Mine water inrush disaster dynamic deduction and intelligent disaster avoidance decision method and system based on digital twin engine

Abstract

The application belongs to the field of digital twin technology, disaster dynamic simulation and intelligent emergency decision-making, and discloses a mine water inrush disaster dynamic deduction and intelligent disaster avoidance decision-making method and system based on a digital twin engine; a direct driving fluid simulation mine tunnel digital twin is generated through a geometric-topology-physical three consistency modeling method; an SPH technology is used to deduce the water inrush spreading process, a disaster state triple State(t) is extracted in real time in a discrete time step iteration mode and stored in a disaster evolution database; a comprehensive risk index γ(t) is used to quantify the risk, a time-varying weighted tunnel network is constructed, a weighted summation scalar quantization mechanism is used to convert the escape time and path risk double targets into a single target, and a D-K algorithm is used to dynamically solve the Pareto optimal disaster avoidance path; the digital twin is used as a carrier, State(t) is used to drive real-time coloring, and interactive three-dimensional dynamic deduction visualization is realized. The core algorithm of the application has the migration property of a general computing platform and is not limited to a specific software platform.

Description

Technical Field

[0001] This invention belongs to the fields of digital twin technology, disaster dynamic simulation, and intelligent emergency decision-making, and particularly relates to a method and system for dynamic simulation and intelligent disaster avoidance decision-making in mine water inrush disasters based on a digital twin engine. The following embodiments are implemented using Python scripts and the Blender open-source platform, but the core algorithms and technical framework are portable to other 3D engines such as Unity and Unreal Engine, and are not limited to a specific software platform. Background Technology

[0002] Mine water inrush disasters are characterized by their suddenness, rapid evolution, and uncertain impact range, making them a major type of disaster that restricts coal mine safety production and emergency rescue efficiency. Statistics show that in the past two decades, there have been approximately 250 coal mine water inrush accidents in my country, resulting in over 1,700 deaths and direct economic losses exceeding 35 billion yuan. The existing emergency rescue system faces common bottlenecks in the following three areas:

[0003] Bottleneck 1 (Visual but not computable): Traditional modeling methods generate 3D models that lack fluid physics boundary conditions, and can only be used for geometric display but cannot directly drive fluid simulation. GIS 2D representation cannot describe true 3D spatial topology, and sophisticated commercial software is costly and has poor reusability.

[0004] Bottleneck 2 (Delta from Simulation to Decision): Existing flood inrush simulation results are mostly output as numerical fields or animated displays, which cannot directly drive path decisions. The offline mode of simulation before planning causes decision results to lag significantly behind the evolution of the disaster, making it difficult to play a role during the golden rescue period.

[0005] Bottleneck 3 (failure of static assumptions): Existing disaster evacuation route planning methods implicitly assume that the roadway capacity remains unchanged, which cannot reflect the dynamic decay of route accessibility over time under the condition of sudden water disaster; most methods take a single distance as the optimization objective and ignore the multi-objective trade-off between escape time and risk. Summary of the Invention

[0006] The core technological contribution of this invention lies in: constructing a disaster state triplet State(t)={h(t),v(t),γ(t)} as a real-time data link between simulation and decision-making, and using the disaster evolution database as a shared bus to achieve real-time closed-loop coupling of the four modules of modeling, inference, decision-making and visualization, thereby fundamentally solving the above three major bottlenecks.

[0007] To address the aforementioned technical problems, this invention provides a method for dynamic simulation and visualization of mine flooding evacuation routes, specifically including the following steps:

[0008] Step S1: Perform automated modeling of the mine's 3D tunnel system: Based on a parametric script-driven 3D modeling engine, automatically reconstruct the coordinates of the tunnel center traverse points into the 3D geometric model.

[0009] Step S2: Establish a simulation model for water inrush propagation based on spatiotemporal dynamics: Establish a water inrush simulation model based on fluid dynamics technology, and use the three-dimensional geometric model generated in step S1 as the fluid constraint boundary to realize the propagation simulation of the entire water inrush process.

[0010] Step S3: Calculate the optimal evacuation route: Taking into account various travel factors, use the DK algorithm to calculate the top N optimal evacuation routes from disaster victims to multiple emergency exits.

[0011] Step S4: Establish a disaster evacuation demonstration model: Based on the optimal disaster evacuation path calculated in step S3, establish the movement trajectory of personnel and drive the human model to move along the trajectory to realize the visualization of emergency disaster evacuation simulation in the event of a disaster.

[0012] Furthermore, in this invention, step S1 specifically includes the following steps:

[0013] Step S11: Collect the coordinates of the center traverse point and cross-sectional parameters of the roadway and store them in the database. At the mine site, underground measurements are conducted using a high-precision total station, or key geometric data is extracted directly from existing digital CAD drawings. The core data collected includes the three-dimensional spatial coordinates of the roadway center traverse point and the corresponding roadway cross-sectional morphological parameters. These cross-sectional parameters include not only the basic roadway width and height, but also key geometric indicators such as the calculated height of the vertical walls and the arch height. To achieve efficient automated calling and management of subsequent Python scripts, all collected heterogeneous data must be cleaned and formatted, and uniformly stored in a structured PostgreSQL database. By establishing a basic information database for mine roadways, the classification management and rapid indexing of massive traverse point data and attribute parameters are achieved, providing standardized data support for 3D reconstruction.

[0014] Step S12: Use Python to write a modeling script, run it on the Blender application interface, and call the data in the database to generate the tunnel cross-sectional profile.

[0015] Step S13: Seamlessly connect and generate single tunnel models and system tunnel models using point-to-surface and surface-to-volume methods. Specifically, through a programmed iterative algorithm, a series of subdivided triangular or quadrilateral geometric planes are automatically generated between two spatially adjacent cross-sectional contour lines based on their corresponding feature point indices. These geometric planes are then seamlessly spliced ​​and closed to construct a single tunnel solid model with volumetric properties. Based on this principle, the system establishes an automated iterative generation mechanism along the extension direction of the tunnel's central guide point: first, the first tunnel model is generated using the first and second cross-sections; then, the second tunnel model is generated using the second and third cross-sections, and so on, achieving continuous construction from beginning to end until all guide points of the tunnel have been traversed, completing the three-dimensional solid generation of the entire tunnel.

[0016] Step S14: Utilize Blender's built-in Boolean splitting module to automatically trim and merge overlapping or intersecting meshes, and employ interpolation algorithms to densify special roadways with large curvature changes to prevent model distortion. Since all individual roadway models are generated based on real geographic coordinates, they naturally possess accurate and fixed spatial relationships in virtual 3D space. After generating all individual roadway models for the entire mine, to address the inevitable model overlap, geometric interpenetration, or patch interference issues at roadway intersections (such as three-way junctions and crossroads), Boolean operations are used to trim the topology of intersecting roadways. This automatically removes redundant overlapping meshes and reconstructs the topology at intersections, ensuring that all intersecting roadways maintain true connectivity in their physical structure, eliminating geometric errors, and providing rigorous and connected physical boundary conditions for subsequent fluid dynamics simulations. To address the problem that the distribution of guide points in the original acquired data is too sparse, leading to an inability to accurately reflect the smooth morphology of roadways at sharp turns, slope changes, or undulating sections, and even resulting in "polyline" distortion, this invention introduces interpolation densification technology. By employing interpolation algorithms (such as Bézier curve interpolation or spline interpolation), the system can automatically identify special road sections with large curvature changes in traverse points, and automatically calculate and add intermediate control points between adjacent traverse points to densify the traverse points. This effectively smooths the turning angles of the roadway, solves the distortion problem of the model in the deformation area, and ensures the smooth transition and accurate restoration of the three-dimensional geometry of the tunnel.

[0017] In a further step of this invention, in step S12, a Python script is used in Blender with each set of central guide points as the origin of spatial coordinates. Based on the cross-sectional dimension parameters of the tunnel, the relative coordinates of each key feature point (such as arch feature points, wall base points, etc.) on the cross-section are calculated. Then, these feature points are connected into a closed curve in a specific order to generate the cross-sectional profile of the tunnel.

[0018] Furthermore, in this invention, step S2 specifically includes the following steps:

[0019] S21: Set the water source parameters, including the location, velocity, and direction of the water inrush, to simulate a disaster occurring at any location.

[0020] S22: Set the generated 3D model of the tunnel system as an "obstacle" as the fluid constraint range.

[0021] S23: Calculate the water flow diffusion process based on smooth particle fluid dynamics technology and cache the results. Render the water body using the rendering engine and call the animation function curve to dynamically demonstrate the disaster situation.

[0022] Furthermore, in this invention, step S3 specifically includes the following steps:

[0023] Step S31: Multiply the actual length of the tunnel by the accessibility coefficient to obtain the equivalent length weight, and use it as the path weight. The accessibility coefficient includes the tunnel type influence factor, the obstacle influence factor, and the slope and terrain consumption coefficient.

[0024] The calculation process for equivalent length is as follows:

[0025] First, the ease of passage through the tunnels is measured. Since different disaster avoidance and escape factors have varying degrees of impact on the speed of evacuation, the calculation of the corresponding influence coefficients for these factors can be based on the influence coefficient of a particular factor on the speed of evacuation. Its mathematical expression is as follows:

[0026]

[0027] In the formula, For a certain section of alleyway; Under the influence of certain factors, people pass through the tunnel The time spent; Under the condition of no influencing factors, personnel pass through the tunnel The time spent. By substituting real-world data into the above formula, the influence coefficient of a certain influencing factor on the speed of disaster evacuation can be calculated. This refers to the difficulty coefficient of passage through the alley.

[0028] Then, the equivalent length of the roadway is defined as the product of the influence coefficients of different influencing factors and the actual length of the roadway. Let the coordinates of node A be... The coordinates of node B are ,but Section length of tunnel Represented as:

[0029]

[0030] The formula for calculating the slope of a tunnel section is as follows:

[0031]

[0032] In the formula, For alleyway The slope (°); when , hour, .

[0033] Let the equivalent length of the tunnel be Its mathematical expression is as follows:

[0034]

[0035] In the formula, Personnel passing through the tunnel under different influencing factors The difficulty level. Representing the The correction coefficient for the common influence factor is obtained through field simulation or experiment.

[0036] Introducing equivalent length instead of object length as a weight in path search has the following significance: In a mine disaster environment, the shortest physical path is not necessarily the fastest. Equivalent length, by converting factors such as slope and obstacles into length units, enables the algorithm to identify some slightly longer but smoother and safer paths. It also unifies various influencing factors with different properties (such as angle and drag coefficient) into length weights, allowing the DK algorithm to perform multi-objective path search within the same directed weighted graph. Furthermore, equivalent length can be dynamically adjusted based on the results of subsequent fluid dynamics simulations, thereby achieving real-time optimization of disaster avoidance paths as the disaster spreads.

[0037] Step S32: Establish a directed weighted graph G(V,E,W) of the mine roadway network, where V is the set of all nodes in the graph, representing the key nodes in the mine roadway; E is the set of all edges in the graph, representing the connection relationship between nodes; and W is the set of equivalent length weights, representing the equivalent length of each edge.

[0038] Step S33: Using the DK algorithm, execute the process from source node V in the directed weighted graph. m To multiple emergency evacuation target nodes V jk The path search is performed by using an edge-removal iteration strategy to obtain the top k shortest paths for each exit branch.

[0039] Step S34: Sort all candidate paths in ascending order of equivalent length cost, and select the shortest path among the top N shortest paths in the global ranking as the final optimal disaster avoidance path result.

[0040] Furthermore, in step S31 of this invention, the method for determining the passage difficulty coefficient includes the following steps:

[0041] S311. By obtaining the ratio of personnel passage time under certain influencing factors to that without influencing factors through on-site simulation, the difficulty coefficient of passage in the alleyway is calculated.

[0042] S312. Calculate the actual slope angle of the roadway using the coordinate difference of the center guide points of adjacent roadways, and determine the terrain consumption coefficient accordingly.

[0043] Furthermore, in step S33 of this invention, for each target node V jk Execute Dijkstra's algorithm to compute the result from the source node V. m To the target node V jk The shortest path is calculated by obtaining the node sequence along the path and storing it in set N, while the edge information of the path is stored in set M. Based on the fluid physics simulation results, the set of edges where the water level at time T is greater than the critical passage height of the tunnel is identified. During the DK search, edges that simulate path blockage caused by a real disaster are deleted first, and the shortest path is recalculated. After each edge deletion, the Dijkstra algorithm is executed to calculate the new path and the result is stored in set L. Then, for each target node V... jk Repeat the above steps to generate the first k shortest paths.

[0044] Furthermore, in this invention, step S4 specifically includes the following steps:

[0045] Step S41: Establish personnel movement trajectories based on the calculated optimal disaster avoidance path.

[0046] Step S42: Generate a human figure model in Blender and use the timeline function to drive the model along the optimal trajectory to visually describe the escape process. By coupling the dynamics of personnel escape with the simulation of sudden water spread, the impact of water flow status on personnel escape at a certain moment can be viewed intuitively, thereby adjusting the disaster avoidance path based on real-time feedback.

[0047] This invention also provides a dynamic simulation and evacuation route optimization visualization system for mine flooding, built on the Python and Blender open-source platform, which executes the aforementioned dynamic simulation and evacuation route optimization visualization method for mine flooding, specifically including the following modules:

[0048] The geometric model generation module is used to read the coordinates of the center traverse points and cross-sectional parameters of the tunnel stored in the database through Python scripts, automatically generate the tunnel cross-sectional contour in Blender, and seamlessly connect adjacent cross-sectional contours in the manner of "point to surface, surface to volume" to construct a three-dimensional tunnel geometric model with real topological relationships.

[0049] The dynamic diffusion simulation module for sudden water flow is used to set the three-dimensional tunnel geometric model constructed by the geometric model generation module as the fluid constraint boundary, and calculate the diffusion process of water flow within the constraint boundary based on smooth particle fluid dynamics technology, generating disaster simulation data for dynamic demonstration.

[0050] The optimal disaster avoidance path search module is used to establish a directed weighted graph with equivalent length as the weight, and use the DK algorithm to solve for the top N optimal disaster avoidance paths from the source node at the current location of the disaster victim to multiple emergency exit target nodes.

[0051] The personnel escape simulation module is used to establish the personnel movement trajectory based on the node coordinates of the optimal disaster avoidance path calculated by the optimal disaster avoidance path search module, generate a character model in Blender, and call the timeline function to drive the character model to move along the movement trajectory.

[0052] Furthermore, in the optimal disaster avoidance path search module of the present invention, the equivalent length is determined by the product of the actual spatial length of the tunnel and at least one of the tunnel type influence factor, obstacle influence factor and terrain consumption coefficient.

[0053] The beneficial effects of this invention are:

[0054] (1) The 3D modeling engine is driven by parametric scripts, which realizes the automated construction of tunnel models. Compared with the traditional manual modeling method, the modeling efficiency is significantly improved. At the same time, Boolean operation and interpolation algorithm are used to ensure the geometric accuracy and topological connectivity of the model, and quickly generate a 3D tunnel model with real undulations and accurate topological relationships.

[0055] (2) A water inrush simulation model was established based on smooth particle fluid dynamics technology, realizing the simulation of water flow diffusion process based on physical mechanism. Compared with the traditional empirical formula numerical simulation method, it can more realistically reflect the motion law of water flow in complex tunnels and provide an intuitive spatiotemporal dynamic demonstration.

[0056] (3) By introducing the concept of equivalent length to comprehensively consider multiple influencing factors such as road type, obstacles, and slope, and by using the combination of Dijkstra's algorithm and K-shortest path algorithm to solve multiple paths to multiple emergency exits, the calculated disaster avoidance paths are more in line with actual traffic conditions and effectively avoid path congestion problems.

[0057] (4) Based on the Python and Blender open source platform, the system development cost has been greatly reduced and the integration has been improved. It provides a scientific and intuitive auxiliary decision-making basis for emergency rescue of mine water inrush disasters and effectively reduces casualties. Attached Figure Description

[0058] The specific embodiments of the present invention will be further explained below with reference to the accompanying drawings.

[0059] Figure 1 This is a flowchart illustrating the overall steps of the method of the present invention.

[0060] Figure 2 Flowchart for modeling the consistency of the three elements in a digital twin of a mine shaft.

[0061] Figure 3 The flowchart shows the calculation of the time-varying weight function ωe(t) and the multi-objective DK optimization.

[0062] Figure 4 This is a system architecture block diagram of the present invention. Detailed Implementation

[0063] Example 1

[0064] Combination Figure 1 and Figure 2 As shown, the dynamic simulation and evacuation route optimization visualization method for mine flooding in this embodiment utilizes Python scripts to drive the Blender open-source platform and, based on measured data stored in a PostgreSQL database, achieves rapid and automatic construction of a three-dimensional geometric model of the mine tunnel system. This method, through standardized data interfaces and parameterized geometric generation logic, realizes automatic reconstruction from two-dimensional traverse data to a three-dimensional solid model. The specific implementation steps are as follows:

[0065] Step S1: Perform automated modeling of the mine's 3D tunnel system: Based on a parametric script-driven 3D modeling engine, automatically reconstruct the coordinates of the tunnel center traverse points into the 3D geometric model.

[0066] In this embodiment, step S1 preferably includes the following steps:

[0067] Step S11: Use a total station or 3D laser scanner to collect the 3D coordinates (X,Y,Z) and cross-sectional parameters (net width W, vertical wall height H, arch parameters, etc.) of each traverse point, and store them in a PostgreSQL database (including a tunnel index table, traverse point data table, and cross-sectional attribute table).

[0068] In this embodiment, the acquisition and storage of modeling data is the foundation of the entire automated modeling process. At the mine site, technicians use total stations and 3D laser scanners to conduct underground measurements, or directly extract key geometric information from existing high-precision digital CAD drawings. The core acquisition target of the system is the central guide point controlling the direction of the roadway. For each guide point, its three-dimensional spatial coordinates (X, Y, Z) in the mine's absolute coordinate system must be acquired to ensure the model possesses realistic geographical location attributes. To accurately reflect the undulations and turns of the roadway, the acquisition density of the guide points needs to be dynamically adjusted according to the actual curvature of the roadway. The sampling frequency can be appropriately reduced in straight sections to decrease data redundancy, while high-density acquisition is required at curves, intersections, or slope change points to ensure accurate morphology. Simultaneously, the cross-sectional morphological parameters corresponding to each section of the roadway must be acquired or defined synchronously. Taking the "semi-circular straight-wall arched roadway" in this embodiment as an example, the cross-sectional parameters to be acquired include: the roadway's net width W, the straight-wall design height H, the arch height parameters, and the lining thickness under different support methods, etc. To support efficient batch calls from subsequent Python scripts, the collected multi-source heterogeneous data needs to be cleaned, validated, and formatted before being uniformly stored in a PostgreSQL relational database. The database design includes a tunnel index table, a traverse point data table, and a cross-sectional attribute table. The tunnel index table stores the unique identifier (ID) and metadata of each tunnel; the traverse point data table stores data containing sequence indexes and precise 3D coordinates in an ordered sequence format; and the cross-sectional attribute table specifically stores geometric construction parameters defining cross-sectional type, width, height, and the number of circular arc fitting segments.

[0069] S12: Geometry Generation and Topology Modeling. Parametric scripts sweep along the centerline cross-section to generate 3D geometry for each segment; a graph structure G=(V,E) (V is the intersection node, E is the tunnel segment) is constructed to achieve automatic topology recognition; Boolean operations are used to eliminate geometric overlap at intersections; cubic spline interpolation is used to densify the traverse points of sparse segments to prevent distortion caused by polyline transformation.

[0070] In this embodiment, preferably, in step S12, a Python script is used in Blender to calculate the relative coordinates of each key feature point (such as arch feature points, wall base points, etc.) on the cross-section based on the cross-sectional size parameters of the tunnel, with each set of center guide points as the origin of the spatial coordinates. Then, these feature points are connected into a closed curve in a specific order to generate the cross-sectional profile of the tunnel.

[0071] In this embodiment, a Python script is used as the core driver to interact with the backend database through Blender's application programming interface (API), read the geometric parameters of each tunnel segment, and automatically calculate and generate a two-dimensional cross-sectional profile in three-dimensional space. For the most typical semi-circular straight-walled arched tunnel in a mine, the specific implementation logic is as follows:

[0072] Step S121: Establish an independent local modeling coordinate system. The script automatically reads the 3D coordinates of the currently processed center traverse point and the next adjacent traverse point, and determines the direction of the tunnel's extension tangent by calculating the vector difference between the two points. Subsequently, the algorithm automatically constructs a 2D plane perpendicular to this tangent direction. All subsequent cross-sectional feature points will be located and solved on this plane, thereby ensuring that the generated cross-section is always perpendicular to the tunnel's orientation and avoiding geometric distortion.

[0073] Step S122: Locate key feature points of the straight wall section. The system reads the net width parameters of the tunnel and the height parameters of the straight wall from the database. Using the current center traverse point as the reference origin, the algorithm first calculates the coordinates of the two wall foot points at the bottom of the straight wall: by offsetting the coordinates of the center traverse point to the left and right by half the net width value of the tunnel in the horizontal axis direction of the local plane, the spatial positions of the left and right wall foot points are accurately locked. Next, based on the straight wall height parameters, the system extends a corresponding distance upward from these two wall foot points along the vertical axis of the local plane to calculate the coordinates of the two arching points at the top of the straight wall. These two arching points are both the end point of the straight wall and the starting point of the upper semi-circular arch, thus completing the geometric definition of the lower rectangular frame of the tunnel.

[0074] Step S123: Perform discretization fitting calculations for the feature points of the upper semicircular arch. To accurately represent the smooth arch structure in a computer 3D environment, the system adopts a piecewise fitting strategy. The program first automatically calculates the radius of the semicircular arch based on the width of the tunnel and reads the preset arc fitting segment parameters. Taking the midpoint of the line connecting the two arch starting points as the center, the system divides the semicircular arc into several small fan-shaped regions through a cyclic iterative algorithm. At each division angle, the program automatically calculates the relative coordinates of the corresponding arc feature nodes using trigonometric function logic. This process starts from the arch starting point on one side and calculates point by point along the arc trajectory until it reaches the arch starting point on the other side, thereby obtaining a series of dense discrete feature point clouds that can realistically fit the semicircular shape.

[0075] Step S124: Construct a closed cross-sectional profile. After the coordinates of all key nodes (including the wall base points, the arch starting point, and all intermediate feature points on the arc) are calculated, the Python script uses a line-drawing function to connect these discrete spatial points sequentially in a counter-clockwise or clockwise topological order. The system first connects the two wall base points at the bottom to form a closed bottom edge, then connects upwards to the arch starting point, and then sequentially connects all the fitted nodes on the arc, finally returning to the starting side. This generates a closed, geometrically defined tunnel cross-sectional profile loop in three-dimensional space, preparing the geometry for subsequent entity generation.

[0076] S13: Physical boundary assignment and three-consistency verification. Assign friction coefficient μ (typical value 0.3~0.5), damping coefficient c, and permeability coefficient k to the tunnel wall; complete the geometric-topological-physical three-consistency verification, and generate a digital twin of the tunnel that can directly drive SPH simulation, fundamentally solving the problem of "visible but not calculable".

[0077] In this embodiment, the system establishes an automated loop mechanism to process each pair of adjacent cross-sectional contours along the tunnel's direction. Assuming the current processing is of the Kth and K+1th cross-sections, since both are generated based on the same parameters, they have identical vertex counts and index orders. The algorithm automatically matches each vertex on the Kth cross-section with its corresponding vertices at the same index and adjacent vertices on the K+1th cross-section, establishing topological connections and generating a series of quadrilateral or triangular facets. These facets enclose the lateral space between the two cross-sections, thus forming a "single-section tunnel solid model" with volumetric properties. The system continues along the extension direction of the traverse points, generating the first solid section using the 1st and 2nd cross-sections, then the second solid section using the 2nd and 3rd cross-sections, and so on. Because adjacent solid sections share a central cross-sectional contour, the generated model is geometrically strictly continuous, without gaps or breaks, achieving seamless connection from beginning to end. After traversing all traverse points of the tunnel, the three-dimensional solid model of the entire tunnel is completed.

[0078] Step S14: Use Blender's built-in Boolean splitting function module to automatically trim and merge overlapping or intersecting mesh parts, and use interpolation algorithms to refine special local tunnels with large curvature changes to prevent model distortion.

[0079] In this embodiment, specifically after generating models of all individual roadways in the entire mine, the system incorporates an automated processing module to ensure physical connectivity, addressing the inevitable model overlap at roadway intersections (such as three-way junctions and crossroads). Since all roadways are generated based on real coordinates, the model entities at intersections will overlap in volume. The system utilizes Blender's built-in Boolean operation module to automatically detect colliding roadway objects in the scene. Once overlap is detected, the script triggers a Boolean "union" operation, automatically calculating the geometric boundaries of the overlapping area, removing redundant internal faces and meshes, and merging the vertices at the intersections. This step requires no manual intervention, ensuring that all intersecting roadways are topologically integrated, forming a realistic physically connected network, providing a smooth internal cavity environment for subsequent water flow simulation. Simultaneously, to address the potential sparse distribution of traverse points in the original measured data—that is, excessively large sampling intervals in straight sections but insufficient data at minute turning points—the system integrates an intelligent interpolation optimization module. During the data reading phase, the system automatically calculates the rate of change of direction between adjacent traverse points. Once a change in direction exceeding a preset threshold is detected over a short distance, or if there are too few sampling points in a long straight section, the system will automatically initiate an interpolation algorithm (such as cubic spline interpolation). This algorithm automatically calculates and inserts several densified control points between two original traverse points, and smoothly transitions the coordinates of these points according to the forward and backward orientation. In this way, the system significantly improves the geometric subdivision density of the model at curves and undulations without changing the original measurement accuracy, effectively eliminating the "polyline" or "angular" distortion caused by insufficient data sampling, and ensuring high-fidelity restoration of both visual and physical properties of the mine tunnel model. Through the above steps, this invention can automatically construct a three-dimensional mine tunnel system model with real geographical coordinates, accurate cross-sectional morphology, and complex connectivity relationships.

[0080] Step S2: Establish a simulation model for water inrush propagation based on spatiotemporal dynamics: Establish a water inrush simulation model based on fluid dynamics technology, and use the three-dimensional geometric model generated in step S1 as the fluid constraint boundary to realize the propagation simulation of the entire water inrush process.

[0081] In this embodiment, step S2 preferably includes the following steps:

[0082] S21: Input the spatial coordinates of the water inrush point, the initial water inrush velocity, the water inrush volume Q (m³ / s), and the initial direction as driving parameters for the SPH simulation.

[0083] S22: Key simulation parameters: particle resolution 0.5m, time step Δt = 0.02s, smoothing kernel radius 1.0m, kinematic viscosity coefficient 1.0 × 10⁻⁶ -6m² / s. The calculation results are cached in real time, and the Cycles rendering engine is used to perform physical rendering of water refraction / reflection.

[0084] S23: Water Level Evolution Iteration: For each roadway segment e, the water level time series change is iteratively calculated using the following explicit finite difference scheme:

[0085]

[0086] wet cross-sectional area A e Calculation of (h): Take A for rectangular cross-section e =W·h;A is taken as the semi-circular straight-walled arched section. e =W·min(h,H wall )+A arch A arch This represents the cross-sectional area of ​​the arch at the corresponding water level (obtained by integrating the central angle). This formula is the Saint-Venant equation. The explicit finite difference approximation under the quasi-static assumption of low Froude number (Fr<<1) meets engineering requirements in terms of accuracy during the initial spreading stage of mine water inrush (flow velocity is usually below 0.5m / s).

[0087] S24 Comprehensive Risk Index Calculation (LaTeX Formula-2). For each roadway segment e:

[0088]

[0089] Typical parameters: H cr =0.5m, V cr =1.2 m / s, β=0.4 (calibrated by field experiment). γ e (t) Simultaneously incorporating both water level and flow velocity, it can distinguish complex working conditions that single-factor models cannot differentiate, such as "high water level and low flow velocity" (relatively safe, with a lower γ value) and "low water level and high flow velocity" (high risk, with a higher γ value).

[0090] Write the S25 disaster evolution database. Set {t, tunnel ID, h} e (t),v e (t),γ e (t),T arr,e The six-tuple is written to the disaster evolution database in real time, where T arr,e From the current water level h e (t) Extrapolation yields T arr,e =t+(H cr -h e (t))·A e / Q in,e .

[0091] Step S3: Calculate the optimal evacuation route: Taking into account various travel factors, use the DK algorithm to calculate the top N optimal evacuation routes from disaster victims to multiple emergency exits.

[0092] In this embodiment, preferably, combined with Figure 3 As shown, step S3 specifically includes the following steps:

[0093] Step S31: Multiply the actual length of the tunnel by the accessibility coefficient to obtain the equivalent length weight. The accessibility coefficient includes the tunnel type influence factor, the obstacle influence factor, and the slope and terrain consumption coefficient.

[0094] For each tunnel section e:

[0095]

[0096] Due to the influence of flow velocity, it has already been in γ e (t) is used to avoid double counting. Coefficient calibration: α1 typical range 1.5~3.0; α3=1 (based on slope angle ρ). e (Geometric determination).

[0097] In this embodiment, preferably, the method for determining the passage difficulty coefficient in step S31 includes the following steps:

[0098] S311. By obtaining the ratio of personnel passage time under certain influencing factors to that without influencing factors through on-site simulation, the difficulty coefficient of passage in the alleyway is calculated.

[0099] S312. Calculate the actual slope angle of the roadway using the coordinate difference of the center guide points of adjacent roadways, and determine the terrain consumption coefficient accordingly.

[0100] Specifically, in this embodiment, the key factors affecting the speed of personnel evacuation underground are first analyzed and extracted, mainly including the spatial distribution of roadways, internal obstacles, and real-time wind speed. Through on-site simulation experiments, the time it takes for personnel to traverse a roadway under certain influencing factors is obtained and compared with the travel time without influencing factors, thereby calculating a coefficient reflecting the ease of passage through the roadway. For example, specific ease-of-passage coefficients are assigned to different types of roadways such as the main track roadway, the main return airway, and the working face, as well as local obstacles such as air doors and belt conveyors.

[0101] In this embodiment, to accurately represent the impact of undulating tunnel systems on the physical energy required for escape, a terrain consumption coefficient is introduced to describe the influence of slope on the speed of passage. The actual slope angle of a tunnel segment is calculated using the coordinate difference between the center guide points of two adjacent tunnels. As the tunnel slope angle increases, the terrain consumption coefficient value for that segment is correspondingly increased, thus enabling the calculation of disaster escape routes to distinguish the difference in escape difficulty between flat and steep sections.

[0102] In this embodiment, the "equivalent length" is defined and calculated as the core weight for disaster avoidance path search. This value is equal to the product of the actual spatial length of the roadway segment and all the above-mentioned accessibility coefficients and terrain consumption coefficients. Based on the roadway type in the local mine, the influence coefficient of roadway type on equivalent length (as shown in Table 1) and the influence coefficient of obstacles on roadway equivalent length (as shown in Table 2) are obtained.

[0103] Table 1:

[0104] 1 Tracks, conveyor belt shafts, and air intake shafts 1.00 2 General Return Lane 1.25 3 working face 1.30 4 Main intake air-mining area connecting road 1.30 5 intake-return air connection roadway in mining area 1.22 6 Impassable alleyways ∞

[0105] Table 2:

[0106] 1 track 1.10 2 Belt conveyor 1.10 3 Fengmen, Fengqiao 1.00

[0107] In this way, physical distance is transformed into a travel cost that better meets the actual needs of escape, ensuring that the calculated path is not only the shortest in distance, but also the optimal solution in terms of escape time or physical exertion.

[0108] S32: Bi-objective weighted summation scalarization.

[0109]

[0110] Constraint: Time feasibility: ∀e∈P, the time when personnel arrive at e satisfies t e <T arr,e Spatial connectivity: Each segment of the path is physically connected at the corresponding time; Risk controllable: R path (P,t)<γ th =0.9. The default values ​​are λ1=0.6 and λ2=0.4; administrators can adjust the λ ratio in real time to balance time and safety, and the system will automatically recalculate.

[0111] S33: DK algorithm execution. On a time-varying weighted directed graph with J as a single weight, for each target exit node V... jk Execute Dijkstra's algorithm to obtain the first optimal path, store the path node sequence in set N, and the edge information in set M; by temporarily deleting the edge with the highest cost on the optimal path and iterating (k times, where k is a preset positive integer), obtain the first k suboptimal paths in turn and store them in set L; after traversing all target nodes, sort all candidate paths in set L according to the J value, and output the first N globally optimal paths.

[0112] S34: Real-time path replanning. Within each time step Δt = 0.02s, the trigger condition is checked according to the above formula for triggering replanning: If any segment γ on path P*(t)... e (t)≥γ th=0.8, or the estimated escape time for the path exceeds the earliest arrival time of the disaster, then a replanning is triggered—excluding γ. e After finding an impassable edge with edge G(t)≥1, re-execute the DK algorithm on the updated graph G'(t).

[0113] Specifically, in this embodiment, Dijkstra's algorithm is first used to calculate and store the first shortest path from the source node to a target node. Then, by temporarily removing edges from this path and iterating again, the first k shortest paths under this branch are calculated sequentially. This process is repeated until all target nodes have been traversed. Finally, a function is used to filter and select the top N shortest paths globally as the final optimal disaster avoidance path results, as shown in the table below.

[0114] 1 52 14→13→42→41→40→39→38→37→36→35→34→33→32→31→30→29→28→27→26→25→24→156→155→52 2911.56 2 52 14→13→42→41→40→39→38→37→36→35→34→57→56→55→54→53→52 2913.34 3 1 14→13→12→11→10→9→8→7→6→5→4→3→2→1 3348.59

[0115] Step S4: Interactive 3D Visualization and Deduction Driven by Digital Twin

[0116] In this embodiment, step S4 preferably includes the following steps:

[0117] S41: Motion trajectory generation. Extract the coordinate sequence of each node of the optimal path {P0, P1, ..., P...} n}, using cubic spline interpolation to generate continuous smooth motion trajectories, according to v person =1.2m / s to calculate the timestamp, and obtain the dataset {(x i ,y i ,z i ,t i )}.

[0118] S42: Three-level real-time shading. Using γ e (t) Drive: <0.3 semi-transparent gray (safe), 0.3~0.8 orange (high risk), ≥0.8 blue (impassable); Red directional arrow for water flow leader; Supports click-to-pop up {h e ,v e ,γ e ,T arr,e Real-time panel.

[0119] S43: Timeline Synchronization Demonstration. Keyframes of the character model are synchronized with the State(t) timeline; supports pausing at any frame to view the flood coverage area and the safe time window (T) of the current optimal path. arr Visualization.

[0120] Example 2

[0121] Combination Figure 4As shown, the mine flood dynamic simulation and evacuation route optimization visualization system of this embodiment is built on the Python and Blender open-source platform, and executes the mine flood dynamic simulation and evacuation route optimization visualization method in Embodiment 1, specifically including the following modules:

[0122] The geometric model generation module is used to read the coordinates of the center traverse points and cross-sectional parameters of the tunnel stored in the database through Python scripts, automatically generate the tunnel cross-sectional contour in Blender, and seamlessly connect adjacent cross-sectional contours in the manner of "point to surface, surface to volume" to construct a three-dimensional tunnel geometric model with real topological relationships.

[0123] The dynamic diffusion simulation module for sudden water flow is used to set the three-dimensional tunnel geometric model constructed by the geometric model generation module as the fluid constraint boundary, and calculate the diffusion process of water flow within the constraint boundary based on smooth particle fluid dynamics technology, generating disaster simulation data for dynamic demonstration.

[0124] The optimal disaster avoidance path search module is used to establish a directed weighted graph with equivalent length as the weight, and use the DK algorithm to solve for the top N optimal disaster avoidance paths from the source node at the current location of the disaster victim to multiple emergency exit target nodes.

[0125] The personnel escape simulation module is used to establish the personnel movement trajectory based on the node coordinates of the optimal disaster avoidance path calculated by the optimal disaster avoidance path search module, generate a character model in Blender, and call the timeline function to drive the character model to move along the movement trajectory.

[0126] In situations where the flow velocity changes drastically (such as steep slopes greater than 15° in the tunnel), linear weighting is insufficient to penalize the path in high-risk areas. This embodiment employs a time-varying weighting function in the form of a nonlinear Sigmoid exponential function:

[0127]

[0128] Where κ1 and κ2 are the exponential growth coefficients (typical values ​​κ1=2.5, κ2=1.5, calibrated through field experiments). This function in γ... e When (t) < 0.5, it approximates the linear form, and at γ e When (t)→1, a stronger path penalty is provided (exp(κ1)≈12.2 times the base value), causing high-risk road segments to be avoided by the system more actively. The bi-objective optimization model, DK algorithm solution process, and replanning mechanism are exactly the same as in Example 1, except that ω e The formula for calculating (t) is different.

[0129] Example 3: System Architecture and Data Bus Description

[0130] The system of this invention uses a disaster evolution database as its core data bus, and its four modules achieve the following real-time data interaction:

[0131] Digital twin modeling module: Writes the roadway's geometric topology physical baseline data (one-time initialization).

[0132] Dynamic simulation module for sudden flood disaster: Writes {h} every Δt = 0.02s. e ,v e ,γ e ,T arr,e (Continuous writing)

[0133] Intelligent disaster avoidance path decision-making module: Read State(t) → Update ω e (t) → DK optimization → Pushing the optimal path (continuous read / write)

[0134] Interactive visualization and deduction module: Read State(t) + path set → Shading and rendering + character animation (continuous reading)

[0135] The three-way real-time coupling ensures a millisecond-level dynamic response across the entire chain, from "change in disaster situation → change in weight → adjustment of path → update of screen". The typical system response delay is less than one simulation time step (0.02s), meeting the real-time requirements of the golden rescue period for sudden flood disasters.

[0136] Many specific details have been set forth in the foregoing description to provide a thorough understanding of the present invention. However, the above description is merely a preferred embodiment of the present invention, and the present invention can be implemented in many other ways different from those described herein. Therefore, the present invention is not limited to the specific embodiments disclosed above. Furthermore, any person skilled in the art can make many possible variations and modifications to the technical solutions of the present invention, or modify them into equivalent embodiments, using the methods and techniques disclosed above, without departing from the scope of the present invention. Any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention, without departing from the content of the present invention, shall still fall within the protection scope of the present invention.

Claims

1. A method for dynamic simulation and intelligent disaster avoidance decision-making in mine water inrush disasters based on a digital twin engine, characterized in that, Includes the following steps: Step S1: Construct a digital twin of the tunnel: Based on the coordinate data of the center traverse point and the cross-sectional morphology parameters of the tunnel, construct a digital twin of the tunnel that takes into account geometric accuracy, spatial topological connectivity and fluid physical boundary consistency, as a computable spatial basis for subsequent water inrush disaster simulation and path decision-making; Step S2: Dynamic simulation and real-time state extraction of water inrush disaster: Using the digital twin generated in step S1 as the fluid constraint boundary, the water inrush spread process is simulated based on fluid dynamics technology; within each simulation time step, the disaster state triplet State(t)={h(t),v(t),γ(t)} of each roadway section is extracted and output in real time, where h(t) is the water level height, v(t) is the water flow velocity, and γ(t) is the comprehensive risk index, and the simulation results are stored in the disaster evolution database; Step S3: Intelligent decision-making for multi-objective dynamic disaster avoidance routes: Based on the disaster status triplet output in real time in Step S2, a time-varying weighted road network driven by the disaster status is constructed; the disaster avoidance route planning is modeled as a multi-objective dynamic optimization problem with escape time and route risk as optimization objectives, and the optimal disaster avoidance route is solved under the constraints of time feasibility, spatial connectivity and risk controllability; when the route feasibility decreases, replanning is automatically triggered. Step S4: Interactive visualization simulation driven by digital twin: Based on the optimal disaster avoidance path output in step S3, extract the three-dimensional coordinate sequence of the path nodes and generate three-dimensional motion trajectory data of the human model; The digital twin is dynamically colored in real time using a disaster status ternary model, and the dynamics of personnel evacuation are synchronously coupled with the evolution of the disaster status over time to achieve three-dimensional interactive visualization simulation.

2. The method according to claim 1, characterized in that, In step S3, the multi-objective dynamic path optimization adopts a weighted summation scalarization mechanism: the escape time objective T is... esc (P,t) and path risk target R path (P,t) are weighted and merged into a single comprehensive cost function J(P,t). Under the meaning of minimizing J, the Dijkstra-K shortest path algorithm (DK algorithm) is used to perform multi-objective path search on the time-varying weighted road network. By adjusting the weight ratio between two objectives, the optimal path set with different escape time-risk trade-off characteristics on the Pareto front is generated, and the top N disaster avoidance paths with the best comprehensive cost are output.

3. The method according to claim 1, characterized in that, In step S1, the construction of the digital twin specifically includes: S11: Collect the three-dimensional coordinates (X,Y,Z) and cross-sectional morphology parameters of the center traverse point of the tunnel and store them in a structured database; S12: Based on a parametric script-driven 3D modeling engine, the engine sweeps the cross-section along the centerline with the central traverse point as the skeleton, automatically generating 3D geometry for each segment; at the same time, it abstracts the tunnel system into a graph structure model G=(V,E), where V is the intersection node and E is the tunnel segment, realizing automatic identification and connectivity verification of the topology of multi-branch interconnected tunnel networks. S13: Boolean operations are used to perform topological fusion of intersecting lanes to eliminate geometric intersections and volume overlaps, ensuring physical connectivity; cubic spline interpolation is used to densify sparse traverse point segments to prevent model distortion due to polyline modeling. S14: Assign physical boundary parameters such as friction coefficient μ, damping coefficient c, and permeability coefficient k to the tunnel wall, complete the geometric-topological-physical consistency verification, and generate a digital twin of the tunnel that can directly drive fluid dynamics simulation.

4. The method according to claim 1, characterized in that, Step S2 specifically includes: S21: Set the parameters of the water inrush source, including the spatial coordinates of the water inrush point, the initial water inrush velocity, the water inrush volume Q, and the initial direction; S22: The digital twin is set as a rigid fluid constraint boundary, and the water flow diffusion process is simulated using smoothed particle hydrodynamics. Key simulation parameters: particle resolution 0.5m, time step Δt=0.02s, smoothing kernel radius 1.0m, kinematic viscosity coefficient 1.0×10⁻⁶. -6 m² / s; S23: Within each simulation time step, perform iterative calculations of water level evolution for each roadway segment e: Q in,e (t) represents the flow rate into tunnel segment e at time t, Q out,e (t) represents the outflow rate, A e (h e (t) represents the wet cross-sectional area of ​​roadway segment e below the current water level he(t); A e (h) Calculation by section type: For rectangular sections, take A e =W·h, A is taken for the semi-circular straight-walled arched cross-section. e =W·min(h,H wall )+A arch (min(hH wall,0 )); S24: Calculate the comprehensive risk index γ for each roadway section e (t): Where H cr V is the critical water level threshold for passage. cr γ is the critical hazard velocity, β is the velocity hazard weighting coefficient; e (t)∈[0,1],γ e (t)=0 indicates complete safety, γ e (t)=1 indicates that the road is completely impassable; S25: {t, tunnel ID, h} e (t),v e (t),γ e The quintuple (t) is written to the disaster evolution database in real time, and the predicted disaster arrival time T is recorded synchronously. arr,e It provides real-time data-driven support for the path decision-making and visualization modules.

5. The method according to claim 2, characterized in that, In step S3, the multi-objective dynamic path optimization specifically includes: S31: Construct a time-varying weighted roadway network. Continuously read State(t) from the disaster evolution database and update the passage cost of each roadway segment according to the following time-varying weight function: Where L e Let α1 be the actual spatial length (m) of roadway segment e, and ρ be the risk weight coefficient. e The slope angle; when γ e When (t)=1, ω e (t)→+∞, this section of road is considered impassable. All coefficients were calibrated through on-site simulation experiments. S32: Establish a dual-objective weighted summation optimization model. The dual objectives of escape time and path risk are scalarized into a single comprehensive cost function: Escape time: Maximum risk level of the route: The set of feasible paths F(t) satisfies the following triple constraints: ① Time feasibility: T esc (P,t0)≤T arr ,e(t0)-t0; ② Spatial connectivity: Each segment on path P is physically connected at the corresponding time; ③ Risk controllability: R path (P,t)<γ th ; S33: Dynamic solution using the Dijkstra algorithm. On a time-varying weighted network with weight J, the Dijkstra algorithm is executed for each target exit node to obtain the first optimal path; by temporarily deleting the edge with the highest cost on the path and iteratively solving, the first k cost-increasing paths are generated; S34: Real-time replanning trigger and output. Path feasibility is checked within each simulation time step; when the trigger condition is met, the DK algorithm is re-executed on the updated graph G'(t) after excluding impassable edges; the set of the top N Pareto avoidance paths with the best overall cost is output.

6. The method according to claim 1, characterized in that, The influence coefficients in the time-varying weighting function were calibrated using the following field simulation experiment method: Where T e,affected Let t be the time (s) taken by personnel to traverse roadway segment e under the influence of the nth influencing factor. e,free Travel time (s) under no influencing factors; α n,e >0 indicates that passage becomes more difficult, α n,e =0 indicates no effect; slope influence coefficient α3=tan(ρ e The coordinates of adjacent traverse points are directly calculated.

7. The method according to claim 1, characterized in that, In step S4, the interactive visualization simulation specifically includes: S41: Extract key nodes P0, P1, ..., P from the optimal disaster avoidance path in sequence. n The three-dimensional coordinate sequence is used to generate a continuous smooth motion trajectory using cubic spline interpolation. The timestamps of each node are calculated according to the preset simulated escape speed, resulting in the motion trajectory dataset {(x i ,y i ,z i ,t i )}; S42: Using the disaster evolution database γ e (t) Drive real-time coloring of the digital twin tunnel: γ e (t) < 0.3 indicates a semi-transparent gray, and 0.3 ≤ γ e (t) < 0.8 indicates a semi-transparent orange color, γ e (t)≥0.8 indicates a semi-transparent blue color, and the water flow leader displays a red directional arrow; clicking on any alleyway will bring up {h e (t),v e (t),γ e (t),T arr,e Real-time information panel; S43: Synchronously couples the keyframes of the character model's motion with the disaster state(t) along the time axis, driving the digital twin to achieve a three-in-one interactive demonstration of disaster simulation and escape process; supports arbitrary frame pause, playback, interactive underground roaming, and T-based... arr,e The path safety time window is visualized based on this.

8. The method according to claim 5, characterized in that, In step S34, the real-time path replanning is triggered by the following conditions: Where P*(t) is the current optimal path at time t, γ th T is the risk trigger threshold. arr,min (t) represents the estimated time when the water flow first reaches the unsubmerged tunnel section in the entire network; replanning is triggered when either of the two conditions is met; during replanning, γ is excluded. e After finding an impassable edge with edge G(t)≥1, re-execute the DK algorithm on the updated graph G'(t).

9. The method according to claim 3, characterized in that, In step S12, a local coordinate system perpendicular to the tangent direction of the tunnel extension is constructed using each set of central guide points as the local coordinate origin. The relative coordinates of each key feature point on the cross section are calculated and connected in sequence to generate a closed cross section profile. For a semi-circular straight-walled arched tunnel, the discretization fitting accuracy of the semi-circular arch is no less than 12 angle segments. After the cross section profile is generated, it is seamlessly connected with adjacent cross sections in the manner of point-to-surface and surface-to-body construction to generate a single-section tunnel solid model.

10. A dynamic simulation and intelligent disaster avoidance decision-making system for mine water inrush disasters based on a digital twin engine, characterized in that: The method described in any one of claims 1 to 9, using the disaster evolution database as a four-module shared data bus, specifically includes: Digital twin modeling module: Based on the coordinate data of the center traverse point, it drives a 3D modeling engine and generates a digital twin of the tunnel through a geometric-topological-physical three-consistency modeling method; Output interface: Writes the reference data of tunnel geometry and topology to the disaster evolution database; The dynamic simulation module for sudden water inrush disasters uses a digital twin as the fluid constraint boundary and employs SPH technology to perform sudden water inrush propagation simulations. It calculates the temporal changes in water level for each roadway section according to the aforementioned water level evolution iterative equation and calculates the comprehensive risk index γ according to the comprehensive risk index formula. e (t); Output interface: Writes State(t) = {h} to the disaster evolution database in real time. e (t),v e (t),γ e (t)} and T arr,e ; Intelligent disaster avoidance path decision-making module: Input interface: continuously reads State(t) from the disaster evolution database; constructs a time-varying weighted road network according to the time-varying weight function, and performs dual-objective weighted summation DK optimization according to formulas -5 to 7; performs real-time verification of path feasibility and triggers replanning according to the real-time replanning trigger condition formula; Output interface: pushes the current optimal disaster avoidance path set to the visualization module; Interactive Visualization Simulation Module: Input Interface: Reads State(t) from the disaster evolution database and the optimal path pushed by the path decision module; Press γ e (t) Drives the digital twin to perform three-level real-time coloring, synchronously couples the disaster evolution with the dynamics of personnel escape along the time axis, realizes three-dimensional interactive visualization simulation, and supports arbitrary frame pause and parameter query.

Images