, a high-precision 3D model is used to visually demonstrate the spatial distribution of strata, the development characteristics of special geological bodies, and the interaction between underground facilities and geological structures. Using the spatial analysis module built into the model, the effective water storage volume of aquifers and key parameters such as fault hydraulic conductivity are calculated. Combined with groundwater dynamic monitoring data, the correlation between water level changes and strata permeability and the distribution of special geological bodies is analyzed.

[0016] The advantages of this invention are: This invention integrates multi-source data from basic geology, geophysical exploration, topography, underground infrastructure data, and groundwater dynamic monitoring. Through preprocessing techniques such as Kriging interpolation, adaptive wavelet thresholding for denoising, and unified coordinate transformation, it accurately extracts key information, avoiding data redundancy or missing crucial elements. After constructing an initial 3D model based on this key data, a particle swarm optimization hybrid algorithm is used to invert aquifer parameters. New data triggers local grid reconstruction and remote sensing dynamically corrects boundary conditions, enabling real-time model updates that accurately reflect the spatiotemporal evolution of hydrogeological structures, overcoming the limitations of traditional static modeling. An improved moving least squares surface fitting method is also applied. The model optimizes the characterization of stratigraphic interfaces and boundaries of special geological bodies; employs multi-scale grid subdivision to balance accuracy and efficiency; and significantly improves model refinement and readability by incorporating a visualization and hierarchical mechanism for special geological bodies. It visually demonstrates the interactions between strata, special geological bodies, and underground infrastructure using a high-precision model. An integrated spatial analysis module calculates key parameters such as the effective water storage volume of aquifers and the hydraulic conductivity of faults. Combined with dynamic groundwater data, it quantitatively analyzes the correlation between water level changes and stratigraphic permeability and the distribution of special geological bodies, achieving a progression from qualitative visualization to quantitative analysis and then to trend prediction, providing precise and comprehensive support for research and engineering applications. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0018] Figure 1 This is a flowchart of the steps of the method of the present invention; Figure 2 Comparison images of geological bodies identified using different methods; Figure 3This is a comparison chart of the accuracy of the dynamic model of aquifer permeability coefficient in this invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0020] Example 1 In this embodiment, as Figure 1 As shown, this invention provides a hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling, the specific steps of which include: S1. Obtain multi-source hydrogeological structure data; the multi-source hydrogeological structure data includes basic geological data, geophysical exploration data, topographic data, underground facility data and groundwater dynamic monitoring data.

[0021] Specifically, the basic geological data includes borehole data and geological mapping data; the borehole data includes the burial depth of the strata and the physical and mechanical parameters of the core samples (such as porosity and permeability) obtained through borehole exploration; the geological mapping data includes surface lithological outcrops, geological boundaries, and strata attitude (strike, dip, dip angle) obtained through geological mapping. The geophysical exploration data includes ground-penetrating radar data and transient electromagnetic method data; the ground-penetrating radar data can provide the electromagnetic reflection characteristics of the underground medium, reflecting different lithological interfaces and cavities, etc.; the transient electromagnetic method data can present the resistivity distribution at different underground depths, which can be used to determine the location of aquifers and fault fracture zones. The topographic data includes digital elevation model data and remote sensing image data; the digital elevation model data reflects the topographic relief, slope, and aspect; the remote sensing image data provides information such as surface vegetation cover, water system distribution, and geomorphic unit division. The underground facility data includes design drawings, construction logs, and completion data of underground facilities (such as tunnels and caverns); the groundwater dynamic monitoring data includes water level monitoring data and water quality testing data.

[0022] S2. Preprocess the multi-source hydrogeological structure data to obtain standardized multi-source hydrogeological structure data. The standardized multi-source hydrogeological structure data meets the requirements of subsequent integration and modeling. For example, standardized transient electromagnetic method data can be directly used to invert resistivity structure, and underground facility data with unified coordinates can be directly embedded into the model to avoid modeling failure or accuracy reduction due to data problems.

[0023] Kriging interpolation is used to fill in the basic geological data to compensate for the spatial discontinuity caused by the sparse borehole locations, thus forming continuous stratigraphic interfaces and attribute distributions. To address the limitation of traditional Kriging interpolation, which only considers spatial correlation, a time weighting factor is introduced to interpolate the basic geological data. The formula is as follows: , in, Representing spacetime coordinates The interpolation result at the location; Represents the spatiotemporal weighting factor, satisfying ; This represents known sample point data; This represents the Lagrange multiplier, used to ensure unbiased estimation; furthermore, statistical analysis is performed on the physical and mechanical parameters of the core samples, using 3... Criteria or box plot methods are used to identify and remove outliers; Terrain correction algorithms and background noise filtering are applied to transient electromagnetic method data to improve the accuracy of resistivity data. An adaptive wavelet thresholding denoising algorithm is used to denoise the ground-penetrating radar data while retaining the effective reflection signal. The adaptive wavelet thresholding algorithm, based on the signal-to-noise ratio, is expressed by the following formula: , in, The threshold value represents the threshold value of the k-th wavelet coefficient in the j-th layer. This represents the standard deviation of the noise at layer j; This represents the number of wavelet coefficients in the j-th layer; This represents the signal-to-noise ratio of the k-th coefficient in the j-th layer; This represents the adjustment factor, with a value range of 0.1-0.5; Represents an exponential function; Radiometric calibration and atmospheric correction are performed on remote sensing image data to eliminate the effects of atmospheric scattering and sensor errors; geometric fine correction is performed by combining ground control points with polynomial fitting or rational function models to match the remote sensing image data with the target coordinate system; Gaussian filtering or median filtering is used to smooth digital elevation model data to remove local abnormal terrain. The design coordinates of underground facilities are converted into a unified Gauss-Kruger projection coordinate system through coordinate transformation parameters; time series interpolation is used to fill short-term missing values ​​in groundwater monitoring data, and abnormal data caused by instrument failure are detected and eliminated based on sliding standard deviation. The final result is standardized multi-source hydrogeological structure data.

[0024] S3. The geological information contained in multi-source data is scattered and redundant. It is necessary to extract key information and effectively integrate it to accurately reflect the characteristics of hydrogeological structure. This invention integrates standardized hydrogeological structure multi-source data to obtain key hydrogeological structure multi-source data. Key hydrogeological structure multi-source data includes: spatial distribution data of special geological bodies, spatial relationship data of underground facilities-strata, and three-dimensional framework data of strata-structure.

[0025] Special geological bodies, such as faults, fracture zones, karst caves, and aquifers, are key factors affecting groundwater seepage and the safety of underground engineering projects. Their spatial location and properties must be accurately determined to avoid overlooking critical hidden dangers during modeling. Spatial distribution data of special geological bodies is used as a hard constraint in step S4 of the model to prevent misclassification of this area as normal rock mass during subsequent mesh assignment.

[0026] The spatial distribution data of the special geological bodies include: the spatial coordinates, strike, and dip angle of faults; the center coordinates and volume of karst caves; and the spatial distribution range and permeability coefficient of aquifers. S31. For transient electromagnetic data in standardized hydrogeological structure multi-source data, resistivity tomography is used to invert the underground resistivity structure and obtain a resistivity profile. A resistivity difference threshold is set to identify resistivity abrupt change zones. Combined with stratigraphic abrupt change points in geological mapping data and core fracture sections in borehole data, cross-validation is performed to determine the fault location and output the spatial coordinates, strike, dip, and dip angle of the fault. A convolutional neural network model was trained using ground-penetrating radar (GPR) data. The model's inputs were the amplitude, frequency, and travel time of the reflected wave, labeled as "cave" or "non-cave." GPR data from standardized multi-source hydrogeological structure data were then input into the trained model to achieve automatic identification and classification of caves. The model was then validated using transient electromagnetic method (TEM) data, which showed that caves exhibited high resistivity anomalies. Finally, the center coordinates and volume of the caves were obtained. Based on the low resistivity anomaly zone derived from transient electromagnetic method data inversion of standardized hydrogeological structure multi-source data, and combined with the groundwater level depth from borehole data and the water level fluctuation range from groundwater dynamic monitoring data, the top and bottom depths of the aquifer are determined, and the spatial distribution range and permeability coefficient of the aquifer are output. In one embodiment, resistivity inversion is performed based on transient electromagnetic method data to generate resistivity profile and resistivity planar distribution map of the study area, recording resistivity of 5-20Ω. The low-resistivity anomaly zone of m is obtained, and the horizontal range and depth range of the anomaly zone are recorded as candidate aquifers. The candidate aquifers are verified to be true aquifers by borehole data: (1) Determine the top plate of the aquifer: the top plate of the unconfined aquifer is the groundwater level, and the water level depth measured by the borehole is the depth of the top plate of the aquifer at the borehole; (2) Determine the bottom plate of the aquifer: combined with the borehole core data, find the depth at which the permeable lithology (such as sand) changes to the impermeable lithology (such as clay, mudstone). For example, if the borehole core shows that 8-15m is medium sand (permeable) and below 15m is clay (impermeable), then 15m is the depth of the bottom plate of the aquifer at the borehole; (3) Verify transient electromagnetic Low resistivity zone data: If the depth of the low resistivity zone (e.g., 10-18m) is close to the actual aquifer depth (8-15m) of the borehole, the transient electromagnetic method data results are reliable; if the deviation is large, the resistivity inversion parameters should be corrected using borehole data; analyze the water level time series curve of the monitoring well to find the depth range where the water level fluctuates with external factors (rainfall, mining). For example, if the monitoring shows that the water level at a depth of 8-15m rises by 0.5m after rain and falls by 1m after mining, while the water level below 15m remains unchanged, it indicates that 8-15m is an effective aquifer and the water-resistant layer below 15m is an impermeable layer; if the water level in a certain low resistivity zone does not fluctuate, it may be a dry sand layer or clay layer, and should be excluded from the aquifer range.

[0027] S32. Existing models are prone to misalignment between underground facilities and strata (e.g., mismatch between the tunnel model and the actual stratum interface) and abnormal topological relationships (e.g., facilities penetrating strata but not being identified), failing to reflect the interaction between the actual project and the geological structure, leading to distorted project risk assessments. This invention unifies the 3D outline of underground facilities and the stratum interface into the same coordinate system through coordinate transformation, calculates vertical distances and topological relationships, clarifies the stratum where the facility is located, its distance from the interface, and whether penetration / overlap exists, providing accurate spatial basis for subsequent model embedding of facilities. Based on CAD format design drawings and actual excavation outline data from construction logs, this invention uses a seven-parameter method to convert the design coordinates of underground facilities into a Gauss-Kruger coordinate system, obtaining the 3D outline of the underground facilities. Based on fundamental geological data, a stratigraphic surface is constructed using borehole data and geological mapping data. In one embodiment, for the stratigraphic depth in the borehole data, such as clay layer at 0-5m and sand layer at 5-12m, Kriging interpolation is used, and a spherical model is selected as the variogram to fit the borehole stratification data and generate a preliminary stratigraphic surface. Combined with surface lithological outcrops in the geological mapping data, such as the boundary between surface clay and sand layers, and stratigraphic attitude, such as sand layer strike at 35°, dip at 125°, and dip angle at 15°, the preliminary stratigraphic surface is corrected to ensure that the stratigraphic surface is consistent with the surface geological characteristics.

[0028] Calculate the vertical distance from each point on the 3D outline of the underground facility to the interface of the adjacent stratum to determine whether the underground facility is located inside the stratum; let the coordinates of the points on the surface of the underground facility be... The formation interface equation is: The formula for calculating vertical distance is as follows: , Where A, B, C, and D represent the first, second, third, and fourth coefficients of the formation interface equation, respectively, which are generated from borehole data through Kriging interpolation. Inspect underground facilities for any abnormal topological relationships such as penetration or overlap with the strata; and set distance tolerance thresholds. Determine whether there is penetration or overlap; if If the underground facility penetrates the upper boundary of the stratum, then the underground facility is considered to be overlapping if the minimum distance between the underground facility point set and the stratum interface is less than or equal to the distance tolerance threshold, and the angle between the local normal vector of the underground facility surface and the normal vector of the stratum interface is greater than 90 degrees. The final result is the spatial relationship data between underground facilities and strata, including the name of the underground facility, the name of the stratum where it is located, the distance between the underground facility and the stratum interface, and the topological relationship status; S33. Generation of Stratigraphic-Structural 3D Framework Data: Based on standardized basic geological data, the initial stratigraphic framework is constructed using the stratigraphic depth information from borehole data and the stratigraphic attitude information from geological mapping data, employing the layer modeling technique in 3D geological modeling software. Specifically, each stratigraphic layer is treated as a spatial surface, with the stratigraphic boundary points exposed by the boreholes as control nodes. Combined with the stratigraphic attitude constraining the extension trend of the surface, a 3D surface model of each individual stratigraphic layer is generated, such as the top interface of the clay layer, the top interface of the sand layer, and the bottom interface.

[0029] Using fault spatial coordinates, strike, dip, and dip angle data as structural constraints, a fault cutting algorithm is employed to correct the initial stratigraphic framework. For faults identified by abrupt changes in stratigraphic attitude and fractured core sections, Boolean operations are used to spatially cut the fault model and the stratigraphic model, causing corresponding displacement or faulting of the strata on both sides of the fault, forming a complex stratigraphic structure influenced by tectonics. Digital elevation model (DEM) data from topographic data is used as surface boundary conditions to trim the top of the generated stratigraphic-structural framework, ensuring that the model's surface morphology matches the actual topographic relief (DEM data smoothed by Gaussian filtering). Finally, the three-dimensional stratigraphic-structural framework data is obtained.

[0030] S4. Construct an initial three-dimensional geological model based on multi-source data of key hydrogeological structures, and dynamically update the initial three-dimensional geological model to obtain a dynamic three-dimensional geological model; Existing mesh generation uses uniform meshes, resulting in insufficient accuracy in critical areas (such as faults and karst caves) and computational redundancy in non-critical areas (such as homogeneous rock masses), making it difficult to balance accuracy and efficiency. This invention adopts a dynamic mesh generation method based on geological complexity, which reduces the computational load in non-critical areas while ensuring accuracy in critical areas, thus balancing accuracy and efficiency.

[0031] S41. Constructing an initial three-dimensional geological model: Based on the stratigraphic-structural three-dimensional framework data, the study area is spatially discretized using a grid partitioning algorithm to generate a three-dimensional grid model containing stratigraphic units and fault structures. The mesh generation algorithm employs a dynamic mesh generation method based on geological complexity, expressed by the following formula: , in, Representing a spatial point Grid size at the location; Indicates the basic grid size; This represents the scaling factor, with a value ranging from 0.8 to 1.2. The weight of the k-th geological element is determined through expert experience or the analytic hierarchy process (AHP), satisfying the following conditions: ; This represents the complexity index of the k-th geological element (fault, karst cave, aquifer); The fault, karst cave, and aquifer attribute parameters (such as fault dip angle, karst cave volume, and aquifer permeability coefficient) in the spatial distribution data of special geological bodies are assigned to the corresponding grid cells through a grid assignment algorithm. Preferably, the grid assignment algorithm uses distance-weighted interpolation. Combined with the spatial relationship data between underground facilities and strata, a three-dimensional solid model of the underground facilities is embedded in the model, and its spatial distance and topological relationship with the surrounding strata are marked to form an initial three-dimensional geological model.

[0032] Existing 3D geological models are mostly static, failing to reflect the spatiotemporal evolution of hydrogeological structures, such as changes in aquifer parameters caused by water level variations. Furthermore, integrating new data is difficult, and the models cannot be updated based on new exploration data, leading to a decline in accuracy over long-term use. This invention uses a particle swarm optimization hybrid algorithm to invert and adjust sensitive parameters, ensuring that the model's calculated water level matches the actual monitored water level. By combining new data-triggered local mesh reconstruction and parameter reassignment, the model's boundary conditions are corrected, enabling dynamic evolution of the model over time and with data updates, thus ensuring long-term model accuracy.

[0033] S42. Dynamically update the initial 3D geological model: Establish a correlation mechanism between groundwater dynamic monitoring data and model parameters, compare the water level monitoring data with the water level calculated by the model, and use a particle swarm optimization hybrid algorithm to invert and adjust sensitive parameters such as aquifer permeability coefficient and porosity; the formula for the particle swarm optimization hybrid algorithm is as follows: , = + , in, , Let represent the velocity and position of the i-th particle in the t-th generation, respectively; Indicates inertia weight; , These represent the first learning factor and the second learning factor, respectively. , These represent the first random number generation function and the second random number generation function, respectively. , These represent the individual's optimal position and the global optimal position, respectively. Indicates the annealing factor; Indicates the amount of temperature change; When new borehole data or geophysical exploration data is acquired, the consistency between the new data and the existing model is determined by root mean square error analysis. If the error exceeds a threshold, the local grid reconstruction of the model (such as densifying the grid around the new data) and parameter reassignment process are triggered. Preferably, the threshold is set to 5%. Combining long-term dynamic information such as vegetation cover change and water system distribution migration in remote sensing image data, the potential impact of surface environment changes on groundwater recharge is identified by machine learning classification algorithms, and the model boundary conditions (such as recharge area range and infiltration coefficient) are corrected to achieve dynamic updating of the initial three-dimensional geological model. Finally, a dynamic three-dimensional geological model that can reflect the spatiotemporal evolution characteristics of hydrogeological structure is obtained.

[0034] S5. Optimize the characterization method of the dynamic three-dimensional geological model to obtain a high-precision three-dimensional model.

[0035] Existing models suffer from blurred stratigraphic interfaces, such as coarse fitting of the clay-sand interface and insufficient precision of special geological body boundaries, such as unclear karst cave boundaries; grid subdivision fails to distinguish key areas, resulting in insufficient local precision or low computational efficiency; and the model has poor visualization effects and unclear information transmission of special geological bodies.

[0036] This invention addresses the problems of blurred stratigraphic interfaces and insufficient accuracy of special geological body boundaries in dynamic 3D geological models. It introduces an improved moving least squares surface fitting algorithm to optimize the dynamic 3D geological model. By increasing the density of control nodes around fault influence zones and karst caves, the smoothness and detail of interface representation are improved. The calculation formula of the improved moving least squares surface fitting algorithm is expressed as follows: , , in, Represents the fitted stratigraphic interface; This represents the global basis function, used to characterize the overall trend of geological interfaces; This represents local basis functions, used to characterize local details of geological interfaces (such as the boundaries of special structures like faults and caves); n and e represent the number of global basis functions and the number of local basis functions, respectively. , Let them represent the first coefficient to be determined and the second coefficient to be determined, respectively. Represents the region weighting function; The indicator function representing the fault influence zone, if point Located within the fault influence zone, Otherwise, it is 0; The function represents the indicator function around the cave, if point Located in the area surrounding the cave, Otherwise, it is 0; , These represent the first and second weighted gain coefficients, respectively, used to adjust the influence intensity of the local basis function in fault and karst regions. Higher node density allows for adjustments by increasing these coefficients. / This makes the contribution of local basis functions to the surface more significant.

[0037] A multi-scale meshing strategy is adopted, using fine meshes (0.5m-2m) for key areas such as aquifers, faults, and fracture zones; and coarse meshes (5m-10m) for homogeneous rock masses. This approach improves local model accuracy while maintaining computational efficiency. A visualization hierarchy mechanism for special geological bodies is established, assigning different display priorities and attribute annotation rules based on cave volume and fault scale, resulting in a high-precision 3D model. This visualization hierarchy mechanism for special geological bodies (classified by volume and scale), with different display priorities and annotation rules, enhances model readability, allowing users to quickly identify key geological information such as large faults and primary caves, thus improving information transmission efficiency.

[0038] S6. Mechanism analysis and correlation mining based on high-precision three-dimensional models.

[0039] Existing models are only used to display geological structures and lack quantitative analysis and correlation mining capabilities. They cannot reveal the correlation between water level changes and geological elements, or the risk of underground engineering disturbing geological structures. This invention uses a high-precision 3D model to intuitively display the spatial distribution of strata, the development characteristics of special geological bodies, and the interaction between underground facilities and geological structures. Utilizing the model's built-in spatial analysis module, it calculates key parameters such as the effective water storage volume of aquifers and the hydraulic conductivity of faults. Combined with dynamic groundwater monitoring data, it analyzes the correlation between water level changes and strata permeability and the distribution of special geological bodies. Through the spatial analysis module, it calculates key parameters such as aquifer water storage volume and fault hydraulic conductivity; Pearson correlation analysis quantifies the correlation between water level and geological elements; and it simulates flow field evolution to predict water level trends, achieving a progressive approach from qualitative display to quantitative analysis to trend prediction.

[0040] Example 2 In this embodiment, as Figure 2 As shown, the construction area of ​​the rail transit system is taken as the study area, and a typical geological profile (200m long, 50m deep) of the study area is used as the base. The horizontal axis represents the horizontal distance of the profile, and the vertical axis represents the underground depth. The figure uses three different lines to represent the results of different identification methods: the black dashed line represents the fault and karst cave boundaries identified by the traditional single-data method (using only borehole data); the blue solid line represents the identification boundary of the traditional dual-data method (borehole + ground-penetrating radar); and the red solid line represents the identification boundary of the multi-source data fusion method of this invention (borehole + ground-penetrating radar + transient electromagnetic + geological mapping). Twelve actual verification points are marked with red dots (fault locations revealed by borehole cores, actual center coordinates of karst caves, and volume measurements). It can be seen from the figure that the method of this invention has the highest overlap with the actual verification points. This invention integrates multi-source data and can solve the problems of incomplete single-data coverage and low identification accuracy of traditional methods.

[0041] Example 3 In this embodiment, as Figure 3As shown, the horizontal axis represents the monitoring time series, totaling 12 months, with one data node per month; the vertical axis represents the calculated permeability coefficient of the aquifer (unit: m / d); the black solid line represents the actual true value of the permeability coefficient, which is the average measured value obtained through 3 pumping tests and 2 pressure tests, serving as the accuracy benchmark; the gray dashed line represents the calculated value of the traditional static model, built only based on the initial borehole data, without subsequent updates; the blue solid line represents the calculated value of the initial 3D geological model, constructed in step S41 of this invention; the red solid line represents the calculated value of the dynamically updated model of this invention, updated after adding new borehole data in the 3rd month, groundwater level monitoring data in the 6th month, and remote sensing vegetation change data in the 9th month. The figure also marks the relative deviation rate at each stage: the deviation rate of the traditional static model increased from 8% initially to 25% in the 12th month; the deviation rate of the initial model stabilized at 7%-9%; the deviation rate of the dynamically updated model decreased to 4% after the update in the 3rd month, decreased to 2.5% after the update in the 6th month, and remained below 2% after the update in the 9th month. By comparing parameter deviations over time, the dynamic update logic of the monitoring data and model parameter correlation mechanism (particle swarm optimization inversion), new data-triggered local reconstruction, and remote sensing correction of boundary conditions established in this invention is clearly demonstrated. This solves the defects of traditional static models that cannot adapt to changes in geological conditions and whose accuracy continuously declines, and shows the role of dynamic updates in ensuring the long-term accuracy of the model.

[0042] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling, characterized in that, Includes the following steps: S1. Obtain multi-source hydrogeological structure data; the multi-source hydrogeological structure data includes basic geological data, geophysical exploration data, topographic and geomorphological data, underground facility data and groundwater dynamic monitoring data; S2. Preprocess the multi-source hydrogeological structure data to obtain standardized multi-source hydrogeological structure data; S3. Integrate the standardized hydrogeological structure multi-source data to obtain key hydrogeological structure multi-source data; Key hydrogeological structure multi-source data include: spatial distribution data of special geological bodies, spatial relationship data of underground facilities and stratigraphy, and three-dimensional framework data of stratigraphy and structure; S4. Construct an initial three-dimensional geological model based on multi-source data of key hydrogeological structures, and dynamically update the initial three-dimensional geological model to obtain a dynamic three-dimensional geological model. S5. Optimize the characterization method of the dynamic three-dimensional geological model to obtain a high-precision three-dimensional model; S6. Mechanism analysis and correlation mining based on high-precision three-dimensional models.

2. The hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling according to claim 1, characterized in that, In step S1, the basic geological data includes borehole data and geological mapping data; the borehole data includes stratum depth and physical and mechanical parameters of core samples; the geological mapping data includes surface lithological outcrops, geological boundaries, and stratum attitude; the geophysical exploration data includes ground-penetrating radar data and transient electromagnetic method data; the topographic data includes digital elevation model data and remote sensing image data; the underground facility data includes design drawings, construction logs, and as-built data of underground facilities; and the groundwater dynamic monitoring data includes water level monitoring data and water quality testing data.

3. The hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling according to claim 2, characterized in that, In step S2, an adaptive wavelet threshold denoising algorithm is used to denoise the ground-penetrating radar data.

4. The hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling according to claim 3, characterized in that, In step S3, the spatial distribution data of the special geological body includes: the spatial coordinates, strike, and dip angle of the fault; the center coordinates and volume of the karst cave; and the spatial distribution range and permeability coefficient of the aquifer. The specific process for obtaining the spatial distribution data of the special geological body is as follows: For transient electromagnetic data in standardized hydrogeological structure multi-source data, resistivity tomography is used to invert the underground resistivity structure and obtain resistivity profile map; a resistivity difference threshold is set to identify resistivity abrupt change zones, and the fault location is determined by cross-validation with stratigraphic abrupt change points in geological mapping data and core fracture sections in borehole data, and the spatial coordinates, strike, dip, and dip angle of the fault are output. A convolutional neural network model was trained using ground-penetrating radar (GPR) data. The model's inputs were the amplitude, frequency, and travel time of the reflected wave, labeled as "cave" or "non-cave." GPR data from standardized hydrogeological structure multi-source data was then input into the trained model to automatically identify and classify caves. The model was then validated using transient electromagnetic method data to ultimately obtain the center coordinates and volume of the caves. Based on the low resistivity anomaly zone derived from transient electromagnetic method data inversion in standardized hydrogeological structure multi-source data, combined with the groundwater level depth in borehole data and the water level fluctuation segment in groundwater dynamic monitoring data, the top and bottom plate depths of the aquifer are determined, and the spatial distribution range and permeability coefficient of the aquifer are obtained.

5. The hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling according to claim 4, characterized in that, In step S3, the specific process of generating underground facility-stratum spatial relationship data is as follows: Based on the design drawings and actual excavation contour data in the construction log of the underground facilities, the design coordinates of the underground facilities are converted into the Gauss-Kruger coordinate system using the seven-parameter method to obtain the three-dimensional contour of the underground facilities. Based on basic geological data, the stratigraphic surface is constructed using borehole data and geological mapping data; the vertical distance from each point on the three-dimensional outline of the underground facility to the interface of the adjacent stratigraphic layer is calculated to determine whether the underground facility is located inside the stratigraphic layer; and the abnormal topological relationship between the underground facility and the stratigraphic layer, such as penetration or overlap, is checked. The final result is the spatial relationship data between underground facilities and strata, including the name of the underground facility, the name of the stratum where it is located, the distance between the underground facility and the stratum interface, and the topological relationship status.

6. The hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling according to claim 5, characterized in that, In step S3, the specific process of generating the stratigraphic-structural three-dimensional framework data is as follows: Based on standardized basic geological data, using the stratigraphic depth information in borehole data and the stratigraphic attitude in geological mapping data, the initial stratigraphic framework is constructed using the layer modeling technology in 3D geological modeling software. Using fault spatial coordinates, strike, dip, and dip angle data as structural constraints, a fault cutting algorithm is employed to correct the initial stratigraphic framework. For faults identified by abrupt changes in stratigraphic attitude and fractured sections in core samples, Boolean operations are used to spatially cut the fault model and stratigraphic model, causing corresponding displacements or faultings of the strata on both sides of the fault, forming a complex stratigraphic structure. Digital elevation model data from topographic data is used as surface boundary conditions to trim the top of the generated stratigraphic-structural framework, ultimately obtaining the three-dimensional stratigraphic-structural framework data.

7. The hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling according to claim 6, characterized in that, Step S4 specifically includes: S41. Constructing an initial 3D geological model: Based on the stratigraphic-structural 3D framework data, the study area is spatially discretized using a grid partitioning algorithm to generate a 3D grid model containing stratigraphic units and fault structures; the fault, karst cave, and aquifer attribute parameters from the spatial distribution data of special geological bodies are assigned to the corresponding grid units using a grid assignment algorithm; combined with the spatial relationship data between underground facilities and stratigraphy, a 3D solid model of the underground facilities is embedded in the 3D grid model, and its spatial distance and topological relationship with the surrounding strata are labeled to form an initial 3D geological model; S42. Dynamically update the initial 3D geological model: Establish a correlation mechanism between groundwater dynamic monitoring data and model parameters, compare the water level monitoring data with the water level calculated by the model, and use a particle swarm optimization hybrid algorithm to invert and adjust the aquifer permeability coefficient and porosity; when new borehole data or geophysical exploration data are acquired, the consistency between the new data and the existing model is judged by root mean square error analysis. If the error exceeds the threshold, the local grid reconstruction and parameter reassignment process of the model is triggered; combined with the long-term dynamic information of vegetation cover change and water system distribution migration in remote sensing image data, the potential impact of surface environment changes on groundwater recharge is identified by machine learning classification algorithm, and the model boundary conditions are corrected to realize the dynamic update of the initial 3D geological model, and finally obtain a dynamic 3D geological model.

8. The hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling according to claim 7, characterized in that, Step S5 specifically includes: An improved moving least squares surface fitting algorithm is introduced to optimize the dynamic 3D geological model. The calculation formula of the improved moving least squares surface fitting algorithm is expressed as follows: ， ， in, Represents the fitted stratigraphic interface; Represents global basis functions; Represents local basis functions; n and e are the number of global basis functions and the number of local basis functions, respectively; , Let them represent the first coefficient to be determined and the second coefficient to be determined, respectively; Represents the region weighting function; Indicator function representing the fault influence zone; , These represent the first weight gain coefficient and the second weight gain coefficient, respectively. By employing a multi-scale grid subdivision strategy, fine grids are used for key areas such as aquifers, faults, and fracture zones, while coarse grids are used for homogeneous rock mass areas. A visualization and hierarchical mechanism for special geological bodies is established, assigning different display priorities and attribute labeling rules based on the volume of karst caves and the scale of faults, resulting in a high-precision 3D model.

9. The hydrogeological structure analysis method based on multi-source data fusion and three-dimensional modeling according to claim 8, characterized in that, Step S6 specifically includes: A high-precision 3D model is used to visually display the spatial distribution of strata, the development characteristics of special geological bodies, and the interaction between underground facilities and geological structures. Using the spatial analysis module built into the model, the effective water storage volume of aquifers and key parameters such as fault hydraulic conductivity are calculated. Combined with groundwater dynamic monitoring data, the correlation between water level changes and strata permeability and the distribution of special geological bodies is analyzed.

Images