Multi-source three-dimensional geological data fusion and management method and related equipment

Abstract

The application discloses a kind of multi-source three-dimensional geological data fusion and management method and related equipment, it is related to geological information technical field, method includes: obtaining standardization geological data;Based on bayesian fusion framework, standardization geological data is fused;Based on the probability constraint provided by posterior probability distribution, multi-scale three-dimensional geological model is constructed using potential field interpolation method;Using Monte Carlo simulation method, the uncertainty quantification is carried out to multi-scale three-dimensional geological model;Multi-scale three-dimensional geological model and uncertainty quantification result are stored to the data management platform in line with OGC standard, and the visualization of geological layer is provided through Web service interface, query and access.This application greatly improves the sharing efficiency of geological achievement data, reuse value and application convenience, and enhances the intelligent level of geological data analysis and decision support capability as a whole.

Description

Technical Field

[0001] This invention relates to the field of geological information technology, and in particular to a method and related equipment for the fusion and management of multi-source three-dimensional geological data. Background Technology

[0002] In the field of geological data processing, geological data management currently faces multiple challenges: diverse data sources (borehole data, geophysical exploration, remote sensing images, etc.), inconsistent data formats, large differences in spatial scales, and discontinuous time series, all of which make data integration difficult. These issues have further triggered a series of chain reactions: First, format heterogeneity prevents direct interaction between different data sources. Manual format conversion is not only time-consuming and labor-intensive, but also prone to introducing data bias due to subjective operation, thus reducing data quality. Second, significant differences in spatial scale make it difficult to achieve high-precision alignment of multi-source data during the spatial registration stage. The fused geological model often exhibits local detail distortion or a lack of overall spatial consistency. Third, the discontinuity of time series makes it difficult for existing technologies to construct continuous geological evolution time series models, failing to meet the high requirements for time dimension continuity in application scenarios such as dynamic early warning of geological disasters and monitoring of underground resource extraction processes. Fourth, the lack of a unified standardized management system leads to scattered data storage, low retrieval efficiency, and a large amount of valuable geological data being idle. At the same time, repeated collection occurs frequently, resulting in a waste of human and material resources. Fifth, traditional fusion methods mostly use static empirical rules, which are difficult to adaptively handle the conflict and complementarity of multi-source information in complex geological environments, resulting in insufficient reliability and robustness of the fusion results. In addition, existing technologies generally lack a mechanism for quantitatively assessing the uncertainty of the fusion process and results, failing to provide a clear confidence reference for subsequent geological decisions and increasing decision-making risks.

[0003] To address these challenges, existing technologies primarily rely on a series of relatively independent processing steps and methods. At the data integration level, manual or semi-automatic tools are typically used to convert data from different formats and unify coordinates, attempting to achieve initial data aggregation. At the data fusion and modeling level, traditional methods often employ multi-source data integration strategies based on empirical weights or simple statistical rules, frequently constructing deterministic 3D geological models primarily from a single dominant data source (such as dense boreholes), while supplementing this with spatial interpolation algorithms to generate continuous geological surfaces or attribute fields. At the data management and service level, a common practice is to establish centralized databases or file systems to store the results data and provide data query and visualization functions through customized client software.

[0004] However, these existing technologies suffer from the following significant drawbacks: severe data silos and a lack of a unified data fusion framework; 3D modeling accuracy limited by a single data source and insufficient uncertainty quantification capabilities; and a lack of intelligent data quality control mechanisms. Specifically, geological data (such as borehole, geophysical, and geochemical data) collected by different departments or projects are often stored in independent systems with inconsistent formats and metadata standards, leading to inefficient cross-source data association and retrieval, creating information barriers. This poses a significant obstacle to "acquiring and standardizing geological data from different sources." In terms of modeling, a single data source (such as relying solely on borehole data) struggles to cover the spatial distribution details of complex geological bodies. The handling of data noise and missing values ​​during modeling relies heavily on empirical assumptions, failing to systematically quantify the probability distribution of model parameters. This results in ambiguous reliability of predictions regarding geological body boundaries and lithological distributions, lacking a robust fusion framework capable of producing "posterior probability distributions characterizing geological parameters." Furthermore, traditional methods often lack probabilistic constraints when "constructing multi-scale 3D geological models using potential field interpolation," leading to models that are insensitive to data conflicts and insufficient utilization of prior knowledge. Furthermore, traditional quality control relies heavily on manual sampling inspections, which struggles to meet the integrity and consistency verification requirements of massive, multi-source data. It is also inefficient and prone to overlooking outliers (such as borehole depth errors and geophysical signal interference), directly impacting the reliability of subsequent fusion results. Most critically, existing technologies generally lack a systematic step of "quantifying uncertainty using Monte Carlo simulation methods," failing to provide probabilistic "uncertainty quantification results" for the model. Moreover, data sharing and reuse are difficult, and there is a lack of "OGC-compliant data management platforms" and standardized "Web service interfaces" to efficiently "provide visualization, querying, and access to geological layers."

[0005] Therefore, there is an urgent need for a technical solution that can systematically address the entire process of multi-source heterogeneous data, from standardized access, probabilistic fusion, high-precision modeling, uncertainty quantification to intelligent data services, in order to overcome the limitations of existing methods and improve the depth of geological data utilization and the reliability of decision support. Summary of the Invention

[0006] The technical problem to be solved by this invention is to address the shortcomings of existing technologies, specifically by providing a method and related equipment for the fusion and management of multi-source three-dimensional geological data, as detailed below: 1) In a first aspect, the present invention provides a method for the fusion and management of multi-source three-dimensional geological data, the specific technical solution of which is as follows: Geological data from different sources are acquired and standardized to obtain standardized geological data. Based on a Bayesian fusion framework, the standardized geological data are fused to obtain a posterior probability distribution representing geological parameters. Based on the probability constraints provided by the posterior probability distribution, a multi-scale three-dimensional geological model is constructed using the potential field interpolation method. The uncertainty of the multi-scale three-dimensional geological model is quantified using the Monte Carlo simulation method to generate uncertainty quantification results corresponding to the multi-scale three-dimensional geological model. The multi-scale three-dimensional geological model and uncertainty quantification results are stored in a data management platform that conforms to the OGC standard, and visualization, querying, and access to geological layers are provided through a Web service interface.

[0007] The beneficial effects of the multi-source three-dimensional geological data fusion and management method provided by this invention are as follows: Geological data from different sources were acquired and standardized, establishing a unified data processing framework. This effectively overcame integration barriers caused by heterogeneous data formats and inconsistent spatial benchmarks, laying a high-quality data foundation for subsequent fusion. A Bayesian fusion framework was introduced for data fusion, enabling the probabilistic acquisition of posterior probability distributions representing geological parameters. This adaptively reconciled conflicts and complementarities among multi-source data, improving the robustness of the fusion process and the reliability of the results. Based on the probabilistic constraints provided by the posterior probability distribution, a multi-scale 3D geological model was constructed using potential field interpolation. This allowed the 3D modeling process to be probabilistically guided by multi-source data, improving the accuracy and geological rationality in depicting the spatial morphology of complex geological structures. Furthermore, Monte Carlo simulation was employed to quantify uncertainty, generating uncertainty quantification results corresponding to the multi-scale 3D geological model. This enabled a quantitative assessment of the model output credibility, providing a clear probabilistic basis for risk decision-making based on the geological model. Finally, the multi-scale three-dimensional geological model and uncertainty quantification results are stored in a data management platform that conforms to the OGC standard, and the visualization, query and access of geological layers are provided through a web service interface. This has built an open and standardized data service system, which has greatly improved the sharing efficiency, reuse value and application convenience of geological data, and enhanced the overall intelligence level and decision support capability of geological data analysis.

[0008] Based on the above scheme, the multi-source three-dimensional geological data fusion and management method of the present invention can be further improved as follows.

[0009] Furthermore, based on the Bayesian fusion framework, standardized geological data are fused to obtain the posterior probability distribution representing the geological parameters. This includes: constructing a prior probability distribution representing the geological parameters, which integrates geological rules and expert experience; establishing a corresponding likelihood function for each data source in the standardized geological data, which describes the degree of matching between the data source and the multi-scale three-dimensional geological model; obtaining a sample set of the posterior probability distribution of the geological parameters using the Markov chain Monte Carlo sampling method based on the prior probability distribution and the likelihood functions of all data sources; and calculating the posterior probability distribution of the geological parameters based on the sample set of the posterior probability distribution.

[0010] The beneficial effects of adopting the above-mentioned further scheme are as follows: By constructing a prior probability distribution that integrates geological rules and expert experience, and establishing a corresponding likelihood function for each standardized geological data source, this step provides a rigorous probabilistic foundation for multi-source data fusion. The Markov chain Monte Carlo sampling method can efficiently handle high-dimensional, nonlinear geological parameter estimation problems, thereby stably obtaining a sample set of posterior probability distributions. The finally calculated posterior probability distribution of geological parameters not only provides the best estimate of the parameters, but more importantly, it fully quantifies the uncertainty of the estimation results, realizing an upgrade of the fusion process from empirical rules to probabilistic inference, significantly improving the objectivity and reliability of the data fusion results.

[0011] Furthermore, based on the probabilistic constraints provided by the posterior probability distribution, a multi-scale three-dimensional geological model is constructed using the potential field interpolation method. This includes: determining the probabilistic information of geological interface feature points and stratigraphic attribute distributions based on the posterior probability distribution; defining a potential field function describing the spatial morphology of the geological interface based on the probabilistic information and borehole layering data; solving the potential field equation corresponding to the potential field function using numerical methods to obtain a continuous potential field distribution in three-dimensional space; extracting equipotential surfaces with preset values ​​from the continuous potential field distribution and using these equipotential surfaces as geological interfaces in the multi-scale three-dimensional geological model; and applying geological rules of stratigraphic stacking and fault cutting to perform topology optimization on the geological interfaces, thus completing the construction of the multi-scale three-dimensional geological model.

[0012] The beneficial effects of adopting the above-mentioned further scheme are: determining the probabilistic information of geological interface feature points and stratigraphic attribute distributions based on posterior probability distributions provides reliable quantitative constraints for subsequent modeling processes. By defining and solving the potential field equations, discrete, multi-source probabilistic information is transformed into a continuous spatial potential field distribution, realizing a mathematical and refined description of the geological interface morphology. Isopotential surfaces are extracted from the continuous potential field as geological interfaces, and topology optimization is performed using stratigraphic stacking and fault cutting rules, ensuring that the constructed multi-scale three-dimensional geological model has a high degree of rationality and consistency in geometric accuracy, spatial relationships, and geological logic.

[0013] Furthermore, the Monte Carlo simulation method is used to quantify the uncertainty of the multi-scale three-dimensional geological model, generating uncertainty quantification results corresponding to the multi-scale three-dimensional geological model. This includes: generating a set of perturbation samples conforming to a preset probability distribution based on the error characteristics of different data sources in the standardized geological data; re-executing the data fusion and three-dimensional modeling process for each sample in the perturbation sample set to obtain the corresponding multi-scale three-dimensional geological model output; performing statistical analysis on the set of multi-scale three-dimensional geological model outputs to calculate the mean, standard deviation, and confidence interval of the geological parameters; and generating a probability distribution function and risk zoning map characterizing the uncertainty of the multi-scale three-dimensional geological model based on the calculated mean, standard deviation, and confidence interval of the geological parameters, as the uncertainty quantification result.

[0014] The beneficial effects of adopting the above-mentioned further scheme are as follows: A set of perturbation samples is generated based on the error characteristics of different data sources in the standardized geological data, systematically simulating the propagation process of uncertainty from the input data to the final geological model. By re-executing the complete fusion and modeling process for each perturbation sample and obtaining a series of possible multi-scale three-dimensional geological model outputs, sufficient statistical samples are provided for uncertainty analysis. Based on this, statistical analysis yields the mean, standard deviation, and confidence intervals of geological parameters, and generates probability distribution functions and risk zoning maps. The final uncertainty quantification results can intuitively and quantitatively reveal the credibility of the model predictions and the spatial distribution of risk, providing a solid probabilistic basis for model-based resource assessment and engineering decision-making.

[0015] 2) Secondly, the present invention also provides a multi-source three-dimensional geological data fusion and management system, the specific technical solution of which is as follows: The system includes an acquisition and processing module, a fusion module, a construction module, a generation module, and a management module. The acquisition and processing module acquires and standardizes geological data from different sources to obtain standardized geological data. The fusion module fuses the standardized geological data using a Bayesian fusion framework to obtain a posterior probability distribution representing geological parameters. Based on the probability constraints provided by the posterior probability distribution, the fusion module constructs a multi-scale three-dimensional geological model using potential field interpolation. The generation module uses Monte Carlo simulation to quantify the uncertainty of the multi-scale three-dimensional geological model, generating uncertainty quantification results corresponding to the multi-scale three-dimensional geological model. The management module stores the multi-scale three-dimensional geological model and uncertainty quantification results in a data management platform compliant with OGC standards and provides visualization, querying, and access to geological layers through a web service interface.

[0016] Based on the above scheme, the multi-source three-dimensional geological data fusion and management system of the present invention can be further improved as follows.

[0017] Furthermore, the fusion module is specifically used for: constructing a prior probability distribution characterizing geological parameters, which integrates geological rules and expert experience; establishing a corresponding likelihood function for each data source in the standardized geological data, which describes the degree of matching between the data source and the multi-scale three-dimensional geological model; obtaining a sample set of the posterior probability distribution of geological parameters using the Markov chain Monte Carlo sampling method based on the prior probability distribution and the likelihood functions of all data sources; and calculating the posterior probability distribution of geological parameters based on the sample set of the posterior probability distribution.

[0018] Furthermore, the construction module is specifically used for: determining the probability information of geological interface feature points and stratigraphic attribute distribution based on posterior probability distribution; defining a potential field function describing the spatial morphology of the geological interface based on the probability information and borehole layering data; solving the potential field equation corresponding to the potential field function using numerical methods to obtain the continuous potential field distribution in three-dimensional space; extracting equipotential surfaces with preset values ​​from the continuous potential field distribution and using the equipotential surfaces as geological interfaces in the multi-scale three-dimensional geological model; and performing topology optimization on the geological interface using geological rules of stratigraphic stacking and fault cutting to complete the construction of the multi-scale three-dimensional geological model.

[0019] Furthermore, the generation module is specifically used to: generate a set of perturbation samples that conform to a preset probability distribution based on the error characteristics of different data sources in the standardized geological data; re-execute the data fusion and 3D modeling process for each sample in the perturbation sample set to obtain the corresponding multi-scale 3D geological model output; perform statistical analysis on the set of multi-scale 3D geological model outputs to calculate the mean, standard deviation, and confidence interval of the geological parameters; and generate a probability distribution function and risk zoning map that characterize the uncertainty of the multi-scale 3D geological model based on the calculated mean, standard deviation, and confidence interval of the geological parameters, as the uncertainty quantification result.

[0020] 3) In a third aspect, the present invention also provides an electronic device, the electronic device including a processor coupled to a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor, so that the electronic device can realize any of the above-mentioned multi-source three-dimensional geological data fusion and management methods.

[0021] 4) In a fourth aspect, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-mentioned methods for the fusion and management of multi-source three-dimensional geological data.

[0022] It should be noted that the beneficial effects of the technical solutions of the second to fourth aspects of the present invention and their corresponding possible implementations can be found in the above description of the technical effects of the first aspect and its corresponding possible implementations, and will not be repeated here. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments of the present invention will be briefly introduced below: Figure 1 This is a flowchart illustrating a multi-source three-dimensional geological data fusion and management method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a multi-source three-dimensional geological data fusion and management system according to an embodiment of the present invention. Detailed Implementation

[0024] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0025] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0026] like Figure 1 As shown in the figure, a multi-source three-dimensional geological data fusion and management method according to an embodiment of the present invention includes the following steps: S1. Obtain and standardize geological data from different sources to obtain standardized geological data. The specific implementation process is as follows: S10. Obtain geological data from different sources. Geological data from different sources refers to the collection of original geological information acquired through various technical means and equipment during geological surveys, explorations, and monitoring activities, exhibiting natural heterogeneity in data format, spatial scale, time series, and semantic connotation. These data sources are extensive and mainly include: borehole data, originating from geological drilling projects, obtained through core logging, well logging instruments, or in-situ testing, recording the stratigraphic depth, lithology, physicochemical properties, and structural information of specific underground points; geophysical exploration data, such as data obtained through seismic, electromagnetic, gravity, or magnetic exploration methods, reflecting the differences in physical properties and structural contours of the underground medium in the form of profiles or volume data; remote sensing imagery, obtained through satellite or airborne sensor observations of the Earth, providing large-scale surface spectral, textural, and topographic information; geographic information system vector data, typically storing spatial objects and their attributes such as geological boundaries, faults, and attitudes in the form of point, line, and area elements; and time-series monitoring data such as groundwater level and geostress, recording the changes in geological environmental parameters at specific locations over time. Acquiring this data is the starting point for the integration work. It is necessary to establish access interfaces that are compatible with multiple transmission protocols and storage media to ensure the integrity of the original data during import.

[0027] S11. Since raw data from different sources may be stored in CSV tables, Shapefile vector files, HDF5 scientific data format, LAS point cloud format, or proprietary binary formats, the format conversion tool needs to parse these heterogeneous formats and uniformly convert them into an intermediate data model for internal system processing. This conversion process not only extracts the geometric, attribute, and metadata information from the data, but also needs to standardize and encapsulate information such as data source, acquisition time, and accuracy description according to predefined metadata standards, forming data objects with a unified structural description, laying the foundation for subsequent coordinate, quality, and semantic processing.

[0028] S12. To ensure that all geological data can be fused and analyzed within a unified three-dimensional spatial reference framework, their spatial coordinate systems must be transformed to a standard coordinate system, such as the 2000 National Geodetic Coordinate System. The coordinate transformation process is based on the original coordinate reference system information of each data source, applying a seven-parameter or four-parameter transformation model to accurately transform the coordinates of all spatial points in the data. For early data lacking a clear coordinate reference, spatial registration is required through methods such as feature point matching. This step eliminates spatial positional deviations caused by inconsistent coordinate benchmarks and forms the geometric basis for realizing spatial overlay and fusion analysis of multi-source data.

[0029] S13. To address the issues of missing values, anomalies, and noise in the raw data, systematic quality control is required. For random noise introduced into geophysical exploration data due to instrument or environmental interference, wavelet transform and other methods are used for filtering. Wavelet transform effectively separates and suppresses noise components by decomposing the signal into sub-bands of different frequencies. For missing attribute values ​​in borehole data due to sampling intervals or recording omissions, spatial interpolation techniques such as Kriging interpolation are used for reasonable filling. Kriging interpolation is an optimal unbiased estimation method based on variogram theory, which can predict attribute values ​​at unknown locations using the structural relationships of known spatial sample points. Simultaneously, outlier detection rules based on statistical regularities and geological logic are established to automatically identify and remove erroneous data records that significantly deviate from reasonable ranges, improving the overall quality and reliability of the dataset.

[0030] S14. After completing the format, coordinate, and quality processing, the semantic information of the data needs to be standardized. This includes semantic mapping and reclassification of geological object types such as lithology, stratigraphy, and structure from different legends or interpretation standards, based on a unified geological classification and coding system, to ensure consistent naming and coding of the same geological entities. Simultaneously, the metadata record for each data point should be improved and standardized to include complete data identification, spatial reference, accuracy evaluation, processing history, and other information. After the above series of processes, the initially acquired heterogeneous geological data is ultimately transformed into "standardized geological data" with unified coordinate benchmarks, standardized data structures, controlled data quality, and consistent semantic descriptions, providing high-quality, computable input for subsequent data fusion engines.

[0031] S2. Based on the Bayesian fusion framework, standardized geological data are fused to obtain the posterior probability distribution representing geological parameters.

[0032] The Bayesian fusion framework is a mathematical framework for probabilistic reasoning and information integration based on Bayes' theorem. In this framework, our understanding of unknown geological models is represented as probability distributions. Starting with the prior probability distribution of geological parameters that integrates prior knowledge and rules, the framework incorporates evidence from different observational data sources in the form of likelihood functions. It then uses Bayes' theorem to probabilistically update all information, ultimately obtaining the posterior probability distribution of geological parameters that reflects the combined influence of multi-source information. This framework provides a rigorous yet flexible mathematical foundation for handling the uncertainties, conflicts, and complementary relationships prevalent in geological data.

[0033] S2 specifically includes: S20. Constructing a priori probability distributions to characterize geological parameters. The prior probability distribution integrates geological rules and expert experience; Among them, the prior probability distribution of geological parameters This is a probabilistic expression of the uncertainty of unknown geological model parameters *m* before any observational data is obtained. The construction process begins with the systematic collection and formalization of existing geological knowledge of the study area. Specifically, it integrates regularities from regional geological surveys, such as the vertical stacking sequence of strata, the lateral facies transition patterns of lithology, and fault development patterns, transforming these geological rules into mathematical constraints. Taking sedimentary sequences as an example, based on Wolfe's law, the vertical relationship of strata can be transformed into a state transition probability matrix using a first-order Markov chain model. Integrating the experience of geological experts, their qualitative judgments on specific geological phenomena (such as the probability of occurrence of a certain lithology or the possible range of fault dip angles) are mapped into a continuous probability distribution form through methods such as fuzzy membership functions or interval probability assignment. Using a limited amount of existing hard data such as boreholes and outcrops within the study area, initial probabilistic models containing information such as lithological category proportions and ranges of physical property parameters are fitted using statistical methods such as kernel density estimation and variogram analysis. Ultimately, information from multiple dimensions, including geological rules, expert experience, and historical data, is integrated to form a comprehensive prior probability distribution that reflects the uncertainty of geological understanding, denoted as... .in, This represents the set of geological model parameters to be estimated.

[0034] Geological rules and expert experience are the core information sources for constructing the prior probability distribution of geological parameters. Geological rules are universal laws summarized from long-term geological practice, such as the stratigraphic superposition law, the cutting law, and the Walter facies law. They provide the basic relationships that geological entities should follow in space and time. Expert experience refers to the profound understanding and intuitive judgment of specific regions or geological phenomena accumulated by geological experts through long-term work, such as the estimation of the intensity of fault activity and the judgment of favorable mineralization areas. In Bayesian fusion, both are formalized as probabilistic constraints, jointly shaping the shape of the prior distribution.

[0035] The prior probability distribution of geological parameters is a probabilistic description of the uncertainty of possible values ​​of geological parameters based on existing geological knowledge, before incorporating specific current observational data. It is a probabilistic encapsulation of geological rules, expert experience, and historical data. This distribution defines the relative probability of different parameter values, providing a reasonable initial estimation range for subsequent Bayesian updates, thus constraining the solution space and guiding the direction of inference.

[0036] S21. Establish a corresponding likelihood function for each data source in the standardized geological data. The likelihood function describes the degree of matching between the data source and the multi-scale three-dimensional geological model; Likelihood function It serves as a bridge connecting geological models and observational data, quantifying the parameters of a given geological model. Under these conditions, current standardized geological data were observed. The probability of this. For each data source, a specific likelihood function needs to be established based on its physical principles and error characteristics. For borehole data, assume the observed formation interface depth. The error follows the mean of the model's predicted depth. variance is The normal distribution has the following likelihood function: , This represents the standard deviation of the observation error for borehole data. For geophysical exploration data such as seismic reflection data, the likelihood function typically involves forward modeling and takes the form of... ,in, It is a forward modeling operator for earthquakes. It is the covariance matrix that characterizes the noise in the data. This represents the observed seismic reflection data under the given geological model parameters m. The probability, Indicates a normal distribution. This represents the seismic forward modeling operator. For geological map data, its likelihood function can be constructed based on the spatial distance between the model prediction boundary and the observation boundary (such as Hausdorff distance). Each likelihood function... Key parameters (such as error variance) can be estimated through historical data statistical analysis, cross-validation, or hierarchical Bayesian methods.

[0037] S22. Based on the prior probability distribution and the likelihood function of all data sources, the posterior probability distribution of geological parameters is obtained using the Markov chain Monte Carlo sampling method. The sample set; Markov chain Monte Carlo sampling is a numerical computation method that constructs a Markov chain with a stationary distribution as the target posterior distribution and draws samples from this chain. The implementation process consists of three stages. The first stage is initialization, which sets an initial state for the Markov chain. It can be from the prior distribution The first stage involves randomly selecting a sample. The second stage is iterative sampling, which uses an improved sampling algorithm (such as the adaptive Metropolis-Hastings algorithm or the Hamiltonian Monte Carlo algorithm) for a large number of iterations. In each iteration... In the middle, according to the proposed distribution Generate a candidate model parameter Calculate the probability of accepting the candidate state. : ,in, Indicates: Candidate model parameters under observed data D. The unnormalized posterior probability, Indicates: Current model parameters The unnormalized posterior probability, Indicates: Given the observed data D, from the candidate model parameters Return to current model parameters The probability density of the proposed distribution, Indicates: from the current model parameters Proposal to candidate model parameters The probability density of the proposed distribution; It is the unnormalized joint posterior probability, which, according to Bayes' theorem, is proportional to the product of the prior and all likelihood functions: . Represents the set of all standardized geological data. This refers to the number of data source types. (In terms of probability) accept As a new state Otherwise, retain the old state. The third stage involves convergence assessment and sample collection, iterating until the Markov chain reaches a stationary state. This can be determined using convergence diagnostic tools such as the Gelman-Rubin statistic. After confirming convergence, samples are collected at regular intervals from subsequent iterations, ultimately yielding a set containing... A set of samples This sample set can be statistically considered as derived from the true posterior probability distribution. Extracted from [the source].

[0038] S23. Calculate the posterior probability distribution of geological parameters based on the sample set of the posterior probability distribution. .

[0039] The sample set is obtained using the Markov chain Monte Carlo sampling method. Afterwards, the posterior probability distribution The characteristics can then be obtained through statistical inference from these samples. For geological parameter vectors... Any scalar parameter in (For example, the porosity of a certain grid point, the thickness of a certain stratum), its sample value is The posterior mean estimate of this parameter The arithmetic mean of the samples is obtained as follows: Posterior standard deviation The formula used to measure the uncertainty of this parameter is as follows: The posterior probability density function of the parameters can be approximated by kernel density estimation of the samples to obtain a smooth curve. Simultaneously, the posterior confidence interval of the parameters can be calculated; for example, a 95% confidence interval can be determined by finding the 2.5% and 97.5% quantiles of the samples. For the entire geological model... Spatial prediction can be based on the integration of all sample models. By calculating the statistics (such as mean and variance) of attribute values ​​at each spatial location, the final probabilistic geological model and the accompanying uncertain spatial distribution field can be obtained.

[0040] The posterior probability distribution of geological parameters, within the Bayesian fusion framework, is an updated and complete probabilistic description of the uncertainty of geological parameters, combining the prior probability distribution with likelihood information from all observational data sources. It integrates all evidence from prior knowledge and current observational data and is the final output of Bayesian inference. This distribution not only provides optimal estimates of geological parameters (such as the posterior mean) but, more importantly, quantifies the uncertainty of the estimates (such as posterior variance and confidence intervals), providing a probabilistic basis for risk decision-making based on geological models.

[0041] S3. Based on the probability constraints provided by the posterior probability distribution, a multi-scale three-dimensional geological model is constructed using the potential field interpolation method, specifically including: S30. Determine the probabilistic information of geological interface feature points and stratigraphic attribute distribution based on posterior probability distribution; Geological interface feature points refer to control points in three-dimensional space that clearly belong to a specific geological interface (such as bedding planes or fault planes), such as stratigraphic boundaries revealed by boreholes or reflection layer sites interpreted by seismic profiles. Stratigraphic property distribution refers to the assignment of physical properties such as lithology and porosity in three-dimensional space. Determining the probabilistic characteristics of these information from the posterior probability distribution of geological parameters is a direct input for probabilistically constrained 3D modeling. Specifically, spatial statistics are performed on the posterior sample set of geological parameters obtained through Markov chain Monte Carlo sampling. For geological interface feature points, each posterior sample corresponds to a set of possible 3D spatial interface locations. By statistically analyzing the frequency with which each spatial location (such as a node in a regular 3D mesh) is identified as belonging to the target geological interface, a probability map of that location being an interface point can be obtained. Simultaneously, statistical quantities such as the mean and standard deviation of the interface burial depth can be calculated. For the distribution of stratigraphic properties, the distribution of each spatial grid cell assigned a certain lithology or physical property parameter value in all posterior samples is statistically analyzed to obtain the probability volume of lithology or the mean and variance fields of physical property parameters (such as porosity). These outputs, namely the spatialized probability map, mean field, and variance field, together constitute the probabilistic information of geological interface feature points and stratigraphic property distribution used to guide subsequent deterministic or probabilistic modeling.

[0042] Among them, the probabilistic information of geological interface feature points and stratigraphic attribute distribution is a data form that spatially expresses the results of Bayesian inference. It is generated by spatial statistics of the posterior probability distribution of geological parameters, specifically including the probability map of the occurrence of geological interfaces at various points in space, the spatial field of statistical quantities (mean, variance) of interface burial depth, and the spatial field of probability distributions or statistical characteristics of the values ​​of various stratigraphic attributes (such as lithology and physical properties). This information is no longer a definite numerical value or boundary, but rather describes the spatial uncertainty of the location of geological interfaces and stratigraphic attributes in the form of probabilistic or statistical quantities, providing a probabilistic constraint framework for subsequent modeling.

[0043] S31. Based on probability information and borehole stratification data, define a potential field function that describes the spatial morphology of geological interfaces. ; Potential function A potential field function is a scalar function defined in three-dimensional space, whose changes implicitly describe the spatial morphology of a geological interface. Typically, a target geological interface is defined as a specific isosurface (e.g., the zero isosurface) of this potential field function. Defining the potential field function. It is necessary to integrate information constraints from multiple sources. The first type is hard constraint points, mainly from borehole stratification data, which provide control points with relatively high certainty regarding the spatial location of geological interfaces. At these points... Above, the forced potential field function takes the value of zero, that is The second type is soft constraints, derived from the probability information generated in the first step. For example, a series of suspected interface points with high probabilities can be extracted from the interface probability map, or the mean field of the interface burial depth can be used. As a trend constraint, this information can be obtained by introducing penalty or regularization terms during the solution of the potential field equations, so that the obtained potential field function is as close as possible to the desired trend. Approaching zero. The potential function itself is usually required to be smooth to satisfy the assumption of continuity of geological interfaces in regions untouched by drastic tectonic alteration. This is achieved by requiring the potential function to satisfy the Laplace equation. To achieve this, in which, It is the Laplace operator. Thus, the definition of the potential function is transformed into a mathematical problem of finding solutions to the Laplace equation under hard constraints while simultaneously satisfying the soft constraint tendency as much as possible.

[0044] Borehole stratification data, derived from direct observations during geological drilling, records the top and bottom depths of each geological stratum traversed by the borehole trajectory, along with corresponding descriptions of lithology and age. It is one of the most reliable and accurate data sources for 3D geological modeling, providing vertical control information on the spatial location of geological interfaces at specific subsurface points. In the construction of potential field functions, borehole stratification points are typically used as hard constraints that must be satisfied, forcing the model to pass through these known points, thereby ensuring consistency between the model and direct observation data.

[0045] S32. Solve the potential field equation corresponding to the potential field function using numerical methods to obtain the continuous potential field distribution in three-dimensional space. Potential field equation, i.e., Laplace equation This is a partial differential equation, which often cannot be solved analytically under complex boundary conditions and internal constraints, and must be solved numerically using discretization methods. The first step is to discretize the three-dimensional mesh of the study area, dividing the entire three-dimensional space into a dense regular hexahedral mesh or an unstructured tetrahedral mesh. The second step is to discretize the differential equation and various constraints onto the mesh system. For the Laplace operator, the finite difference method is commonly used for discretization on regular meshes, while the finite element method is used on irregular meshes. After discretization, the original partial differential equation is transformed into a large system of linear equations: .in, It is a coefficient matrix, the structure of which is determined by the discrete format; It consists of the potential function values ​​to be determined at all grid nodes. A column vector arranged in a specific order; This is the right-hand side vector, which contains information from boundary conditions and internal constraints. Hard constraint points are provided by borehole layer data. The Dirichlet boundary conditions are directly substituted into the system of equations. Soft constraints derived from probabilistic information may be addressed by adding additional equations or modifying the right-hand side. The third step is to solve this large sparse linear system of equations. Due to the large number of grid cells and the high dimensionality of the equations, an efficient iterative method is needed, such as the preprocessed conjugate gradient method. After solving, the potential function values ​​at each grid node are obtained. These discrete values ​​together constitute a discrete approximate representation of the "continuous potential field distribution in three-dimensional space" covering the entire study area.

[0046] Among them, the potential field function of the geological interface spatial morphology is a scalar function defined in the three-dimensional space of the study area. The special isosurfaces of this function (often the zero isosurface) are used to implicitly represent a geological interface. The magnitude and sign of the function value can be understood as the "potential" or "sign distance" of a spatial point from the target interface; the potential at points on the interface is zero, while the potential values ​​above and below the interface are positive and negative (or opposite) respectively. By solving for the potential field function that satisfies certain constraints (such as passing through borehole points or meeting smoothness requirements), a mathematical description of complex geological interfaces can be obtained indirectly but accurately.

[0047] In this context, the continuous potential field distribution in three-dimensional space refers to the potential field function value obtained at every spatial point (usually represented as nodes in a dense mesh) within the entire three-dimensional modeling region after numerically solving the potential field equations. The set. It realizes the potential field function from a discrete but fully covered perspective. The approximation is that the distribution is continuous, meaning that the potential value at any point can be obtained by interpolation within the grid scale, providing a complete data foundation for the next step of accurately extracting equipotential surfaces (geological interfaces).

[0048] S33. Extract equipotential surfaces with preset values ​​from the continuous potential field distribution, and use the equipotential surfaces as geological interfaces in the multi-scale three-dimensional geological model. An equipotential surface is a surface in three-dimensional space formed by points with equal potential function values. This is defined after the potential function is... Under the convention of representing the target geological interface with a zero value, it is necessary to obtain discrete three-dimensional potential field data. Extract The isosurface is extracted. This process is implemented using isosurface extraction algorithms, most commonly the moving cube algorithm. The moving cube algorithm processes each voxel (a small cube composed of eight adjacent grid points). Based on the sign combination of the potential field values ​​at the eight vertices of the voxel, it calculates the intersection points of the isosurface with the edges of the voxel through linear interpolation, and connects these intersection points with a series of triangular facets to approximate the isosurface segment passing through the voxel. After traversing all voxels, these triangular facets are pieced together to form a continuous surface expressed by a triangular mesh. This surface is the geometry of the extracted geological interface. For cases with multiple strata, it is necessary to define and solve the potential field function for each stratum interface separately, and extract their respective zero isosurfaces. The extracted initial triangular mesh surface may have some irregularities due to data noise or discretization errors, requiring smoothing and topology checks to ensure that the surface is manifold and does not self-intersect. Ultimately, these geometrically processed equipotential triangular meshes were officially recognized as key geological structural surfaces such as stratigraphic interfaces and fault planes in multi-scale three-dimensional geological models.

[0049] S34. Apply geological rules of stratigraphic stacking and fault cutting to perform topological optimization of geological interfaces and complete the construction of a multi-scale three-dimensional geological model.

[0050] Multiple geological interfaces extracted directly from the potential field are independent geometric surfaces. Their spatial relationships may violate basic geological laws, requiring topology optimization based on geological rules. The first rule is the stratigraphic superposition rule, which states that in normal strata sequences undamaged by tectonic activity, the interfaces of overlying strata should always be vertically above the interfaces of underlying strata, without intersections or inversions. The optimization algorithm needs to detect the spatial relationships between all pairs of stratigraphic interfaces. For any discovered local overlaps, corrections are made by adjusting the local vertex elevations of one or two interfaces to ensure the overall superposition order is correct. The second rule is the fault cutting rule. Fault interfaces cut and discontinuate stratigraphic interfaces. The optimization process requires cutting the stratigraphic interfaces based on the fault model's attitude and displacement. First, the intersection line between the fault plane and the ground plane is determined. Then, based on the fault type (normal fault, reverse fault, etc.) and displacement, the portions of the ground plane on both sides of the fault are relatively moved to form a reasonable fault contact relationship, leaving fault traces on the ground plane. After these topological adjustments are completed, each stratigraphic interface and fault plane jointly encloses an independent three-dimensional region, corresponding to different geological units (such as a specific stratum or lithological body). Finally, the probabilistic information of stratigraphic attribute distribution (such as lithological probability volume) is assigned to the corresponding three-dimensional geological units, or representative attribute parameters are assigned to each unit. Thus, a "multi-scale three-dimensional geological model" integrating geometry, topological relationships, and attribute information is constructed. This model can output expressions at different scales or levels of detail as needed.

[0051] Among these, the geological rules governing stratigraphic superposition and fault cutting are principles in geology that describe the fundamental spatial relationships between geological bodies. The stratigraphic superposition rule states that, without tectonic disturbance, new strata overlay older strata, forming a normal, vertically stacked sequence. The fault cutting rule states that faults, as destructive structures, cut and disrupt existing stratigraphic sequences, causing relative displacement of the strata on either side of the fault plane. In the final stage of constructing multi-scale three-dimensional geological models, applying these rules to logically check and adjust the spatial relationships of the initially generated geometric surfaces is a crucial step in ensuring that the model conforms to geological common sense and possesses geological rationality.

[0052] Among them, multi-scale 3D geological models are 3D digital geological models capable of integrating and expressing everything from regional macrostructures to local details. They are not single models, but rather an organic combination of a series of 3D models with different spatial scales and content precisions, constructed through level-of-detail techniques or layered modeling strategies. For example, regional-scale models (e.g., 1:250,000) display the main structural framework and stratigraphic units; mining-scale models (e.g., 1:10,000) refine engineering geological stratification and structural morphology; and site-scale models (e.g., 1:5,000) meticulously depict the internal structure and property variations of rock and soil masses. Such models can meet the geological analysis needs of different stages and tasks.

[0053] S4. Using the Monte Carlo simulation method, the uncertainty of the multi-scale three-dimensional geological model is quantified, generating uncertainty quantification results corresponding to the multi-scale three-dimensional geological model, specifically including: S40. Based on the error characteristics of different data sources in the standardized geological data, generate a set of perturbation samples that conforms to a preset probability distribution. The error characteristics of different data sources refer to the inherent uncertainties and error patterns of each geological observation data due to its acquisition principle, instrument accuracy, environmental interference, and interpretation process. The purpose of generating a perturbation sample set is to simulate the potential impact of these errors on the final 3D geological model. The implementation process begins with a quantitative description of the error characteristics of each data source. For borehole stratification data, its depth error is typically assumed to follow a mean of zero and a standard deviation of [missing value]. normal distribution For geophysical exploration data (such as seismic velocity), the errors may have spatial correlations, requiring a spatial covariance matrix. To describe the elements of this matrix Indicates position and The correlation of errors is often analyzed using an exponential model. Calculate, where, It is the standard deviation of the error. It is the distance between two points. This refers to the relevant length. For vectorized geological map data, the positioning error can be characterized by a uniform or normal distribution. The preset probability distribution is the aforementioned distribution model selected for each data source to describe its error. When generating the perturbation sample set, for each original observation data point in the standardized geological data... An error value is randomly selected from its corresponding preset probability distribution. This results in a perturbed data point. Repeat this process. Next (for example) By synchronizing all data points, a dataset containing... A “perturbation sample set” of a complete dataset copy, each copy representing a data reality affected by random error.

[0054] The error characteristics of different data sources refer to the inherent features in terms of magnitude, statistical distribution, and spatial correlation of errors contained in standardized geological data from different sources such as boreholes, geophysical exploration, and remote sensing, due to factors such as observation methods, instrument accuracy, environmental noise, and differences in human interpretation. For example, borehole depth errors are usually unbiased and random and independent; while seismic velocity inversion errors have strong spatial correlations. Accurately quantifying these characteristics is a prerequisite for conducting reliable uncertainty analysis.

[0055] In Monte Carlo simulations, a pre-defined probability distribution is a mathematical model used to simulate the error characteristics of different data sources. It is a mathematical description of the random behavior of data errors. Commonly used pre-defined probability distributions include the normal distribution (describing random independent errors), the multivariate normal distribution (describing errors with spatial correlation), and the uniform distribution (describing a well-defined range of systematic errors). The choice of distribution depends on the understanding of the mechanism of data error generation and the analysis of historical data.

[0056] S41. For each sample in the disturbed sample set, re-execute the data fusion and 3D modeling process to obtain the corresponding multi-scale 3D geological model output. The data fusion and 3D modeling process refers to the complete sequence of steps defined from S2 to S34. For the perturbation sample set... Sample This requires using it as new input and completely re-executing the process. This includes: re-executing Bayesian fusion in step S2 based on the perturbed data to construct a new posterior probability distribution of geological parameters; then, using this new posterior distribution as a constraint, re-executing the potential field interpolation method for 3D modeling in step S3 to construct a completely new multi-scale 3D geological model. This model It includes all geological interface geometries, spatial distributions of geological bodies, and attribute assignments inferred under this specific error condition. Because the perturbation sample set contains... There are only a few samples, so this process needs to be repeated. Next. To handle the massive computational load, this step must be performed in a high-performance computing environment using a parallel computing strategy, distributing different perturbation samples across different computing nodes to execute the modeling process simultaneously. Ultimately, a model will be obtained from... A collection of possible three-dimensional geological models This "collection of outputs from multi-scale three-dimensional geological models" reflects the range of uncertainties caused by the propagation of input data errors to the final model.

[0057] In this context, the perturbation sample set refers to a complete, new copy of the dataset generated during the uncertainty quantification process in Monte Carlo simulations. This copy is created by simultaneously applying a random perturbation to all observation points in the original standardized geological data, based on the error characteristics of their respective data sources. This sample represents a specific reality among all possible error combinations of the input data. In the workflow, each such sample is independently input into the subsequent data fusion and 3D modeling process to generate a corresponding 3D geological model affected by that specific error reality.

[0058] S42. Perform statistical analysis on the set of outputs from the multi-scale three-dimensional geological model, and calculate the mean, standard deviation, and confidence interval of the geological parameters. Geological parameters are quantities that describe the characteristics of a model, such as a specific spatial location. The burial depth of the strata above lithology category probability or porosity value For the model set Statistical analysis involves analyzing each geological parameter of interest from... Extracting this parameter from the model Each possible value is used to form a sample sequence, and then its statistical characteristics are calculated. For continuous parameters (such as burial depth)... First, calculate its mean. As the best estimate: ,in, It is the first The burial depth at that location in the model is then calculated. Next, the standard deviation is calculated. To measure the uncertainty of the estimate: Then, based on this For each sample value, its confidence interval can be calculated. For example, a 95% confidence interval is calculated by finding the 2.5th percentile of the sample. and 97.5th percentile This indicates that there is a 95% probability that the true value falls within this interval. For discrete parameters (such as lithology), the frequency of occurrence of each lithology can be calculated as its probability. This step requires calculation of all key locations and parameters in the model space, ultimately generating statistical results such as the geological parameter mean field, standard deviation field, and confidence interval range field.

[0059] The set of outputs from the multi-scale 3D geological models comprises a series of possible 3D geological models obtained through Monte Carlo simulation. Each model in this set is the output of a "data reality" generated by applying random perturbations to the original input data, and is produced after independently executing a complete 3D modeling process using Bayesian data fusion and potential field interpolation. Essentially, this set of models represents a statistical sample set of all reasonable possibilities for the final 3D geological model, taking into account input data errors.

[0060] S43. Based on the mean, standard deviation and confidence interval of the calculated geological parameters, generate a probability distribution function and risk zoning map that characterize the uncertainty of the multi-scale three-dimensional geological model, as the result of uncertainty quantification.

[0061] The probability distribution function is the most complete mathematical description of the uncertainty of a single geological parameter. For a continuous parameter at a point in space (such as porosity)... ), according to its Sample values An approximate probability density function can be fitted using the kernel density estimation method. This function This visually demonstrates the relative probability of different porosity values ​​at a given point, representing a "probability distribution function characterizing the uncertainty of a multi-scale three-dimensional geological model." The risk zoning map visualizes the uncertainty space to support decision-making. The generation process involves selecting a key geological parameter (such as the coal seam roof depth) and its risk threshold (e.g., a depth prediction error greater than 5 meters may pose a high risk to mining design). The standard deviation field of this parameter across the entire model space is then used. Alternatively, the confidence interval width can be used to set risk level classification criteria. For example, defining the standard deviation. The area is a low-risk zone (green). The area is a medium-risk area (yellow). The area is designated as a high-risk zone (red). Then, based on the standard deviation calculated for each spatial location, the area is colored according to this standard to generate a "risk zoning map" in a two-dimensional plane or three-dimensional space. The probability distribution function and the risk zoning map, together with the statistical field data generated in the third step, constitute a complete "uncertainty quantification result," clearly revealing the reliability of the model and the spatial distribution of risk.

[0062] Among these, the mean, standard deviation, and confidence interval of a geological parameter are core indicators used to describe the statistical characteristics of a specific geological parameter within the set of models generated by Monte Carlo simulation. The mean is the arithmetic average of the parameter values ​​across all possible models, representing the best estimate of the parameter. The standard deviation measures the dispersion of the parameter value relative to the mean across all possible models, directly quantifying the uncertainty of the parameter estimate. A confidence interval is a numerical range representing the probability that the true parameter value falls within that range at a given confidence level. For example, a 95% confidence interval provides the upper and lower bounds where the parameter value has a 95% probability of falling within that range.

[0063] Among them, the probability distribution function and risk zoning map, which characterize the uncertainty of multi-scale three-dimensional geological models, are two core expressions of the uncertainty quantification results of Monte Carlo simulations. The probability distribution function, in the form of a mathematical function, fully describes the probability of a single geological parameter value. The risk zoning map, based on the uncertainty index of geological parameters, classifies risk levels according to the degree of impact on engineering or geological decisions, and maps the results onto a geographic space visualization map for intuitive guidance on risk avoidance and resource deployment.

[0064] The uncertainty quantification result is the final output of the Monte Carlo simulation analysis, a comprehensive information package containing numerical values, functions, and maps. It systematically answers the questions of "how reliable the model is" and "where the risks lie," specifically including the statistical characteristic field of geological parameters, the probability distribution functions of key location parameters, and global or local risk zoning maps. This result serves as a crucial bridge connecting geological models with decision-making applications such as engineering risk assessment and resource probability estimation.

[0065] S5. Store the multi-scale 3D geological model and uncertainty quantification results in a data management platform that conforms to the OGC standard, and provide visualization, querying, and access to the geological layers through a web service interface. The specific implementation process is as follows: S50. Before storage, the multi-scale 3D geological model and uncertainty quantification results produced by the modeling process must be converted into an open, standard format, and a clear data organization structure must be established. The geometric parts of the multi-scale 3D geological model (such as stratigraphic interfaces and fault planes represented by triangular meshes) are usually converted to CityGML, 3D Tiles, or a dedicated GeoPackage format for storage; the attribute parts of the model (such as lithological codes and physical property parameters for each geological unit) are stored as attribute tables associated with geometric objects. Uncertainty quantification results, including the mean field, standard deviation field, confidence interval boundaries, and risk zoning raster map of geological parameters, are processed according to data type: continuous scalar fields (such as the mean porosity) can be stored as a raster data stack in GeoTIFF or NetCDF format; probability distribution functions can be stored in a relational database in parametric form (such as the mean and variance of a normal distribution) or as a pre-computed JSON file; risk zoning maps are stored in raster or vector format. All data must be described in accordance with a unified metadata standard (such as ISO 19115) and logically organized according to a hierarchical structure of "project-region-scale-data type" to lay the foundation for subsequent data entry and publication.

[0066] The S51 OGC-compliant data management platform adopts a layered architecture, with data stored in different storage components based on its characteristics. For massive amounts of 3D model triangulation data and rasterized uncertain field data, due to their large file sizes, they are stored in distributed file systems such as HDFS or object storage such as MinIO to achieve highly reliable and scalable storage. For point, line, and polygon vector data with spatial characteristics (such as borehole locations, geological boundaries, and risk zoning vectors), as well as metadata requiring complex spatial queries, they are stored in spatial databases such as PostGIS. For time-series monitoring data (if included in the model), they are stored in time-series databases such as InfluxDB. During storage, the system automatically creates spatial indexes for all spatial data, B-tree or hash indexes for attribute data, and records complete data traceability information. After storage, the data management platform forms a comprehensive geological data warehouse containing raw observation data, intermediate processed data, and final model results.

[0067] S52. In the service layer of the data management platform, dedicated server software is configured. For two-dimensional raster data in geological layers, such as geological plans and risk zoning raster maps, it is published as a Web map service through GeoServer or MapServer. WMS receives HTTP requests containing parameters such as layer name, spatial extent, and output image size, dynamically renders and returns map images in PNG or JPEG format. For vector geological features, such as fault lines, contour lines, and high-risk area polygons, they are published as Web feature services. WFS supports users to perform spatial and attribute queries using the standard Filter Encoding language and returns vector data in GML or GeoJSON format, supporting CRUD operations. For three-dimensional geological model data and multidimensional uncertain raster field data, it is published as a Web overlay service. WCS allows users to request raw data values ​​according to specified spatial extent, resolution, and band (for multidimensional data), returning data slices in GeoTIFF or NetCDF format for professional clients to perform in-depth analysis.

[0068] S53. To meet customized application requirements, a set of descriptive state transition application programming interfaces (APIs) are developed at the service layer. These RESTful APIs are designed around resources and use standard HTTP methods. For example, an interface can be designed to list all multi-scale 3D geological models of a project; an interface can be designed to obtain all stratigraphic interface geometry data of a specified model at the site scale; and an interface can be designed to accept a JSON request body containing spatial extent, attribute conditions, and model version, perform complex cross-queries, and return information on geological bodies that meet the conditions and their associated uncertainty indicators. The API response format is uniformly JSON and is described and documented in accordance with the OpenAPI specification to facilitate integration and invocation by third-party systems, such as engineering design software and decision support systems.

[0069] S54. At the application layer, develop a browser-based 3D visualization portal. This portal utilizes WebGL technology and frameworks such as Cesium or Three.js, using a map provided by WMS as the base map and overlaying and rendering multi-scale 3D geological models provided by WCS and dedicated 3D data services. This enables seamless scaling and browsing from macroscopic regional structures to microscopic site stratigraphic profiles. Uncertainty quantification results, such as risk zoning maps, are displayed overlaid through WMS, and probability distribution functions are dynamically displayed in the sidebar as charts after data is obtained by calling a RESTful API. The portal provides rich interactive tools: a query tool allows users to click on any location in the 3D model, and the portal backend requests multi-dimensional information such as stratigraphic attributes and uncertainty values ​​at that location via API; a profile generation tool allows users to draw arbitrary survey lines, and the system obtains and renders geological profile maps and uncertainty curves via API. Furthermore, the portal provides a data download module, allowing authorized users to download vector data or model tiles of selected areas by calling WFS or specific API interfaces.

[0070] The OGC-compliant data management platform is a software system that manages, publishes, and shares spatial data in accordance with a series of international standards and protocols established by the Open Geospatial Consortium. A core feature of this platform is its interoperability, ensuring that systems developed by different vendors can access and process geospatial information through a unified standard interface. In this invention, such a platform not only stores multi-source 3D geological data and model results, but more importantly, by implementing standard services such as WMS, WFS, and WCS, it enables geological layers to be seamlessly integrated and used by a wide range of map browsers, GIS desktop software, and custom applications.

[0071] In this context, a geological layer is a concept that logically organizes geospatial data according to its subject matter or data type. Within the context of this invention, a geological layer can refer to geological maps, lithological distribution maps, and risk zoning maps expressed in two-dimensional form, or it can refer to stratigraphic interfaces, fault surfaces, geological body models, etc., expressed in three-dimensional form. In the data management platform, each geological layer corresponds to a dataset that can be independently published, displayed, queried, and analyzed. These datasets are provided externally through OGC standard services or API interfaces, and users can select and load one or more layers for overlay analysis and visualization as needed.

[0072] In another embodiment, the following steps are included: ① Acquiring and standardizing geological data from different sources (i.e., data acquisition and preprocessing): acquiring geological data from multiple sources, performing coordinate unification, format conversion, and quality control; ② Performing data fusion based on a Bayesian fusion framework (i.e., data fusion): integrating multi-source data using a Bayesian framework to construct a prior probability distribution characterizing geological parameters. Establish a corresponding likelihood function for each data source. The posterior probability distribution of geological parameters was obtained using the Markov chain Monte Carlo sampling method. The process involves: ① Calculating the posterior probability distribution of geological parameters using a sample set; ② Constructing a multi-scale three-dimensional geological model using potential field interpolation (i.e., three-dimensional modeling): Based on the fused data, a multi-scale three-dimensional geological model is constructed using potential field interpolation; ③ Quantifying uncertainty using Monte Carlo simulation (i.e., uncertainty quantification): Based on the error characteristics of different data sources in the standardized geological data, a set of perturbation samples conforming to a preset probability distribution is generated, and the data fusion and three-dimensional modeling process (i.e., Monte Carlo simulation) is re-executed for each sample to calculate the model uncertainty index; ④ Storing the model and results in a data management platform conforming to OGC standards and providing services through a Web service interface (i.e., data management and services): The model results are stored in a distributed platform to provide standardized data services.

[0073] Furthermore, Bayes' theorem is used to integrate multi-source geological data into a unified probabilistic framework, achieving effective fusion of different data sources. The formula used for data fusion based on the Bayesian fusion framework is as follows: ,in, For geological model parameters, For the first Various data sources This represents the posterior probability distribution of geological parameters (i.e., the posterior probability distribution). Indicates: Geological model parameters, representing the geological attributes to be estimated (such as lithology, structural parameters, etc.). Indicates: the Various data sources (such as borehole data, seismic data, geological maps, etc.). The posterior probability distribution of geological parameters represents the updated understanding of model parameters after all data has been observed. It represents the likelihood function established for the data source, which indicates the probability of observing these data given the model parameters. Representation: Characterizes the prior probability distribution of geological parameters, representing the initial understanding of model parameters before observation data. Represents the marginal likelihood, as a normalization constant.

[0074] Furthermore, a multi-scale three-dimensional geological model is constructed using the potential field interpolation method, specifically including the following sub-steps: ① Based on probability information and borehole stratification data, define a potential field function to describe the spatial morphology of the geological interface (i.e., define the potential field function of the geological interface). Construct the potential field function based on geological interface feature points, contour lines, and borehole stratification data from multi-source data. ,in Represents three-dimensional spatial coordinates, function values Quantitatively describes the relative distance and trend intensity from a spatial point to the target geological interface; the potential field function satisfies the Laplace equation: .

[0075] Set the location of the known geological interface The function value is positive in the upper region of the interface and negative in the lower region (or vice versa, determined according to the geological sequence logic). The absolute value of the function value is positively correlated with the actual distance from the spatial point to the interface, so as to accurately depict the spatial distribution pattern and extension trend of the geological interface.

[0076] ② Numerical methods are used to solve the potential field equations corresponding to the potential field functions, obtaining the continuous potential field distribution in three-dimensional space (i.e., solving the potential field equations based on borehole data and geophysical constraints); specifically, the finite difference method is used to discretize the three-dimensional space into a grid, and the known geological interface locations in the borehole layer data are then analyzed. The values ​​are substituted into the discrete equation system as Dirichlet hard constraints, while the continuous stratigraphic trend information obtained from geophysical exploration (such as seismic reflection horizons and resistivity inversion profiles) is transformed into L2 regularized soft constraint terms to balance the spatial smoothness of the potential field function with the consistency of multi-source data. The discrete linear equation system is then solved iteratively using the conjugate gradient method to obtain the potential field function values ​​of all grid nodes in the study area. This forms a continuous potential field distribution in three-dimensional space covering the target area (i.e., a continuous potential field distribution field), providing a basis for the accurate extraction of subsequent geological interfaces.

[0077] ③ Extract equipotential surfaces with preset values ​​from the continuous potential field distribution, and use these equipotential surfaces as geological interfaces in the multi-scale 3D geological model (i.e., extract equipotential surfaces as geological interfaces). Based on the continuous potential field distribution in 3D space (i.e., the continuous potential field distribution field), the Marching Cubes algorithm is used to process all grid nodes. Values ​​are extracted using isosurfaces, and selected... The corresponding zero-point isosurface serves as the initial geometric contour of the geological interface. A topological consistency check is performed on the extracted initial isosurface to correct anomalies such as interface breaks and overlaps caused by mesh discretization or data noise. Simultaneously, the initial isosurface is locally fine-tuned using known geological interface depth information from borehole data to ensure that the spatial deviation between the interface and the borehole stratification points is less than a preset threshold. Finally, the corrected isosurface is transformed into a triangular patch model or voxel representation, forming structured three-dimensional geological interface data. This interface data can be directly integrated into a multi-scale three-dimensional geological model as boundary elements for dividing different geological strata, supporting subsequent spatial modeling and attribute assignment of geological bodies.

[0078] ④ Apply geological rules based on stratigraphic stacking and fault cutting to perform topological optimization of geological interfaces, completing the construction of a multi-scale 3D geological model (i.e., applying geological rules for model optimization). Specifically, this includes temporal constraints based on stratigraphic sedimentary sequences to ensure that the stacking relationship of stratigraphic units in the model conforms to the geological principle of "newer above, older below"; for fault cutting scenarios, based on the fault's attitude parameters (strike, dip, dip angle) and displacement, topological reconstruction of the stratigraphic interfaces in the fault-affected area is performed to maintain a reasonable slip contact relationship between the stratigraphic units on the hanging wall and footwall of the fault; simultaneously, the spatial overlap areas of adjacent stratigraphic units in the model are automatically detected and precisely corrected, eliminating stratigraphic interpenetration anomalies by adjusting the interface vertex coordinates or triangular facet connection methods; furthermore, combined with the regional geological tectonic background, the contact relationship between the upper and lower strata of the unconformity surface is verified and optimized to ensure the model's dual consistency in topological structure and geological logic, improving the realism of the multi-scale 3D geological model to actual geological bodies, and meeting the accuracy and reliability requirements of engineering surveys, resource assessments, and other scenarios.

[0079] Furthermore, the process of quantifying uncertainty using the Monte Carlo simulation method includes: ① Based on the error characteristics of different data sources in standardized geological data, a set of perturbation samples conforming to a preset probability distribution is generated (i.e., a set of perturbation samples is generated based on the data error distribution); each perturbation sample is substituted into the three-dimensional geological modeling process to sequentially complete operations such as stratigraphic temporal constraint verification, fault topology reconstruction, stratigraphic overlap correction, and unconformity surface contact relationship optimization, thereby obtaining the multi-scale three-dimensional geological model output corresponding to each sample (i.e., the three-dimensional geological model output result); statistical analysis is performed on the output results of all samples to calculate the probability distribution characteristics of key geological parameters of the model (such as stratigraphic unit thickness, fault dip angle, unconformity surface burial depth, etc.), including mean, standard deviation, confidence interval, and probability density curve, thereby realizing the quantitative characterization of model uncertainty caused by data errors in the three-dimensional geological modeling process.

[0080] Generating the perturbation sample set is the first step in Monte Carlo simulation, requiring the construction of a reasonable probability distribution model based on the error characteristics of different data sources. In multi-source geological data fusion, errors from different data sources can influence and propagate through the modeling process: for the function Its output variance can be approximated by a first-order Taylor expansion, yielding the error propagation formula: .

[0081] For spatially relevant data (such as geophysical observations), it is necessary to construct the covariance matrix: ,in, Distance The relevant functions at a given location, commonly used models include: ① Exponential model: ② Gaussian model: ③ Spherical model: (when In data fusion, the weights of each data source should reflect its reliability. This weighting scheme based on the inverse of variance ensures that high-precision data receives greater weight.

[0082] ② For each sample in the disturbed sample set, the data fusion and 3D modeling process is re-executed to obtain the corresponding multi-scale 3D geological model output (i.e., the modeling process is re-executed for each disturbed sample); the corresponding model output value is calculated, and after collecting the output results of all disturbed samples, the variance estimate of the output is obtained through statistical methods. Based on the spatial correlation function model (such as exponential, Gaussian or spherical model), a covariance matrix reflecting the spatial dependence between sample points is constructed. This matrix needs to accurately characterize the degree of correlation of samples at different locations under disturbance, providing a key statistical basis for subsequent multi-source data weight optimization and uncertainty quantification.

[0083] ③ Perform statistical analysis on the output set of the multi-scale three-dimensional geological model, and calculate the mean, standard deviation, and confidence interval of the geological parameters (i.e., statistically analyze the distribution characteristics of the model parameters, and calculate the mean, standard deviation, and confidence interval); specifically, for the model parameters... (Such as formation interface depth, physical parameters), the mean estimate is: ,in For Monte Carlo sample size, For the first The parameter values ​​for each sample. The standard deviation reflects the degree of dispersion of the parameter: The coefficient of variation (CV) is used to compare the relative uncertainty of different parameters. .

[0084] ④ Based on the statistical analysis results, a probability distribution function and risk zoning map representing the uncertainty of the multi-scale three-dimensional geological model are generated as uncertainty quantification results (i.e., uncertainty visualization maps are generated). By mapping the uncertainty quantification results of the parameters to the spatial coordinate system, the probability cloud map can intuitively present the probability distribution density of the parameters at different locations and clearly show the spatial clustering characteristics of high uncertainty areas. The risk zoning map, based on the preset risk threshold (such as standard deviation multiples or probability intervals) and combined with the influence weight of the parameters on the engineering objectives, divides the study area into low, medium, and high risk levels, providing intuitive and accurate spatial decision support for the subsequent optimization and adjustment of engineering plans, key area exploration, and the formulation of risk prevention and control measures.

[0085] Furthermore, storing the model and results on an OGC-compliant data management platform and providing services (i.e., data services) via a Web service interface conforms to OGC international standards and supports the following service types: ① WMS (Web Map Service): provides rasterized display of geological layers; ② WFS (Web Feature Service): provides query and download of vector geological features; ③ WCS (Web Coverage Service): provides access to slices of 3D geological models; ④ SensorThings API: supports access and distribution of real-time monitoring data.

[0086] Example 1: The application scenario is: the underground space planning of the mining area requires the construction of a multi-scale three-dimensional geological model at a scale of 1:10,000. Data sources: 2,500 borehole data, 15 seismic profiles, geological maps (1:50,000), gravity anomaly data, etc. The implementation steps include: (1) acquiring and standardizing geological data from different sources (i.e., data preprocessing): unifying the coordinates of all data to the independent coordinate system of the mining area, and performing quality checks to remove abnormal data points. (2) performing data fusion based on the Bayesian fusion framework (i.e., data fusion): using the Bayesian fusion framework (i.e., Bayesian framework) to integrate borehole data and geophysical exploration data, updating the prior probability distribution of the geological parameters (i.e., the stratigraphic interface prior model) through Markov chain Monte Carlo sampling, and obtaining the posterior probability distribution of the geological parameters of the stratigraphic interface depth. (3) Constructing a multi-scale three-dimensional geological model (i.e., three-dimensional modeling) using the potential field interpolation method: Based on the probability constraints provided by the posterior probability distribution, the potential field function is defined according to the probability information and borehole layer data, and the potential field equation is solved using numerical methods to obtain the continuous potential field distribution. Then, equipotential surfaces are extracted from the continuous potential field distribution as geological interfaces, and finally, eight main stratigraphic interfaces are constructed. The model spatial resolution is 50 m (X) × 50 m (Y) × 5 m (Z). (4) Quantifying uncertainty using the Monte Carlo simulation method (i.e., uncertainty analysis): Based on the error characteristics of different data sources in the standardized geological data, a set of disturbance samples conforming to the preset probability distribution is generated. The number of Monte Carlo simulations is set to 10,000. The data fusion and modeling process is re-executed for each disturbance sample. Finally, the set of outputs of the multi-scale three-dimensional geological model is statistically analyzed, and the standard deviation of the location of each stratigraphic interface is calculated. (5) Store the model and results in a data management platform that conforms to OGC standards and provide services (i.e., output results) through a Web service interface: generate a 3D model file (GeoPackage format) containing geometry and attributes, a detailed uncertainty analysis report, and provide online visualization and query access by publishing OGC Web services (such as WMS, WFS, WCS). Compared with traditional methods, the modeling accuracy is improved, reducing the interface position error from ±8 meters to ±3 meters; multi-source data fusion increases the effective data points by 40%, achieving full-area probabilistic assessment for the first time, and the automated process shortens the modeling cycle from 6 months to 1 month. Compared with subsequent engineering boreholes, the model prediction accuracy reaches 92%, the groundwater level prediction error is less than 1.5 meters, and the accuracy of identifying geological disaster-prone areas is 85%.

[0087] Example 2: The application scenario is: three-dimensional prediction and resource estimation of deep ore bodies in a certain mining area; data sources: 150 borehole core data (including grade analysis), airborne magnetic survey data, induced polarization sounding profile data, and geological structure interpretation map. Key technologies include: (1) Three-dimensional convolutional neural network (as a supplementary advanced method for attribute prediction), the input is: spatialized geological parameter (lithology, structure, geochemistry) volume data. The output is: three-dimensional distribution probability volume of ore body grade. The training samples are: generated based on spatial data (borehole data, geophysical inversion results) of known ore body areas. (2) Data fusion strategy: geophysical inversion results (such as magnetic and induced polarization) are used as part of standardized geological data and incorporated as soft constraints when constructing the prior probability distribution of geological parameters and establishing the corresponding likelihood function. Geological rules (metallogenic theory) are used as hard constraints and formalized and integrated into the prior probability distribution of geological parameters. Through the data fusion process based on the Bayesian fusion framework, the Bayesian update integration of multi-source information is realized. (3) Uncertainty assessment (i.e., the specific application of uncertainty quantification using Monte Carlo simulation method): Statistical analysis is performed on the set of outputs from the multi-scale three-dimensional geological model to calculate the mean, standard deviation, and confidence interval of geological parameters, thereby obtaining the confidence interval for resource estimation, the probability distribution of deep blind ore prediction, and the risk assessment results for exploration target selection. The prediction results are as follows: In resource estimation, the controllable resource quantity is 5 million tons (95% confidence interval: 4.8-5.2 million tons), the inferred resource quantity is 3 million tons (uncertainty ±25%), there are 3 deep prediction target areas, and the high probability area is (P>0.7). Verification drilling: 5 verification holes were drilled in the prediction target area, 4 of which encountered ore, with an ore-to-ore rate of 80% and an average grade prediction error of 12%.

[0088] This invention aims to provide a multi-source three-dimensional geological data fusion and management technology to solve the problems of data integration difficulties, low modeling accuracy, and insufficient uncertainty quantification in the existing technology. In this invention, (1) by acquiring and standardizing geological data from different sources, the coordinates, format and quality are unified to provide a consistent data foundation for fusion; by using data fusion based on the Bayesian fusion framework and using the potential field interpolation method to construct multi-scale three-dimensional geological models, reliable geological structure models at different scales from the region to the site are generated; by using the Monte Carlo simulation method to quantify uncertainty, the reliability of the model output is probabilistically evaluated, and confidence intervals and risk maps are generated; by storing the model and results in a data management platform that conforms to the OGC standard and providing services through the Web service interface, efficient data management, sharing and visualization applications are realized. Meanwhile, it provides reliable technical support for key application scenarios such as underground space development in mining areas, precise exploration of mineral resources, and geological disaster risk assessment. By establishing a unified geological data management system and intelligent analysis framework, it enhances the sharing and reuse value of geological data, reduces the cost of manual intervention in the data integration process, strengthens the adaptability of models to incremental data, ensures the timeliness and credibility of geological models, and helps to make geological engineering decisions more scientific and accurate.

[0089] This invention is applicable to scenarios such as geological surveys, mineral resource exploration, underground space planning in mining areas, and geological disaster early warning. With the rapid development of geological exploration technology, multi-source data such as remote sensing images, geophysical exploration, drilling sampling, and groundwater level monitoring exhibit characteristics of heterogeneous formats, large scale differences, and varying spatiotemporal resolutions. Traditional geological data processing methods struggle to achieve efficient integration and in-depth data mining. Addressing this pain point, this invention organically combines key technologies such as data standardization, multi-scale feature fusion, and dynamic model updating to construct a complete technical system covering data access, fusion processing, model building, and intelligent management. This not only improves the utilization efficiency and modeling accuracy of geological data but also provides reliable technical support for intelligent decision-making in this field, promoting the transformation of geological information technology from decentralized and static to integrated and dynamic. Its technological achievements can widely serve the precision operation needs of various sub-fields of geological engineering.

[0090] This invention discloses a multi-source 3D geological data fusion and management technology, the implementation of which corresponds to a multi-source 3D geological data fusion and management method. This method includes the following steps: acquiring and standardizing geological data from different sources (i.e., data acquisition and preprocessing module) to obtain standardized geological data; performing data fusion based on a Bayesian fusion framework (i.e., data fusion engine) to obtain the posterior probability distribution of geological parameters; constructing a multi-scale 3D geological model using potential field interpolation (i.e., 3D modeling module); quantifying uncertainty using Monte Carlo simulation (i.e., uncertainty quantification module) to generate uncertainty quantification results; and storing the model and results in a data management platform conforming to OGC standards and providing services through a Web service interface (i.e., data management platform). This technology integrates heterogeneous data from multiple sources such as boreholes, geophysical exploration, and remote sensing using a Bayesian fusion framework, constructs a multi-scale 3D geological model using potential field interpolation, quantifies model uncertainty using Monte Carlo simulation, and provides data services conforming to OGC standards. This invention solves the problems of difficult data integration, low modeling accuracy, and insufficient uncertainty quantification in traditional methods, and has broad application prospects in fields such as underground space planning, mineral resource exploration, and geological disaster early warning. Examples show that the present invention can improve modeling accuracy by 30-50% and increase processing efficiency by more than 5 times.

[0091] In the above embodiments, although the steps are numbered S1, S2, etc., they are only specific embodiments given by the present invention. Those skilled in the art can adjust the execution order of S1, S2, etc. according to the actual situation. The scheme after adjusting the order is also within the protection scope of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0092] like Figure 2 As shown in the figure, a multi-source three-dimensional geological data fusion and management system according to an embodiment of the present invention includes an acquisition and processing module, a fusion module, a construction module, a generation module, and a management module. The acquisition and processing module is used to acquire and standardize geological data from different sources to obtain standardized geological data. The fusion module is used to fuse the standardized geological data based on a Bayesian fusion framework to obtain a posterior probability distribution characterizing geological parameters. The fusion module is used to construct a multi-scale three-dimensional geological model using a potential field interpolation method based on the probability constraints provided by the posterior probability distribution. The generation module is used to perform uncertainty quantification on the multi-scale three-dimensional geological model using a Monte Carlo simulation method to generate uncertainty quantification results corresponding to the multi-scale three-dimensional geological model. The management module is used to store the multi-scale three-dimensional geological model and uncertainty quantification results in a data management platform conforming to the OGC standard, and provide visualization, querying, and access to geological layers through a Web service interface.

[0093] Optionally, in the above technical solution, the fusion module is specifically used to: construct a prior probability distribution characterizing geological parameters. The prior probability distribution integrates geological rules and expert experience; a corresponding likelihood function is established for each data source in the standardized geological data. The likelihood function describes the degree of matching between the data source and the multi-scale 3D geological model. Based on the prior probability distribution and the likelihood functions of all data sources, the posterior probability distribution of the geological parameters is obtained using the Markov chain Monte Carlo sampling method. The sample set; the posterior probability distribution of geological parameters is calculated from the sample set based on the posterior probability distribution. .

[0094] Optionally, in the above technical solution, the construction module is specifically used for: determining the probability information of geological interface feature points and stratigraphic attribute distributions based on posterior probability distributions; and defining a potential field function describing the spatial morphology of the geological interface based on the probability information and borehole layering data. Numerical methods are used to solve the potential field equations corresponding to the potential field functions to obtain the continuous potential field distribution in three-dimensional space. Equipotential surfaces with preset values ​​are extracted from the continuous potential field distribution and used as geological interfaces in the multi-scale three-dimensional geological model. The geological interfaces are topologically optimized by applying geological rules of stratigraphic stacking and fault cutting to complete the construction of the multi-scale three-dimensional geological model.

[0095] Optionally, in the above technical solution, the generation module is specifically used to: generate a set of perturbation samples conforming to a preset probability distribution based on the error characteristics of different data sources in the standardized geological data; re-execute the data fusion and 3D modeling process for each sample in the perturbation sample set to obtain the corresponding multi-scale 3D geological model output; perform statistical analysis on the set of multi-scale 3D geological model outputs to calculate the mean, standard deviation, and confidence interval of the geological parameters; and generate a probability distribution function and risk zoning map characterizing the uncertainty of the multi-scale 3D geological model based on the calculated mean, standard deviation, and confidence interval of the geological parameters, as the uncertainty quantification result.

[0096] In another embodiment, the system of the present invention includes: (1) Data Acquisition and Preprocessing Module (corresponding to step S1): The data acquisition and preprocessing module supports the access of multi-source heterogeneous data such as borehole data, geophysical exploration data (earthquake, electromagnetic, gravity), remote sensing images, and GIS vector data. It achieves unified processing of different data formats (such as CSV, Shapefile, HDF5, LAS) through format conversion tools. In the standardization process, wavelet transform is used to filter signal noise in geophysical exploration data, Kriging interpolation method is used to fill missing attribute values ​​in borehole data, and the coordinate system of all data is transformed to the 2000 National Geodetic Coordinate System (CGCS2000) or the unified coordinate system required by the project. This completes the process of acquiring and standardizing geological data from different sources to obtain standardized geological data.

[0097] (2) Data Fusion Engine (corresponding to step S2): The data fusion engine performs data fusion based on a Bayesian fusion framework. This framework first constructs a prior probability distribution representing geological parameters. An initial model is formed by integrating geological rules and expert knowledge. Then, a corresponding likelihood function is established for each data source in the standardized geological data. This describes the degree of matching between each data source and the geological model. Subsequently, based on the prior probability distribution and the likelihood function of all data sources, a Markov chain Monte Carlo sampling method is used to obtain the posterior probability distribution of the geological parameters. The sample set is used to calculate the posterior probability distribution of geological parameters. This process utilizes Bayes' theorem to achieve probabilistic fusion of multi-source data. In response to data conflicts, the reliability of different data sources is indirectly reflected through error parameters (such as variance) in their likelihood functions. The smaller the variance, the higher the weight, thereby achieving the organic integration of heterogeneous data.

[0098] (3) 3D Modeling Module (corresponding to step S3): The 3D modeling module adopts a multi-scale modeling strategy. Specifically, it constructs a multi-scale 3D geological model by using potential field interpolation based on the probability constraints provided by the posterior probability distribution. This process includes: determining the probability information of geological interface feature points based on the posterior distribution; defining the potential field function based on this information and borehole layering data; obtaining the 3D continuous potential field distribution by solving the potential field equation using numerical methods; extracting equipotential surfaces as geological interfaces; and finally applying stratigraphic stacking and fault cutting rules for topology optimization. This achieves multi-scale expression from microscopic (such as core composition), mesoscopic (stratigraphic units) to macroscopic (regional tectonics), and combines level of detail technology to achieve seamless switching and real-time rendering of models at different scales.

[0099] (4) Uncertainty Quantification Module (corresponding to step S4): The uncertainty quantification module performs the process of quantifying the uncertainty of a multi-scale three-dimensional geological model using the Monte Carlo simulation method. This module generates a set of perturbation samples conforming to a preset probability distribution based on the error characteristics of different data sources in the standardized geological data. Then, for each perturbation sample, the data fusion and three-dimensional modeling process is re-executed to obtain a possible geological model. Next, statistical analysis is performed on the generated model set to calculate the mean, standard deviation, and confidence interval of the geological parameters. Finally, based on the statistical analysis results, a probability distribution function and risk zoning map characterizing the model's uncertainty are generated. This module calculates and visually displays the uncertainty index of the model's prediction results based on Monte Carlo simulation and Bayesian posterior probability distribution.

[0100] (5) Data Management Platform (corresponding to step S5): The data management platform is responsible for storing multi-scale 3D geological models and uncertainty quantification results in a data management platform that conforms to OGC standards, and provides visualization, querying, and access to geological layers through a Web service interface. The platform adopts distributed storage architectures such as HDFS to support PB-level data storage, builds an intelligent retrieval system, and publishes WMS, WFS, WCS services and RESTful API interfaces that conform to OGC standards for third-party systems to call, so as to realize data sharing and reuse.

[0101] The system in this embodiment includes: a data acquisition and preprocessing module for acquiring and standardizing geological data from different sources; a data fusion engine for acquiring geological data from multiple sources such as boreholes, geophysical exploration, and remote sensing images, and performing coordinate unification, format conversion, quality control, and metadata extraction; and an engine for integrating multi-source heterogeneous data using a Bayesian framework and constructing a prior probability distribution characterizing geological parameters. (i.e., prior knowledge), and establish corresponding likelihood functions for each data source. (i.e., likelihood function) and the Markov chain Monte Carlo sampling method (i.e., Markov chain Monte Carlo sampling) are used to obtain the posterior probability distribution of geological parameters. The system includes: a module for generating a sample set and calculating the posterior probability distribution of geological parameters; a 3D modeling module for constructing multi-scale 3D geological models using potential field interpolation, which, based on fused data, uses potential field interpolation, implicit surface reconstruction, and geological rule constraints to construct multi-scale 3D geological models of regions, mining areas, and sites; an uncertainty quantification module for quantifying uncertainty using Monte Carlo simulation, which generates a set of perturbation samples conforming to a preset probability distribution based on the error characteristics of different data sources in standardized geological data, and re-executes the data fusion and 3D modeling process (i.e., Monte Carlo simulation) and sensitivity analysis for each sample, and performs statistical analysis on the generated model set (i.e., calculating the confidence interval of model parameters, the probability distribution of prediction results, and the standard deviation of key interface locations); and a data management platform for storing models and results to a data management platform conforming to OGC standards and providing services through a Web service interface, which uses distributed storage, spatial database, and object storage technologies, provides RESTful API and OGC standard services, and supports Web 3D visualization, mobile data acquisition, and desktop GIS integration.

[0102] The system architecture is designed as follows: The multi-source 3D geological data fusion and management system adopts a modular and layered architecture design, realizing full-process management from acquiring and standardizing geological data from different sources, performing data fusion based on a Bayesian fusion framework, constructing multi-scale 3D geological models using potential field interpolation, quantifying uncertainties using Monte Carlo simulation methods, to storing the models and results on an OGC-compliant data management platform and providing services through a Web service interface. The system architecture includes five core layers: data layer, processing layer, model layer, service layer, and application layer. Each layer achieves loose coupling and integration through standardized interfaces.

[0103] The core design principles include: ① Scalability: Supporting dynamic access to new data sources and modeling methods. ② Interoperability: Adhering to international standards such as OGC and ISO. ③ High performance: Employing distributed computing and parallel processing technologies. ④ Reliability: Providing data backup, fault tolerance mechanisms, and version control.

[0104] Furthermore, the posterior probability distribution of geological parameters obtained using the Markov chain Monte Carlo sampling method is utilized. The sample set is used to obtain the posterior distribution characteristics of the model parameters (such as mean, median, variance, etc.) through statistical inference, and the core parameters of the model are updated accordingly. At the same time, uncertainty indicators are calculated based on the posterior distribution, including but not limited to the confidence interval of the parameters, the probability distribution of the prediction results, the standard deviation and coefficient of variation of the key geological interface locations. The uncertainty indicators are used to quantify the reliability of the model output and provide a risk assessment basis for subsequent geological analysis or decision-making.

[0105] This paper presents a visualization of the uncertainty quantification results generated using the Monte Carlo simulation method. Specifically, it utilizes probability cloud plots as a graphical method to represent probability distributions. The definition of probability cloud plots is achieved by using color depth to indicate probability density and employing shape ambiguity to express the magnitude of uncertainty. The paper also elucidates the concept of confidence intervals, which are interval estimates of population parameters in statistics. These intervals are calculated based on sample data and predetermined confidence levels using statistical software or mathematical formulas. Confidence intervals are visualized by using shaded areas and different colors to distinguish different confidence levels in the chart. Furthermore, the paper explains that the purpose of risk zoning is to classify risks into different levels for management and decision-making. This classification is based on the probability and impact of risk occurrence, referencing historical data and expert experience. Risk zoning is represented by using different colors or patterns to distinguish risk levels and clearly labeling the meaning of each risk level with legends. These visualization techniques adhere to visualization design principles to ensure effective communication: To enhance interactivity, they provide data exploration tools (e.g., clicking on a section of a chart to display detailed data) and allow users to filter data using sliders, drop-down menus, and other controls to achieve multi-dimensional analysis and comparison, while also considering dynamic displays to utilize animations or interactive elements to showcase data changes; to improve readability, they simplify chart design (removing unnecessary decorative elements to maintain chart simplicity and avoid information overload) and use clear labels and legends (ensuring each data point is clearly labeled and the legend clearly explains the symbols and colors used in the chart); to accurately convey information, they ensure visual elements precisely correspond to the data (ensuring the correct interpretation of visual elements such as colors and shapes) and select the chart type that best expresses the information based on the data characteristics (e.g., bar charts are suitable for displaying categorical data, while line charts are suitable for displaying trends), thereby avoiding misleading the audience's understanding of the data.

[0106] The data fusion process based on the Bayesian fusion framework includes the following steps: (1) Constructing a prior probability distribution representing geological parameters. (i.e., the prior probability distribution for constructing geological models), integrating geological rules and expert experience; (2) for each data source Establish the corresponding likelihood function (3) Using the Markov chain Monte Carlo sampling method, the data is analyzed from the joint posterior distribution. (4) Based on the sampling results, update the model parameters and calculate the uncertainty index.

[0107] Multi-source 3D geological data fusion is a key technical step in constructing high-precision geological models. This technology integrates geological data from different sources, with varying precision and formats, achieving data complementarity and synergy, thereby improving the reliability and accuracy of the geological model.

[0108] The basic principle of data fusion is as follows: Based on Bayesian theory and information theory, data fusion achieves optimal combination of multi-source data by establishing correlations between data sources. The fusion process needs to consider the following aspects: quantifying data uncertainty and assessing the reliability and error range of different data sources; unifying the coordinate systems and spatial references of different data sources in spatial registration; handling time differences in data collected at different times during time synchronization; and ensuring a unified geological interpretation framework across different data sources in terms of semantic consistency. Based on the fusion level and strategy, data fusion methods can be divided into: data-level fusion, which directly fuses the original observation data, retaining the maximum amount of information but with high computational complexity; feature-level fusion, which extracts features from each data source before fusion, thus balancing information retention and computational efficiency; and decision-level fusion, which fuses the interpretation results of each data source, suitable for integrating heterogeneous data sources. Data fusion based on the Bayesian fusion framework (i.e., the Bayesian fusion algorithm) is one of the core methods for multi-source geological data fusion. It achieves optimal combination of different data sources through a probabilistic reasoning framework, and this algorithm is particularly suitable for handling the uncertainty problems commonly found in geological data.

[0109] The application of Bayes' theorem in geological data fusion is as follows: Bayes' theorem provides a mathematical framework for inferring geological model parameters from observational data. ,in, Let posterior probability distribution (i.e., posterior probability) represent the probability distribution of geological parameters given observation data. Model under conditions The probability of; The likelihood function (i.e., the likelihood function) is established for the data source and describes the model's ability to interpret the observed data. To characterize the prior probability distribution (i.e., prior probability) of geological parameters, geological rules and expert knowledge are integrated. This is an evidence factor used for normalization. It is relevant for multiple independent data sources. The joint posterior distribution can be expressed as: This formula enables the natural fusion of multi-source data, with each data source contributing to the estimation of model parameters through its likelihood function. The weights of the data sources are automatically determined by the shape of their likelihood functions, eliminating the need for manual setting of weight coefficients.

[0110] The complete data fusion process based on the Bayesian fusion framework is theoretically rooted in Bayesian theory, including basic concepts of probability theory (such as random events, probability spaces, and probability distributions) and the mathematical expression of Bayes' theorem, which is expressed as a formula. With the core as the core, among which, It is the prior probability distribution (i.e., prior probability) that characterizes geological parameters. It is a likelihood function established for the data source. This refers to the posterior probability distribution (i.e., posterior probability) of geological parameters. This formula forms the mathematical foundation for the update process from prior to posterior, embodying the Bayesian inference principle of introducing and fusing prior information and accumulating and updating evidence. The process begins with data preprocessing, a crucial step in acquiring and standardizing geological data from different sources. This includes data cleaning and standardization (specific operations such as missing value handling, outlier detection, and data normalization) and data transformation and feature extraction (involving feature selection, feature construction, and dimensionality reduction techniques), aiming to address the issues of data consistency and integrity in integrated data processing. The preprocessed data enters the core fusion stage. First, the prior probabilities are determined, i.e., constructing the prior probability distribution representing the geological parameters. This is based on domain knowledge, statistical analysis of historical data, assumptions based on data distribution, and parameter estimation. Simultaneously, the likelihood function is constructed, i.e., establishing a corresponding likelihood function for each data source to describe the matching degree between each data source and the geological model. These two steps together complete model building and training. The process then proceeds to the Bayesian inference stage, namely probability update. Algorithms such as Markov chain Monte Carlo sampling are used to calculate the posterior probability, thereby obtaining a sample set of posterior probability distributions for the geological parameters and calculating their posterior probability distributions. The resulting posterior distributions are then used in the interpretation and application stage for risk assessment and decision support. This is also a crucial input for uncertainty quantification using Monte Carlo simulation methods, aiming to improve the accuracy and reliability of decision-making. The entire fusion model needs to be validated and optimized to ensure its effectiveness. Methods include cross-validation, using model evaluation metrics, and ultimately serving result analysis and decision-making. The flowchart also illustrates the data fusion methodology, including the classification of multi-source data (such as structured and unstructured data, discrete and continuous data) and fusion strategies and techniques (such as data-level fusion, feature-level fusion, decision-level fusion, and fusion algorithms such as weighted average, Kalman filtering, and Bayesian networks). The process demonstrates the complementarity and uncertainty management of multi-source information, and achieves multi-objective fusion through the comprehensive analysis of multiple indicators and image data, reflecting how data fusion in intelligent systems addresses the challenges of data fusion in real-world problems. Despite its advantages, the process also faces challenges in terms of computational complexity and data quality requirements. These are summarized in the section on the advantages and challenges of Bayesian fusion, and demonstrated through application examples and case studies.

[0111] Constructing prior probability distributions representing geological parameters is a key step in data fusion based on a Bayesian fusion framework. It formalizes geological expert knowledge and geological rules into probability distributions. In practice, this construction requires integrating multiple geological information sources: First, regularized priors based on regional geological survey results, such as the vertical stacking sequence of strata and the lateral facies transition laws of lithology, can be transformed into state transition probability matrices using a first-order Markov chain model. Second, statistical priors based on existing borehole and geophysical profile data can be fitted with probability density functions for parameters such as lithological distribution and burial depth range using methods such as kernel density estimation and spatial autocorrelation analysis. Third, qualitative knowledge based on geological expert experience, such as the probability intervals of fault development and favorable lithofacies zones for ore body occurrence, can be mapped into continuous probability distributions using fuzzy membership functions. Furthermore, for geological units of different scales (such as regional structures and local ore bodies), a hierarchical prior probability structure needs to be constructed. Upper-level priors constrain the parameter space of lower-level models to ensure spatial consistency. The constructed prior probability distribution will serve as the initial model. The sampling basis provides a reasonable initial parameter range for the subsequent Bayesian update process, effectively reducing invalid iterations during sampling and improving model convergence efficiency. Regarding sedimentary sequence rules, based on Walther's law and vertical evolution constraints, the stratigraphic superposition relationship can be transformed into a state transition probability matrix using a first-order Markov chain model, quantifying the transition probability of adjacent stratigraphic units. In terms of tectonic constraints, for geometric parameters such as fault strike, dip angle, and fold axial plane morphology defined by the regional tectonic framework, combined with the qualitative judgment of geologists on the intensity of tectonic development, a fuzzy membership function is used to map these parameters into a continuous probability distribution, clarifying the spatial possibility interval of tectonic elements. Regarding lithological distribution patterns, based on sedimentary facies models (such as the lithological assemblage characteristics of delta fronts and fluvial facies), kernel density estimation and spatial autocorrelation analysis are performed on lithological samples from existing borehole and geophysical profile data to fit the probability density function of lithological types in three-dimensional space, accurately characterizing the statistical laws of lateral facies transitions in lithology. The multi-source prior distributions formed by the above probabilistic expressions collectively constitute the initial model. The sampling basis provides a reasonable initial range of parameters for the subsequent Bayesian update process, effectively reduces invalid iterations of adaptive MCMC sampling, improves model convergence efficiency and spatial consistency, and provides a reliable prior constraint basis for uncertainty quantification.

[0112] Among them, the a priori distribution types include: ① Gaussian prior: applicable to continuously varying geological parameters, such as porosity or permeability, with the corresponding formula as follows: ② Multinomial prior: Applicable to discrete geological categories, such as lithology or facies zones, with the corresponding formula as follows: ,in, This reflects the experts' preference for different geological categories.

[0113] Among them, prior information can be obtained from the following channels: (1) Borehole core analysis data: covering the measured results of physical parameters such as lithology sequence, porosity, and permeability, which is the direct basis for the quantitative constraint of geological models. (2) Well logging curve interpretation results: information on lithology, thickness, and fluid properties of strata inferred from well logging response characteristics such as resistivity, sonic transit time, and density. (3) Sedimentary facies evolution model: based on the regularity of lithological combination and spatial distribution in different sedimentary environments summarized by sedimentological theory, providing theoretical guidance for facies zone division. (4) Remote sensing image interpretation results: macroscopic geological information such as surface faults, folds, and lithological outcrops identified by satellite or airborne remote sensing data, which helps to constrain the spatial morphology of underground geological bodies. (5) Engineering geological survey report: including the stratigraphic structure and geotechnical parameters exposed by engineering projects such as foundation pits and tunnels, providing high-precision verification and supplementation for local regional geological models.

[0114] Specifically, a corresponding likelihood function (i.e., likelihood function) is established for the data source. The establishment of the model describes the geological model's ability to interpret observational data and serves as a bridge connecting theoretical models and actual observations. Specifically, ① the likelihood function of borehole data, specifically, for the depth of the stratigraphic interface observed in the borehole... Assume that the observation error follows a normal distribution: ,in, (m) represents the interface depth predicted by the model. ① Standard deviation of observation error. ② Likelihood function for geophysical data; specifically, for seismic reflection data, the likelihood function needs to consider forward modeling: ,in, For earthquake forward modeling operators, This is the noise covariance matrix. ③ The likelihood function of geological map data; specifically, for observations of surface geological boundaries, a distance metric can be used: ,in, The Hausdorff distance between the boundary lines.

[0115] The parameters in the likelihood function (such as the error variance) can be estimated by the following methods: ① Empirical estimation: based on statistical analysis of historical data; ② Cross-validation: evaluating the prediction error by methods such as leave-one-out method; ③ Hierarchical Bayesian method: treating the error parameter as a random variable for inference.

[0116] Due to the high dimensionality of geological models and the complexity of posterior distributions, it is usually impossible to directly calculate the posterior distribution. Therefore, the Markov chain Monte Carlo sampling method (i.e., the Markov chain Monte Carlo (MCMC) method) is required for sampling. The basic principle of the MCMC method is: the MCMC method constructs a Markov chain so that its stationary distribution is equal to the posterior probability distribution of the target geological parameters (i.e., the posterior distribution P(m|D)). After a sufficiently long iteration, the samples of the Markov chain can be approximated as samples drawn from the posterior distribution. The Metropolis-Hastings algorithm is the most basic MCMC method. Its steps are as follows: (1) Initialization: Select the initial model. (2) Propose a new state: from the proposal distribution Mid-sampling candidate model (3) Calculate the acceptance probability: (4) Accept or reject: based on probability accept Otherwise, keep (5) Iteration: Repeat steps (2)-(4) until convergence.

[0117] To address the high dimensionality and strong spatial correlation of geological model parameters, the adaptive MCMC algorithm dynamically optimizes the proposal distribution by analyzing the parameter distribution characteristics (such as mean, variance, and covariance matrix between parameters) in historical sampling trajectories in real time. The algorithm considers the shape and scale of geological parameters. For example, when a set of geological parameters (such as formation permeability and porosity) exhibits significant spatial clustering characteristics during sampling, the algorithm automatically adjusts the proposed distribution covariance of these parameters to enhance the correlation matching between parameters and reduce invalid sampling. Simultaneously, combining the multi-scale hierarchical structure of the geological model, it employs differentiated proposed distribution adjustment strategies for parameters at different scales—larger adjustment step sizes are used for macro-regional parameters to quickly explore the distribution space, while finer step sizes are used for micro-local parameters to ensure sampling accuracy. Furthermore, the algorithm incorporates trajectory analysis results from convergence diagnosis into adaptive logic: if the trajectory plot shows that parameter fluctuations are persistently large and unstable, the algorithm increases the exploration range of the proposed distribution; if the trajectory shows a stable trend, the step size is reduced to decrease sample autocorrelation. Through this dynamic adjustment mechanism, the improved adaptive MCMC algorithm can increase the convergence speed by more than 30% in geological model sampling, while effectively reducing the autocorrelation coefficient between samples, providing a more efficient and reliable sampling basis for subsequent geological model parameter updates and uncertainty quantification using Monte Carlo simulation methods (such as calculating confidence intervals for resource estimation).

[0118] Multiple Markov chains at different temperatures are run, and sampling difficulties of multi-peak posterior distributions are overcome through inter-chain exchanges. The settings of different temperature chains need to be adapted to the multi-scale hierarchical structure of the geological model: chains corresponding to macroscopic regional parameters adopt a wider temperature gradient to quickly break through the local optimum of the regional geological structure; chains corresponding to microscopic local parameters adopt a narrower temperature gradient to balance exploration and sampling accuracy. For multi-peak posterior distributions commonly found in geological models (such as the combined distribution of permeability and porosity under different lithological combinations), high-temperature chains can escape local extreme regions by virtue of their higher energy states, while low-temperature chains focus on high-probability density areas of the target distribution for fine sampling; the inter-chain state exchange follows the Metropolis-Hastings acceptance criterion, attempting to exchange the current states of adjacent temperature chains every fixed number of iterations (e.g., 10-20 times), with the acceptance probability determined by the temperature difference between the two chains and the ratio of their corresponding posterior probabilities. Furthermore, this strategy can work in conjunction with the aforementioned adaptive proposal distribution adjustment mechanism: when the trajectory plot of the convergence diagnosis shows large fluctuations in the low-temperature chain parameters, the temperature difference between adjacent temperature chains can be dynamically increased to enhance the exploration capability; if the low-temperature chain trajectory tends to be stable, the temperature difference can be reduced to improve sampling efficiency. Through this collaborative optimization, the parallel tempering strategy can significantly reduce sampling bias in multi-peak distribution scenarios. Combined with the adaptive MCMC algorithm, it can further improve the convergence speed of geological model sampling by 15%-20%, ensuring that the results of subsequent uncertainty quantification (such as confidence intervals for resource estimation) using Monte Carlo simulation methods are more robust.

[0119] Gradient-guided sampling is particularly suitable for high-dimensional continuous parameter spaces. In geological model parameter estimation, it effectively avoids the inefficient exploration problem of random walk sampling in high-dimensional geological parameter spaces (such as three-dimensional permeability fields and porosity distribution of multiple lithological combinations) by introducing a potential function coupled with geological priors (such as the continuity constraints of stratigraphic interfaces and the spatial correlation of lithological distribution). Compared with the traditional random walk MCMC, this algorithm can quickly cross low-probability regions by utilizing the gradient information of geological parameters (such as the lithological boundary gradient based on seismic inversion data and the parameter change rate of well logging curves), significantly reducing the autocorrelation of sampling trajectories and improving mixing efficiency. In the geological model construction of this invention, Hamiltonian Monte Carlo is combined with an adaptive proposal distribution mechanism, which can dynamically adjust the step size according to the spatial heterogeneity of geological parameters: when there is strong spatial correlation of geological parameters (such as the permeability distribution of continuous sand bodies), the step size is increased to accelerate global exploration; when there are drastic local changes in geological parameters (such as lithological abrupt changes near faults), the step size is reduced to ensure sampling accuracy. With this adjustment, the convergence speed of the algorithm is improved by 25%-30% compared with traditional methods when dealing with complex geological structures. This provides a more efficient and stable sampling basis for updating model parameters after multi-source data fusion and for uncertainty quantification using Monte Carlo simulation methods (such as calculating confidence intervals for resource estimation).

[0120] Sampling convergence can be evaluated using the Gelman-Rubin statistic, which is interpreted as comparing the variances of multiple chains; a value close to 1 indicates convergence.

[0121] Based on MCMC sampling results, model parameters can be updated and uncertainties quantified, providing probabilistic support for geological decision-making, specifically including: ① Posterior mean estimation: ,in, The first sample obtained by MCMC ① Sample size. ② Maximum a posteriori (MAP) estimate: MAP estimation can select the model with the highest posterior probability from the sampled data.

[0122] Uncertainty quantification indicators include: ① Posterior variance: This reflects the degree of uncertainty in the model parameters. ② Prediction uncertainty: For predicting geological attributes at unobserved locations, the prediction distribution can be calculated: .

[0123] The design principle of the three-level coding system for bedrock strata is as follows: standardized coding of bedrock strata is the foundation for realizing multi-source data fusion and multi-scale modeling. The three-level coding system proposed by Qingdao City systematically expresses the weathering degree, genetic age, and lithological characteristics of bedrock, providing a unified framework for the digital management of geological information. The design objectives of the coding system include: ① Uniqueness: Each geological unit corresponds to a unique code, avoiding ambiguity; ② Hierarchy: The coding structure reflects the hierarchical relationship of geological classification; ③ Scalability: Reserving coding space to adapt to newly discovered geological types; ④ Compatibility: Coordinating with existing geological survey and engineering exploration specifications; ⑤ Readability: The codes possess certain semantic information, facilitating manual identification.

[0124] The three-level coding adopts the format "XX-XX-XX" or "XXY-XXY-XXY". The first level (weathering degree) reflects the physical weathering state of the bedrock; the main code (XX) indicates the weathering level (e.g., unweathered, slightly weathered, moderately weathered, strongly weathered, completely weathered); the auxiliary code (Y) further subdivides the weathering characteristics (e.g., the degree of development of weathering fissures, the type of weathering products); the second level (genesis and age) expresses the geological genesis and formation age of the bedrock; the main code (XX) indicates the geological age and rock genetic type (e.g., Yanshanian intrusive rocks, Cretaceous volcanic rocks); the auxiliary code (Y) refines the rock assemblage or lithofacies characteristics; the third level (lithological characteristics) describes the mineral composition and structure of the bedrock; the main code (XX) identifies the main rock type (e.g., granite, basalt, gneiss); the auxiliary code (Y) supplements the minor mineral or special structural information.

[0125] The flexibility and simplification rules of the coding include: ① Optional auxiliary codes: When detailed information is not required, the auxiliary codes can be omitted and simplified to the “XX-XX-XX” format; ② Hierarchical aggregation: In large-scale modeling, the third-level coding can be ignored and units with similar lithology can be merged; ③ Dynamic expansion: The auxiliary codes can be increased by 1-3 digits as needed to adapt to special geological conditions.

[0126] Furthermore, the 3D modeling module supports model construction at the following three scales: ① Regional scale (1:50000-1:250000): tectonic unit division, reconstruction of major stratigraphic interfaces, and 3D representation of the fault system; ② Mining area scale (1:10000-1:50000): engineering geological stratification, groundwater system simulation, and identification of geological hazard-prone areas; ③ Site scale (1:500-1:10000): detailed soil and rock stratification, 3D distribution of geological parameters, and engineering suitability evaluation.

[0127] Furthermore, the uncertainty quantification module employs the Monte Carlo simulation method, with at least 10,000 iterations (i.e., generating a perturbation sample set based on the data error distribution, typically no smaller than 10,000 samples), outputting the following uncertainty indicators: 95% confidence intervals for model parameters, probability distribution function of prediction results, standard deviation of key geological interface locations, and risk level zoning map. Specific technical requirements for uncertainty quantification using the Monte Carlo simulation method (i.e., the uncertainty quantification module) are specified, with the Monte Carlo simulation method used as the core technical approach. This method assesses the uncertainty of the geological model through extensive random sampling, requiring at least 10,000 iterations to ensure the reliability of the statistical results. Technical parameter requirements: Iteration count standard, minimum iteration count: 10,000; recommended iteration count: 50,000-100,000 (for complex geological conditions); Convergence test: determining whether to increase the iteration count through variance stability. Computational resource requirements: Each iteration requires regenerating a complete multi-scale 3D geological model, necessitating parallel computing capabilities to improve efficiency; storage space requirements increase linearly with the number of iterations.

[0128] This paper further elaborates on multi-scale three-dimensional geological models. First, it introduces the concept and its importance, clarifying the definition of a multi-scale three-dimensional geological model as a three-dimensional model constructed by combining geological data at different scales. It also discusses the classification of models (based on different classification methods according to data sources and application fields) and the importance of models, reflecting their application in geological research (to interpret geological phenomena and predict resource distribution) and their value in engineering applications (in fields such as mineral mining and civil engineering). Its construction relies on technical principles and methods. First, it involves data acquisition and processing. Data acquisition technology covers data acquisition methods such as geological exploration and remote sensing. Data processing methods correspond to the process of acquiring and standardizing geological data from different sources, including data preprocessing, correction, and fusion. The core is model building technology, which uses potential field interpolation to build multi-scale three-dimensional geological models. This technology involves mesh generation technology (metamorphism and optimization of meshes at different scales) and model interpolation and fitting (interpolating and fitting geological interfaces using algorithms). For example, geological interfaces are constructed by solving potential field equations. The constructed models play a key role in application fields and case studies, including resource exploration and development (such as using models to optimize exploration strategies and resource assessment in mineral resource exploration, and application examples of models in oil and gas field development) and environmental protection and disaster prediction (such as the role of models in geological disaster monitoring and early warning, and assessing the impact of human activities on the geological environment in environmental impact assessment). The development of this field is influenced by both trends and challenges. Technological trends include advancements in high-precision data acquisition technologies (such as high-resolution remote sensing and 3D seismic exploration) and improved computing power (the application of high-performance computing in geological model construction). Simultaneously, challenges exist, such as data integration and management difficulties (the integration of data from different sources and scales) and model validation and updating (the establishment of model validation methods and continuous update mechanisms). Finally, conclusions are presented, including a comprehensive evaluation of multi-scale 3D geological models (the models' overall contribution to geological science and their role in promoting geological science theory and practice) and future prospects (the potential development directions of models in geological research and applications).

[0129] Furthermore, the data management platform adopts a layered architecture, including: a storage layer consisting of a distributed file system (HDFS), a spatial database (PostGIS), a time-series database (InfluxDB), and object storage (MinIO); a service layer comprising RESTful API interfaces, OGC standard services (WMS, WFS, WCS), and real-time data stream processing (Kafka); and an application layer comprising Web 3D visualization (WebGL), a mobile data acquisition app, a desktop GIS integration plugin, and a data analysis workbench. This architecture embodies the core technical characteristics of a modern geological data management system. Employing a classic three-tier design pattern, the architecture organically combines the storage, service, and application layers to achieve a complete technology stack from data storage to business applications. This architectural design aligns perfectly with the data management requirements of the BIM-GIS integration framework, particularly in handling multi-source heterogeneous geological data, where the layered architecture effectively addresses challenges such as diverse data formats, complex access patterns, and varying performance requirements.

[0130] The platform is a complete system that stores models and results in an OGC-compliant data management platform and provides services through a web service interface (i.e., the data management platform). As the system terminal, the platform is responsible for receiving and managing standardized geological data, posterior probability distributions of geological parameters, multi-scale three-dimensional geological models, and uncertainty quantification results generated from the following processes: acquiring and standardizing geological data from different sources; performing data fusion based on a Bayesian fusion framework; constructing multi-scale three-dimensional geological models using potential field interpolation; and quantifying uncertainties using Monte Carlo simulation. The platform's functionality begins with data collection and integration. Its data access module supports access from various data sources (such as databases, APIs, log files, etc.) and implements data flow management for real-time data flow monitoring and management. Subsequently, data cleaning and preprocessing ensure data quality control (removing duplicate, erroneous, and incomplete data) and data format unification (converting data formats to meet platform requirements). The processed data enters the core stage of data storage and management. Based on scientific database design, data model construction (defining data entities, relationships, and attributes) and database optimization are carried out to improve query efficiency and data processing speed. Strict data security and backup strategies are implemented, including access control mechanisms (setting different levels of user permissions) and data backup strategies (regularly backing up data to prevent loss or damage). The stored data empowers users through data visualization and display modules, providing dashboards and charts for real-time data visualization (displaying key data indicators through charts) and interactive data exploration (users can customize queries and analyses). It also supports reporting and data export functions, including multi-format report generation (supporting PDF, Excel, and other formats) and data export functionality (facilitating in-depth data analysis). To uncover the deeper value of the data, the platform provides data analysis and mining capabilities, covering data querying and reporting (such as complex query processing supporting SQL and NoSQL queries, and automated generation of various statistical reports using report generation tools) and data mining algorithm applications (such as predictive model building using machine learning algorithms for predictive analysis, and user behavior analysis analyzing user data to optimize products and services). The platform's security and collaborative operation are ensured by the user management and collaboration module, which enables user permission and role management (assigning roles with different permissions based on responsibilities through role customization and allocation) and user behavior monitoring (monitoring user activities to ensure data security). It also provides team collaboration tools (for task assignment and progress tracking to manage team workflows and project progress) and a knowledge sharing platform to promote the exchange of knowledge and experience among team members.To ensure the system's adaptability and viability, the platform possesses integration and scalability, supporting third-party service integration (such as seamless integration with other systems and platforms via application programming interfaces (APIs), and integration of social media data to enhance analytical capabilities) and system expansion and upgrade capabilities (based on a modular design, it facilitates on-demand expansion of system functions, and a system upgrade strategy is regularly updated to maintain the platform's advanced nature). Ultimately, through the synergy of these functions, the platform provides a robust data infrastructure and intelligent service support for geological model-based risk assessment and decision support.

[0131] The overall architecture of the multi-source 3D geological data fusion and management system has the following data flow and processing logic: The data acquisition layer gathers raw geological data from multiple sources, including borehole data acquisition (such as core scanning, well logging data, and stratigraphic information), geophysical exploration (such as seismic data, gravity and magnetic data, and electrical resistivity data), remote sensing data (such as satellite imagery, aerial photography, and DEM data), geological survey field observations (such as sample analysis and structural measurement), laboratory data (such as rock physical properties, chemical composition, and microstructure), and historical data (such as archival materials, reports, and map data). These heterogeneous data are the starting point for all processing flows of the system. The data then enters the data fusion layer. First, it undergoes standardization processing (i.e., data preprocessing) of geological data from different sources. This step specifically includes data cleaning, format conversion, quality inspection, and coordinate system unification (covering projection transformation, benchmark unification, and accuracy registration) to form standardized geological data. Next, multi-source data registration is performed, including spatial registration, temporal registration, and attribute association, ensuring that the data is fusionable in time, space, and semantics. The core processing is completed by data fusion based on a Bayesian fusion framework (i.e., the fusion algorithm engine), which constructs prior probability distributions representing geological parameters. And establish a corresponding likelihood function for each data source. Inference is performed using the Markov chain Monte Carlo sampling method, which involves weight allocation, conflict resolution, and uncertainty handling, ultimately obtaining the posterior probability distribution of geological parameters. The sample set is used to calculate the posterior probability distribution of geological parameters. This layer then performs a quality assessment, including accuracy evaluation, integrity check, and consistency verification. The fused probabilistic information is input into the 3D modeling layer. This layer first performs geological structure modeling, using potential field interpolation to construct a multi-scale 3D geological model, specifically including stratigraphic modeling, fault modeling, and fold modeling. Simultaneously, attribute modeling, such as physical property distribution, composition modeling, and porosity modeling, is performed, assigning attributes to the geological body. The modeling process involves voxelization (such as mesh generation, interpolation calculation, and resolution optimization) to construct a digital 3D body. Model verification is then performed to ensure the geological rationality and data consistency of the model, and the results of uncertainty quantification using Monte Carlo simulation methods are integrated. This layer also integrates a visualization engine to achieve 3D rendering, interactive operations, and profile generation for model inspection and display. Finally, the completed multi-scale 3D geological model and uncertainty quantification results are persisted and published in the data management layer. The data storage module utilizes distributed storage technology and provides backup, recovery, and version control functions. The metadata management module maintains the data dictionary, lineage, and standard specifications. The access control module ensures security mechanisms for user authentication, permission management, and audit logs. The core service is provided by the data service module, which stores the model and results in an OGC-compliant data management platform and provides services through a web service interface. This module exposes API interfaces to facilitate data sharing and format conversion and relies on workflow management (including process orchestration, task scheduling, and status monitoring) to automate the processing chain. The entire system operates under system monitoring (including performance monitoring, anomaly alerts, and resource management) to ensure stability and efficiency. This system architecture achieves a closed-loop process from multi-source heterogeneous data acquisition, probabilistic fusion, 3D modeling to intelligent management services.

[0132] It should be noted that the beneficial effects of the multi-source 3D geological data fusion and management system provided in the above embodiments are the same as those of the multi-source 3D geological data fusion and management method described above, and will not be repeated here. Furthermore, the system provided in the above embodiments is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to the actual situation to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiments, and will not be repeated here.

[0133] An electronic device according to an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned multi-source three-dimensional geological data fusion and management methods.

[0134] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the above-described methods for the fusion and management of multi-source three-dimensional geological data.

[0135] The above description is merely a preferred embodiment of the present invention and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of disclosure in this invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this invention.

[0136] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for multi-source three-dimensional geological data fusion and management, characterized in that, include: Obtain and standardize geological data from different sources to obtain standardized geological data; Based on the Bayesian fusion framework, the standardized geological data are fused to obtain the posterior probability distribution characterizing the geological parameters. Based on the probability constraints provided by the posterior probability distribution, a multi-scale three-dimensional geological model is constructed using the potential field interpolation method. The Monte Carlo simulation method is used to quantify the uncertainty of the multi-scale three-dimensional geological model, generating uncertainty quantification results corresponding to the multi-scale three-dimensional geological model. The multi-scale three-dimensional geological model and the uncertainty quantification results are stored in a data management platform that conforms to the OGC standard, and the visualization, query and access of the geological layers are provided through a Web service interface; Based on a Bayesian fusion framework, the standardized geological data are fused to obtain the posterior probability distribution characterizing the geological parameters, including: A prior probability distribution characterizing geological parameters is constructed, which integrates geological rules and expert experience; A corresponding likelihood function is established for each data source in the standardized geological data, and the likelihood function describes the degree of matching between the data source and the multi-scale three-dimensional geological model; Based on the prior probability distribution and the likelihood function of all data sources, a sample set of the posterior probability distribution of the geological parameters is obtained using the Markov chain Monte Carlo sampling method. The posterior probability distribution of the geological parameters is calculated based on the sample set of the posterior probability distribution. Based on the probability constraints provided by the posterior probability distribution, a multi-scale three-dimensional geological model is constructed using the potential field interpolation method, including: Based on the aforementioned posterior probability distribution, the probability information of geological interface feature points and stratigraphic attribute distribution is determined. Based on the probability information and borehole stratification data, a potential field function describing the spatial morphology of the geological interface is defined. The potential field equation corresponding to the potential field function is solved by numerical method to obtain the continuous potential field distribution in three-dimensional space. Equipotential surfaces with preset values ​​are extracted from the continuous potential field distribution, and the equipotential surfaces are used as geological interfaces in the multi-scale three-dimensional geological model. The geological interface is topologically optimized by applying geological rules of stratigraphic stacking and fault cutting to complete the construction of the multi-scale three-dimensional geological model; The Monte Carlo simulation method is used to quantify the uncertainty of the multi-scale three-dimensional geological model, generating uncertainty quantification results corresponding to the multi-scale three-dimensional geological model, including: Based on the error characteristics of different data sources in the standardized geological data, a set of perturbation samples conforming to a preset probability distribution is generated; For each sample in the disturbed sample set, the data fusion and 3D modeling process is re-executed to obtain the corresponding multi-scale 3D geological model output; Statistical analysis is performed on the set of outputs from the multi-scale three-dimensional geological model to calculate the mean, standard deviation, and confidence interval of the geological parameters; Based on the calculated mean, standard deviation, and confidence interval of the geological parameters, a probability distribution function and risk zoning map characterizing the uncertainty of the multi-scale three-dimensional geological model are generated as the uncertainty quantification result.

2. A multi-source three-dimensional geological data fusion and management system, characterized in that, It includes an acquisition and processing module, a fusion module, a construction module, a generation module, and a management module; The acquisition and processing module is used to: acquire and standardize geological data from different sources to obtain standardized geological data; The fusion module is used to: fuse the standardized geological data based on a Bayesian fusion framework to obtain a posterior probability distribution characterizing the geological parameters; The fusion module is used to: construct a multi-scale three-dimensional geological model based on the probability constraints provided by the posterior probability distribution and using the potential field interpolation method. The generation module is used to: use the Monte Carlo simulation method to perform uncertainty quantification on the multi-scale three-dimensional geological model, and generate uncertainty quantification results corresponding to the multi-scale three-dimensional geological model; The management module is used to: store the multi-scale three-dimensional geological model and the uncertainty quantification results to a data management platform that conforms to the OGC standard, and provide visualization, query and access to the geological layers through a Web service interface; The fusion module is specifically used for: A prior probability distribution characterizing geological parameters is constructed, which integrates geological rules and expert experience; A corresponding likelihood function is established for each data source in the standardized geological data, and the likelihood function describes the degree of matching between the data source and the multi-scale three-dimensional geological model; Based on the prior probability distribution and the likelihood function of all data sources, a sample set of the posterior probability distribution of the geological parameters is obtained using the Markov chain Monte Carlo sampling method. The posterior probability distribution of the geological parameters is calculated based on the sample set of the posterior probability distribution. The building module is specifically used for: Based on the aforementioned posterior probability distribution, the probability information of geological interface feature points and stratigraphic attribute distribution is determined. Based on the probability information and borehole stratification data, a potential field function describing the spatial morphology of the geological interface is defined. The potential field equation corresponding to the potential field function is solved by numerical method to obtain the continuous potential field distribution in three-dimensional space. Equipotential surfaces with preset values ​​are extracted from the continuous potential field distribution, and the equipotential surfaces are used as geological interfaces in the multi-scale three-dimensional geological model. The geological interface is topologically optimized by applying geological rules of stratigraphic stacking and fault cutting to complete the construction of the multi-scale three-dimensional geological model; The generation module is specifically used for: Based on the error characteristics of different data sources in the standardized geological data, a set of perturbation samples conforming to a preset probability distribution is generated; For each sample in the disturbed sample set, the data fusion and 3D modeling process is re-executed to obtain the corresponding multi-scale 3D geological model output; Statistical analysis is performed on the set of outputs from the multi-scale three-dimensional geological model to calculate the mean, standard deviation, and confidence interval of the geological parameters; Based on the calculated mean, standard deviation, and confidence interval of the geological parameters, a probability distribution function and risk zoning map characterizing the uncertainty of the multi-scale three-dimensional geological model are generated as the uncertainty quantification result.

3. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the multi-source three-dimensional geological data fusion and management method of claim 1.

4. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the multi-source three-dimensional geological data fusion and management method of claim 1.

Images