Three-dimensional visualization and numerical simulation coordination method for heavy metal migration in soil

Abstract

This invention provides a collaborative method for three-dimensional visualization and numerical simulation of heavy metal migration in soil, relating to the field of computer data processing technology. The method acquires the original borehole concentration dataset of a contaminated site, imports it into EVS to construct a three-dimensional concentration and geological model; extracts the model mesh and interface data, performs format conversion and spatial mapping, and imports it into a three-dimensional porous media numerical simulation engine; configures parameters to solve the transformation equations for water movement and solute convection dispersion migration, outputting a time-series dataset at multiple time nodes; eliminates spatial location bias through an affine transformation matrix and uses linear time interpolation to obtain interpolated concentration values ​​for spatiotemporal consistency calibration of heterogeneous data; imports the calibration data back into EVS to drive the rendering of a dynamic pollution concentration time-series model; and runs an automated script to extract spatial distribution rendering data at fixed time steps to generate a dynamic time-series simulation video file. This invention achieves dynamic visualization and simulation of soil heavy metal pollution diffusion under spatiotemporal consistency conditions for heterogeneous data.

Description

Technical Field

[0001] This invention relates to the field of computer data processing technology, and in particular to a collaborative method for three-dimensional visualization and numerical simulation of heavy metal migration in soil. Background Technology

[0002] Traditional site surveys and soil heavy metal migration and diffusion models often employ static models, which are insufficient to effectively reflect the spatiotemporal evolution of pollution. Environmental management and remediation decisions require accurate understanding of the dynamic migration characteristics of pollutants in complex strata. EVS software possesses 3D geological modeling, spatial analysis, and visualization capabilities. Simultaneously, the 3D porous media numerical simulation engine, as a tool for simulating water and solute migration in porous media, can characterize water movement and solute migration processes based on relevant equations. Both tools play crucial roles in their respective fields.

[0003] To gain a deeper understanding of the dynamic migration patterns of pollutants and to provide a more intuitive visualization, the industry has attempted to combine 3D visualization modeling software with numerical simulation software. However, due to the differences in the underlying architecture and data structures of these two systems, challenges arise in the actual collaborative processing process, including complex heterogeneous data conversion and inaccurate spatiotemporal matching.

[0004] Such cross-software collaborative operations typically involve cumbersome steps such as grid node mapping and time step calibration. Existing workflows have limitations in terms of coherence and automated data processing, making it difficult to efficiently achieve seamless integration of time-series numerical simulations of soil heavy metal pollution diffusion with high-fidelity dynamic 3D visualization and extrapolation. Summary of the Invention

[0005] This invention provides a collaborative method for three-dimensional visualization and numerical simulation of heavy metal migration in soil, aiming to solve the technical problems of inaccurate spatiotemporal matching of heterogeneous data and limited coherence of dynamic inference in the collaborative process of three-dimensional visualization and numerical simulation software in the prior art.

[0006] This invention provides a collaborative method for three-dimensional visualization and numerical simulation of heavy metal migration in soil, the method comprising the following steps: Step S1: Obtain the original borehole concentration dataset of the contaminated site, which includes soil heavy metal concentration values ​​at different depths and their corresponding spatial coordinates. Step S2: Import the original borehole concentration dataset into the environmental visualization system EVS to generate the EVS target coordinate system and construct a three-dimensional heavy metal concentration distribution model. Combine the site stratigraphic information to construct a three-dimensional geological model containing stratigraphic physical properties and interface spatial coordinates. Step S3: Extract the computational grid node data and stratigraphic interface data from the three-dimensional heavy metal concentration distribution model and the three-dimensional geological model, perform format conversion and spatial mapping, and import them as initial concentration boundary conditions and stratigraphic files into the three-dimensional porous media numerical simulation engine. Step S4: Configure model parameters and scene boundary conditions in the three-dimensional porous media numerical simulation engine, solve the equations for water movement and solute convection-dispersion migration and transformation, and output a time-coordinate-concentration time series dataset file based on the spatial coordinate system of the numerical simulation output at multiple time nodes. Step S5: Perform spatiotemporal consistency calibration on heterogeneous data: extract the geometric origin coordinates of the spatial coordinate system output by numerical simulation and the EVS target coordinate system to construct a three-dimensional affine transformation matrix to eliminate spatial position deviation; and use linear time interpolation to obtain the interpolated concentration values ​​between the discrete time step nodes output by adjacent simulation solvers to fill the data gaps caused by time step differences, and generate a core control file that spatiotemporally maps and associates the three-dimensional mesh boundary, stratigraphic features and the interpolated concentration values. Step S6, drive and render the dynamic pollution concentration time series model: import the calibrated time series data back to EVS to drive model construction; read the planar two-dimensional red line data of the target polluted plot, calculate the shortest spatial distance from the three-dimensional pollution feather mesh node to the polygon red line boundary, perform data mask operation for mesh nodes that exceed the red line range, and forcibly invalidate the color channel and alpha opacity channel in the rendering pipeline of the dynamic pollution concentration time series model to stop the three-dimensional rendering of the node; Step S7, Automated simulation video generation: Run an automated script to extract spatial distribution rendering data and generate static images at fixed time steps; create a high-dimensional tensor containing horizontal and vertical pixels, color channels and total frame count dimensions as a three-dimensional matrix array in system memory; write the RGB pixel value array of discrete static images into the address of consecutive frames in timestamp order; encapsulate and output a dynamic temporal simulation video file, thereby realizing dynamic visualization simulation of soil heavy metal pollution diffusion under the condition of spatiotemporal consistency of heterogeneous data.

[0007] Optionally, in step S2, the model construction process specifically includes: Outliers were removed from the imported original borehole concentration dataset using the 3σ criterion, and missing values ​​were filled in using Kriging interpolation. The three-dimensional heavy metal concentration distribution model was generated by calling the three-dimensional interpolation module. The stratigraphic modeling method was used, and the continuous elevation surfaces at the top and bottom of each generalized stratigraphic unit were fitted using the Kriging interpolation algorithm. The three-dimensional geological model was established in a closed three-dimensional space using a hexahedral mesh division rule.

[0008] Optionally, in step S3, the specific operations of format conversion and space mapping are as follows: The spatial coordinates and concentration values ​​of the exported nodes are generated into an ASCII-encoded comma-separated text document as the initial concentration boundary conditions. The hexahedral mesh data output by EVS is reconstructed into a Delaunay tetrahedral finite element mesh, and the coordinates of the EVS mesh nodes are mapped one-to-one with the tetrahedral finite element nodes constructed by the calculation engine through the nearest neighbor search algorithm.

[0009] Optionally, in step S4, when configuring model parameters, a feedforward artificial neural network topology is used to map the soil particle size distribution data and bulk density obtained from the field survey into hydraulic parameters: Its input layer receives scalar inputs of the mass percentages of sand, silt, and clay particles, as well as soil bulk density. The hidden layer uses the Sigmoid function as the activation function, and the output layer uses a linear activation function to output the predicted values ​​of effective porosity, saturated permeability coefficient, and residual water content.

[0010] Optionally, in step S4, the equations for water movement and solute convection-dispersion migration are solved using adaptive time step control logic based on the Courant number. Set the initial time step and the maximum time step. Calculate the global maximum Courant number after each iteration. If the global maximum Courant number is greater than 1.0, the next iteration step size is automatically halved. If the global maximum Courant number is less than 0.5 for three consecutive iterations, the next iteration step size is multiplied by a magnification factor of 1.2.

[0011] Optionally, in step S4, before solving the equations for water movement and solute convection-dispersion migration and transformation, a comparative simulation scenario is also constructed: Intervention scenarios including electroremediation, chemical leaching, or phytoremediation were established, and root extraction rate equations with ion electromobility, leaching pseudo-first-order kinetic reaction rate constant, or exponential decay with soil depth were configured respectively; at the same time, a natural decay scenario considering only the natural migration law of heavy metals was established.

[0012] Optionally, in step S5, the dataset file also includes a normalization process: missing node values ​​are marked as non-numeric string placeholders; for values ​​below the conventional detection limit, the exact value is assigned as half of the method detection limit value of the detection instrument in this batch of experiments.

[0013] Optionally, in step S5, the specific calculation logic for obtaining the interpolated concentration value using linear time interpolation is as follows: For the concentration value C1 at the start time T1 and the concentration value C2 at the end time T2, calculate the time difference between the target time T and T1 and divide it by the total time difference between T2 and T1 to obtain a dimensionless scaling factor; multiply this scaling factor by the concentration difference between C2 and C1 and add it to the base concentration value C1 to obtain the interpolated concentration value of the grid node at the target time T.

[0014] Optionally, in step S6, the calculation of the shortest spatial distance and the data masking operation specifically involve: The two-dimensional Delaunay triangulation algorithm is used to transform the two-dimensional red line into triangular facet geometry. The mesh nodes of the three-dimensional pollution plume are projected onto a horizontal two-dimensional plane. The Euclidean distance from each projected node to each boundary line segment of the two-dimensional red line polygon is calculated and the minimum value is extracted as the shortest spatial distance. When the shortest spatial distance is greater than 0 meters, its RGBA sequence is assigned to be completely transparent (0,0,0,0).

[0015] Optionally, in step S7, the generation of the automated simulation video specifically involves: The video writing module of the computer vision processing library is invoked. The width dimension of the three-dimensional matrix array composed of the high-dimensional tensor is fixed at 1920 pixels, and the height dimension is fixed at 1080 pixels. After traversing and writing the static image file, the video writing module releases the video stream pointer at a set frame rate of 25 frames per second, performs stream data encapsulation, and outputs a video file in the standard H.264 encoding format.

[0016] The present invention has achieved the following beneficial effects: This invention extracts computational grid node data and stratigraphic interface data from a 3D geological and concentration model, performs format conversion and spatial mapping, and then imports the data into numerical simulation software to obtain a time-series dataset file. A 3D affine transformation matrix is ​​used to eliminate spatial location biases, and linear time interpolation is employed to fill data gaps caused by differences in time steps. This spatiotemporal consistency calibration mechanism for heterogeneous data overcomes data matching problems during the collaboration of different software. The calibrated coherent time-series data is imported back into a visualization system for data masking operations, and automated scripts and tensor matrix encapsulation technology are used to write continuous frame addresses for video output. This method enables dynamic time-series extrapolation under spatiotemporal consistency conditions for heterogeneous data, improving the visualization accuracy and automation level of environmental remediation decision-making.

[0017] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 A flowchart illustrating a collaborative method for three-dimensional visualization and numerical simulation of heavy metal migration in soil, provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the data flow and pipeline topology within the EVS system for the construction of the basic three-dimensional spatial model and the generation of the core control file in this embodiment of the invention. Figure 3 This is a schematic diagram of the TimeSettings parameter setting interface for the core control file (TCF file) generated internally by the 3destimation module of the EVS system in this embodiment of the invention. Figure 4 This is a schematic diagram of the complete pipeline topology within the EVS visualization system for the dynamic pollution concentration time series model in this embodiment of the invention. Figure 5 This is a three-dimensional spatial distribution diagram of the As heavy metal concentration at a certain site at a certain estimated time (18:00 on June 20, 2024), output by the method of this embodiment of the invention. Figure 6 This is a three-dimensional spatial distribution diagram of the As heavy metal concentration at a certain site at another estimated time (00:00 on May 20, 2024), output by the method of the embodiment of the present invention. Detailed Implementation

[0020] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0021] Traditional methods for displaying soil heavy metal migration, diffusion, and remediation results often employ static models, which fail to effectively reflect the spatiotemporal evolution of pollution. EVS is a 3D visualization modeling and analysis software with 3D geological modeling, spatial analysis, and visualization capabilities. A 3D porous media numerical simulation engine is a software for simulating water and solute migration in porous media. It describes soil water movement based on the Richards equation and uses convection-dispersion equations to characterize the solute migration process. This invention provides a collaborative method for 3D visualization and numerical simulation of soil heavy metal transport, integrating EVS's 3D visualization with the solute migration numerical simulation of the 3D porous media numerical simulation engine to achieve time-series numerical simulation and dynamic 3D visualization extrapolation of soil heavy metal pollution diffusion.

[0022] This invention discloses a collaborative method for three-dimensional visualization and numerical simulation of heavy metal migration in soil, such as... Figure 1 As shown, it includes the following steps: Step S1: Obtain the original borehole concentration dataset of the contaminated site. The original borehole concentration dataset contains soil heavy metal concentration values ​​at different depths and their corresponding spatial coordinates.

[0023] Specifically, boreholes are laid out within the contaminated site according to a pre-defined sampling grid. For each borehole, soil samples are collected at different depths based on the site's geological strata. Specifically, the pre-defined sampling grid size is determined according to the "Technical Guidelines for Site Environmental Investigation," with the grid side length set to a square unit ranging from 20 to 50 meters. The sampling standards for different depths are as follows: within the range from the surface to 3 meters below ground, one physical sample is collected every 0.5 meters; between 3 meters below ground and the groundwater level, one physical sample is collected every 1.0 meter; and denser sampling must be conducted 0.2 meters above and below the boundary lines crossing different lithological strata. Before detecting heavy metal concentrations, the collected soil samples must undergo pretreatment steps, specifically: the soil samples are air-dried naturally, gravel and plant / animal remains larger than 2 mm are removed, and the samples are ground and passed through a 100-mesh nylon sieve. Then, 0.2000 g of soil sample is weighed and placed in a digestion vessel, and a mixed acid solution consisting of nitric acid, hydrofluoric acid, and hydrogen peroxide is added for microwave digestion. After cooling, acid removal, and volume adjustment, the sample is introduced into the detection instrument. The collected soil samples are sent to the laboratory, where the concentration of the target heavy metals is detected using inductively coupled plasma mass spectrometry or atomic absorption spectrophotometry. The three-dimensional spatial coordinates (including x-axis, y-axis, and depth elevation) of each soil sample are recorded and mapped to their corresponding heavy metal concentration values, forming a raw borehole concentration dataset containing spatial location information and pollution concentration attributes.

[0024] Step S2: Import the original borehole concentration dataset into the Environmental Visualization System (EVS) to generate the EVS target coordinate system and construct a three-dimensional heavy metal concentration distribution model. Combine the site stratigraphic information to construct a three-dimensional geological model that includes stratigraphic physical properties and interface spatial coordinates.

[0025] Specifically, the original borehole concentration dataset is imported into the environmental visualization system EVS. The 3σ criterion is used to identify and remove outliers, and Kriging interpolation is used to complete missing values. Specifically, the 3σ criterion is implemented as follows: for the same heavy metal element, the arithmetic mean and standard deviation of the concentration values ​​of all samples in the original borehole concentration dataset are calculated; sampling points with concentration values ​​greater than the arithmetic mean plus three times the standard deviation, or less than the arithmetic mean minus three times the standard deviation, are identified as outliers and removed from the dataset. The execution steps of the Kriging interpolation method are as follows: First, calculate the experimental semivariogram between sample points and construct a scatter plot of distance versus variance; then, use a spherical model or an exponential model to fit the scatter plot with a theoretical semivariogram function to extract three spatial autocorrelation parameters: nugget value, sill value, and range; subsequently, set the interpolation search neighborhood, specifying a minimum of 4 sample points and a maximum of 12 sample points for each grid node interpolation calculation, and determine the weight coefficients of each sample point based on the fitted semivariogram function; finally, perform a weighted summation of the heavy metal concentrations of the target grid nodes to obtain the completed value. The numerical determination method for the three-dimensional grid resolution is as follows: calculate the maximum length and maximum width of the bounding rectangle of the site's planar polygon, divide them by 100 respectively to obtain the grid node spacing for the X and Y axes; the elevation node spacing for the Z axis is fixed at 0.1 meters. The specific execution logic of the stratigraphic modeling method is as follows: Elevation values ​​of stratigraphic lithological boundary points recorded in the borehole logging data are extracted. Kriging interpolation is then used to fit the top and bottom continuous elevation surfaces of each generalized stratigraphic unit. Subsequently, within the closed three-dimensional space formed by the top and bottom surfaces, a solid geological volume element is generated by filling it with a hexahedral mesh, thereby establishing the geological model. Based on the preprocessed dataset, the three-dimensional interpolation module is called in EVS, the three-dimensional mesh resolution is set, and a three-dimensional heavy metal concentration distribution model of the site is generated through interpolation calculations. In this model, each node of the divided computational mesh contains spatial coordinate information and the corresponding heavy metal concentration value.Meanwhile, based on the stratigraphic lithology distribution, stratigraphic thickness, and stratigraphic interface depth information recorded in the field survey, different stratigraphic units were defined in EVS using the stratigraphic modeling method. For each type of generalized stratigraphic unit, corresponding physical property parameters (including porosity, permeability, and unit weight) were defined and assigned. The acquisition method and specific definition of the above physical property parameters are as follows: Porosity is defined as the ratio of pore volume to total soil volume, and its value is determined by the indoor ring cutter method. The porosity value range for surface soil is set to 0.40 to 0.55, and the porosity value range for sand layer is set to 0. The permeability ranges from 25 to 0.40; the permeability reflects the ability of porous media to allow fluid to pass through, and its value is derived from the observed drawdown data of in-situ pumping tests or micro-water tests. This data is converted into a permeability coefficient and input into the model. The permeability coefficient for silty clay layers is set between 1.0 × 10⁻⁵ m / s and 1.0 × 10⁻³ m / s per day, and the permeability coefficient for sand layers is set between 1.0 and 10.0 m / s per day. The bulk density is defined as the mass of dry soil per unit volume, measured by drying and weighing undisturbed soil samples on-site, and its value range is set between 1.1 and 1.6 g / cm³. A three-dimensional geological model corresponding to the geological conditions of the contaminated site is generated, defining the three-dimensional spatial coordinates of different strata and the spatial distribution range of geological structures.

[0026] Step S3: Extract the computational grid node data and stratigraphic interface data from the three-dimensional heavy metal concentration distribution model and the three-dimensional geological model, perform format conversion and spatial mapping, and import them as initial concentration boundary conditions and stratigraphic files into the three-dimensional porous media numerical simulation engine.

[0027] Specifically, regarding the extraction of concentration model data, based on the three-dimensional heavy metal concentration distribution model constructed in EVS, the attribute data of each computational grid node is extracted by calling the node export module. The computational nodes in the model are traversed, and the spatial three-dimensional coordinates of the nodes and their corresponding heavy metal concentration values ​​are extracted, generating a text document containing coordinates and concentration data. This text document serves as the initial concentration boundary condition for solute transport calculations in the three-dimensional porous media numerical simulation engine. Specifically, the underlying data structure of the text document is limited to an ASCII-encoded comma-separated value file, with its matrix columns strictly arranged in the order of X-space coordinates, Y-space coordinates, Z-space coordinates, and heavy metal concentration values, without a header. The steps for converting the data into a layered geometric file supported by the three-dimensional porous media numerical simulation engine are as follows: reconstructing the hexahedral mesh data output by EVS into a Delaunay tetrahedral finite element mesh; generating a node geometric tensor containing the global coordinates of all finite element nodes, and an element topology matrix describing the connection relationships of the tetrahedral vertices; and mapping the EVS mesh node coordinates to the finite element nodes of the three-dimensional porous media numerical simulation engine through a nearest neighbor search algorithm. The specific logical judgment conditions and tolerance constraints of the nearest neighbor node search algorithm are as follows: For each tetrahedral finite element node generated in the three-dimensional porous media numerical simulation engine model, calculate the Euclidean distance between its three-dimensional spatial coordinates and the coordinates of all imported EVS hexahedral mesh nodes; extract the EVS node corresponding to the minimum Euclidean distance; if the minimum Euclidean distance is less than or equal to a preset spatial search tolerance threshold (the tolerance threshold is dynamically set to the maximum spatial diagonal length of the EVS basic mesh unit), then... If the heavy metal concentration value corresponding to the EVS node exceeds the tolerance threshold, then the exact heavy metal concentration value is assigned to the tetrahedral finite element node constructed by the calculation engine as the initial concentration boundary condition. If the extracted minimum Euclidean distance exceeds the tolerance threshold, it is determined that the 3D porous media numerical simulation engine node is not physically covered by the initial effective data field of EVS. To avoid distortion of the pollution field, its initial concentration parameter is forcibly assigned to the heavy metal background value or detection limit value in the environmental survey report of the target site. For the extraction and conversion of geological model data, the spatial 3D coordinate point set of the upper and lower interfaces of each major stratum is extracted from the 3D geological model generated by EVS, and the extracted stratum interface data is converted into a layered geometric file supported by the 3D porous media numerical simulation engine. On this basis, the physical property parameters such as porosity and permeability coefficient defined in the geological modeling stage are assigned to the corresponding computational grid cells after conversion. After completing the conversion of coordinate and concentration text documents and layered files, they are imported into the 3D porous media numerical simulation engine project to complete the basic data initialization configuration of the numerical model.

[0028] Step S4: Configure model parameters and scene boundary conditions in the three-dimensional porous media numerical simulation engine, solve the equations for water movement and solute convection-dispersion migration and transformation, and output a time-coordinate-concentration time series dataset file based on the spatial coordinate system of the numerical simulation output at multiple time nodes.

[0029] Specifically, a three-dimensional simulation region with the same geometric dimensions and spatial orientation as EVS is constructed in the three-dimensional porous media numerical simulation engine; soil physicochemical parameters (including effective porosity, saturated permeability coefficient, residual water content, and soil distribution coefficient or adsorption coefficient) are assigned to the imported stratum units. The effective porosity, saturated permeability coefficient, and residual water content are determined as follows: particle size distribution data (including the mass percentage of sand, silt, and clay) of field soil samples are extracted, and the particle size distribution data and the measured soil bulk density are input into a soil transformation function. Hydraulic parameters are then derived through a neural network algorithm. The specific network structure and internal composition of the neural network algorithm are as follows: A feedforward artificial neural network topology is adopted. Its input layer contains four neurons, receiving scalar inputs of sand mass percentage, silt mass percentage, clay mass percentage, and soil bulk density, respectively. The hidden layer is a single-layer structure containing 15 neurons, with each node using the Sigmoid function as a non-linear activation function. The output layer contains three neurons, using a linear activation function, outputting predicted values ​​of effective porosity, saturated permeability coefficient, and residual water content, respectively. The weight matrices and bias vectors between each layer are obtained through pre-training and convergence using the Levenberg-Marquardt backpropagation algorithm based on a known soil database. Specifically, the known soil database uses the publicly available UNSODA soil hydraulics database, extracting at least 500 sample sets containing complete measured particle size distribution, bulk density, and corresponding hydraulic parameters as the training set. The convergence judgment logic and hyperparameter configuration for pre-training are as follows: the initial learning rate is set to 0.01, and the target mean squared error (MSE) threshold is... The maximum number of iterations is set to 1000, and an early stopping mechanism is introduced. That is, when the mean square error calculated on the validation set no longer decreases after 10 consecutive iterations, the algorithm determines that it has reached the global optimum and terminates the iteration. At this point, the final network weight matrix and bias vector are solidified and output. The adsorption coefficient characterizes the distribution ratio of heavy metals between the solid and liquid phases. Its determination and derivation steps are as follows: prepare a series of target heavy metal solutions with different concentration gradients, add quantitative dry soil samples respectively, and conduct isothermal oscillation batch equilibrium experiments; measure the heavy metal concentration in the supernatant after adsorption equilibrium, and calculate the soil solid-phase adsorption amount; plot isothermal adsorption curves with liquid phase concentration as the abscissa and solid-phase adsorption amount as the ordinate, and use a linear adsorption isotherm model for fitting. The slope of the fitted line is the value of the adsorption coefficient, in liters per kilogram. Define the model boundary conditions, including the top boundary, bottom boundary, lateral groundwater recharge and discharge, and the source strength release rate of the pollution source; set the simulation time attributes, including initial conditions, iteration time step, and overall simulation prediction period. Specifically, the core calculation parameters in the convection-dispersion migration and transformation equation also include longitudinal dispersion and lateral dispersion. The longitudinal dispersion reflects the mechanical dispersion effect of heavy metals along the direction of water flow, and its value range is set empirically to be 0.5 to 5.0 meters based on the maximum spatial scale of the simulation area; the lateral dispersion reflects the dispersion effect perpendicular to the direction of water flow and is set to one-tenth of the value of the longitudinal dispersion. Regarding the setting of the iteration time step, the model adopts an adaptive time step control logic based on the Courant number: the initial time step is set to 0.01 days, and the maximum time step is 1.0 days. After each iteration, the program calculates the Courant number of all grid cells. This value is calculated by multiplying the actual water flow velocity of the grid cell by the current time step and then dividing by the spatial characteristic size of the grid cell. The logical judgment condition is as follows: if the global maximum Courant number is greater than 1.0, it is determined that the calculation may diverge, and the next iteration step is automatically halved; if the global maximum Courant number is less than 0.5 for three consecutive iterations, the next iteration step is multiplied by an amplification factor of 1.2 to ensure the convergence and stability of the numerical calculation throughout the entire simulation period. Furthermore, when configuring the model parameters, intervention scenarios using repair techniques and natural decay scenarios without repair techniques are constructed respectively. In scenarios employing remediation technologies, parameters are configured according to the preset remediation scheme: For electrokinetic remediation, electric field distribution parameters and ion mobility are added; for chemical leaching, the injection rate, extraction rate, and reaction kinetic parameters of the leaching solution are set. The aforementioned parameters are determined as follows: The ion mobility is derived by obtaining the free molecular diffusion coefficient of the target heavy metal ion in water, dividing it by the product of the ideal gas constant and the absolute temperature, and then multiplying it by the Faraday constant and the valence charge of the heavy metal ion. Its mathematical derivation relationship is defined as follows: ,in, Ion mobility (unit: ), The free molecular diffusion coefficient of this heavy metal ion in water (unit: ), This is the absolute value (dimensionless) of the valence charge of the heavy metal ion. Faraday constant (with a fixed value) ), The ideal gas constant (with a fixed value) ), The absolute temperature of the system (unit: The reaction kinetic parameters adopt a pseudo-first-order kinetic reaction model, with the core parameter, the reaction rate constant k, set to a range of 0.01 to 0.15 per day. This constant was determined through indoor batch leaching kinetic experiments, using the absolute value of the slope after fitting the natural logarithmic linear regression of heavy metal desorption concentration data at different time points. For phytoremediation technology, an equation relating plant root distribution depth to extraction rate was established. The parameter configuration standards for the aforementioned remediation scenarios are as follows: In the electroremediation scenario, the set electric field distribution parameters include the spatial coordinates of the cathode and anode, and the applied DC voltage gradient is set to a range of 1 to 2 volts per centimeter; in the chemical leaching scenario, the leaching solution injection rate is determined based on the site's permeability coefficient, and the injection flow rate is set to 0.5 to 2.0 pore volumes per day; in the phytoremediation scenario, the root distribution depth is set to 0.2 to 0.6 meters below the surface based on the average root length of the selected remediation plants, and the extraction rate equation sets the phytoabsorption process of heavy metals as a sink term that decays exponentially with soil depth, with the sink term coefficient showing a positive linear relationship with root density and potential transpiration rate. The specific mathematical formula is defined as follows: .in, The dimensionless coefficient for empirical absorption reduction (within the range of 0.5 to 1.0) is used. To measure the potential transpiration rate (unit: ), The root density distribution function decays exponentially with depth, and its specific functional form is defined as follows: (in This represents the maximum root density on the earth's surface. The empirical attenuation coefficient is set between 0.02 and 0.05. ), Maximum root distribution depth (unit: ), This represents the soil heavy metal concentration at the current grid depth. In a scenario without remediation techniques, the convection, dispersion, and degradation parameters of heavy metals under hydrogeological conditions are defined. A three-dimensional porous media numerical simulation engine is run to simultaneously solve the partial differential equations for water movement in the porous media and the convection-dispersion migration transformation equations for heavy metal transport within discrete time steps. After calculation, the heavy metal concentration calculation results for each spatially discrete node are output according to the preset observation time nodes and exported as a time-series dataset file containing spatial coordinates, corresponding time nodes, and concentration values.

[0030] Step S5: Perform spatiotemporal consistency calibration on heterogeneous data: extract the geometric origin coordinates of the spatial coordinate system output by numerical simulation and the EVS target coordinate system to construct a three-dimensional affine transformation matrix to eliminate spatial position deviation; and use linear time interpolation to obtain the interpolated concentration values ​​between the discrete time step nodes output by adjacent simulation solvers to fill the data gaps caused by time step differences and generate the core control file.

[0031] Specifically, the multi-time-node concentration dataset exported from the 3D porous media numerical simulation engine undergoes format conversion and coordinate system alignment; simulated outliers are removed; missing node values ​​are marked as missing placeholders, and values ​​below the conventional detection limit are marked as undetected placeholders, forming a time-series data table. As a specific implementation, the underlying data structure of the standardized time-series data table is a multi-dimensional matrix, as shown in Table 1. This table strictly aligns the absolute spatial coordinates (X, Y, Z) of each calculation node with its corresponding heavy metal (e.g., As) concentration values ​​at multiple prediction time nodes (e.g., May 20, 2024, February 14, 2027). This structured high-dimensional data format provides a standard data reading interface for subsequent EVS core control files.

[0032] Table 1

[0033] Specifically, the coordinate system alignment operation is as follows: extract the geometric origin coordinates of the spatial coordinate system output by the numerical simulation and the EVS target coordinate system, construct a three-dimensional affine transformation matrix, and perform a translation transformation to eliminate spatial position deviations. The specific matrix operation of this three-dimensional affine transformation is defined as follows: construct a 4th-order homogeneous coordinate translation transformation matrix at the system bottom layer. The specific data structure generation rules of this matrix are as follows: set all diagonal elements of the matrix to a constant of 1; set the four elements of the last row of the matrix to constants 0, 0, 0, and 1 respectively; the numerical differences in the three spatial dimensions between the absolute coordinates of the EVS origin of the environmental visualization system and the local origin coordinates of the three-dimensional porous media numerical simulation engine, namely the X-axis spatial offset, Y-axis spatial offset, and Z-axis spatial offset, are precisely assigned to the 1st, 2nd, and 3rd element positions of the fourth column of the matrix, respectively. During coordinate system alignment calculations, the system extracts the 3D spatial coordinates of any grid node exported from the 3D porous media numerical simulation engine and converts them into a homogeneous coordinate vector containing four elements. This vector is then multiplied by the previously constructed translation transformation matrix using standard matrix multiplication. The resulting vector represents the fully aligned EVS absolute spatial coordinates. This precise underlying logic step completes the seamless binding of coordinate systems between the two software programs. The logical condition for simulating extreme values ​​is that the concentration exceeds the theoretical maximum solubility limit of the target heavy metal salt in water at standard atmospheric pressure and 25 degrees Celsius. The missing placeholder is specifically defined in the database as a non-numeric (NaN) string placeholder; the concentration value of the undetected placeholder is not zero, but is precisely assigned half of the method detection limit (MDL) value of the detection instrument in this batch of experiments. Using half of the method detection limit as a substitute for Not Detected (ND) data is a widely accepted statistical processing criterion in the fields of environmental monitoring and contaminated site risk assessment (e.g., conforming to data processing guidelines of authoritative environmental agencies such as the US EPA). Compared to directly assigning zero or the original detection limit, taking half can effectively reduce the systematic underestimation or overestimation bias caused by the overall concentration field distribution, making the spatial interpolation results closer to the real physical environment, while ensuring the mathematical continuity of subsequent logarithmic or interpolation operations. The standardized time-series data table is imported back to EVS, and a core control file is created using the time-series control tool: the grid and ground layer control module extracts the three-dimensional grid boundary, grid subdivision accuracy, and stratigraphic structure characteristics of the site, and passes them to the three-dimensional spatial interpolation calculation module; this module receives spatial data and standardized time-series concentration data, sets the start and end time domains, selects the time interpolation algorithm, and generates a core control file that correlates the time step with the concentration of each node.When generating the core control file, in order to ensure the visual continuity and geological fidelity of the dynamic time-series simulation model and the initial static model in terms of three-dimensional spatial distribution, the three-dimensional spatial interpolation calculation module is forced by the system to reuse the same set of kriging spatial autocorrelation parameters (including the same nugget value, sill value, range, and selected semi-variogram model) fitted in step S2 during the spatial interpolation process of the discrete node dataset imported by the three-dimensional porous media numerical simulation engine, and locks the same upper and lower bound rules of the search neighborhood (the minimum number of participants in a single calculation). Number, maximum (Sampling points), thus avoiding non-physical volume expansion or jumps in the pollution plume boundary morphology on the time-axis extrapolation slices caused by inconsistencies in the underlying mesh interpolation algorithms or weight allocation parameters between heterogeneous models. Further, the specific execution logic of the selected time interpolation algorithm is as follows: the system uses linear time interpolation to obtain the intermediate time concentration values ​​between the discrete time step nodes output by two adjacent simulation solvers. For any given mesh node in 3D space, its concentration value C1 at the start time T1 and its concentration value C2 at the end time T2 are extracted; for any target prediction time T between T1 and T2, the interpolation module first calculates the time difference between T and T1, and divides it by the total time difference between T2 and T1 to obtain a dimensionless scaling factor; then, this scaling factor is multiplied by the concentration difference between C2 and C1, and the product is added to the base concentration value C1 to obtain the interpolated concentration value of the mesh node at the target time T. The system traverses all mesh nodes to perform this operation, thereby filling the data gaps between discrete simulation time nodes and achieving continuous and smooth rendering of the 3D pollution plume.

[0034] The specific data flow in the EVS system for generating the core control file (TCF file) is as follows: the ground plane, grid range, accuracy, and spatial interpolation method parameters from the gridding and horizons module are unidirectionally passed to the 3d estimation module to construct the spatial basis data for heavy metal pollution; such as Figure 3 As shown, in the Time Settings panel within the 3D estimation module, the interpolation method and pollutant types are set, and finally, the TCF core control file is packaged and output. Subsequently, the data stream passes through... Figure 2 The split display and scaling module, as shown, assigns specific attributes to the model. After being processed by the pollution feather module, it is finally imported into the viewer module for initial static spatial rendering.

[0035] Step S6, drive and render the dynamic pollution concentration time series model: import the calibrated time series data back to EVS to drive model construction; read the planar two-dimensional red line data of the target polluted plot, calculate the shortest spatial distance from the three-dimensional pollution feather mesh node to the polygon red line boundary, perform data mask operation for mesh nodes that exceed the red line range, and forcibly invalidate the color channel and alpha opacity channel in the rendering pipeline of the dynamic pollution concentration time series model to stop the three-dimensional rendering of the node.

[0036] Furthermore, a soil heavy metal migration visualization model is constructed based on the core control file. The time-series data reading module parses the time-series data table and the core control file; this data is then passed to the attribute assignment module to add displacement and scaling attributes to the 3D spatial model. The drawing import module reads the planar 2D redline data of the target contaminated site and converts it into triangular facet geometry. This conversion uses a 2D Delaunay triangulation algorithm, specifically: connecting discrete coordinate points of the redline boundary within the same plane into a non-intersecting network of triangles, ensuring that the circumcircle of any triangle in the network does not contain any other discrete data points. The specific parameterization operations for adding displacement and scaling attributes to the 3D spatial model are as follows: the scale factor of the X and Y axes is fixed at 1.0, and the elevation scale factor of the Z axis is enlarged to 5.0 to highlight the vertical migration characteristics of heavy metals; for different stratigraphic interfaces, a 2.0-meter vertical translation offset is forcibly applied in the Z-axis direction to achieve visual physical separation and analysis of the stratigraphic model. The spatial distance calculation module is invoked to calculate the spatial distance from the boundary point of the three-dimensional pollution plume to the two-dimensional red line of the land parcel; the spatial intersection filtering module is synchronously passed to filter out data outside the red line range that exceeds the distance threshold; the calculation steps for the shortest spatial distance from the boundary point of the three-dimensional pollution plume to the two-dimensional red line of the land parcel are as follows: project the grid nodes of the three-dimensional pollution plume onto the horizontal two-dimensional plane, calculate the Euclidean distance from each projected node to each boundary line segment of the two-dimensional red line polygon in turn, and extract the minimum value as the shortest spatial distance. The specific logical judgment condition of the intersection filtering module is as follows: The distance buffer threshold is set to 0 meters. Since simple Euclidean absolute distance cannot effectively distinguish whether a node is located inside or outside the red-line polygon in the plane, the intersection filtering module introduces a point-polygon topological relationship judgment rule based on ray casting as a prerequisite before performing filtering: A virtual ray is emitted from the 3D mesh projection node in the positive direction of the horizontal 2D plane (e.g., the positive X-axis), and the number of intersections between the ray and the boundary of the 2D red-line polygon is counted. If the number of intersections is odd, the node is determined to be inside the red-line polygon, and the system directly retains the node data and maintains its original RGBA rendering attributes. If the number of intersections is even, the node is determined to be outside the red-line polygon. At this time, a data mask operation is triggered, forcibly invalidating its RGBA sequence and assigning it a completely transparent value (0,0,0,0), thereby stopping the 3D rendering of nodes outside the red-line range. If the calculated shortest spatial distance is equal to 0 meters, the node data is retained; if the calculated distance is greater than 0 meters, a data mask operation is performed to remove it. The specific execution steps of the data mask operation are as follows: For mesh nodes with a distance greater than 0 meters, the attribute values ​​of their color channel and Alpha opacity channel are forcibly invalidated in the rendering pipeline and assigned a completely transparent RGBA sequence (0,0,0,0), thereby stopping the 3D rendering of the node.Furthermore, the color mapping mechanism of the legend module follows these rules: extract the global minimum and global maximum values ​​of the target heavy metal concentration at the current simulation time; establish a linear correspondence table between the concentration value range and the RGB color space; map the global minimum value to pure blue and the global maximum value to pure red; for concentration data between the maximum and minimum values, perform linear interpolation of RGB values ​​between blue and red according to their percentage of the entire value range to generate a continuously transitioning color band, thereby rendering the visual attributes of each spatial node. The processed 3D model data and distance analysis results are then imported into the integrated 3D viewer module for rendering. Additionally, the current time information in the time-series data reading module is passed to the formatted string module (corresponding). Figure 4 Generate date and time title text (corresponding) Figure 4 ); Legend module (corresponding) Figure 4 Based on the set concentration threshold, a corresponding legend is generated, and all the above output results are simultaneously transmitted to the 3D viewer.

[0037] The pipeline topology within the visualization system for the dynamic pollution concentration time series model is as follows: Figure 4 As shown, specifically: the TCF (Time Series Control File) module reads standardized time series data TCF control files and connects its output to the explode display and scaling module to assign explode and scaling attributes to the 3D spatial model; simultaneously, the CAD (CAD drawing data) module reads the plot plan red line data, which is then converted into triangular patches by the polygon triangulation module; subsequently, the distance module from the 2D region is used to calculate the spatial distance from the pollution node to the red line boundary; the distance data and 3D concentration data are synchronously transmitted to the intersection module to filter and retain the valid pollution node data within the red line range, and finally the data stream is imported into the viewer module for dynamic rendering.

[0038] Step S7, Automated simulation video generation: Run an automated script to extract spatial distribution rendering data and generate static images at fixed time steps; create a high-dimensional tensor containing horizontal and vertical pixels, color channels and total frame count dimensions as a three-dimensional matrix array in system memory; write the RGB pixel value array of the discrete static image into the address of consecutive frames in timestamp order; encapsulate and output the dynamic time-series simulation video file.

[0039] Understandably, when executing the automated output of visualization results, the automated processing script is invoked to control the EVS system to advance the timeline according to a preset time step, synchronously extracting the spatial distribution rendering data in the model at the corresponding time point, and generating static image files reflecting the distribution of heavy metal pollution. Subsequently, by batch reading the generated static image files, the video writing module of the computer vision processing library is invoked, and the video encoding format is explicitly specified as H.264; a blank three-dimensional matrix array with the same pixel resolution as the static image is created in the system memory as a video container; the specific dimensions of the blank three-dimensional matrix array are defined as: a high-dimensional tensor containing a horizontal width dimension, a vertical height dimension, three color channel dimensions (red, green, and blue), and a total frame count dimension; wherein the width dimension is fixed at 1920 pixels, the height dimension is fixed at 1080 pixels, and the total frame count dimension is equal to the total number of static image files sorted in the input folder. The static image files are sorted in ascending order based on the timestamp information contained in their names. The RGB pixel value arrays of the static images are read frame by frame according to the sorting results and written sequentially into the consecutive frame addresses of the blank three-dimensional matrix array. After traversing all the static image files in the specified folder, the video writing module releases the video stream pointer at a set frame rate of 25 frames per second, performs stream data encapsulation, and finally outputs a dynamic timing simulation video file in standard MP4 format.

[0040] In practice, the automated processing script is written in Python. Its core execution logic is as follows: set the step size variable for time extrapolation (e.g., set step_size_hours=6, meaning slicing every 6 hours); the script first calls the interface to obtain the start and end times, and then enters a conditional loop (while current_date<=end_date); in each loop, by calling the system's underlying API instructions (such as evs.set_module), the current timestamp string is dynamically overwritten to the corresponding attribute of the time control module (readtcf) to drive the EVS model state update; after the model update is completed, the view output interface is called to generate a static image, and the current timestamp is incremented by a set time step, until the entire simulation prediction cycle is traversed.

[0041] To verify the effectiveness of this method, the following were performed respectively: Figure 5 and Figure 6 As shown, the three-dimensional spatial distribution projection results of the As heavy metal concentration time series model for a certain site, output using the method of this embodiment, are presented. Through high-fidelity image rendering, the three-dimensional diffusion path of the pollution plume and the distribution of the core high-concentration area can be intuitively depicted, demonstrating the efficiency and accuracy of this collaborative method in dynamic visualization projection.

[0042] In summary, the method provided in this application is no longer limited to the display of traditional static models or the fragmented calculations of a single software. Instead, it constructs a three-dimensional geological and concentration model by acquiring the original borehole concentration dataset; and solves the solute migration equation by using a neural network to map hydraulic parameters and adaptive time steps in a three-dimensional porous media numerical simulation engine. More importantly, this solution innovatively introduces a spatiotemporal consistency calibration mechanism for heterogeneous data (affine transformation and linear time interpolation), and performs RGBA channel transparency masking operations based on the shortest spatial distance in the EVS rendering pipeline. Finally, relying on automated scripts and tensor matrix encapsulation technology, it realizes continuous, high-fidelity dynamic time-series simulation and automated video output of soil heavy metal pollution migration, greatly improving the visualization accuracy and automation level of environmental remediation decision-making.

[0043] It is particularly important to note that the finite element simulation module for solving the water movement and solute transport in porous media described in this embodiment of the invention is essentially a computer algorithm logic for numerically solving specific partial differential equations (such as the Richards equation and the convection-dispersion equation). In specific engineering practices and industrial deployments, the underlying computing power of this numerical simulation engine can be achieved through a self-developed finite element / finite difference solver, or it can seamlessly connect to and call mainstream open-source scientific computing libraries in the industry, such as MODFLOW, MT3DMS, and OpenGeoSys. The core innovation of this invention lies in the spatiotemporal consistency coordination of heterogeneous data and the reconstruction of the dynamic visualization rendering pipeline, which does not depend on or limit itself to any specific closed-source commercial numerical simulation software.

[0044] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A collaborative method for three-dimensional visualization and numerical simulation of heavy metal migration in soil, characterized in that, The method includes the following steps: Step S1: Obtain the original borehole concentration dataset of the contaminated site, which includes soil heavy metal concentration values ​​at different depths and their corresponding spatial coordinates. Step S2: Import the original borehole concentration dataset into the environmental visualization system EVS to generate the EVS target coordinate system and construct a three-dimensional heavy metal concentration distribution model. Combine the site stratigraphic information to construct a three-dimensional geological model containing stratigraphic physical properties and interface spatial coordinates. Step S3: Extract the computational grid node data and stratigraphic interface data from the three-dimensional heavy metal concentration distribution model and the three-dimensional geological model, perform format conversion and spatial mapping, and import them as initial concentration boundary conditions and stratigraphic files into the three-dimensional porous media numerical simulation engine. Step S4: Configure model parameters and scene boundary conditions in the three-dimensional porous media numerical simulation engine, solve the equations for water movement and solute convection-dispersion migration and transformation, and output a time-coordinate-concentration time series dataset file based on the spatial coordinate system of the numerical simulation output at multiple time nodes. Step S5: Perform spatiotemporal consistency calibration on heterogeneous data: extract the geometric origin coordinates of the spatial coordinate system output by numerical simulation and the EVS target coordinate system to construct a three-dimensional affine transformation matrix to eliminate spatial position deviation; and use linear time interpolation to obtain the interpolated concentration values ​​between the discrete time step nodes output by adjacent simulation solvers to fill the data gaps caused by time step differences, and generate a core control file that spatiotemporally maps and associates the three-dimensional mesh boundary, stratigraphic features and the interpolated concentration values. Step S6, drive and render the dynamic pollution concentration time series model: import the calibrated time series data back to EVS to drive model construction; Read the planar two-dimensional red line data of the target contaminated site, calculate the shortest spatial distance from the three-dimensional pollution feather mesh node to the polygon red line boundary, perform a data mask operation for mesh nodes that exceed the red line range, and forcibly invalidate the color channel and alpha opacity channel in the rendering pipeline of the dynamic pollution concentration time series model to stop the three-dimensional rendering of the node. Step S7, Automated video generation: Run an automated script to extract spatial distribution rendering data and generate static images at fixed time steps; A high-dimensional tensor containing horizontal and vertical pixels, color channels, and total frame count is created in system memory as a three-dimensional matrix array. The RGB pixel value arrays of discrete static images are written into the address of consecutive frames in timestamp order, and the dynamic temporal simulation video file is encapsulated and output, thereby realizing the dynamic visualization simulation of soil heavy metal pollution diffusion under the condition of spatiotemporal consistency of heterogeneous data.

2. The method according to claim 1, characterized in that, In step S2, the model construction process specifically includes: Outliers were removed from the imported original borehole concentration dataset using the 3σ criterion, and missing values ​​were filled in using Kriging interpolation. The three-dimensional heavy metal concentration distribution model was generated by calling the three-dimensional interpolation module. The stratigraphic modeling method was used, and the continuous elevation surfaces at the top and bottom of each generalized stratigraphic unit were fitted using the Kriging interpolation algorithm. The three-dimensional geological model was established in a closed three-dimensional space using a hexahedral mesh division rule.

3. The method according to claim 1, characterized in that, In step S3, the specific operations of format conversion and space mapping are as follows: The spatial coordinates and concentration values ​​of the exported nodes are generated into an ASCII-encoded comma-separated text document as the initial concentration boundary conditions. The hexahedral mesh data output by EVS is reconstructed into a Delaunay tetrahedral finite element mesh, and the coordinates of the EVS mesh nodes are mapped one-to-one with the tetrahedral finite element nodes constructed by the calculation engine through the nearest neighbor search algorithm.

4. The method according to claim 1, characterized in that, In step S4, when configuring model parameters, a feedforward artificial neural network topology is used to map the soil particle size distribution data and bulk density obtained from the field survey into hydraulic parameters: Its input layer receives scalar inputs of the mass percentages of sand, silt, and clay particles, as well as soil bulk density. The hidden layer uses the Sigmoid function as the activation function, and the output layer uses a linear activation function to output the predicted values ​​of effective porosity, saturated permeability coefficient, and residual water content.

5. The method according to claim 1, characterized in that, In step S4, the equations for water movement and solute convection-dispersion migration are solved using adaptive time step control logic based on the Courant number. Set the initial time step and the maximum time step, and calculate the global maximum Courant number after each iteration; if the global maximum Courant number is greater than 1.0, the step size of the next iteration will be automatically halved. If the global maximum Courant number is less than 0.5 for three consecutive iterations, the next iteration step size is multiplied by a magnification factor of 1.

2.

6. The method according to claim 1, characterized in that, In step S4, before solving the equations for water movement and solute convection-dispersion migration and transformation, a comparative simulation scenario is also constructed: Intervention scenarios including electroremediation, chemical leaching, or phytoremediation were established, and root extraction rate equations with ion electromobility, leaching pseudo-first-order kinetic reaction rate constant, or exponential decay with soil depth were configured respectively; at the same time, a natural decay scenario considering only the natural migration law of heavy metals was established.

7. The method according to claim 1, characterized in that, In step S5, the dataset file also includes a normalization process: missing node values ​​are marked as non-numeric string placeholders; for values ​​below the conventional detection limit, the exact value is assigned as half of the detection limit value of the detection instrument method.

8. The method according to claim 1, characterized in that, In step S5, the specific calculation logic for obtaining the interpolated concentration value using linear time interpolation is as follows: For the concentration value C1 at the start time T1 and the concentration value C2 at the end time T2, calculate the time difference between the target time T and T1 and divide it by the total time difference between T2 and T1 to obtain a dimensionless scaling factor; multiply this scaling factor by the concentration difference between C2 and C1 and add it to the base concentration value C1 to obtain the interpolated concentration value of the grid node at the target time T.

9. The method according to claim 1, characterized in that, In step S6, the calculation of the shortest spatial distance and the data masking operation are specifically as follows: The two-dimensional Delaunay triangulation algorithm is used to transform the two-dimensional red line into triangular facet geometry. The mesh nodes of the three-dimensional pollution plume are projected onto a horizontal two-dimensional plane. The Euclidean distance from each projected node to each boundary line segment of the two-dimensional red line polygon is calculated and the minimum value is extracted as the shortest spatial distance. When the shortest spatial distance is greater than 0 meters, its RGBA sequence is assigned to be completely transparent (0,0,0,0).

10. The method according to claim 1, characterized in that, In step S7, the generation of the automated simulation video specifically involves: The video writing module of the computer vision processing library is invoked. The width dimension of the three-dimensional matrix array composed of the high-dimensional tensor is fixed at 1920 pixels, and the height dimension is fixed at 1080 pixels. After traversing and writing the static image file, the video writing module releases the video stream pointer at a set frame rate of 25 frames per second, performs stream data encapsulation, and outputs a video file in the standard H.264 encoding format.

Images