Coal mine high-precision three-dimensional geological modeling method and system based on multi-source heterogeneous data fusion

By constructing a unified semantic reference model and multi-evidence inversion, and combining discontinuity constraints and anisotropy estimation, the semantic and scale differences in multi-source data fusion are solved, achieving high-precision three-dimensional geological modeling of coal mines, improving the stability and credibility of the model, and supporting applications such as gas control and ground pressure management.

CN121415003APending Publication Date: 2026-01-27GENERAL PROSPECTING INSTITUTE OF CHINA NATIONAL ADMINISTRATION OF COAL GEOLOGY
View PDF 0 Cites 4 Cited by

Patent Information

Application Number
CN202511591049.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-11-03
Publication Date
2026-01-27

AI Technical Summary

Technical Problem

Existing coal mine geological modeling methods suffer from several problems in mining areas with multiple structures, strong disturbances, or data-intensive conditions. These problems include significant differences in semantics and scale between different data sources, easy over-identification or omission of discontinuous structures in fracture identification, failure to consider the directionality of geological structures in interpolation, and lack of adaptive mechanisms in model verification. These issues result in insufficient modeling accuracy and reliability.

Method used

By constructing a unified semantic reference model, integrating multi-source heterogeneous data, and introducing discontinuity constraints and anisotropy estimation mechanisms, integrated modeling of geological structure and properties is achieved. This includes semantic system construction, multi-evidence consistency inversion, implicit stratigraphic function solution, and anisotropy estimation, which are then verified in conjunction with production monitoring data.

Benefits of technology

It significantly improves the accuracy, stability, and interpretability of three-dimensional geological models in complex tectonic zones, outputs high-precision three-dimensional geological models of coal mines, and provides uncertain results to support risk identification and decision-making.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121415003A_ABST
    Figure CN121415003A_ABST
Patent Text Reader

Abstract

The invention provides a coal mine high-precision three-dimensional geological modeling method and system based on multi-source heterogeneous data fusion, and relates to the technical field of coal mine geological modeling. The method comprises the following steps: establishing a unified semantic reference system and a time-space coordinate reference, and generating an alignment anchor point set; mapping the multi-source observation data to form an alignment feature set and calculating a source weight; according to a multi-evidence consistency rule and geomechanical prior inversion, a discontinuous geological structure is obtained; constructing an implicit stratigraphic function and applying a boundary jump condition to solve a stratigraphic boundary set; and taking the stratum boundary as a framework, and executing anisotropic joint estimation to obtain a property distribution model and an uncertainty result. According to the method, integrated modeling of the geological structure and the property of the complex structure area is achieved, and the precision, the stability and the credibility of the three-dimensional geological model are remarkably improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of coal mine geological modeling technology, and in particular to a high-precision three-dimensional geological modeling method and system for coal mines based on multi-source heterogeneous data fusion. Background Technology

[0002] Currently, coal mine geological modeling mainly relies on data from borehole logging, seismic exploration, roadway exposure, and a limited amount of in-situ testing. Traditional modeling processes typically focus on stratigraphic interpretation and structural classification, supplemented by geological attribute interpolation and volume model reconstruction, to form a three-dimensional model reflecting the distribution and structural characteristics of the ore body. While this approach can meet design and production needs when the geological conditions in the mining area are relatively simple and the data types are limited, its modeling accuracy and stability are significantly restricted in mining areas with multiple structures, strong disturbances, or data-intensive conditions.

[0003] With the advancement of digitalization and intelligentization in mines, coal mine geological modeling is gradually developing towards higher precision, real-time performance, and multi-source fusion. New data sources, such as microseismic monitoring, tunnel laser scanning, InSAR deformation observation, ground-penetrating radar, and core experiments, are increasingly being introduced into geological modeling systems. Spatiotemporal collaborative processing and uncertainty measurement of multi-source data have become key directions for improving model reliability. Simultaneously, geological models not only serve resource estimation but are also gradually integrating multi-field coupled analysis for gas control, ground pressure monitoring, and ventilation safety, achieving dynamic updates and closed-loop management of mine geological information.

[0004] Despite the increasing variety of methods for acquiring multi-source data, existing modeling methods still suffer from several prominent problems: First, the semantic and scale differences between different data sources are significant, making it difficult to achieve unified description and effective fusion; second, discontinuous structures such as fractures and fissures often rely on single data interpretations, easily leading to over- or under-identification; third, the interpolation and modeling processes primarily rely on spatial distance as a constraint, failing to fully consider the directionality and anisotropy of geological structures; and fourth, the model verification process lacks an adaptive mechanism, making it difficult to correct in real time based on new observations. These problems result in decreased model accuracy and insufficient reliability in complex geological environments, hindering the further development of 3D geological information systems for coal mines. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a high-precision three-dimensional geological modeling method and system for coal mines based on multi-source heterogeneous data fusion. By constructing a unified semantic reference, fusing multi-source observation information, and introducing discontinuity constraints and anisotropy estimation mechanisms, the integrated modeling of coal mine geological structure and properties is realized, significantly improving the accuracy, stability, and interpretability of three-dimensional geological models in complex tectonic zones.

[0006] To achieve the above objectives, the present invention provides the following solution: A high-precision 3D geological modeling method for coal mines based on multi-source heterogeneous data fusion includes: A semantic system covering geological object elements, structural elements, geological property attributes and observation elements in the mining area is constructed. Standardized description rules are determined, a unified time and space coordinate benchmark is determined, alignment anchor point generation rules are formulated, and an alignment anchor point set is generated based on geological objects with the same name, landmark interfaces and structural signs, thus obtaining a semantic reference model and an alignment anchor point set. Multi-source heterogeneous geological observation data are mapped according to the semantic reference model and the unified time and space coordinate benchmark to form an alignment feature set. The source weight is calculated based on the source accuracy and reliability to obtain the alignment feature set and the source weight. Based on the alignment feature set, inversion is performed according to the multi-evidence consistency rule and combined with geomechanical priors to generate a candidate set of discontinuous geological structures. An implicit stratigraphic function is established, and the geological contact points in the semantic reference model are combined with the marker interface. Boundary jump conditions are set with the candidate set of discontinuous geological structures as geometric constraints. The implicit stratigraphic function is solved to obtain the stratigraphic boundary set and structural geometric relationship that satisfy geometric and topological consistency. Using the stratigraphic boundary set as a spatial framework, an anisotropic estimation model considering structural connectivity and directionality is constructed to jointly estimate the target geological properties and output uncertainty results, forming a property distribution model and uncertainty results, and obtaining a three-dimensional geological model of the coal mine that includes the stratigraphic boundary set, the property distribution model and the uncertainty results.

[0007] Preferably, it further includes: A consistency verification mechanism is established based on production monitoring data and physical constraints to verify the stratigraphic boundary set, the property distribution model, and the uncertainty results. When the verification difference exceeds the preset threshold, the source weight and local model parameters are adaptively adjusted and incremental updates are implemented until the convergence requirement is met, and the verified and updated three-dimensional geological model of the coal mine and the model source record are output.

[0008] Preferably, the unified time and space coordinate reference consists of a unified spatial coordinate reference for the mining area, a surface elevation reference, and a time reference reference; the parameter set corresponding to the unified time and space coordinate reference includes coordinate transformation parameters and time alignment parameters, and the parameter set is used to uniformly map the multi-source heterogeneous geological observation data to the unified time and space coordinate reference.

[0009] Preferably, the alignment anchor point generation rule consists of a spatial tolerance threshold, a temporal tolerance threshold, and a consistency scoring rule; the alignment anchor point set includes geological contact point anchor points, landmark interface anchor points, and tectonic sign anchor points; the screening condition corresponding to the alignment anchor point set is a conflict threshold, which is used to remove conflicting anchor points to obtain a stable alignment anchor point set.

[0010] Preferably, the determination of the source weight is based on the observation spatial resolution, observation noise level, spatial coverage, acquisition timeliness, and consistency score of the alignment anchor set; the normalization rule corresponding to the source weight is used to limit the source weight within a preset numerical range so that the multi-evidence consistency rule and the anisotropy estimation model can be invoked.

[0011] Preferably, the generation constraints of the candidate set of discontinuous geological structures consist of a multi-evidence consistency rule and a connectivity constraint; the regularization term corresponding to the generation constraints is a sparsity regularization, which is used to suppress isolated responses and improve the reliability of the candidate set of discontinuous geological structures.

[0012] Preferably, the energy functional corresponding to the implicit stratigraphic function consists of a fitting term, a smoothing term, and a discontinuity penalty term; the boundary condition set corresponding to the discontinuity penalty term is a boundary jump condition set, the position of which is determined by the candidate set of discontinuous geological structures, and the boundary jump condition set is used to ensure the geometric and topological consistency of the stratigraphic boundary set at the fracture location.

[0013] Preferably, the expression for the implicit stratigraphic function is: exist : , exist , ,in ; in, For modeling regions; The implicit stratigraphic function is defined as the implicit stratigraphic function, and the zero level set of the implicit stratigraphic function corresponds to the stratigraphic boundary. This refers to the signed distance data item generated from the geological contact points and marker interfaces in the semantic reference model; A positive regularization coefficient is used to balance the data terms and the anisotropic smoothing terms; This refers to the set of discontinuous surfaces corresponding to the candidate set of discontinuous geological structures. and These are the function lateral limits across the discontinuous surface, respectively; n represents the function jump amplitude determined by the discontinuous indication intensity. Let be the unit normal vector of the discontinuous surface; For anisotropic metric tensors; The unit direction field is determined by the strike and dip angle of the strata; It is the anisotropy ratio, used to distinguish the smoothness intensity along the structural direction from that in the transverse direction; It is a unit array.

[0014] Preferably, the directional information of the anisotropy estimation model is jointly determined by the structural orientation field and the structural connectivity relationship; the structural orientation field is composed of stratigraphic strike information, stratigraphic dip information and normal information of the candidate set of discontinuous geological structures; the joint estimation object includes at least one of coal seam thickness, porosity, gas content and rock mass strength.

[0015] A high-precision 3D geological modeling system for coal mines based on multi-source heterogeneous data fusion includes: The semantic reference construction unit is used to construct a semantic system covering geological object elements, structural elements, geological property attributes and observation elements in the mining area, determine standardized description rules, determine unified time and space coordinate benchmarks, formulate alignment anchor point generation rules, and generate an alignment anchor point set based on geological objects with the same name, landmark interfaces and structural signs, thus obtaining the semantic reference model and the alignment anchor point set. The multi-source alignment and feature normalization unit is used to map multi-source heterogeneous geological observation data according to the semantic reference model and the unified time and space coordinate benchmark to form an alignment feature set, and calculate the source weight based on the source accuracy and reliability to obtain the alignment feature set and the source weight. The discontinuous structure identification and inversion unit is used to perform inversion based on the alignment feature set, according to the multi-evidence consistency rule and combined with geomechanical priors, to generate a candidate set of discontinuous geological structures. Implicit stratigraphic reconstruction unit is used to establish implicit stratigraphic functions, combine the geological contact points in the semantic reference model with the marker interface, and set boundary jump conditions with the candidate set of discontinuous geological structures as geometric constraints to solve the implicit stratigraphic functions, thereby obtaining a set of stratigraphic boundaries and structural geometric relationships that satisfy geometric and topological consistency. The anisotropy estimation and uncertainty analysis unit is used to construct an anisotropy estimation model that considers structural connectivity and directionality, using the stratigraphic boundary set as a spatial framework. It performs joint estimation of the target geological properties and outputs uncertainty results, forming a property distribution model and uncertainty results, and obtaining a three-dimensional geological model of the coal mine that includes the stratigraphic boundary set, the property distribution model and the uncertainty results.

[0016] The present invention discloses the following technical effects: (1) This invention establishes a semantic reference model and a set of alignment anchor points, and completes multi-source mapping under the unified time and space coordinate reference. The method classifies heterogeneous observations such as borehole, seismic, scanning, deformation and experiment into the same semantic and coordinate system, effectively eliminating semantic drift and spatial misalignment caused by inconsistent naming, scale and coordinate in the prior art, ensuring the comparability and transferability of subsequent inversion and estimation links, and improving the fusion accuracy and model stability from the source.

[0017] (2) This invention introduces a multi-evidence consistency rule on the alignment feature set and combines it with geomechanical priors for joint inversion. The method identifies fractures, fissures and weak interlayers as a whole discontinuous geological structure, avoiding over-interpretation and under-interpretation caused by existing technologies relying on single data. At the same time, the source weight and consistency rule jointly constrain the evidence contribution, improving the reliability and traceability of the candidate set of discontinuous structures, and providing robust constraints for geometric and property modeling.

[0018] (3) In the implicit stratigraphic function solution, the present invention sets boundary jump conditions for the candidate set of discontinuous geological structures and combines geological contact points and marker interfaces in the semantic reference model. The method simultaneously characterizes the relationship between continuous strata and fault dislocation under the same implicit expression, ensuring the geometric and topological consistency of the stratigraphic boundary set. It overcomes the unreasonable smoothing and topological errors generated by traditional interpolation near faults and significantly improves the authenticity of structural geometry.

[0019] (4) The present invention uses the set of stratigraphic boundaries as a spatial framework and adopts an anisotropic estimation model that considers structural connectivity and directionality to perform joint property estimation. The method incorporates the differences along the layer and along the structural direction into the constraints, avoiding the "pulling effect" caused by the existing technology of Euclidean distance interpolation. The multi-property joint modeling can simultaneously utilize the co-occurrence and physical correlation of cross-properties to improve the spatial consistency and prediction accuracy of the distribution of properties such as thickness, porosity, and gas content.

[0020] (5) The present invention outputs uncertainty results while generating property distribution models, and together with the set of strata boundaries, it constitutes a three-dimensional geological model of coal mine. The method makes up for the problems of weak model credibility and traceability in the prior art, so that risk identification, gas control, ground pressure management and mining design can be made based on confidence intervals. The model output structure is clear and the elements are complete, which is convenient for subsequent access to monitoring and simulation, and realizes direct implementation from modeling to application. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart of the method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the system structure provided in an embodiment of the present invention. Detailed Implementation

[0023] 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.

[0024] The purpose of this invention is to provide a high-precision three-dimensional geological modeling method and system for coal mines based on multi-source heterogeneous data fusion. Through semantic unification, multi-evidence inversion, fracture constraint reconstruction, and joint anisotropy estimation, it achieves high-precision fusion and uncertainty quantification of multi-source heterogeneous data, significantly improving the accuracy, continuity, and reliability of the three-dimensional geological model of coal mines.

[0025] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] Figure 1 The method flowchart provided in the embodiments of the present invention is as follows: Figure 1 As shown, this invention provides a high-precision three-dimensional geological modeling method for coal mines based on multi-source heterogeneous data fusion, including: Step 100: Construct a semantic system covering geological object elements, structural elements, geological property attributes and observation elements in the mining area, determine standardized description rules, determine unified time and space coordinate benchmarks, formulate alignment anchor point generation rules, and generate an alignment anchor point set based on geological objects with the same name, landmark interfaces and structural signs to obtain a semantic reference model and an alignment anchor point set. Step 200: Map the multi-source heterogeneous geological observation data according to the semantic reference model and the unified time and space coordinate benchmark to form an alignment feature set, and calculate the source weight based on the source accuracy and reliability to obtain the alignment feature set and source weight; Step 300: Based on the aligned feature set, perform inversion according to the multi-evidence consistency rule and combined with geomechanical priors to generate a candidate set of discontinuous geological structures; Step 400: Establish implicit stratigraphic functions, combine geological contact points and marker interfaces in the semantic reference model, and set boundary jump conditions with candidate sets of discontinuous geological structures as geometric constraints to solve the implicit stratigraphic functions and obtain stratigraphic boundary sets and structural geometric relationships that satisfy geometric and topological consistency. Step 500: Using the stratigraphic boundary set as the spatial framework, construct an anisotropic estimation model that considers structural connectivity and directionality, jointly estimate the target geological properties and output uncertainty results, form a property distribution model and uncertainty results, and obtain a three-dimensional geological model of the coal mine that includes the stratigraphic boundary set, property distribution model and uncertainty results.

[0027] In step 100 of this embodiment, a semantic system oriented towards the mining area is first constructed. The semantic system consists of four parts: geological object elements, structural elements, geological property attributes, and observational elements, used to uniformly describe stratigraphic structure, structural morphology, and observational characteristics. Specifically, geological object elements define objects such as coal seams, interbedded rock, ore-bearing bodies, and bedrock, and their attribute fields; structural elements define structural types such as faults, fissures, and folds, and their spatial relationships; geological property attributes define the dimensions and value ranges of parameters such as thickness, porosity, gas content, and intensity level; and observational elements define observation locations, time labels, and quality evaluation methods. Based on this, this embodiment establishes a semantic reference model, specifying field names, value dictionaries, unit conversions, and default processing rules, and establishing inclusion and adjacency relationships between objects. To ensure coverage and consistency, the semantic reference model contains no fewer than 60 field entries, including no fewer than 12 categories of geological object elements, no fewer than 8 categories of structural elements, and no fewer than 10 categories of geological property attributes; the units are uniformly adopted using the International System of Units (SI), and the default values ​​are set with upper and lower limits based on the 5th and 95th percentiles of the same type of sample.

[0028] This embodiment establishes a unified time and spatial coordinate benchmark, which consists of a unified spatial coordinate benchmark for the mining area, a surface elevation benchmark, and a time reference benchmark. The unified spatial coordinate benchmark for the mining area is used to project various types of data onto the same Cartesian coordinate system; the surface elevation benchmark is used to unify elevation values ​​from different sources to the same reference surface; and the time reference benchmark is used to uniformly correct the recording time of different acquisition devices. This embodiment sets a parameter set corresponding to this benchmark, including a coordinate transformation parameter set and a time alignment parameter set. The coordinate transformation parameter set is used to realize the conversion between local coordinates and the unified coordinates, including the central meridian, scale factor, translation, and rotation; the time alignment parameter set is used to realize time standardization, including equipment clock deviation, acquisition time zone, and time correction parameters. In this embodiment, the translation range from local coordinates to the unified coordinates is -5000 to 5000 meters, the rotation range is -5 to 5 milliradians, the scale factor relative error is no greater than one part per million, the equipment clock deviation correction limit is no greater than 2 seconds, and the residual error after time alignment is no greater than 1 second. Through this benchmark and parameter set, the consistency of all observation data in time and space is ensured.

[0029] This embodiment establishes alignment anchor point generation rules to build correspondences between multi-source data. The alignment anchor points include three categories: geological contact point anchors, landmark interface anchors, and structural indicator anchors. Geological contact point anchors originate from the stratigraphic boundaries of boreholes or exposed areas; landmark interface anchors originate from the locations of marker layers or significant reflective interfaces; and structural indicator anchors originate from fault traces or fracture zone centerlines. This embodiment uses three constraints for anchor point generation: spatial tolerance threshold, temporal tolerance threshold, and consistency scoring rules. The spatial tolerance threshold is 0.5 meters in the near field of the tunnel and 5 meters in the far field of the mine field; the temporal tolerance threshold is 30 minutes; and the consistency scoring threshold is set to 70 points, below which no anchor points are generated. When multiple candidates exist within the same spatial neighborhood and the score difference is less than 10 points, this embodiment selects the record with the smallest spatial residual and temporal residual as the priority anchor point. To ensure a reasonable distribution of anchor point types, this embodiment controls the proportion of three types of anchor points during the generation process: geological contact point anchors account for no less than 40%, marker interface anchors account for no less than 30%, and tectonic indicator anchors account for no less than 20%, with the remaining proportion reserved for dynamic supplementation. Through the above rules and threshold settings, this embodiment achieves accurate spatial and temporal alignment of multi-source data and provides a consistent reference benchmark for subsequent geological modeling.

[0030] After completing step 100, this embodiment performs field verification and unit unification on the multi-source heterogeneous geological observation data according to the semantic reference model, and then performs standardized mapping of position and time based on the unified time and spatial coordinate benchmark. Position mapping includes the conversion between plane coordinates and elevation, and time mapping includes the conversion between equipment time and reference time. After mapping, each record is completed with required semantic fields and time tags. To ensure consistency, this embodiment performs sampling verification on samples before data entry, with a sampling ratio of no less than 5%. The allowable upper limit for position residuals is 0.5 meters in the near field of roadways and 5 meters in the far field of mine fields; the allowable upper limit for time residuals is 1 second. Records exceeding the upper limit are marked as requiring verification in this embodiment and are not included in the alignment feature generation.

[0031] This embodiment uses the aforementioned set of alignment anchor points as a reference, and performs spatial and temporal registration between the standardized observation records and adjacent anchor points to form an alignment feature set. Alignment features refer to a unified description used for subsequent inversion and estimation, based on the degree of conformity between a single observation and the anchor points in the semantic, spatial, and temporal dimensions. In this embodiment, the spatial neighborhood radius and temporal neighborhood window are set during registration: a near-field radius of 2 meters for tunnels and a far-field radius of 20 meters for mining areas; the time window is 30 minutes. When a single observation matches more than 3 anchor points in its neighborhood, this embodiment selects the combination with the smallest spatial difference and the smallest temporal difference for feature generation; when an observation is missing in any dimension, a 5%–95% quantile interval of similar samples is used as a placeholder to avoid features becoming unusable due to missing data.

[0032] This embodiment calculates source weights based on source accuracy and reliability. The weights are determined by five factors: spatial resolution, observation noise level, spatial coverage, acquisition timeliness, and consistency score with the set of alignment anchor points. Spatial resolution is graded according to the actual sampling interval: intervals no greater than 0.1 meters are considered high, 0.1–1 meters are considered medium, and more than 1 meter are considered low. Observation noise level is based on the statistical difference of repeated observations: standard deviation no greater than 1% is considered low noise, 1%–5% is considered medium noise, and more than 5% is considered high noise. Spatial coverage is calculated based on the percentage of effective observation voxels: no less than 0.6 is considered high, 0.3–0.6 is considered medium, and less than 0.3 is considered low. Acquisition timeliness is graded according to the interval between "data and current modeling time": no more than 30 days is considered high, 30–180 days is considered medium, and more than 180 days is considered low. The consistency score follows the percentage score from step 100, with a threshold of 70 points. Observations below the threshold are not included in the weight calculation in this embodiment.

[0033] This embodiment applies interval stretching to the five criteria mentioned above, ensuring each score falls within the 0-1 range, with higher scores representing better quality. Initial source weights are then synthesized using equal weighting, and these weights are limited to a preset range to ensure robustness. To avoid extreme source monopolies, this embodiment sets the lower limit of source weights to 0.05 and the upper limit to 0.60. When a source's initial weight exceeds the upper limit, it is truncated to the upper limit, and the excess is proportionally distributed to the remaining sources. When a source's initial weight is below the lower limit, it is raised to the lower limit, and the weights are proportionally reduced from the source with the highest weight. The normalized source weights are then standardized again across all sources, resulting in a total weight of 1.00. This normalization rule ensures that quality differences between different sources can be distinguished without causing any single source to have an unstable impact on subsequent calculations.

[0034] In this embodiment, after generating the alignment feature set and source weights, two quality controls are performed: First, at least 10% of the observations are randomly selected as a validation set to test the spatial and temporal consistency of the alignment features, requiring a spatial consistency pass rate of at least 95% and a temporal consistency pass rate of at least 98%; second, a sensitivity test is performed on the robustness of the source weights, perturbing any one of the five criteria by ±10% within its allowable range and observing the weight changes, requiring that the weight change of any source does not exceed 0.10. Through the above process, this embodiment finally obtains an alignment feature set and source weights that can be directly used by the multi-evidence consistency rule and the anisotropy estimation model, providing stable input for subsequent discontinuous structure inversion and implicit stratigraphic solution.

[0035] Furthermore, in step 300 of this embodiment, based on the alignment feature set, a multi-evidence consistency rule is established to uniformly incorporate evidence reflecting discontinuities into the evaluation, including seismic discontinuities, attribute mutations and geometric curvature, spatial convergence of microseismic event clusters, gradient mutations in deformation observations, and roughness and reflection intensity mutations in near-field scanning. To ensure operability, this embodiment performs standardized scoring on each piece of evidence within the neighborhood of each candidate voxel or measuring point (the near-field radius of the tunnel is 5 meters, and the far-field radius of the mine field is 30 meters), and sets "at least 3 pieces of evidence simultaneously reaching 0.60 or higher and the overall score not lower than 0.70" as the selection threshold. The geomechanical prior is used to eliminate spurious responses incompatible with the stress environment. This embodiment uses the regional principal stress direction and the allowable fault surface orientation as constraints, requiring the strike and dip angle of the candidate discontinuity surface to enter the prior tolerance zone (strike tolerance not greater than 20 degrees, dip angle tolerance not greater than 15 degrees), and requiring a continuous extension length along the strike or dip direction not less than 50 meters.

[0036] This embodiment applies connectivity constraints to the response points selected by the multi-evidence consistency rule, based on three-dimensional adjacency relationships. Connectivity constraints are used to suppress discrete isolated points and strengthen discontinuous structural morphologies such as bands and surfaces: the adjacency criterion uses three-dimensional 26-adjacency, with a minimum connected unit size of no less than 200 voxels or an equivalent length of no less than 30 meters; when the gap between two cluster responses is less than 10 meters and their orientation and dip angle differ by no more than 10 degrees, they are determined to be the same structure and bridged; when the local orientation of any point within a cluster deviates from the cluster's average orientation by more than 25 degrees and the continuous length is less than 15 meters, it is determined to be branching noise and discarded. To enhance geometric consistency, this embodiment performs local planar fitting and banded fitting within each connected cluster, requiring the goodness of fit to reach a preset threshold (planar fitting goodness of fit no less than 0.85, banded fitting goodness of fit no less than 0.80). Clusters that do not reach the threshold are not included in the candidate set of discontinuous geological structures.

[0037] This embodiment applies sparsity regularization to response clusters that satisfy the connectivity constraints. Sparsity regularization is used to suppress weak evidence tails while preserving the main geometric shape. Specifically, a dual-threshold hysteresis suppression is set: a high threshold of 0.85 and a low threshold of 0.65. Voxels within a continuous region that are below the low threshold and not connected to the high threshold region are deleted; voxels between the low and high thresholds but whose neighborhood average score is below 0.70 are contracted. To improve stability, this embodiment performs repeated sampling tests on the input observations for at least 100 rounds. Only those identified as discontinuous structures with a frequency of at least 0.70 during repeated tests are included in the candidate set of discontinuous geological structures. Simultaneously, the candidate's type (fault, fracture zone, or breccia zone), spatial orientation, and confidence level (confidence levels are divided into three levels: above 0.85, between 0.75 and 0.85, and below 0.75) are output as geometric constraints in subsequent implicit stratigraphic solutions.

[0038] Furthermore, in step 400 of this embodiment, using the geological contact points and marker interfaces in the semantic reference model as constraints, a signed distance data item is generated for each spatial location according to the proximity principle and normal consistency criterion: positive for locations above the interface and negative for locations below, with the distance being the shortest distance from the point to the local fitting plane; outliers are suppressed by limiting their influence through neighborhood quantiles, with the neighborhood radius being 2 meters in the near field of the tunnel and 20 meters in the far field of the mine field, and the upper and lower limits of the limit being the 5th percentile and the 95th percentile, respectively. In this embodiment, the strike and dip angle are obtained by performing least-squares plane fitting on the contact points and marker interfaces, and a unit direction field is constructed accordingly to characterize the direction along the layer; the direction field is smoothed in sparse regions using tensor interpolation, with an interpolation radius of 10 meters to ensure direction continuity without excessive smoothing.

[0039] This embodiment uses the candidate set of discontinuous geological structures as geometric constraints, defining a set of boundary jump conditions for each discontinuous surface: quantitative jumps of function values ​​are set on both sides of the surface while maintaining the continuity of anisotropic flux, thereby forming a clear discontinuity relationship at the fracture location. The jump amplitude is obtained by mapping the discontinuity indicator intensity classification, with weak, medium, and strong levels corresponding to small, medium, and large values ​​respectively, and a typical value ratio of 1:2:3; jump conditions are only arranged at the location of the candidate structure when the connectivity length is not less than 50 meters, the strike tolerance does not exceed 20 degrees, and the dip tolerance does not exceed 15 degrees. Adjacent candidates less than 10 meters apart and with an azimuth difference of less than 10 degrees are bridged to ensure the geometric continuity of the fracture surface in segments and zones.

[0040] This embodiment constructs an energy functional consisting of a fitting term, a smoothing term, and a discontinuity penalty term: the fitting term approximates signed distance data, the smoothing term anisotropically constrains gradients to pass more easily along the directional field and is more suppressed vertically, and the discontinuity penalty term strengthens jump conditions at fracture locations. The computational domain is discretized using a tetrahedral mesh, with mesh edge lengths ranging from 0.5 to 1 meter in the near field and 2 to 5 meters in the far field; natural boundary conditions are used at the outer boundary to avoid additional traction on layers near the boundary. To avoid over-smoothing of discontinuous neighborhoods, this embodiment reduces the smoothing intensity for elements less than 3 meters from the fracture surface, with the reduction coefficient ranging from 0.5 to 0.8 of the original intensity, to preserve geometric abrupt changes.

[0041] In this embodiment, the anisotropy metric tensor is set as a weighted combination of the dominant direction of the orientation field and its orthogonal direction. The anisotropy ratio is used to control the difference in smoothing intensity along the construction direction and laterally, and its value is preferably between 0.1 and 0.5. The regularization coefficient is used to maintain stability in sparse regions and enhance fitting in dense regions, and is preferably set in partitions between 0.01 and 0.2. The large sparse linear equation system is solved using the preconditioned conjugate gradient method, and the preconditioner is either incomplete decomposition or geometric multigrid. The convergence criterion is set as either a relative residual not exceeding one ten-thousandth or a maximum iteration step not exceeding 2000 steps, whichever is satisfied first. To avoid local oscillations, this embodiment performs an orientation field smoothing update every 100 iterations, with a smoothing window radius of 5 meters.

[0042] This embodiment extracts the zero-level set of implicit stratigraphic functions after the solution is obtained, generates a stratigraphic boundary set using an isosurface tracking algorithm, and calculates the normal continuity and patch connectivity indices. At fault surfaces, it checks whether jumps reach a set threshold, requiring the median absolute value of jump errors to not exceed 0.2 meters. To ensure topological consistency, this embodiment performs porosity and self-intersection detection on each stratigraphic interface, with a minimum porosity threshold of 2 meters and a self-intersection allowance threshold of zero. When a detection fails, it recalculates locally with a refinement not exceeding twice the precision of the original mesh to eliminate pseudo-intersections. The final output stratigraphic boundary set exhibits clear discontinuity at faults, maintains a smooth transition in continuous regions, and is consistent with the aforementioned direction field and candidate discontinuous geometry, providing a reliable spatial framework for subsequent property estimation and uncertainty analysis.

[0043] Furthermore, in step 500 of this embodiment, an anisotropic estimation model considering structural connectivity and directionality is constructed using the stratigraphic boundary set as a spatial framework. The directional information of the anisotropic estimation model is jointly determined by the structural direction field and the structural connectivity. The structural direction field is composed of stratigraphic strike information, stratigraphic dip information, and normal information of the candidate set of discontinuous geological structures, and is used to limit the search neighborhood and smoothing direction under the principle of prioritizing strata-direction and weakening lateral dependence. To ensure a balance between resolution and computational cost, this embodiment takes a voxel side length of 1 meter in the main mining zone and 2–5 meters in the periphery; the neighborhood radius along the structural direction is taken as 30–60 meters, and the lateral neighborhood radius is taken as 10–20 meters, with the ratio of the two preferably between 2 and 4. The structural connectivity is conditional on the intersection, superposition, and strata-connection of stratigraphic boundaries, requiring that the shortest path length between any two points on the connectivity map is no greater than 100 meters to participate in the same neighborhood estimation.

[0044] This embodiment's joint estimation targets include at least one of coal seam thickness, porosity, gas content, and rock mass strength, and allows simultaneous estimation of multiple properties to utilize cross-property statistical correlations. Sample inclusion follows three conditions: first, the number of samples within a spatial neighborhood consistent with the tectonic orientation field must be no less than 30; second, the coefficient of variation for any property must not exceed 1.0; if it does, partitioned estimation is used for that property; and third, the correlation strength across properties must reach a threshold of at least 0.40 to be used for joint estimation. When a property is locally sparse, this embodiment only uses auxiliary properties with a correlation strength of at least 0.50 to participate in the joint estimation, and limits the contribution of auxiliary properties to no more than 50% of the target property's information content to avoid over-migrating.

[0045] This embodiment performs a directional search within the anisotropic neighborhood, prioritizing extension along the structural direction field with lateral attenuation to ensure interlayer continuity and suppress interlayer leakage. To control geological rationality, this embodiment applies three types of hard and soft constraints during estimation: hard constraints include non-negative thickness, porosity between 0 and 0.4, gas content between 0 and 30 cubic meters per ton, and rock mass strength between 0 and 200 MPa; soft constraints include upper limits for gradients along the interlayer direction and gradients across the interlayer direction, with the former not exceeding a 10% variation per 10 meters and the latter not exceeding a 5% variation per 10 meters; the soft constraints are relaxed by half within 3 meters of the candidate set of discontinuous geological structures to preserve geometric abrupt changes. For cases where samples are adjacent to a fault and there are significant differences between the samples on both sides, this embodiment estimates independently by fault side and prohibits cross-fault sampling.

[0046] This embodiment also outputs uncertainty results to characterize the combined effect of observation errors and model structural errors. The uncertainty results consist of two parts: observation uncertainty, converted to an observation confidence interval per voxel based on the source weights and observation noise levels; and model uncertainty, converted to a structural confidence interval per voxel based on the estimated residuals and sensitivity assessment to small perturbations (not exceeding 10 degrees of directional angle) in the constructed orientation field, using sensitivity coefficients. The two types of confidence intervals are merged at the voxel scale into a total confidence interval, and 5% and 95% quantile estimates are given simultaneously. When the width of the total confidence interval exceeds 30% of the typical magnitude of this property, this embodiment marks this region as a high uncertainty region and suggests prioritizing the addition of samples in subsequent iterations.

[0047] This embodiment employs a combination of stratified retention and cross-validation for quality control: at least 10% of the samples are extracted as a validation set according to geological zoning and structural zones, and 50% cross-validation is performed. Evaluation indicators include mean absolute error and root mean square error, with property-related thresholds set: mean absolute error for thickness not exceeding 0.3 meters, mean absolute error for gas content not exceeding 3.0 cubic meters per ton, mean absolute error for porosity not exceeding 0.03, and mean absolute error for rock mass strength not exceeding 15 MPa. If any property fails to meet the standard in any zone, the lateral neighborhood radius for that zone is reduced by 20%, and the sample inclusion quantity is increased by 20%, followed by recalculation. The coverage of the predicted intervals in the validation set is compared with the nominal confidence level, requiring that the actual coverage of 90% of the nominal intervals be no less than 85%.

[0048] This embodiment assembles the stratigraphic boundary set, property distribution model, and uncertainty results to obtain an integrated three-dimensional geological model of a coal mine containing both geometric and property information. The output uses a combination of layered voxels and boundary surfaces: the boundary surfaces are stored as a triangular mesh, with side lengths not exceeding 1 meter in the main mining zone and not exceeding 3 meters in the periphery; the voxel resolution remains consistent with the aforementioned values, and high uncertainty areas are marked. To facilitate subsequent applications, this embodiment adds a directional gradient index and a total confidence interval width to each property voxel, enabling direct screening of high-risk areas in gas control, ground pressure management, and mining design. Thus, the three-dimensional geological model of the coal mine possesses comprehensive characteristics of continuous strata, separable fractures, grouped properties, and usable uncertainty, directly supporting subsequent simulation and decision-making processes.

[0049] In step 600 of this embodiment, a consistency verification mechanism is established based on production monitoring data and physical constraints. Production monitoring data includes mine pressure monitoring, microseismic event location, roadway and working face deformation measurement points, ventilation volume and pressure, gas concentration and extraction rate, etc.; physical constraints include mechanical equilibrium constraints, ventilation network conservation constraints, and gas material balance constraints. The verification objects are the stratigraphic boundary set, property distribution model, and uncertainty results. To ensure operability, this embodiment sets in-situ quantification criteria: boundary geometry verification uses the distance from the newly added contact point to the boundary as the measure, with the median error in the near field of the roadway not exceeding 0.5 meters and in the far field of the mine field not exceeding 1.0 meter; topological verification requires that the boundary surfaces of the same layer have no self-intersections and the minimum size of the pores is not less than 2 meters; property verification uses the mean absolute error of the reserved samples as the measure, with thickness not exceeding 0.3 meters, porosity not exceeding 0.03, gas content not exceeding 3.0 cubic meters per ton, and rock mass strength not exceeding 15 MPa; uncertainty verification uses the interval coverage rate as the measure, with the actual coverage rate of the nominal 90% interval not less than 85%. If any criterion is not met, it is determined that there is a verification discrepancy.

[0050] This embodiment locates and clusters the verification differences on a spatial grid, calculating the difference intensity and coverage area of ​​each grid block. When the difference intensity of a continuous region exceeds a threshold and the area is not less than the designed minimum repair unit, subsequent adaptive adjustments are triggered. The difference intensity threshold is 0.5 meters in the geometric dimension, 50% of the corresponding dimensional threshold in the property dimension, and 10% of the nominal coverage gap in the uncertainty dimension. The minimum repair unit is 20×20×10 meters in the mining zone and 50×50×20 meters in the outer area. To avoid frequent fluctuations, this embodiment uses a hysteresis criterion: adjustment is triggered only if the same area fails to meet the standard consecutively in two independent verifications; if the difference intensity change between two adjacent verifications is less than 10%, triggering is temporarily suspended, and observation intensification is prioritized.

[0051] In step 700 of this embodiment, adaptive adjustments are implemented within the trigger area in an outward-to-inward order. First, the source weights are re-estimated, and fine-tuned based on the improvement in consistency of the monitoring data. Each adjustment does not exceed 20% of the original value and is subject to upper and lower limits: a lower limit of 0.05 and an upper limit of 0.60. Any excess weight is proportionally allocated to other sources. Then, minimal necessary adjustments are made to the local model parameters: directional information is corrected within a small angle range, with the angle change not exceeding 10°; the fracture jump amplitude is fine-tuned according to the strength of monitoring evidence, with each adjustment not exceeding 30% of the original level; the anisotropy ratio is limited to between 0.1 and 0.5; and the regularization intensity is set in zones between 0.01 and 0.2. These adjustments only take effect within the differential positioning area and its 30-meter buffer zone, and a local recalculation is performed after each round of adjustments. The mesh refinement during local recalculation does not exceed twice the original resolution.

[0052] This embodiment employs an incremental update strategy: only the triggered region is recalculated and smoothly spliced ​​with adjacent regions at the boundary. A rapid verification is performed immediately after each update, recalculating geometric and property errors and re-statistically calculating uncertainty coverage. Convergence is considered complete when all triggered regions simultaneously meet the quantitative criteria in step 600 during two consecutive verifications, and the changes in core indicators do not exceed 5% (e.g., the average absolute error decrease in thickness and gas content is less than 5%). This embodiment records key points of each update, including the adjusted data source and its weight range, the updated parameter type and value range, the affected spatial range, and the update time, to support subsequent technical reproduction and safety assessment. After convergence, a verified and updated 3D geological model of the coal mine is output for subsequent production decisions and risk assessment.

[0053] Corresponding to the above methods, such as Figure 2 As shown, this embodiment also provides a high-precision three-dimensional geological modeling system for coal mines based on multi-source heterogeneous data fusion, including: The semantic reference construction unit is used to construct a semantic system covering geological object elements, structural elements, geological property attributes and observation elements in the mining area, determine standardized description rules, determine unified time and space coordinate benchmarks, formulate alignment anchor point generation rules, and generate an alignment anchor point set based on geological objects with the same name, landmark interfaces and structural signs, thus obtaining the semantic reference model and the alignment anchor point set. The multi-source alignment and feature normalization unit is used to map multi-source heterogeneous geological observation data according to the semantic reference model and the unified time and space coordinate benchmark to form an alignment feature set, and calculate the source weight based on the source accuracy and reliability to obtain the alignment feature set and the source weight. The discontinuous structure identification and inversion unit is used to perform inversion based on the alignment feature set, according to the multi-evidence consistency rule and combined with geomechanical priors, to generate a candidate set of discontinuous geological structures. Implicit stratigraphic reconstruction unit is used to establish implicit stratigraphic functions, combine the geological contact points in the semantic reference model with the marker interface, and set boundary jump conditions with the candidate set of discontinuous geological structures as geometric constraints to solve the implicit stratigraphic functions, thereby obtaining a set of stratigraphic boundaries and structural geometric relationships that satisfy geometric and topological consistency. The anisotropy estimation and uncertainty analysis unit is used to construct an anisotropy estimation model that considers structural connectivity and directionality, using the stratigraphic boundary set as a spatial framework. It performs joint estimation of the target geological properties and outputs uncertainty results, forming a property distribution model and uncertainty results, and obtaining a three-dimensional geological model of the coal mine that includes the stratigraphic boundary set, the property distribution model and the uncertainty results.

[0054] The beneficial effects of this invention are as follows: (1) This invention establishes a semantic reference model covering geological object elements, structural elements, geological property attributes, and observation elements, and adopts a unified time and space coordinate benchmark and alignment anchor point generation rule. This invention eliminates naming differences, dimensional differences, and coordinate differences between different sources from the data entry point, achieving source consistency of "same name, same meaning, same domain, same scale". Compared with the traditional approach that relies on post-processing, this mechanism solidifies semantic, spatiotemporal, and quality elements into standards at the very beginning of the model chain, avoiding mismatches and drifts in the subsequent inversion and estimation stages. At the same time, since the semantic reference model is an open field system, it can incorporate new data and new attributes without breaking existing rules, demonstrating excellent scalability and full life cycle maintainability, which is a systematic innovation relative to the existing technology.

[0055] (2) Based on the construction of the alignment feature set, this invention introduces quantitative determination and normalization rules for source weights. It uses observation spatial resolution, observation noise level, spatial coverage, acquisition timeliness, and consistency score with the alignment anchor set as common criteria to form a comparable, transferable, and controllable "evidence contribution". Unlike existing schemes that use empirical weights or single signal-to-noise ratio weighting, this multi-factor, cross-dimensional weighting framework can dynamically suppress the influence of low-quality sources on the fusion results and maintain stable inversion input in scenarios of regional heterogeneity and sample sparsity. This significantly improves the robustness and stability of multi-source fusion and solves the long-standing problem of "strong evidence being submerged by weak evidence" in complex mining areas.

[0056] (3) This invention integrates evidence such as seismic discontinuous response, attribute mutation, curvature, deformation and near-field scanning using the multi-evidence consistency rule, and generates a candidate set of discontinuous geological structures by combining geomechanical priors and connectivity constraints; further, in the implicit reconstruction stage, discontinuous geometry is injected into the solution using boundary jump conditions. Compared with traditional single-evidence interpretation or post-fault cutting, this invention takes fractures, fissures and breccia zones as the first type of modeling objects, and through the closed loop of "evidence consistency - structural connectivity - physical priors - implicit jump", it makes geometric identification and model solution identical, significantly reducing over-interpretation and under-interpretation, especially in the fault neighborhood, effectively avoiding the false continuity caused by isotropic smoothing, which constitutes a key improvement over the existing technology.

[0057] (4) This invention uses implicit stratigraphic functions to express stratigraphic boundaries and applies jump conditions at discontinuous locations and anisotropic smoothing at continuous regions. The estimation neighborhood and smoothing direction are jointly defined by the tectonic direction field and tectonic connectivity relationship. Within the spatial framework of the stratigraphic boundary set, thickness, porosity, gas content, rock mass strength, etc. are jointly estimated and uncertainty results are output. This "geometry-property synergy" approach makes the continuity along the stratigraphic direction and the restriction across the stratigraphic direction the endogenous properties of the estimation, avoiding the "pulling effect" caused by isotropic interpolation based on Euclidean distance, and improving the overall prediction accuracy and consistency through the co-occurrence relationship across properties. At the same time, the uncertainty is explicitly output, providing a confidence interval level decision basis for mining design, gas control and ground pressure management, and has significant engineering applicability and interpretability advantages.

[0058] (5) In the preferred embodiment, the present invention sets up a consistency verification mechanism based on production monitoring data and physical constraints, and performs controlled fine-tuning and incremental updates on source weights and local model parameters when the difference exceeds the threshold, recalculating only locally in the triggering region until convergence is achieved. Unlike one-time "static modeling", this "observation-model-verification-update" engineering closed loop transforms the model from a static result into an iterative digital foundation: it ensures the rapid absorption of new evidence and continuous correction of the model, while avoiding the cost and instability risks caused by full-domain recalculation, significantly improving the timeliness and credibility of the model. This closed-loop mechanism complements the aforementioned semantic reference, evidence weighting, fracture constraints and anisotropy estimation, forming a creative combination of the overall method, reflecting comprehensive progress compared to existing technologies.

[0059] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0060] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A high-precision three-dimensional geological modeling method for coal mines based on multi-source heterogeneous data fusion, characterized in that, include: A semantic system covering geological object elements, structural elements, geological property attributes and observation elements in the mining area is constructed. Standardized description rules are determined, a unified time and space coordinate benchmark is determined, alignment anchor point generation rules are formulated, and an alignment anchor point set is generated based on geological objects with the same name, landmark interfaces and structural signs, thus obtaining a semantic reference model and an alignment anchor point set. Multi-source heterogeneous geological observation data are mapped according to the semantic reference model and the unified time and space coordinate benchmark to form an alignment feature set. The source weight is calculated based on the source accuracy and reliability to obtain the alignment feature set and the source weight. Based on the alignment feature set, inversion is performed according to the multi-evidence consistency rule and combined with geomechanical priors to generate a candidate set of discontinuous geological structures. An implicit stratigraphic function is established, and the geological contact points in the semantic reference model are combined with the marker interface. Boundary jump conditions are set with the candidate set of discontinuous geological structures as geometric constraints. The implicit stratigraphic function is solved to obtain the stratigraphic boundary set and structural geometric relationship that satisfy geometric and topological consistency. Using the stratigraphic boundary set as a spatial framework, an anisotropic estimation model considering structural connectivity and directionality is constructed to jointly estimate the target geological properties and output uncertainty results, forming a property distribution model and uncertainty results, and obtaining a three-dimensional geological model of the coal mine that includes the stratigraphic boundary set, the property distribution model and the uncertainty results.

2. The method for high-precision three-dimensional geological modeling of coal mines based on multi-source heterogeneous data fusion according to claim 1, characterized in that, Also includes: A consistency verification mechanism is established based on production monitoring data and physical constraints to verify the stratigraphic boundary set, the property distribution model, and the uncertainty results. When the verification difference exceeds the preset threshold, the source weight and local model parameters are adaptively adjusted and incremental updates are implemented until the convergence requirement is met, and the verified and updated three-dimensional geological model of the coal mine and the model source record are output.

3. The method for high-precision three-dimensional geological modeling of coal mines based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The unified time and space coordinate benchmark consists of a unified spatial coordinate benchmark for the mining area, a surface elevation benchmark, and a time reference benchmark; the parameter set corresponding to the unified time and space coordinate benchmark includes coordinate transformation parameters and time alignment parameters, and the parameter set is used to uniformly map the multi-source heterogeneous geological observation data to the unified time and space coordinate benchmark.

4. The method for high-precision three-dimensional geological modeling of coal mines based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The alignment anchor point generation rule consists of a spatial tolerance threshold, a temporal tolerance threshold, and a consistency scoring rule; the alignment anchor point set includes geological contact point anchor points, landmark interface anchor points, and tectonic sign anchor points; the screening condition corresponding to the alignment anchor point set is a conflict threshold, which is used to remove conflicting anchor points to obtain a stable alignment anchor point set.

5. The method for high-precision three-dimensional geological modeling of coal mines based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The determination of the source weights is based on the observation spatial resolution, observation noise level, spatial coverage, acquisition timeliness, and the consistency score of the alignment anchor set. The normalization rule corresponding to the source weights is used to limit the source weights within a preset numerical range so that the multi-evidence consistency rule and the anisotropy estimation model can call them.

6. The method for high-precision three-dimensional geological modeling of coal mines based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The generation constraints of the candidate set of discontinuous geological structures consist of a multi-evidence consistency rule and a connectivity constraint; the regularization term corresponding to the generation constraints is a sparsity regularization, which is used to suppress isolated responses and improve the reliability of the candidate set of discontinuous geological structures.

7. The method for high-precision three-dimensional geological modeling of coal mines based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The energy functional corresponding to the implicit stratigraphic function consists of a fitting term, a smoothing term, and a discontinuity penalty term; the boundary condition set corresponding to the discontinuity penalty term is the boundary jump condition set, the position of which is determined by the candidate set of discontinuous geological structures, and the boundary jump condition set is used to ensure the geometric and topological consistency of the stratigraphic boundary set at the fracture location.

8. The method for high-precision three-dimensional geological modeling of coal mines based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The expression for the implicit stratigraphic function is: exist : , exist , ,in ; in, For modeling area; The implicit stratigraphic function is defined as the implicit stratigraphic function, and the zero level set of the implicit stratigraphic function corresponds to the stratigraphic boundary. This refers to the signed distance data item generated from the geological contact points and marker interfaces in the semantic reference model; A positive regularization coefficient is used to balance the data terms and the anisotropic smoothing terms; This refers to the set of discontinuous surfaces corresponding to the candidate set of discontinuous geological structures. and These are the function lateral limits across the discontinuous surface, respectively; n represents the function jump amplitude determined by the discontinuous indication intensity. Let be the unit normal vector of the discontinuous surface; For anisotropic metric tensors; The unit direction field is determined by the strike and dip angle of the strata; It is the anisotropy ratio, used to distinguish the smoothness intensity along the structural direction from that in the transverse direction; It is a unit array.

9. The method for high-precision three-dimensional geological modeling of coal mines based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The directional information of the anisotropy estimation model is jointly determined by the structural orientation field and the structural connectivity relationship; the structural orientation field is composed of stratigraphic strike information, stratigraphic dip information and normal information of the candidate set of discontinuous geological structures; the joint estimation object includes at least one of coal seam thickness, porosity, gas content and rock mass strength.

10. A high-precision three-dimensional geological modeling system for coal mines based on multi-source heterogeneous data fusion, characterized in that, include: The semantic reference construction unit is used to construct a semantic system covering geological object elements, structural elements, geological property attributes and observation elements in the mining area, determine standardized description rules, determine unified time and space coordinate benchmarks, formulate alignment anchor point generation rules, and generate an alignment anchor point set based on geological objects with the same name, landmark interfaces and structural signs, thus obtaining the semantic reference model and the alignment anchor point set. The multi-source alignment and feature normalization unit is used to map multi-source heterogeneous geological observation data according to the semantic reference model and the unified time and space coordinate benchmark to form an alignment feature set, and calculate the source weight based on the source accuracy and reliability to obtain the alignment feature set and the source weight. The discontinuous structure identification and inversion unit is used to perform inversion based on the alignment feature set, according to the multi-evidence consistency rule and combined with geomechanical priors, to generate a candidate set of discontinuous geological structures. Implicit stratigraphic reconstruction unit is used to establish implicit stratigraphic functions, combine the geological contact points in the semantic reference model with the marker interface, and set boundary jump conditions with the candidate set of discontinuous geological structures as geometric constraints to solve the implicit stratigraphic functions, thereby obtaining a set of stratigraphic boundaries and structural geometric relationships that satisfy geometric and topological consistency. The anisotropy estimation and uncertainty analysis unit is used to construct an anisotropy estimation model that considers structural connectivity and directionality, using the stratigraphic boundary set as a spatial framework. It performs joint estimation of the target geological properties and outputs uncertainty results, forming a property distribution model and uncertainty results, and obtaining a three-dimensional geological model of the coal mine that includes the stratigraphic boundary set, the property distribution model and the uncertainty results.

Citation Information

Cited By

  • Green mine construction planning method based on three-dimensional modeling

    CN121639962A

  • Regional geological three-dimensional model construction method based on multi-source data fusion

    CN121837532A

  • Coal field aquifer spatial distribution analysis method based on surveying and mapping fusion

    CN122115769A

  • Coalfield aquifer spatial distribution analysis method based on surveying and mapping fusion

    CN122115769B