Visual dynamic processing method and system for geological survey data

Abstract

The invention provides a geological exploration data visualization dynamic processing method and system, and relates to the technical field of intelligent geological exploration, and the method comprises the steps: constructing a prior three-dimensional geological structure model and a forward substitution model; constructing a multi-source joint loss function based on forward modeling output of the forward modeling substitution model and geological survey data; based on a multi-source joint loss function, performing layer-by-layer reasoning reconstruction on geological point cloud parameters in the prior three-dimensional geological structure model to obtain a dynamic three-dimensional geological parameter set; performing fusion rendering and interaction analysis on the dynamic three-dimensional geological parameter set to obtain a visualization result of a fusion geological structure and multi-field evolution information, and constructing a scene deduction component based on the dynamic three-dimensional geological parameter set and a forward substitution model; according to the geological situation sensing and deduction early warning method and device, high-precision geological situation sensing and deduction early warning can be achieved through deep fusion of the forward modeling model constrained by physical information and the layered dynamic inversion algorithm.

Description

Technical Field

[0001] This application relates to the field of intelligent geological exploration technology, and more specifically, to a method and system for dynamic processing of geological exploration data visualization. Background Technology

[0002] With the development of the Internet of Things, high-performance computing, and artificial intelligence technologies, the field of geophysical exploration is gradually entering a new stage of intelligent geological exploration. Intelligent geological exploration aims to integrate multi-source exploration data (such as seismic, electromagnetic, borehole, and remote sensing data) and utilize advanced data fusion, numerical simulation, and machine learning technologies to construct high-fidelity three-dimensional subsurface geological models, thereby achieving digital and intelligent perception and prediction of geological structures, properties, and evolution processes.

[0003] Against this backdrop, dynamic processing methods for geological exploration data visualization have become a key supporting technology. By combining geological modeling, geophysical data inversion, and 3D visualization, these methods attempt to present underground conditions through an intuitive graphical interface, assisting engineering decision-making and disaster early warning. However, existing geological exploration data visualization still suffers from significant technical shortcomings, failing to meet the engineering requirements of high precision, real-time performance, and interpretability. Specifically: while pure physical simulations offer interpretability, they are computationally expensive, difficult to update in real time, and heavily reliant on precise initial parameters and boundary conditions; while pure data-driven models are computationally fast, they lack physical constraints and are prone to producing physically unreasonable predictions when training data is insufficient or in out-of-distribution scenarios, resulting in low generalization ability and reliability; furthermore, existing visualizations are mostly static 3D displays or only provide simple attribute queries, failing to deeply integrate and contextualize real-time updated geological parameters, multiphysics evolution predictions, and risk warning indicators, making it difficult for visualization systems to directly support forward-looking decision-making. Therefore, how to achieve high-precision geological situation awareness and early warning through the deep integration of physically constrained forward models and hierarchical dynamic inversion algorithms is a challenge facing the industry. Summary of the Invention

[0004] This application provides a method and system for dynamic processing of geological exploration data visualization, which can achieve high-precision geological situation perception and early warning through the deep integration of a forward model constrained by physical information and a hierarchical dynamic inversion algorithm.

[0005] In a first aspect, this application provides a dynamic processing method for visualizing geological exploration data, the dynamic processing method comprising the following steps: A priori three-dimensional geological structure model is constructed based on a geological knowledge rule base and planar geological maps. A forward substitution model is obtained by training a graph neural network architecture constrained by physical information based on multi-physics field control equations. Obtain geological exploration data, and construct a multi-source joint loss function based on the forward modeling output of the forward substitution model and the geological exploration data; The hierarchical recursive inversion optimization algorithm is used to reconstruct the geological point cloud parameters in the prior three-dimensional geological structure model layer by layer based on the multi-source joint loss function to obtain a dynamic three-dimensional geological parameter set. The dynamic three-dimensional geological parameter set is fused, rendered, and interactively analyzed to obtain a visualization result that integrates geological structure and multi-field evolution information. A scenario inference component is constructed based on the dynamic three-dimensional geological parameter set and the forward substitution model. Dynamic early warning decisions are made based on the visualization results and the scenario simulation components for geological exploration data.

[0006] In this embodiment, the construction of a priori three-dimensional geological structure model based on a geological knowledge rule base and planar geological maps specifically includes: Geological element analysis is performed on the aforementioned planar geological map to obtain an element feature set; A set of three-dimensional control points is determined based on the geological knowledge rule base and the feature set of the elements. By spatially mapping the three-dimensional control point set through the spatial topological relationships between geological features, a priori three-dimensional geological structure model is obtained.

[0007] In this embodiment, the forward substitution model obtained by training a graph neural network architecture constrained by physical information based on multiphysics control equations specifically includes: A graph neural network architecture is constructed based on the connection relationships of geological structure nodes; The physical regularization loss term is determined based on the multiphysics control equations, and then the physical regularization loss term and the data fitting loss term are combined to obtain the joint loss function. The graph neural network architecture is optimized and trained using geological structure parameters based on the joint loss function to obtain a forward substitution model.

[0008] In this embodiment, geological survey data is acquired using a ground-based three-dimensional laser scanning device.

[0009] In this embodiment, constructing a multi-source joint loss function based on the forward modeling output of the forward substitution model and the geological exploration data specifically includes: The spatial distribution loss term is determined by the different data types of the geological exploration data; The forward modeling output of the forward substitution model and the spatial distribution loss term are matched to obtain the multi-source loss value; A multi-source joint loss function is constructed based on the multi-source loss value and the physical constraint penalty term of the forward substitution model.

[0010] In this embodiment, a hierarchical recursive inversion optimization algorithm is used to perform layer-by-layer inference and reconstruction of the geological point cloud parameters in the prior 3D geological structure model based on the multi-source joint loss function, resulting in a dynamic 3D geological parameter set, specifically including: The geological point cloud parameters in the prior three-dimensional geological structure model are used as the parameter set to be optimized. Using the multi-source joint loss function as the objective, the optimized geological point cloud parameters are obtained by inverting and iterating the set of parameters to be optimized and updating the gradient through the ensemble smoother algorithm. The optimized geological point cloud parameters are used as spatial constraints for layer-by-layer inversion to obtain a dynamic three-dimensional geological parameter set.

[0011] In this embodiment, the fusion rendering and interactive analysis of the dynamic three-dimensional geological parameter set to obtain a visualization result that integrates geological structure and multi-field evolution information specifically includes: The dynamic three-dimensional geological parameter set is visualized and mapped, and then the geological structure geometry, physical field attributes and deterministic information in the dynamic three-dimensional geological parameter set are converted into geometric primitives and color mapping textures, respectively. The geometric primitives and the color mapping texture are fused and drawn to obtain an initial three-dimensional scene; In the initial 3D scene, parameterized analysis instructions are generated in response to user operations; The initial 3D scene and the parameterized analysis instructions are encapsulated to obtain a visualization result that integrates geological structure and multi-field evolution information.

[0012] In this embodiment, the scenario inference component constructed based on the dynamic three-dimensional geological parameter set and the forward modeling substitution model specifically includes: The initial state of the model is determined by the dynamic three-dimensional geological parameter set, and a forward evolution operator is established based on the forward substitution model. An inference engine for constructing scenario inference components is built based on the initial state of the model and the forward evolution operator; A scenario-driven interface is constructed, and then a scenario inference component is constructed through the inference engine and the scenario-driven interface.

[0013] In this embodiment, the dynamic early warning decision-making based on the visualization results and the scenario inference component for geological exploration data specifically includes: The analysis instructions in the visualization results are parsed to generate early warning analysis tasks; Based on the aforementioned early warning analysis task, the scenario simulation component is invoked to calculate the probability of exceeding the limits and the rate of evolution of key early warning indicators; A comprehensive risk assessment is conducted on the probability of exceeding the limit and the rate of evolution based on preset risk level classification rules to obtain an early warning decision report.

[0014] Secondly, this application provides a dynamic processing system for visualizing geological exploration data, used to execute a dynamic processing method for visualizing geological exploration data, the dynamic processing system comprising: The model building module is used to construct a priori three-dimensional geological structure model based on a geological knowledge rule base and planar geological maps. The forward substitution model is obtained by training a graph neural network architecture constrained by physical information based on multi-physics field control equations. The data fusion module is used to acquire geological exploration data and construct a multi-source joint loss function based on the forward modeling output of the forward substitution model and the geological exploration data. The recursive inversion module is used to use a hierarchical recursive inversion optimization algorithm to perform layer-by-layer reasoning and reconstruction of the geological point cloud parameters in the prior three-dimensional geological structure model based on the multi-source joint loss function, so as to obtain a dynamic three-dimensional geological parameter set. The visualization and simulation module is used to perform fusion rendering and interactive analysis on the dynamic three-dimensional geological parameter set to obtain a visualization result that integrates geological structure and multi-field evolution information. Based on the dynamic three-dimensional geological parameter set and the forward modeling substitution model, a scenario simulation component is constructed. The early warning decision module is used to make dynamic early warning decisions based on the visualization results and the scenario simulation component.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: A priori 3D geological structure model is constructed based on a geological knowledge rule base and planar geological maps. A forward substitution model is obtained by training a graph neural network architecture constrained by physical information based on multiphysics field control equations. Geological exploration data is acquired, and a multi-source joint loss function is constructed based on the forward output of the forward substitution model and the geological exploration data. A hierarchical recursive inversion optimization algorithm is used to reconstruct the geological point cloud parameters in the priori 3D geological structure model layer by layer based on the multi-source joint loss function, resulting in a dynamic 3D geological parameter set. The dynamic 3D geological parameter set is fused, rendered, and interactively analyzed to obtain a visualization result that integrates geological structure and multi-field evolution information. A scenario inference component is constructed based on the dynamic 3D geological parameter set and the forward substitution model. Dynamic early warning decisions are made based on the visualization result and the scenario inference component using geological exploration data.

[0016] Therefore, this application demonstrates that it can achieve deep integration of a forward model constrained by physical information and a hierarchical dynamic inversion algorithm. Firstly, by constructing a priori 3D geological structure model based on a geological knowledge rule base and planar geological maps, and training a forward substitution model using a graph neural network constrained by physical information based on multi-physics control equations, prior geological knowledge and physical mechanisms are deeply embedded into a machine learning framework. This constructs a simulation foundation that combines expert experience, physical interpretability, and efficient computational capabilities, solving the problems of pure data-driven models lacking physical constraints and having weak generalization ability. Furthermore, by acquiring geological exploration data and constructing a multi-source joint loss function based on the forward output of the forward substitution model and the exploration data, a data fitting term and a physical regularization term can be integrated, ensuring that subsequent optimization processes simultaneously pursue both the fit to measured data and adherence to physical laws. Secondly, by using a hierarchical recursive inversion optimization algorithm based on the multi-source joint loss function, the prior model... The geological point cloud parameters in the model are reconstructed layer by layer to obtain a dynamic three-dimensional geological parameter set. A recursive optimization strategy from global to local and from coarse to fine is adopted, which significantly improves the computational efficiency and stability of complex geological parameter inversion. This is conducive to realizing dynamic, online, and high-precision updates of the geological model with the input of new exploration data. Then, by fusing, rendering and interactively analyzing the dynamic three-dimensional geological parameter set, a visualization result of the fused structure and multi-field information is obtained. Based on the dynamic parameter set and the forward modeling substitution model, a scenario inference component is constructed, which can complete the real-time transformation of the geological situation from abstract data to intuitive three-dimensional scenes and visualize and infer future scenarios such as groundwater seepage and rock stress evolution. Finally, based on the visualization results and the scenario inference component, dynamic early warning decisions are made for geological exploration data. The system can parse interactive analysis commands, automatically call the inference engine to calculate the probability of exceeding limits and the rate of evolution of key indicators, and conduct a comprehensive evaluation in combination with risk rules.

[0017] In summary, the technical solution adopted in this application can achieve high-precision geological situation perception and early warning by deeply integrating the forward model constrained by physical information with the hierarchical dynamic inversion algorithm. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a dynamic processing method for visualizing geological exploration data provided in this application; Figure 2 This is an exemplary flowchart for determining a multi-source joint loss function according to the present application; Figure 3 This is an exemplary flowchart for determining a dynamic three-dimensional geological parameter set according to the present application; Figure 4 This is a module structure diagram of a dynamic processing system for visualizing geological exploration data provided in this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] This application provides a method and system for dynamic processing of geological exploration data visualization. Its core is to construct a priori three-dimensional geological structure model based on a geological knowledge rule base and planar geological maps; to obtain a forward substitution model through a graph neural network architecture constrained by physical information and trained on multi-physics field control equations; to acquire geological exploration data; to construct a multi-source joint loss function based on the forward output of the forward substitution model and the geological exploration data; to use a hierarchical recursive inversion optimization algorithm to perform layer-by-layer reasoning and reconstruction of the geological point cloud parameters in the priori three-dimensional geological structure model based on the multi-source joint loss function, thereby obtaining a dynamic three-dimensional geological parameter set; to perform fusion rendering and interactive analysis on the dynamic three-dimensional geological parameter set, obtaining a visualization result that integrates geological structure and multi-field evolution information; to construct a scenario inference component based on the dynamic three-dimensional geological parameter set and the forward substitution model; and to make dynamic early warning decisions based on the visualization results and the scenario inference component.

[0022] Example 1: To better understand the above technical solution, the following will provide a detailed description of the technical solution in conjunction with the accompanying drawings and specific implementation methods. (Refer to...) Figure 1 As shown in the figure, this is a flowchart of a dynamic processing method for visualizing geological exploration data according to this embodiment of the present application. The dynamic processing method includes the following steps: In step S1, a priori three-dimensional geological structure model is constructed based on the geological knowledge rule base and planar geological maps. A forward substitution model is obtained by training a graph neural network architecture constrained by physical information based on multi-physics field control equations.

[0023] In this embodiment, the construction of a priori three-dimensional geological structure model based on a geological knowledge rule base and planar geological maps can be carried out in the following manner: Geological element analysis is performed on the aforementioned planar geological map to obtain an element feature set; A set of three-dimensional control points is determined based on the geological knowledge rule base and the feature set of the elements. By spatially mapping the three-dimensional control point set through the spatial topological relationships between geological features, a priori three-dimensional geological structure model is obtained.

[0024] In practice, the process begins by reading a planar geological map file (e.g., Shapefile format). An open-source geographic information processing library (e.g., GDAL) is used to parse the point, line, and polygon features, extracting their coordinate information and geological attribute fields (e.g., "strata name" and "fault type"). These extracted feature sets are then combined into a feature set. Next, a predefined geological knowledge rule base file (e.g., JSON format) is loaded. This rule base file contains entries such as "stratigraphic attitude rules" and "fault geometry rules." The program iterates through the feature set, and for each "stratigraphic boundary" feature, based on the corresponding "stratigraphic attitude rule" in the rule base (e.g., dip 120°, dip angle 30°), the two-dimensional coordinate points are aligned along the dip... The orientation and dip angle are extrapolated to different depths to generate a set of three-dimensional spatial points. For the "fault" element, three-dimensional points controlling the fault plane are generated according to the "fault geometry rules". All three-dimensional spatial points generated after rule transformation of all elements are summarized to form a point cloud data set. This point cloud data set is then used as a three-dimensional control point set. Finally, using a three-dimensional geological modeling library (such as GemPy), the three-dimensional control point set is grouped and input according to its geological interface. At the same time, the spatial topological relationship between geological bodies is used as a modeling constraint. Based on all scattered data and topological constraints, the three-dimensional geological modeling library can generate continuous three-dimensional surfaces (triangular meshes) through interpolation algorithms. The three-dimensional geological framework model composed of triangular mesh surfaces is then used as a priori three-dimensional geological structure model.

[0025] It should be noted that the geological knowledge rule base in this application refers to a set of geological experiences and rules stored in a structured form and parsable by a computer, which can transform abstract two-dimensional map information into specific constraints for three-dimensional spatial construction; the feature set is a set of basic graphic units with geological semantics extracted from the original map; the three-dimensional control point set is a discrete skeleton point that represents the spatial morphology of geological interfaces after applying knowledge rules, and is the skeleton for constructing a continuous three-dimensional model; the prior three-dimensional geological structure model is an initial three-dimensional geometric framework that is reconstructed based on control points and topological relationships and conforms to geological laws, which can provide a definite spatial carrier for subsequent physical attribute inversion and dynamic updates.

[0026] In this embodiment, the forward substitution model obtained by training a graph neural network architecture constrained by physical information based on multiphysics control equations can be specifically achieved in the following manner: A graph neural network architecture is constructed based on the connection relationships of geological structure nodes; The physical regularization loss term is determined based on the multiphysics control equations, and then the physical regularization loss term and the data fitting loss term are combined to obtain the joint loss function. The graph neural network architecture is optimized and trained using geological structure parameters based on the joint loss function to obtain a forward substitution model.

[0027] In specific implementation, firstly, the grid data of the prior 3D geological structure model is read, and each grid node of the prior 3D geological structure model is used as a graph node. The edges connecting the nodes in the grid are used as edges of the graph. Then, a graph data object is created using a graph data processing library (such as PyTorchGeometric). The feature vector of each node is initialized with its 3D spatial coordinates and physical property values ​​(such as density). A graph convolutional network model is selected as the basic structure. That is, a network class containing multiple graph convolutional layers, nonlinear activation functions and output layers can be defined using a deep learning framework (such as PyTorch). Then, this defined neural network class that can receive and process the above geological grid graph data is used as a graph neural network architecture. Then, the governing equation (such as the Poisson equation) of the physical field to be simulated (such as the gravity field) is determined, and automatic differentiation technology is used as a differentiable numerical solver for the governing equation. During training, geological structure parameters are simultaneously input into a graph neural network architecture and a differentiable numerical solver. The graph neural network architecture outputs predicted physical field values, while the differentiable numerical solver outputs physical field values ​​that follow the physical equations. The mean square error (MSE) between the predicted and the physical field values ​​is then calculated and used as the physical regularization loss term. Introducing the physical regularization loss term ensures that the forward substitution model satisfies basic physical conservation laws, thus improving its interpretability and generalization ability. Finally, real observation data is obtained, and the MSE between the network's predicted and actual observation values ​​is calculated. This MSE is used as the data fitting loss term. The joint loss function is obtained by weighting the physical regularization loss term and the data fitting loss term. The mathematical expression of the joint loss function is: Total Loss = α × Physical Regularization Loss + β × Data Fitting Loss (where α and β are preset coefficients, α + β = 1). Finally, different three-dimensional physical property parameter models (i.e., geological structure parameter sets) can be generated using geostatistical techniques. A high-fidelity numerical simulator is then used to calculate the corresponding physical field response for each parameter model as the true value. An optimizer (such as Adam) and a joint loss function are then used to train the graph neural network architecture. By adjusting the network weights to minimize the loss, the final network weights are saved when the training loss value approaches its minimum. This neural network model with the final network weights can then be used as a forward modeling substitute. It should be noted that geological bodies are essentially non-Euclidean spatial data, and the interactions between their units (such as faults and strata) are determined by spatial topology. In this application, a graph neural network is used to characterize the interactions between these units (such as faults and strata).

[0028] It should be noted that the graph neural network architecture in this application refers to a deep learning model that abstracts the grid structure of a geological model into graph data and is specifically used to learn the physical field response on this grid structure. It can be used to establish an efficient, nonlinear mapping function from complex geological structures to complex physical fields. The physical regularization loss term refers to the error term generated by comparing the neural network prediction with the solution results of physical equations. It can embed physical laws as strong constraints into the learning process, which is beneficial to improving the model's generalization ability in areas with scarce data. The data fitting loss term refers to the error term between the neural network prediction and the actual observed data. It can drive the network to learn to reproduce the observed phenomena. The joint loss function is a composite optimization objective that integrates the prior knowledge of physical laws and the facts of data. It is the command feature for training the graph neural network architecture. The forward substitution model is a data-driven model built through physical constraint training, numerical simulation, and differentiability, used to perform calculations from geological parameters to physical fields.

[0029] In step S2, geological exploration data is acquired, and a multi-source joint loss function is constructed based on the forward modeling output of the forward modeling substitution model and the geological exploration data.

[0030] It should be noted that in this embodiment, geological exploration data is acquired through a ground-based three-dimensional laser scanning device. Specifically, multiple scanning stations are deployed around the area to be explored (such as landslides, open-pit mine slopes, and tunnel faces). By setting up the scanning device on the station and starting the device, the laser transmitter emits laser pulses towards the target area, and the receiver records the return time and intensity of the reflected pulses. The point cloud dataset containing spatial point coordinates and reflection intensity information can then be used as geological exploration data. In addition, it should be noted that the ground-based three-dimensional laser scanning device in this application refers to a non-contact active detection equipment that integrates laser ranging, high-speed scanning, and high-precision positioning and attitude determination technologies. It is used to quickly and accurately acquire high-density, high-precision three-dimensional geometric morphology information (i.e., point cloud) of the surface of targets such as geological outcrops and engineering rock masses.

[0031] Preferably, in this embodiment, reference Figure 2 As shown, this figure is an exemplary flowchart for determining the multi-source joint loss function according to the present application. In this embodiment, the construction of the multi-source joint loss function based on the forward modeling output of the forward substitution model and the geological exploration data can be achieved by the following steps: In step S21, the spatial distribution loss term is determined based on the different data types of the geological exploration data; In step S22, loss matching is performed on the forward modeling output of the forward substitution model and the spatial distribution loss term to obtain the multi-source loss value; In step S23, a multi-source joint loss function is constructed based on the multi-source loss value and the physical constraint penalty term of the forward substitution model.

[0032] In practice, firstly, borehole data from the geological exploration data is read, and the coordinates of each borehole are spatially matched with the grid nodes of the prior 3D geological structure model to find the nearest nodes. This allows the discrete point data to be associated with the continuous field model, which is beneficial for the spatial registration and fusion of multi-source heterogeneous data. The square of the difference between the measured physical property values ​​of the borehole and the current predicted values ​​of the matched nodes is calculated, and then the average of the squared differences of all boreholes is calculated. This average is used as the spatial distribution loss term based on the borehole data, which helps to improve the reliability of the inversion results at key locations. In addition, for the geophysical profile data from the geological exploration data, the coordinates of each observation point are matched with the grid of the forward substitution model, and the forward substitution model is called to calculate the forward prediction value of the matched points. The mean square error between the observed value and the forward prediction value is then used as the spatial distribution loss term based on the geophysical data, which helps to enhance the characteristics of the geophysical data, which has a wide coverage and is sensitive to changes in physical properties. Then, weighting coefficients can be preset for the spatial distribution loss items corresponding to different types of data (e.g., borehole data weight 0.7, geophysical data weight 0.3), and calculated according to the weighting formula, that is: total weighted loss = weight 1 × borehole data spatial distribution loss item + weight 2 × geophysical data spatial distribution loss item. Here, intelligent weighting based on the physical meaning and confidence level of different data sources can reduce the imbalance of the contribution of multi-source data. The total weighted loss value calculated above is then used as the multi-source loss value. This multi-source loss value can quantify the degree of mismatch between the prediction results of the forward substitution model under the current geological model parameters and all exploration data as a whole. Finally, a physical constraint penalty term is defined to introduce prior knowledge. For example, the sum of the squares of the differences between the parameter values ​​of all adjacent grid nodes in the forward model is used as the physical constraint penalty term. This helps to suppress false anomalies caused by data noise or ill-posedness, thereby enhancing the stability and geological rationality of the solution. A regularization coefficient λ (e.g., λ=0.01) is then set to control the strength of the smoothing constraint. The multi-source loss value is then added to the weighted physical constraint penalty term, i.e., final loss = multi-source loss value + λ × physical constraint penalty term. This mathematical expression can then be used as the multi-source joint loss function, which unifies data-driven fitting and physical constraints within an optimizable framework. This allows the multi-source joint loss function to both drive the model to match the observation data and ensure that the inversion process always evolves in a geologically credible and physically reasonable direction through the physical constraint penalty term, thereby improving the interpretability and generalization ability of the inversion results.

[0033] It should be noted that the spatial distribution loss term in this application is an indicator that quantifies the difference between observed values ​​and model predictions at corresponding spatial locations. This loss term acts on the data source, forcing the inversion process to prioritize fitting the local features revealed by this type of data. The multi-source loss value is a scalar value obtained by weighting and combining different individual losses. It describes the overall fitting deviation between the forward substitution model and all available data, and is the core hub for achieving multi-source information fusion and collaborative constraints. The physical constraint penalty term is a mathematical term introduced in the inversion objective to characterize prior physical laws or geological knowledge (such as parameter smoothness and range constraints). It prevents the solution process from producing geologically discontinuous and physically unreasonable results due to simply fitting the data. The multi-source joint loss function is an overall optimization objective that integrates the data-driven fitting objective (i.e., the multi-source loss value) and the knowledge-driven constraint objective (i.e., the physical constraint penalty term).

[0034] In step S3, a hierarchical recursive inversion optimization algorithm is used to perform layer-by-layer reasoning and reconstruction of the geological point cloud parameters in the prior three-dimensional geological structure model based on the multi-source joint loss function, thereby obtaining a dynamic three-dimensional geological parameter set.

[0035] Preferably, in this embodiment, reference Figure 3 As shown, this figure is an exemplary flowchart for determining a dynamic three-dimensional geological parameter set according to the present application. In this embodiment, the hierarchical recursive inversion optimization algorithm is used to perform layer-by-layer reasoning and reconstruction of the geological point cloud parameters in the prior three-dimensional geological structure model based on the multi-source joint loss function to obtain the dynamic three-dimensional geological parameter set. This can be achieved through the following steps: In step S31, the geological point cloud parameters in the prior three-dimensional geological structure model are used as the parameter set to be optimized; In step S32, with the multi-source joint loss function as the target, the set of parameters to be optimized is inverted and iterated and the gradient is updated through the ensemble smoother algorithm to obtain the optimized geological point cloud parameters. In step S33, the optimized geological point cloud parameters are used as spatial constraints for layer-by-layer inversion to obtain a dynamic three-dimensional geological parameter set.

[0036] In practice, firstly, the geophysical attribute values ​​(such as density) stored at each node of the prior 3D geological structure model are extracted, and then a one-dimensional list containing all node attribute values ​​is used as the parameter set to be optimized. Then, in an automatic differentiation framework (such as PyTorch), a calculation function is written to assign the values ​​of the parameter set to be optimized back to the corresponding nodes of the 3D geological model. The forward substitution model and the multi-source joint loss function are then called to calculate the total loss value. An optimizer (such as the Adam optimizer) is then initialized. By introducing Gaussian filtering, the stability and convergence speed of the inversion process can be improved. The parameter set to be optimized is passed to the optimizer for iterative updates, and a Gaussian filter is applied to the updated parameter values ​​to apply a smoothing constraint. When the iteration reaches loss convergence, the parameter set in the optimizer at this time is used as the optimized geological point cloud parameters. By utilizing the differentiability of the forward substitution model, a complete and automatically computed gradient graph can be established between the geological parameters and the multi-source joint loss function, which can use gradient information to guide the parameter update direction. Finally, the mesh of the prior three-dimensional geological structure model can be coarsened to generate a coarse mesh model. The above inversion process can be repeated on the coarse mesh model to obtain the parameter distribution at the coarse mesh scale. Then, a three-dimensional linear interpolation algorithm is used to interpolate the parameter distribution at the coarse mesh scale onto all nodes of the original fine mesh. The parameter set of the prior three-dimensional geological structure model at this time can be used as the dynamic three-dimensional geological parameter set.

[0037] It should be noted that the parameter set to be optimized in this application is a mathematical object directly manipulated and updated by the inversion algorithm; the optimized geological point cloud parameters are the output of the single-scale inversion iteration; the ensemble smoother algorithm refers to the spatial filtering operation introduced after gradient update, which can serve as a built-in stabilizer for inversion iteration, suppressing non-physical high-frequency oscillations that may occur during parameter update, and ensuring that each iteration moves closer to a geologically more reasonable smooth solution; spatial constraints refer to the guidance provided by the hierarchical inversion strategy for fine-scale inversion, which can be achieved through the information transmission channel between multi-scale inversions; the dynamic three-dimensional geological parameter set includes multi-source data information and prior physical constraints, has spatial consistency, and is the core data foundation for subsequent visualization.

[0038] In step S4, the dynamic three-dimensional geological parameter set is fused, rendered, and interactively analyzed to obtain a visualization result that integrates geological structure and multi-field evolution information. Based on the dynamic three-dimensional geological parameter set and the forward modeling substitution model, a scenario inference component is constructed.

[0039] In this embodiment, the fusion rendering and interactive analysis of the dynamic three-dimensional geological parameter set to obtain a visualization result that integrates geological structure and multi-field evolution information can be performed in the following manner: The dynamic three-dimensional geological parameter set is visualized and mapped, and then the geological structure geometry, physical field attributes and deterministic information in the dynamic three-dimensional geological parameter set are converted into geometric primitives and color mapping textures, respectively. The geometric primitives and the color mapping texture are fused and drawn to obtain an initial three-dimensional scene; In the initial 3D scene, parameterized analysis instructions are generated in response to user operations; The initial 3D scene and the parameterized analysis instructions are encapsulated to obtain a visualization result that integrates geological structure and multi-field evolution information.

[0040] In practical implementation, firstly, for geological structure geometry, a 3D visualization library is used. The node coordinates and grid cell connections of the model are taken as input, and the resulting unstructured grid data objects can be used as geometric primitives representing geological structures, enabling precise conversion from numerical results to 3D geometric forms. For physical field attributes, a color mapping table is used. The physical field values ​​of all nodes are first normalized to between 0 and 1. Then, based on the normalized value of each node, the corresponding color component is looked up from the color mapping table. This entire set of color data associated with the geometric primitives is used as a color mapping texture. Visual encoding is performed through intuitive color gradients, allowing users to intuitively perceive complex and multidimensional field information, which helps enhance the understandability of the data. For deterministic information, the transparency of the geometric primitives can be adjusted to facilitate the simultaneous display of core data and metadata, which helps to quickly focus on high-confidence areas. Then, using the plotter in the 3D visualization library, unstructured mesh objects are added to the plotter. The plotter's add mesh function is called, and the rendering mode is specified as "vertex color shading." Simultaneously, auxiliary display elements such as 3D coordinate axes and color legends are added to obtain the initial 3D scene. This fully utilizes the parallel computing power of the GPU to achieve high frame rate, real-time rendering of massive geological data, providing performance assurance for the dynamic visualization of large-scale, high-precision geological models. Finally, in the initial 3D scene, an event listener function is set for the interactive window. For example, when the mouse clicks on a location on the 3D model, the listener function is triggered. This function obtains the 3D coordinates of the clicked point, then uses these coordinates to look up the dynamic 3D geological parameter set, obtaining all attribute values ​​for those coordinates. Input commands are then allowed in the initial 3D scene, such as "Analyze the trend of physical property changes from point A to point B." The data object combining the coordinates and text commands is then used as a parameterized analysis command. The composite object obtained by combining the initial 3D scene and the currently generated parameterized analysis command serves as the visualization result that integrates geological structure and multi-field evolution information.

[0041] It should be noted that, in this application, visualization mapping refers to the process of converting numerical geological model data into renderable basic graphic elements and visual attributes in computer graphics; geometric primitives specifically refer to the collection of basic graphic elements such as points, lines, and faces used to construct the visual morphology of three-dimensional geological bodies; color-mapped textures refer to data that, based on physical field values, can assign a specific color to each vertex or face of a geometric primitive using a color lookup table; the initial three-dimensional scene is a three-dimensional visualization that integrates geometric and color information; parametric analysis instructions refer to converting user interactions in the three-dimensional scene into computational instructions with clearly defined operation types and spatial parameters. Furthermore, in this application, the visualization result integrating geological structure and multi-field evolution information is a composite data package that integrates static visualization scenes and dynamic user interaction instructions. It can display the current state of the underground structure and provide a directly operable interactive interface for subsequent scenario simulation and early warning decision-making.

[0042] In this embodiment, the scenario inference component constructed based on the dynamic three-dimensional geological parameter set and the forward modeling substitution model can be implemented in the following manner: The initial state of the model is determined by the dynamic three-dimensional geological parameter set, and a forward evolution operator is established based on the forward substitution model. An inference engine for constructing scenario inference components is built based on the initial state of the model and the forward evolution operator; A scenario-driven interface is constructed, and then a scenario inference component is constructed through the inference engine and the scenario-driven interface.

[0043] In specific implementation, firstly, the parameter values ​​in the 3D geological parameter set are arranged into a one-dimensional array according to node order. This one-dimensional array is used as the initial state of the scenario simulation model. A forward evolution operator is designed based on the forward substitution model: a function is initialized, whose input is the current state array and output is the next state array. Inside the function, the forward substitution model is called to calculate the change in each parameter value, which is then added to the current state to obtain the state for the next time step. This function is then used as the forward evolution operator. Next, the core loop program of the simulation engine is written, where the input is the initial state and the forward evolution operator. A total number of simulation time steps is set within the program. This loop can then be executed again. The operation involves passing the current state to the forward evolution operator, which returns the state for the next time step. The new state is then saved, and the new state is used as the current state to continue the next loop. This allows the program to function as the inference engine for a scenario inference component. It can simulate the complete dynamic process of a geological model evolving from its current state to the future under the drive of physical laws, and fully record the historical trajectory. Finally, the scenario-driven interface is a module that receives external commands and configures inference parameters. This interface can pass the updated parameters and the received start command to the inference engine. The inference engine uses the updated parameters to reinitialize the operators and then starts the inference calculation. Thus, a program module containing the inference engine and the scenario-driven interface can be used as a scenario inference component.

[0044] It should be noted that the forward evolution operator in this application is a mathematical function or calculation process that encapsulates specific physical or engineering evolution laws. It represents the calculation of the state of the geological model at the current moment to determine the state at the next moment, and is the core calculation rule for inference and prediction. The inference engine is the control program responsible for executing iterative calculations, used to simulate the process of the geological model gradually changing over time. The scenario-driven interface is the interaction channel between the user or upper-level system and the inference core, used to drive the generation of inference results under different conditions. The scenario inference component is a complete functional unit integrated by the inference engine and the scenario-driven interface. It can quickly simulate and predict the future dynamic changes of underground structures and physical fields based on current geological knowledge and under different assumptions, providing analytical capabilities for early warning decision-making.

[0045] In step S5, dynamic early warning decisions are made based on the visualization results and the scenario simulation component for geological exploration data.

[0046] In this embodiment, the dynamic early warning decision-making based on the visualization results and the scenario inference component for geological exploration data can be carried out in the following manner: The analysis instructions in the visualization results are parsed to generate early warning analysis tasks; Based on the aforementioned early warning analysis task, the scenario simulation component is invoked to calculate the probability of exceeding the limits and the rate of evolution of key early warning indicators; A comprehensive risk assessment is conducted on the probability of exceeding the limit and the rate of evolution based on preset risk level classification rules to obtain an early warning decision report.

[0047] In practical implementation, firstly, the encapsulated parametric analysis instructions are extracted from the visualization results. These instructions are structured data, including, for example, operation type and target area coordinates. The parametric analysis instructions include fields such as task type, target area parameters, and a list of indicators to be calculated for the early warning analysis task. Then, multiple possible scenario parameters are input into the scenario inference component through the scenario-driven interface. For each set of scenario parameters, the inference engine is invoked to infer a period of time in the future, and the sequence value of the indicator over time is calculated according to the indicator's definition formula. The proportion of scenarios in all scenario inference results where the indicator value exceeds a preset safety threshold is statistically analyzed. For each indicator... The indicator can be defined as the probability of exceeding the limit. In all scenarios where indicators exceed the limit, the average time from the start of the simulation to the first time the indicator exceeds the threshold is calculated. Then, 1 divided by this average time is used as the evolution rate of the indicator. Finally, for each key early warning indicator calculated in the early warning analysis task, its calculated probability of exceeding the limit and evolution rate are substituted into the rules for judgment to determine its individual risk level. The highest risk level among all indicators is selected as the comprehensive risk level of the target area, following the principle of choosing the highest. A text report is then generated based on the pre-set risk level classification rules according to actual needs. This text report can then serve as the early warning decision report.

[0048] It should be noted that the early warning analysis task in this application refers to the specific set of instructions that concretize the analytical intent generated by the user through visual interaction into a series of calculable and executable quantitative analysis objectives; key early warning indicators refer to core physical or engineering parameters used to quantitatively characterize the stability state of geological bodies, engineering risks, or environmental effects; the probability of exceeding limits refers to the likelihood that the value of a key early warning indicator will exceed its safety threshold under various uncertain future scenarios, and can be used to quantify the probability of risk occurrence; the rate of evolution refers to the average speed at which a key early warning indicator deteriorates from its current state to exceeding the threshold under risk scenarios, and can be used to quantify the urgency of risk development; the risk level classification rule is the judgment logic or standard that maps quantitative indicators such as the probability of exceeding limits and the rate of evolution to qualitative risk levels, and is the basis for realizing the transformation from data to decision-making, which can be preset according to actual risk early warning needs.

[0049] In summary, the technical solution adopted in this application can achieve high-precision geological situation perception and early warning by deeply integrating the forward model constrained by physical information with the hierarchical dynamic inversion algorithm.

[0050] Example 2: This application provides a dynamic processing system for visualizing geological exploration data, referencing... Figure 4As shown, this figure is a module structure diagram of a dynamic processing system for visualizing geological exploration data provided in this application. The dynamic processing system includes: The model building module 100 is used to build a priori three-dimensional geological structure model based on a geological knowledge rule base and planar geological maps. The forward substitution model is obtained by training a graph neural network architecture constrained by physical information based on multi-physics field control equations. The data fusion module 200 is used to acquire geological exploration data and construct a multi-source joint loss function based on the forward modeling output of the forward substitution model and the geological exploration data. The recursive inversion module 300 is used to perform layer-by-layer reasoning and reconstruction of the geological point cloud parameters in the prior three-dimensional geological structure model based on the multi-source joint loss function using a hierarchical recursive inversion optimization algorithm to obtain a dynamic three-dimensional geological parameter set. The visualization and simulation module 400 is used to perform fusion rendering and interactive analysis on the dynamic three-dimensional geological parameter set to obtain a visualization result that integrates geological structure and multi-field evolution information, and to construct a scenario simulation component based on the dynamic three-dimensional geological parameter set and the forward modeling substitution model. The early warning decision module 500 is used to make dynamic early warning decisions based on the visualization results and the scenario simulation component for geological exploration data.

[0051] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0052] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compactdisc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0053] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

Claims

1. A method for dynamic visualization and processing of geological exploration data, characterized in that, The dynamic processing method includes the following steps: A priori three-dimensional geological structure model is constructed based on a geological knowledge rule base and planar geological maps. A forward substitution model is obtained by training a graph neural network architecture constrained by physical information based on multi-physics field control equations. Obtain geological exploration data, and construct a multi-source joint loss function based on the forward modeling output of the forward substitution model and the geological exploration data; The hierarchical recursive inversion optimization algorithm is used to reconstruct the geological point cloud parameters in the prior three-dimensional geological structure model layer by layer based on the multi-source joint loss function to obtain a dynamic three-dimensional geological parameter set. The dynamic three-dimensional geological parameter set is fused, rendered, and interactively analyzed to obtain a visualization result that integrates geological structure and multi-field evolution information. A scenario inference component is constructed based on the dynamic three-dimensional geological parameter set and the forward modeling substitution model. Dynamic early warning decisions are made based on the visualization results and the scenario simulation components for geological exploration data.

2. The method for dynamic visualization processing of geological exploration data as described in claim 1, characterized in that, The construction of a priori three-dimensional geological structure models based on geological knowledge rule bases and planar geological maps specifically includes: Geological element analysis is performed on the aforementioned planar geological map to obtain an element feature set; A set of three-dimensional control points is determined based on the geological knowledge rule base and the feature set of the elements. By spatially mapping the three-dimensional control point set through the spatial topological relationships between geological features, a priori three-dimensional geological structure model is obtained.

3. The method for dynamic visualization processing of geological exploration data as described in claim 1, characterized in that, The forward substitution model obtained by training a graph neural network architecture constrained by physical information based on multiphysics control equations specifically includes: A graph neural network architecture is constructed based on the connection relationships of geological structure nodes; The physical regularization loss term is determined based on the multiphysics control equations, and then the physical regularization loss term and the data fitting loss term are combined to obtain the joint loss function. The graph neural network architecture is optimized and trained using geological structure parameters based on the joint loss function to obtain a forward substitution model.

4. The method for dynamic visualization and processing of geological exploration data as described in claim 1, characterized in that, Geological survey data is obtained using ground-based three-dimensional laser scanning equipment.

5. The method for dynamic visualization processing of geological exploration data as described in claim 1, characterized in that, The construction of a multi-source joint loss function based on the forward modeling output and the geological exploration data specifically includes: The spatial distribution loss term is determined by the different data types of the geological exploration data; The forward modeling output of the forward substitution model and the spatial distribution loss term are matched to obtain the multi-source loss value; A multi-source joint loss function is constructed based on the multi-source loss value and the physical constraint penalty term of the forward substitution model.

6. The method for dynamic visualization and processing of geological exploration data as described in claim 1, characterized in that, The hierarchical recursive inversion optimization algorithm is used to reconstruct the geological point cloud parameters in the prior 3D geological structure model layer by layer based on the multi-source joint loss function, resulting in a dynamic 3D geological parameter set, which specifically includes: The geological point cloud parameters in the prior three-dimensional geological structure model are used as the parameter set to be optimized. Using the multi-source joint loss function as the objective, the optimized geological point cloud parameters are obtained by inverting and iterating the set of parameters to be optimized and updating the gradient through the ensemble smoother algorithm. The optimized geological point cloud parameters are used as spatial constraints for layer-by-layer inversion to obtain a dynamic three-dimensional geological parameter set.

7. The method for dynamic visualization processing of geological exploration data as described in claim 1, characterized in that, The dynamic three-dimensional geological parameter set is fused, rendered, and interactively analyzed to obtain a visualization result that integrates geological structure and multi-field evolution information. Specifically, this includes: The dynamic three-dimensional geological parameter set is visualized and mapped, and then the geological structure geometry, physical field attributes and deterministic information in the dynamic three-dimensional geological parameter set are converted into geometric primitives and color mapping textures, respectively. The geometric primitives and the color mapping texture are fused and drawn to obtain an initial three-dimensional scene; In the initial 3D scene, parameterized analysis instructions are generated in response to user operations; The initial 3D scene and the parameterized analysis instructions are encapsulated to obtain a visualization result that integrates geological structure and multi-field evolution information.

8. The method for dynamic visualization processing of geological exploration data as described in claim 1, characterized in that, The scenario inference component constructed based on the dynamic three-dimensional geological parameter set and the forward modeling substitution model specifically includes: The initial state of the model is determined by the dynamic three-dimensional geological parameter set, and a forward evolution operator is established based on the forward substitution model. An inference engine for constructing scenario inference components is built based on the initial state of the model and the forward evolution operator; A scenario-driven interface is constructed, and then a scenario inference component is constructed through the inference engine and the scenario-driven interface.

9. The method for dynamic visualization processing of geological exploration data as described in claim 1, characterized in that, Dynamic early warning decision-making based on the visualization results and the scenario simulation component specifically includes: The analysis instructions in the visualization results are parsed to generate early warning analysis tasks; Based on the aforementioned early warning analysis task, the scenario simulation component is invoked to calculate the probability of exceeding the limits and the rate of evolution of key early warning indicators; A comprehensive risk assessment is conducted on the probability of exceeding the limit and the rate of evolution based on preset risk level classification rules to obtain an early warning decision report.

10. A dynamic processing system for visualizing geological exploration data, used to execute the dynamic processing method for visualizing geological exploration data as described in any one of claims 1 to 9, characterized in that, The dynamic processing system includes: The model building module is used to construct a priori three-dimensional geological structure model based on a geological knowledge rule base and planar geological maps. The forward substitution model is obtained by training a graph neural network architecture constrained by physical information based on multi-physics field control equations. The data fusion module is used to acquire geological exploration data and construct a multi-source joint loss function based on the forward modeling output of the forward substitution model and the geological exploration data. The recursive inversion module is used to use a hierarchical recursive inversion optimization algorithm to perform layer-by-layer reasoning and reconstruction of the geological point cloud parameters in the prior three-dimensional geological structure model based on the multi-source joint loss function, so as to obtain a dynamic three-dimensional geological parameter set. The visualization and simulation module is used to perform fusion rendering and interactive analysis on the dynamic three-dimensional geological parameter set to obtain a visualization result that integrates geological structure and multi-field evolution information. Based on the dynamic three-dimensional geological parameter set and the forward modeling substitution model, a scenario simulation component is constructed. The early warning decision module is used to make dynamic early warning decisions based on the visualization results and the scenario simulation component.

Images