A simulation method for the spread of hazardous chemical leaks in complex terrain

Abstract

This invention discloses a simulation method for the diffusion of hazardous chemicals leaks in complex terrain, belonging to the interdisciplinary field of hazardous chemicals and computer simulation. It aims to solve the technical problems of existing technologies in accurately characterizing terrain properties, rapidly calculating terrain wind field distortion, and precisely simulating the ground-level flow and accumulation effects of heavy gas particles in complex micro-scale terrain. The method includes the following steps: S1, obtaining a high-precision micro-scale terrain mesh characterization based on UAV oblique photography; S2, constructing a micro-scale three-dimensional wind field diagnostic model by integrating terrain geometric effects; S3, performing dynamic flow field calculations using an improved heavy gas Lagrange particle algorithm; and S4, using spatiotemporal heterogeneity to dynamically extrapolate the three-dimensional concentration field, completing the simulation of hazardous chemical leakage and diffusion in complex terrain. This application achieves accurate simulation of hazardous chemical leakage and diffusion in complex terrain and can be applied to emergency decision support for hazardous chemical leakage accidents in complex terrains such as chemical industrial parks and mountain valleys.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of hazardous chemicals and computer simulation. Specifically, it relates to a simulation method for the leakage and diffusion of hazardous chemicals in complex terrain. Background Technology

[0002] With the clustered development of modern chemical industries, the scale of production, storage, and transportation of various hazardous chemicals has increased dramatically, leading to a surge in sudden leaks. Due to the highly uncertain, rapid-spreading, and hazardous nature of hazardous chemical accidents, leaks pose a serious threat to the safety of people and property in surrounding areas. Therefore, developing high-precision hazardous chemical leak and diffusion simulation technology is of significant practical importance for accident prevention and emergency response decision-making.

[0003] Existing methods for simulating the diffusion of hazardous chemicals mainly include Gaussian plume models and conventional Lagrange smoke cloud models. These methods are mostly based on the idealized assumption of flat terrain or use conventional digital elevation model data with low resolution for terrain characterization. However, in complex micro-scale terrains such as mountains, canyons, hills, or chemical industrial parks containing complex structures, wind fields can generate intense local backflow, shielding, bypassing, and acceleration effects. Traditional simulation methods based on the assumption of flat terrain are difficult to accurately describe the gas diffusion behavior under such complex wind field environments.

[0004] Meanwhile, when typical hazardous chemical gases such as chlorine, liquefied natural gas, and heavy petroleum gas leak, they are classified as heavy gases because their density is greater than that of the surrounding air. Heavy gases exhibit a significant negative buoyancy effect during diffusion, manifesting as gravitational settling, accumulation along low-lying areas, and ground-level spread. This unique diffusion behavior places higher demands on the accuracy of simulation methods.

[0005] However, existing simulation techniques have significant limitations in handling these complex problems. On the one hand, current techniques cannot quickly acquire and convert high-precision microscale terrain boundaries that support physical topology calculations, resulting in insufficient accuracy in terrain representation. On the other hand, traditional Lagrange particle models do not consider the spatiotemporal correction of particle flow by the negative buoyancy of heavy air, and employ simple disappearance or fully elastic reflection methods when particles collide with terrain boundaries, failing to accurately reflect the ground-level flow and aggregation behavior of heavy air particles on complex terrain surfaces. These technical deficiencies lead to serious deviations in the prediction of diffusion boundaries and concentration distribution under complex microscale terrain, failing to meet the needs of high-precision and high-timeliness emergency rescue decision-making. Summary of the Invention

[0006] The purpose of this invention is to provide a simulation method for the leakage and diffusion of hazardous chemicals in complex terrain. It mainly solves the technical problems of existing technologies in accurately characterizing terrain properties, rapidly calculating terrain wind field distortion, and accurately simulating the flow and accumulation effects of heavy gas particles along the ground in complex terrain at the microscale.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A method for simulating the diffusion of hazardous chemicals in complex terrain includes the following steps:

[0009] S1, based on UAV oblique photography, obtains microscale high-precision terrain mesh physical property characterization;

[0010] S2, integrating terrain geometry effects, constructs a microscale three-dimensional wind field diagnostic model;

[0011] S3 uses an improved heavy gas Lagrange particle algorithm for dynamic flow field calculation;

[0012] S4 employs spatiotemporal heterogeneity to perform dynamic simulation of three-dimensional concentration fields, completing the simulation of hazardous chemical leakage and diffusion under complex terrain.

[0013] Furthermore, the specific process of step S1 is as follows:

[0014] S11 uses a drone equipped with a multi-lens oblique photography system to take cross-flight photos and obtain multi-view aerial images with an overlap greater than a set threshold.

[0015] S12 uses a multi-view vision 3D geometric reconstruction algorithm to perform aerial triangulation operator analysis, generate dense point cloud data, and construct a continuous 3D surface mesh model.

[0016] S13 uses a point cloud feature classification algorithm to extract the geometric texture features of the land surface, vegetation and structures, and constructs an aerodynamic roughness length matrix of spatial heterogeneity.

[0017] S14, the three-dimensional surface mesh model is transformed into a spatially oriented bounding box tree structure with hierarchical topological relationships.

[0018] Furthermore, the specific process of step S2 is as follows:

[0019] S21, Obtain macroscopic meteorological parameters of the outer boundary of the simulation area;

[0020] S22, the Gaussian objective analysis method is used to perform three-dimensional flow field spatial interpolation of the benchmark wind speed along the spatially oriented bounding box tree boundary to generate the initial estimated wind field;

[0021] S23 introduces the mass conservation condition of a three-dimensional incompressible fluid as a constraint, establishes the Poisson equation, and corrects the stream function of the initial estimated wind field through the relaxation iteration method, outputting a three-dimensional microscale diagnostic wind field.

[0022] Furthermore, the specific process of step S3 is as follows:

[0023] S31, obtain the physical parameters of the hazardous chemical leak source, and discretize the source strength mass flow rate into a virtual Lagrange particle beam;

[0024] S32, introduce heavy air negative buoyancy and gravity settling correction terms to calculate the instantaneous velocity of Lagrange particles;

[0025] S33, Establish an adaptive boundary collision mechanism based on terrain roughness dissipation, and perform real-time intersection collision detection between the particle trajectory line and the spatially oriented bounding box tree;

[0026] S34 uses an inelastic momentum dissipation model to update the particle velocity after the collision.

[0027] Furthermore, the specific process of step S4 is as follows:

[0028] S41 uses a spatiotemporal heterogeneous computing architecture to encapsulate wind field diagnosis and particle dynamics calculations into the underlying computing kernel.

[0029] S42 updates the three-dimensional coordinates of the entire particle space in real time through multi-threaded parallel computing, and introduces a spatiotemporal adaptive kernel density estimation method to transform the discrete particle mass spatial distribution into a continuous three-dimensional spatial concentration field.

[0030] S43 sets a safety protection concentration threshold and dynamically renders the distribution of excessive smoke plumes in real time within a three-dimensional digital twin terrain.

[0031] Furthermore, the operator parsing method in step S12 is as follows: extract feature points in the image and perform image matching to establish the relative relationship between each image; use bundle adjustment technique to perform overall adjustment optimization on all images, and use dense point cloud generation algorithm to perform dense matching on the images to obtain high-density three-dimensional point cloud data of the target area.

[0032] Furthermore, in step S12, the method for constructing a continuous three-dimensional surface mesh model is as follows: the dense point cloud is triangulated to generate a set of mesh elements with topological connections; point cloud filtering is performed to remove noise points and outliers; and hole filling is performed to ensure the integrity and continuity of the mesh model.

[0033] Further, in step S32, within time step Δt, the instantaneous velocity vector (up, vp, wp) of each Lagrange particle is expressed as follows:

[0034] up = u + u'

[0035] vp = v + v'

[0036] wp = w + w' - wg

[0037] Where (u, v, w) is the three-dimensional microscale diagnostic wind speed of the corresponding spatial grid; (u', v', w') is the atmospheric turbulent random fluctuation velocity calculated using the Markov chain random walk model; Let be the gravitational settling velocity of heavy gas particles, and its calculation formula is:

[0038]

[0039] In the formula, Where g is the drag coefficient and g is the acceleration due to gravity. This represents the real-time density of heavy gas particles. For ambient air density, It is the characteristic scale of particles.

[0040] Furthermore, in step S33, if a collision occurs, the terrain normal vector n at the collision point is calculated, and the particle's velocity vector is decomposed into normal velocities. and tangential velocity The particle velocities after the collision are updated using an inelastic momentum dissipation model.

[0041]

[0042]

[0043] In the formula, The normal restitution coefficient; The tangential friction dissipation function is related to the aerodynamic roughness length, thereby constraining heavy air particles to flow or accumulate along the low-lying areas of complex terrain surfaces after collision.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] (1) This invention utilizes a multi-lens oblique photography system to acquire highly overlapping multi-view aerial images. A continuous three-dimensional surface mesh model is generated through aerial triangulation and dense point cloud reconstruction. Furthermore, geometric texture features of the land surface, vegetation, and structures are extracted to construct a spatially heterogeneous aerodynamic roughness length matrix. Based on this, the three-dimensional surface mesh model is transformed into a hierarchical spatially oriented bounding box tree structure, providing an accurate digital twin terrain boundary base for subsequent wind field diagnosis and particle collision detection. This significantly improves the accuracy and computational efficiency of terrain property characterization under microscale complex terrain.

[0046] (2) This invention uses a spatially oriented bounding box tree structure as the boundary constraint, employs Gaussian objective analysis to interpolate macroscopic meteorological parameters as the initial estimated wind field, then introduces the mass conservation condition of a three-dimensional incompressible fluid to establish the Poisson equation, and corrects the stream function of the initial wind field through a relaxation iteration method, finally outputting a three-dimensional microscale diagnostic wind field that strictly satisfies the continuity equation. This model can accurately capture the wind field distortion characteristics induced by complex terrain such as mountain backflow, valley acceleration, and structure obstruction and flow around, overcoming the technical defects of wind field simulation distortion under the traditional flat terrain assumption, and providing high-precision wind field driving conditions for hazardous chemical diffusion simulation.

[0047] (3) This invention introduces a gravity settling velocity correction term into the traditional Lagrange particle model, giving heavy air particles a physical tendency to settle towards the ground during their motion. Simultaneously, a real-time collision detection mechanism based on a spatially oriented bounding box tree is established. When a particle collides with the terrain boundary, an inelastic momentum dissipation model is used to correct the normal and tangential velocities, respectively, where the attenuation of the tangential velocity is related to the aerodynamic roughness length. This mechanism enables heavy air particles to spread and accumulate along the low-lying areas of the terrain surface after a collision, rather than simply disappearing or completely elastically rebounding, truly reflecting the diffusion behavior characteristics of heavy air under complex terrain.

[0048] (4) This invention employs multi-threaded parallel computing to update the spatial three-dimensional coordinates of all particles in real time, and transforms the discrete particle mass distribution into a continuous three-dimensional concentration field through a spatiotemporal adaptive kernel density estimation method. A smaller bandwidth is used in dense particle regions to preserve detailed features, while a larger bandwidth is used in sparse regions to ensure continuity. Simultaneously, the distribution of excessive smoke plumes is dynamically rendered in real time in a three-dimensional digital twin terrain, providing intuitive and efficient visualization support for emergency decision-making in hazardous chemical leakage accidents in complex terrains such as chemical industrial parks and mountain valleys. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating the method of the present invention. Detailed Implementation

[0050] The present invention will be further described below with reference to the accompanying drawings and embodiments. The embodiments of the present invention include, but are not limited to, the following embodiments.

[0051] like Figure 1 As shown, this invention discloses a simulation method for the diffusion of hazardous chemicals in complex terrain. This method acquires high-precision terrain data using an unmanned aerial vehicle (UAV) oblique photography system and combines it with an improved Lagrange particle dynamics model to achieve accurate simulation of the diffusion of heavy gases containing hazardous chemicals in complex terrain. Figure 1 As shown, the method mainly includes four technical steps: microscale high-precision terrain grid physical property characterization based on UAV oblique photography, construction of microscale three-dimensional wind field diagnostic model integrating terrain geometric effects, calculation of improved heavy gas Lagrange particle dynamics flow field, and spatiotemporal heterogeneous integration and dynamic extrapolation of three-dimensional concentration field.

[0052] In the microscale high-precision terrain mesh physical property characterization stage based on UAV oblique photogrammetry, the first step is to use a UAV oblique photogrammetry system to collect terrain data of the simulated hazardous chemical area. The UAV oblique photogrammetry system includes a multi-lens oblique camera, which contains multiple lenses mounted at a certain angle, capable of simultaneously acquiring orthophoto and oblique images of the target area. During actual operation, the UAV flies and takes pictures according to a preset intersecting flight path, maintaining a certain degree of overlap between adjacent flight paths to ensure that the acquired multi-view aerial images have sufficient coverage and stereo matching accuracy. The aerial image data acquired by the multi-lens oblique camera is transmitted to a post-processing computer for 3D reconstruction calculations.

[0053] Aerial triangulation operators were used to analyze the acquired multi-view aerial imagery using a multi-view vision 3D geometric reconstruction algorithm. The aerial triangulation operator extracts feature points from the images and performs image matching to establish relative relationships between images, thereby recovering the exterior orientation elements. Based on this, bundle adjustment techniques were used to perform overall adjustment optimization on all images, improving the overall accuracy of the points. After adjustment, a dense point cloud generation algorithm was used to perform dense matching on the images, obtaining high-density 3D point cloud data of the target area. The dense point cloud data contains the 3D coordinate information of the target area's surface and features, accurately reflecting the terrain's geometry.

[0054] A continuous 3D surface mesh model is constructed based on the generated dense point cloud data. The 3D surface mesh model uses a triangulation algorithm to transform discrete point cloud data into a set of mesh elements with topological connections. Adjacent elements form a continuous spatial surface by sharing edges and vertices. During the construction of the 3D surface mesh model, point cloud filtering is performed to remove noise and outliers, and hole filling is performed to ensure the integrity and continuity of the mesh model.

[0055] A point cloud feature classification algorithm was used to analyze the features of a 3D surface mesh model, extracting geometric texture features of the land surface, vegetation, and structures. The point cloud feature classification algorithm includes a boundary detection method based on normal vector changes to identify abrupt terrain changes and feature edges; a vegetation identification method based on elevation difference to distinguish between surface points and vegetation points; and a structure differentiation method based on reflection intensity to identify different types of artificial structures. An aerodynamic roughness length matrix representing spatial heterogeneity was constructed based on the feature classification results. This matrix records surface roughness information at different locations. Aerodynamic roughness length is an important parameter characterizing the degree of airflow obstruction by the underlying surface; a larger roughness length indicates a stronger obstruction effect on airflow. In practical applications, the roughness length in vegetated areas is typically 0.5 to 2 meters, in densely built-up areas it is typically 2 to 10 meters, and in bare ground or water surfaces it is typically 0.1 to 0.5 meters.

[0056] A continuous 3D surface mesh model is transformed into a spatially oriented bounding box tree structure with hierarchical topological relationships. The spatially oriented bounding box tree structure is constructed using a hierarchical sequence of axis-aligned bounding boxes, with outer bounding boxes containing inner bounding boxes, and inner bounding boxes further containing more subdivided bounding boxes. During collision detection, the intersection of the particle trajectory line with the outer bounding box is first checked; if they do not intersect, a collision is directly determined; if they intersect, the intersection with the inner bounding box is checked, and so on recursively until the bottom bounding box. This hierarchical bounding box structure significantly improves the query efficiency of collision detection, meeting the computational requirements of real-time simulation. The spatially oriented bounding box tree structure serves as a digital twin boundary basis for Lagrange particle collision detection, providing accurate terrain constraints for subsequent particle dynamics calculations.

[0057] In the construction phase of the microscale three-dimensional wind field diagnostic model incorporating topographic geometric effects, macroscopic meteorological parameters of the outer boundary of the simulated area are first acquired. These macroscopic meteorological parameters include baseline wind speed, baseline wind direction, and atmospheric stability parameters. Baseline wind speed and direction are obtained from meteorological station observation data or numerical weather prediction model outputs. Atmospheric stability parameters are determined using the Pasquale stability classification method based on meteorological elements such as temperature lapse rate and cloud cover. The Pasquale stability classification divides atmospheric stability into seven levels: strongly unstable, unstable, weakly unstable, neutral, weakly stable, stable, and strongly stable. Different stability levels correspond to different atmospheric diffusion capacities.

[0058] A Gaussian objective analysis method is used to interpolate the reference wind speed along the spatially oriented bounding box tree boundary in three-dimensional flow field, generating an initial estimated wind field (uc, vc, wc). The Gaussian objective analysis method uses a Gaussian weighting function to interpolate discrete meteorological station data in three-dimensional space; the weighting function decays exponentially with increasing distance. In the actual calculation, the three-dimensional spatial distance between the interpolation point and each meteorological station is first determined, then the weighting coefficients are calculated based on the distance, and finally, the wind speed value at the interpolation point is obtained through weighted averaging. The wind field generated by the Gaussian objective analysis method maintains consistency with the observed data and exhibits continuous spatial variation characteristics.

[0059] A Poisson equation is established by introducing the mass conservation condition of a three-dimensional incompressible fluid as a constraint to correct the initial estimated wind field. Under the assumption of incompressible fluid, the fluid density is constant, and the velocity field satisfies the continuity equation. Since the initial estimated wind field usually does not strictly satisfy the mass conservation condition, it needs to be corrected using the Poisson equation. The Poisson equation is solved using a relaxation iteration method, which continuously adjusts the wind field distribution through iterative solutions to gradually approach a solution that satisfies the mass conservation constraint. In each iteration, the divergence value of the velocity field is calculated, and the correction amount is calculated based on the divergence value and superimposed on the velocity field. The iteration process continues until the divergence value decreases below a preset convergence threshold. The three-dimensional microscale diagnostic wind field corrected by the Poisson equation strictly satisfies the first-order continuity equation.

[0060]

[0061] It can accurately reflect the wind field characteristics under complex terrain conditions such as mountain backflow, valley acceleration, and structure obstruction. The leeward side of the mountain forms a backflow zone, and the wind direction is opposite to the incoming flow direction; in the canyon area, the wind speed increases due to the funneling effect; the windward side of the building creates a stagnation zone, where the wind speed decreases and the pressure increases.

[0062] In the calculation stage of the improved heavy gas Lagrange particle dynamics flow field, the physical parameters of the hazardous chemical leak source are first obtained. These parameters include the type of leaked substance, the location of the leak source, the source mass flow rate, the leak temperature, and the leak pressure. Based on these parameters, the source mass flow rate is discretized into a virtual Lagrange particle beam continuously released per unit time. Each Lagrange particle represents a certain mass of hazardous chemical substance and has independent mass and composition characteristics. During the simulation, particles are released from the leak source location at regular time intervals, forming a continuous particle stream.

[0063] The velocity of the Lagrange particles is calculated by introducing the negative buoyancy of the heavy air and the correction term for gravity settling. Within the time step Δt, the instantaneous velocity vector (up, vp, wp) of each Lagrange particle is expressed as:

[0064] up = u + u'

[0065] vp = v + v'

[0066] wp = w + w' - wg

[0067] Where (u, v, w) is the three-dimensional microscale diagnostic wind speed of the corresponding spatial grid; (u', v', w') is the atmospheric turbulent random fluctuation velocity calculated using the Markov chain random walk model; Let be the gravitational settling velocity of heavy gas particles, and its calculation formula is:

[0068]

[0069] In the formula, Where g is the drag coefficient and g is the acceleration due to gravity. This represents the real-time density of heavy gas particles. For ambient air density, The characteristic scale of the particles is denoted by . This formula takes into account the difference between the density of heavy gas particles and the density of ambient air, the characteristic scale of the particles, and the drag coefficient. Because the density of heavy gas particles is greater than that of ambient air, their gravitational settling velocity is downward, causing them to tend to settle towards the ground. This is a key difference between the diffusion of heavy and light gases; heavy gases exhibit significant gravitational settling characteristics during diffusion.

[0070] Subsequently, an adaptive boundary collision mechanism based on terrain roughness dissipation is established. When calculating the new spatial position of particles at each time step, the particle trajectory line is intersected and collision detected in real time with the spatially oriented bounding box tree. Specifically, the starting and ending positions of the particles at the current time step are first calculated, and then it is detected whether the particle trajectory line intersects with any bounding box in the bounding box tree. If an intersection is detected, the specific location of the intersection and the terrain geometry information at the collision point are further determined.

[0071] If a collision occurs, calculate the terrain normal vector n at the collision point, and decompose the particle's velocity vector into normal velocities. and tangential velocity The particle velocities after the collision are updated using an inelastic momentum dissipation model.

[0072]

[0073]

[0074] In the formula, The normal restitution coefficient; The tangential friction dissipation function is related to the aerodynamic roughness length. The normal velocity component reverses direction after the collision and is multiplied by the normal restitution coefficient. Due to the use of an inelastic collision model, the normal restitution coefficient is less than one, meaning that some normal kinetic energy is lost. The tangential velocity component is multiplied by the tangential friction dissipation function related to the aerodynamic roughness length, which attenuates the tangential velocity according to the terrain roughness. The greater the terrain roughness, the more significant the friction dissipation, and the more obvious the attenuation of the tangential velocity. Through this adaptive boundary collision mechanism, heavy air particles are constrained to flow or accumulate along the low-lying areas of complex terrain surfaces after the collision, rather than simply disappearing or undergoing complete elastic reflection. This truly reflects the motion characteristics of heavy air particles on complex terrain surfaces; when encountering terrain obstacles, heavy air will spread along the surface and accumulate in low-lying areas.

[0075] In the spatiotemporal heterogeneous integration and three-dimensional concentration field dynamic extrapolation stages, a spatiotemporal heterogeneous computing architecture is adopted to achieve efficient parallel computing. This architecture encapsulates wind field diagnostic calculations and particle dynamics calculations into independent computing kernels, leveraging the parallel computing capabilities of modern multi-core processors to execute multiple computational tasks simultaneously. Wind field diagnostic calculations and particle dynamics calculations communicate via a data transfer interface; the particle dynamics calculations use the wind speed data output from the wind field diagnostics to update the particle motion state.

[0076] A spatiotemporally adaptive kernel density estimation method is introduced to transform the discrete spatial distribution of particle mass into a continuous three-dimensional spatial concentration field. The kernel density estimation method constructs a three-dimensional Gaussian kernel function centered on each Lagrange particle, and the bandwidth of the three-dimensional Gaussian kernel function is adaptively adjusted according to the local particle density. A smaller bandwidth is used in densely populated regions to preserve the detailed features of the concentration field, while a larger bandwidth is used in sparsely populated regions to ensure the continuity of the concentration field. The kernel functions of all particles are superimposed to form the three-dimensional spatial concentration field C(x,y,z,t), where the concentration value represents the mass of hazardous chemicals per unit volume. The spatiotemporally adaptive kernel density estimation method overcomes the spatial resolution limitations of traditional grid statistical methods and can generate a smooth and continuous concentration field distribution.

[0077] A safety protection concentration threshold is set, and the distribution of hazardous chemical gas plumes exceeding the standard is dynamically rendered in real time within a 3D digital twin terrain. The distribution of plumes exceeding the standard indicates areas where the concentration exceeds the safety protection threshold, and is visually displayed on the 3D terrain model using color coding. Simultaneously, the coverage area of ​​the exceeding area is calculated and displayed, providing a reference for emergency evacuation routes. The real-time diffusion front represents the current diffusion boundary of the hazardous chemical gas, reflecting the impact range of a leak. The dynamic rendering module updates the distribution of plumes exceeding the standard, the coverage area, and the real-time diffusion front in the 3D digital twin terrain in real time, providing intuitive visualization support for emergency response decision-making.

[0078] In a practical application scenario, taking a chemical industrial park as an example, this park is located in a hilly area and contains multiple storage tank areas and production facilities. Assume a liquid chlorine storage tank in the park leaks. Liquid chlorine is a heavy gas with a density greater than air. Traditional methods typically use the assumption of flat ground for diffusion simulation, which cannot accurately reflect the impact of terrain on the wind field. This invention first acquires high-precision three-dimensional terrain data of the park using an UAV oblique photography system to construct a digital twin terrain model. Then, a microscale wind field diagnostic model is used to calculate the terrain-induced wind field distortion effect, obtaining a detailed wind field distribution within the park. In particle dynamics calculations, the heavy gas characteristics of liquid chlorine are considered, a gravity settling correction term is introduced, and an adaptive boundary collision mechanism is used to simulate the ground-level flow and accumulation behavior of liquid chlorine. Finally, a three-dimensional concentration field distribution is generated using a kernel density estimation method, dynamically rendering the diffusion range of the excessive plume. Emergency commanders can determine the area requiring evacuation and evacuation routes based on the simulation results, and formulate scientific emergency rescue plans.

[0079] This invention, through the implementation of the above-mentioned technical solution, effectively solves the technical problems of existing technologies in accurately characterizing terrain properties, rapidly calculating terrain wind field distortion, and precisely simulating the ground-level flow and accumulation effects of heavy gas particles in complex micro-scale terrain. This method can be widely applied to emergency decision support for hazardous chemical spill accidents in complex terrains such as chemical industrial parks, mountain valleys, and port terminals.

[0080] The above embodiments are merely one of the preferred embodiments of the present invention and should not be used to limit the scope of protection of the present invention. Any modifications or refinements made to the main design concept and spirit of the present invention that are not of substantial significance, but solve the same technical problem as the present invention, should be included within the scope of protection of the present invention.

Claims

1. A method for simulating the leakage and diffusion of hazardous chemicals in complex terrain, characterized in that, Includes the following steps: S1, based on UAV oblique photography, obtains microscale high-precision terrain mesh physical property characterization; S2, integrating terrain geometry effects, constructs a microscale three-dimensional wind field diagnostic model; S3 uses an improved heavy gas Lagrange particle algorithm for dynamic flow field calculation; S4 employs spatiotemporal heterogeneity to perform dynamic simulation of three-dimensional concentration fields, completing the simulation of hazardous chemical leakage and diffusion under complex terrain.

2. The simulation method for hazardous chemical leakage and diffusion under complex terrain according to claim 1, characterized in that, The specific process of step S1 is as follows: S11 uses a drone equipped with a multi-lens oblique photography system to take cross-flight photos and obtain multi-view aerial images with an overlap greater than a set threshold. S12 uses a multi-view vision 3D geometric reconstruction algorithm to perform aerial triangulation operator analysis, generate dense point cloud data, and construct a continuous 3D surface mesh model. S13 uses a point cloud feature classification algorithm to extract the geometric texture features of the land surface, vegetation and structures, and constructs an aerodynamic roughness length matrix of spatial heterogeneity. S14, the three-dimensional surface mesh model is transformed into a spatially oriented bounding box tree structure with hierarchical topological relationships.

3. The simulation method for hazardous chemical leakage and diffusion under complex terrain according to claim 2, characterized in that, The specific process of step S2 is as follows: S21, Obtain macroscopic meteorological parameters of the outer boundary of the simulation area; S22, the Gaussian objective analysis method is used to perform three-dimensional flow field spatial interpolation of the benchmark wind speed along the spatially oriented bounding box tree boundary to generate the initial estimated wind field; S23 introduces the mass conservation condition of a three-dimensional incompressible fluid as a constraint, establishes the Poisson equation, and corrects the stream function of the initial estimated wind field through the relaxation iteration method, outputting a three-dimensional microscale diagnostic wind field.

4. The simulation method for hazardous chemical leakage and diffusion under complex terrain according to claim 3, characterized in that, The specific process of step S3 is as follows: S31, obtain the physical parameters of the hazardous chemical leak source, and discretize the source strength mass flow rate into a virtual Lagrange particle beam; S32, introduce heavy air negative buoyancy and gravity settling correction terms to calculate the instantaneous velocity of Lagrange particles; S33, Establish an adaptive boundary collision mechanism based on terrain roughness dissipation, and perform real-time intersection collision detection between the particle trajectory line and the spatially oriented bounding box tree; S34 uses an inelastic momentum dissipation model to update the particle velocity after the collision.

5. The simulation method for hazardous chemical leakage and diffusion under complex terrain according to claim 4, characterized in that, The specific process of step S4 is as follows: S41 uses a spatiotemporal heterogeneous computing architecture to encapsulate wind field diagnosis and particle dynamics calculations into the underlying computing kernel. S42 updates the three-dimensional coordinates of the entire particle space in real time through multi-threaded parallel computing, and introduces a spatiotemporal adaptive kernel density estimation method to transform the discrete particle mass spatial distribution into a continuous three-dimensional spatial concentration field. S43 sets a safety protection concentration threshold and dynamically renders the distribution of excessive smoke plumes in real time within a three-dimensional digital twin terrain.

6. The simulation method for hazardous chemical leakage and diffusion under complex terrain according to claim 5, characterized in that, The operator parsing method in step S12 is as follows: extract feature points in the image and perform image matching to establish the relative relationship between each image; use bundle adjustment technique to perform overall adjustment optimization on all images; use dense point cloud generation algorithm to perform dense matching on the images to obtain high-density three-dimensional point cloud data of the target area.

7. The simulation method for hazardous chemical leakage and diffusion under complex terrain according to claim 6, characterized in that, In step S12, the method for constructing a continuous three-dimensional surface mesh model is as follows: the dense point cloud is triangulated to generate a set of mesh elements with topological connections, point cloud filtering is performed to remove noise points and outliers, and hole filling is performed to ensure the integrity and continuity of the mesh model.

8. The simulation method for hazardous chemical leakage and diffusion under complex terrain according to claim 7, characterized in that, In step S32, within time step Δt, the instantaneous velocity vector (up, vp, wp) of each Lagrange particle is expressed as follows: up = u + u' vp = v + v' wp = w + w' - wg Where (u, v, w) is the three-dimensional microscale diagnostic wind speed of the corresponding spatial grid; (u', v', w') is the atmospheric turbulent random fluctuation velocity calculated using the Markov chain random walk model; Let be the gravitational settling velocity of heavy gas particles, and its calculation formula is: In the formula, Where g is the drag coefficient and g is the acceleration due to gravity. This represents the real-time density of heavy gas particles. For ambient air density, It is the characteristic scale of particles.

9. The simulation method for hazardous chemical leakage and diffusion under complex terrain according to claim 8, characterized in that, In step S33, if a collision occurs, the terrain normal vector n at the collision point is calculated, and the particle's velocity vector is decomposed into normal velocities. and tangential velocity The particle velocities after the collision are updated using an inelastic momentum dissipation model. In the formula, The normal restitution coefficient; The tangential friction dissipation function is related to the aerodynamic roughness length, thereby constraining heavy air particles to flow or accumulate along the low-lying areas of complex terrain surfaces after collision.

Images