Rail transit fire emergency evacuation simulation analysis method and system

Abstract

The invention discloses a rail transit fire emergency evacuation simulation analysis method and system, and relates to the technical field of fire evacuation simulation analysis, and the method comprises the steps: collecting three-dimensional space structure data, fire scene parameters and initial passenger distribution data of a target rail transit station; constructing an evacuation simulation model of the intelligent body passengers; generating dynamic fire risk field data; coupling the dynamic fire risk field data with the evacuation simulation model, and constructing an evacuation simulation operation result; according to the method, the technical problems that in the prior art, the rail transit fire emergency evacuation simulation scene restoration is not accurate, the path decision is not combined with dynamic risk and personnel interaction, and consequently the safety and efficiency of emergency evacuation are insufficient are solved; the technical effects that accurate simulation of the evacuation process and optimization of the evacuation path in the rail transit fire scene are achieved, and the safety and efficiency of emergency evacuation are improved are achieved.

Description

Technical Field

[0001] This invention relates to the field of fire evacuation simulation analysis technology, specifically to a method and system for emergency evacuation simulation analysis in rail transit fires. Background Technology

[0002] With the rapid expansion of rail transit networks and the continuous growth of passenger flow, the safety and efficiency of emergency evacuation during fires at rail transit stations have become key issues in urban public safety. Existing rail transit fire evacuation simulation technologies largely rely on simplified spatial models and fixed risk parameters, making it difficult to accurately reproduce the three-dimensional spatial structural characteristics of stations, the dynamic development of fires, and personnel interaction behaviors. This results in significant discrepancies between simulation results and actual scenarios. Furthermore, traditional methods do not fully integrate dynamic factors such as real-time fire risk and personnel density distribution in path decision-making, and lack in-depth analysis of evacuation trajectories. This leads to insufficient optimization of the generated evacuation plans, making it difficult to effectively address emergency evacuation needs in complex fire scenarios, resulting in inaccurate evacuation guidance and incomplete safety hazard identification.

[0003] Existing technologies suffer from technical problems such as inaccurate simulation of emergency evacuation scenarios in rail transit fires, failure to incorporate dynamic risks and personnel interaction into route decisions, resulting in insufficient safety and efficiency in emergency evacuation. Summary of the Invention

[0004] This application provides a simulation analysis method and system for emergency evacuation in rail transit fires, which addresses the technical problems in existing technologies where the simulation of emergency evacuation scenarios in rail transit fires is inaccurate and path decisions do not incorporate dynamic risks and personnel interactions, resulting in insufficient safety and efficiency of emergency evacuation.

[0005] In view of the above problems, this application provides a simulation analysis method and system for emergency evacuation in rail transit fires.

[0006] The first aspect of this application provides a simulation analysis method for emergency evacuation in rail transit fires, the method comprising:

[0007] The system collects three-dimensional spatial structure data, fire scene parameters, and initial passenger distribution data of the target rail transit station. Based on the three-dimensional spatial structure data and the initial passenger distribution data, it constructs an evacuation simulation model for intelligent passengers. It loads the fire scene parameters to dynamically simulate the fire development process and generates dynamic fire risk field data. It couples the dynamic fire risk field data with the evacuation simulation model to drive the intelligent passengers to make path decisions and construct evacuation simulation results. Based on the evacuation simulation results, it identifies the trajectory of the intelligent passengers, generates a passenger individual trajectory dataset for dynamic path analysis, and constructs evacuation path optimization suggestions.

[0008] A second aspect of this application provides a simulation analysis system for emergency evacuation in rail transit fires, the system comprising:

[0009] The system includes a data acquisition module for collecting 3D spatial structure data, fire scene parameters, and initial passenger distribution data of the target rail transit station; an evacuation simulation model construction module for constructing an evacuation simulation model of the intelligent passenger based on the 3D spatial structure data and the initial passenger distribution data; a dynamic fire risk field data generation module for loading the fire scene parameters to dynamically simulate the fire development process and generate dynamic fire risk field data; an evacuation simulation result construction module for coupling the dynamic fire risk field data with the evacuation simulation model to drive the intelligent passenger to make path decisions and construct evacuation simulation results; and an evacuation path optimization suggestion construction module for identifying the trajectory of the intelligent passenger based on the evacuation simulation results, generating a passenger individual trajectory dataset for dynamic path analysis, and constructing evacuation path optimization suggestions.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] This method involves collecting 3D spatial structure data, fire scenario parameters, and initial passenger distribution data of the target rail transit station; constructing an intelligent passenger evacuation simulation model; dynamically simulating the fire development process by loading the fire scenario parameters, generating dynamic fire risk field data; coupling the dynamic fire risk field data with the evacuation simulation model to drive the intelligent passenger to make path decisions and construct evacuation simulation results; and using the evacuation simulation results to identify the trajectory of the intelligent passenger, generating individual passenger trajectory datasets for dynamic path analysis, and constructing evacuation path optimization suggestions. This achieves the technical effect of accurately simulating the evacuation process and optimizing evacuation paths in rail transit fire scenarios, improving the safety and efficiency of emergency evacuation. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A schematic flowchart of a simulation analysis method for emergency evacuation in rail transit fires, provided as an embodiment of this application;

[0014] Figure 2 This is a schematic diagram of the structure of a railway fire emergency evacuation simulation analysis system provided in an embodiment of this application.

[0015] Figure labeling: Data acquisition module 10, evacuation simulation model construction module 20, dynamic fire risk field data generation module 30, evacuation simulation operation result construction module 40, evacuation route optimization suggestion construction module 50. Detailed Implementation

[0016] This application provides a simulation analysis method and system for emergency evacuation in rail transit fires, which addresses the technical problems in existing technologies such as inaccurate simulation of emergency evacuation scenarios in rail transit fires, and the failure to incorporate dynamic risks and personnel interaction into route decisions, resulting in insufficient safety and efficiency of emergency evacuation.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] Example 1, as Figure 1 As shown, this application provides a simulation analysis method for emergency evacuation in rail transit fires, the method comprising:

[0019] Step S100: Collect three-dimensional spatial structure data, fire scene parameters, and initial passenger distribution data of the target rail transit station.

[0020] Specifically, a panoramic scan of the target rail transit station is performed using a 3D laser scanner to acquire 3D point cloud data. After denoising and registration processing, a 3D point cloud model is constructed. Subsequently, the station is traversed to complete building identification and extract key structural elements, which are then mapped onto the 3D point cloud model to form accurate 3D spatial structure data. Simultaneously, a fire risk assessment is conducted based on this 3D point cloud model, pre-setting multiple potential fire locations and performing fire source power analysis, plotting fire source power growth curves, and clarifying the station's environmental ventilation conditions through ventilation and smoke extraction signs. Combining the fire source power growth curves and ventilation conditions, fire scenario parameters suitable for multiple scenarios are constructed. In addition, based on the 3D spatial structure data, accessibility analysis is performed to delineate accessible spaces. Historical passenger flow data of the target station is introduced and mapped onto the accessible spaces for reasonable spatial distribution, ultimately generating initial passenger distribution data that closely matches the actual passenger flow situation.

[0021] Step S200: Based on the three-dimensional spatial structure data and the initial passenger distribution data, construct an evacuation simulation model for intelligent passenger agents.

[0022] Specifically, when constructing an evacuation simulation model of intelligent passenger agents based on the collected 3D spatial structure data and initial passenger distribution data, the 3D spatial structure data is first discretized into a raster to build an evacuation network topology map that accurately reflects the spatial layout of the station. Based on this topology map, the environment of the target rail transit station is initialized, dividing it into several basic environmental spatial units. Then, according to the basic environmental spatial units, the initial passenger distribution data is mapped one by one to the evacuation network topology map, constructing multiple intelligent passenger agents corresponding to the actual passenger distribution. Simultaneously, all intelligent passenger agents are traversed to configure individual attribute parameters, such as movement speed and evacuation priority, to ensure... Each agent possesses a unique behavioral foundation. Next, multiple simulation time steps are set, and agents are matched to these time steps. The coordinates of each agent in the basic environmental spatial unit are determined by combining individual attribute parameters. A global distance matrix is ​​constructed by calculating the Euclidean distance between coordinates, and the spatial proximity relationships between agents are analyzed. A neighborhood list is then built based on these relationships. Finally, according to the set simulation time steps, dynamic interaction analysis is performed on the agents in the neighborhood list to build a dynamic interaction network capable of simulating passenger avoidance and following behaviors. This network is then integrated into the model, ultimately forming a passenger evacuation simulation model containing multiple agents with dynamic interaction behavior rules.

[0023] Step S300: Load the fire scene parameters to dynamically simulate the fire development process and generate dynamic fire risk field data.

[0024] Specifically, the process begins by conducting physical field simulations based on pre-constructed fire scenario parameters. This involves building a fire physics engine capable of accurately reproducing the fire spread patterns and characteristics of smoke diffusion and temperature changes. The engine is then matched with the spatial layout of the target rail transit station to generate a computational domain network covering the entire station. Next, multiple potential ignition locations are mapped one by one into the computational domain network. The engine drives the simulation of the entire fire process from ignition to spread, capturing in real-time changes in smoke diffusion range, concentration distribution, and temperature gradient at different times and spatial locations. This yields spatiotemporal distribution data of fire source hazards, including dynamic data on smoke and temperature. Based on this spatiotemporal distribution data, a risk quantification assessment is conducted, extracting dynamic fire risks that change over time. Factors such as the poisoning risk coefficient corresponding to smoke concentration and the burn risk coefficient corresponding to temperature are used. Then, spatial coordinate correlation analysis is performed on the computational domain network and the evacuation network topology to establish a spatial coordinate mapping relationship between the two. According to this mapping relationship, the dynamic fire risk factors are allocated to each node of the evacuation network topology using spatial interpolation to obtain spatial interpolation results. At the same time, the accessibility of each area of ​​the site is evaluated based on the evacuation network topology to construct initial accessibility attribute data. Finally, the spatially interpolated dynamic fire risk factors are correlated and fused with the initial accessibility attribute data to generate dynamic fire risk field data that can reflect the fire risk level and accessibility feasibility of each area of ​​the site in real time due to smoke diffusion and temperature changes.

[0025] Step S400: Couple the dynamic fire risk field data with the evacuation simulation model to drive the intelligent passenger to make path decisions and construct the evacuation simulation results.

[0026] Specifically, when coupling dynamic fire risk field data with the intelligent agent passenger evacuation simulation model, the personnel density of each area of ​​the station is first calculated based on the initial passenger distribution data, and a personnel density influence factor is constructed. Then, according to multiple set simulation time steps, the dynamic fire risk field data is mapped to the evacuation network topology, and the initial travel cost of each path is calculated in combination with the regional traffic conditions. By weighted summing of the personnel density influence factor, the initial travel cost, and the dynamic fire risk factor, a comprehensive travel cost that reflects the safety and smoothness of the path is obtained, thereby driving each intelligent agent passenger to generate a movement intention that fits its own attributes and the current environment. Next, conflict analysis is performed on the movement intentions of all intelligent agents, discrete conflict events are extracted, and a conflict arbitrator is constructed. After calculating and adjudicating the conflict events, a movement intention modification instruction is generated. Based on this instruction, multiple candidate target units are matched for the intelligent agent and the passenger's target movement intention is updated. Finally, the validity of the updated passenger target movement intention is verified. After successful verification, the intention is executed to perform path simulation, and the temporal positions of each intelligent agent passenger at different time steps are recorded. All temporal positions are integrated to construct a complete evacuation simulation result.

[0027] Step S500: Based on the evacuation simulation results, identify the trajectory of the intelligent passenger, generate a passenger individual trajectory dataset for dynamic path analysis, and construct evacuation path optimization suggestions.

[0028] Specifically, when constructing evacuation route optimization suggestions based on evacuation simulation results, the temporal positions of each agent passenger are first extracted from the simulation results. This is combined with information such as movement speed and dwell time recorded during the simulation to perform spatiotemporal state analysis, obtaining complete spatiotemporal state record data for each agent passenger. Subsequently, trajectory structuring is performed based on the temporal positions and spatiotemporal state record data, connecting discrete location points to construct multiple complete trajectory chains. Data augmentation techniques are used to supplement trajectory details and remove outlier data, forming a comprehensive dataset of individual passenger trajectories. Based on this dataset, a spatiotemporal clustering algorithm is executed to group passenger trajectories with similar characteristics. Traces are categorized into multiple spatiotemporal clusters, and then the mainstream evacuation route patterns, secondary route patterns, and abnormal route patterns are dynamically identified. Next, by combining the evacuation network topology map and dynamic fire risk field data, safety assessments are conducted on the three types of route patterns from dimensions such as risk level, traffic efficiency, and congestion degree, generating a safety diagnostic report that includes potential hazards and optimization directions for each route. Finally, based on the safety diagnostic report, differentiated guidance strategies are formulated for the characteristics of different route patterns, such as optimizing the signage and guidance of mainstream routes, improving the traffic capacity of secondary routes, and avoiding risk areas of abnormal routes, ultimately constructing reasonable and feasible evacuation route optimization suggestions.

[0029] In one possible implementation, step S100 further includes:

[0030] Step S110: Perform a panoramic scan of the target rail transit station using a 3D laser scanner to obtain 3D point cloud data.

[0031] Step S120: Denoise and register the three-dimensional point cloud data to construct a three-dimensional point cloud model.

[0032] Step S130: Traverse the target rail transit stations to identify buildings, extract key structural elements, map the key structural elements to the three-dimensional point cloud model, and construct three-dimensional spatial structure data.

[0033] Step S140: Based on the three-dimensional point cloud model, conduct a fire risk assessment, preset multiple potential ignition locations for fire source power analysis, and plot the fire source power growth curve.

[0034] Step S150: Based on the three-dimensional point cloud model, identify ventilation and smoke exhaust conditions to determine the site's environmental ventilation conditions.

[0035] Step S160: Based on the site's environmental ventilation conditions and the fire source power growth curve, map to multiple potential ignition locations to construct fire scenario parameters.

[0036] Step S170: Based on the three-dimensional spatial structure data, perform accessibility analysis, construct an accessible space, introduce historical passenger flow data of the target rail transit station and map it to the accessible space for spatial distribution, and generate the initial passenger distribution data.

[0037] Specifically, a high-precision 3D laser scanner is used to perform a comprehensive, all-around panoramic scan of all functional areas and spatial structures of the target rail transit station, including the station hall, platform, passageway, staircase, entrance and exit. The laser ranging principle is used to quickly capture detailed information such as the spatial position, size, shape and relative layout of building components, facilities and equipment within the station. Simultaneously, data such as distance and angle during the scanning process are recorded and converted into 3D coordinate points, ultimately forming high-density, high-precision 3D point cloud data that can completely restore the physical spatial characteristics of the station.

[0038] For the acquired 3D point cloud data of the target rail transit station, professional denoising algorithms such as statistical filtering and radius filtering are first used to remove isolated noise points, redundant data, and outliers caused by environmental interference and equipment errors during the scanning process, ensuring the purity of the point cloud data. Then, registration technology based on feature matching or ICP iterative nearest point is used to align and fuse the scattered point cloud data obtained from multi-view and multi-region scanning, eliminating coordinate deviations under different scanning perspectives and achieving spatial consistency stitching of point cloud data. Through the collaborative processing of denoising and registration, the discrete 3D coordinate points are integrated into a complete, accurate 3D point cloud model that can realistically restore the spatial morphology, building components, and facility layout of the station, providing a reliable spatial data foundation for subsequent processes such as extraction of key structural elements and construction of fire scene parameters.

[0039] By traversing the 3D point cloud model and actual spatial scene of the target rail transit station, a building structure recognition algorithm is used to accurately identify the functional areas and core components within the station. Key structural elements such as staircases, passageways, entrances and exits, safety exits, fire compartment boundaries, equipment rooms, and obstacles are extracted. At the same time, the spatial location, size, functional attributes, and interconnections of each element are recorded. Subsequently, the extracted key structural elements are accurately mapped to the denoised and registered 3D point cloud model according to the real spatial coordinates and topological relationships. Through data association and integration, the discrete structural elements and the point cloud model are deeply fused, and finally, a 3D spatial structure data that can completely and accurately reflect the station's spatial layout, structural features, and traffic logic is constructed.

[0040] Based on the established 3D point cloud model, and leveraging its accurate reconstruction of the spatial layout of rail transit stations, the distribution of building components, and the location of facilities and equipment, a comprehensive fire risk assessment is conducted. Combining the characteristics of high-risk fire scenarios within the station, such as areas with dense electrical equipment, flammable material storage areas, and ventilation dead zones, multiple typical and representative potential ignition locations are pre-defined. For each pre-defined potential ignition location, dynamic analysis and calculation of the fire source power are performed by simulating different types and quantities of combustibles and environmental conditions, accurately quantifying the energy release intensity of the fire source at different combustion stages. Based on the multiple sets of fire source power data obtained from the analysis and calculation, a fire source power growth curve is plotted, which can intuitively reflect the power change pattern of the fire source from ignition, development, peak, to decay.

[0041] Using the constructed 3D point cloud model as a spatial reference, and leveraging the detailed information such as the station's building structure and facility layout accurately reproduced in the model, the system comprehensively identifies and precisely marks ventilation and smoke exhaust-related facilities and components within the station, including ventilation openings, smoke exhaust ducts, air ducts, fan equipment, and fireproof smoke exhaust valves. This clarifies the spatial location, specifications, distribution density, and connectivity of various ventilation and smoke exhaust facilities. Simultaneously, considering the airtightness of the station's building structure and the orientation of passageways, the system analyzes airflow paths and potential obstruction areas. By quantitatively calculating key indicators such as ventilation volume, smoke exhaust efficiency, and airflow organization, the system ultimately determines the ventilation conditions of the target rail transit station environment, including its ventilation and smoke exhaust capacity, airflow circulation patterns, and smoke exhaust effects under fire scenarios.

[0042] Based on the established site environment ventilation conditions, including key information such as ventilation volume, smoke extraction efficiency, and airflow path, and combined with fire source power growth curves plotted for multiple potential ignition locations, a spatial correlation algorithm is used to map ventilation conditions and fire source power growth patterns to the corresponding potential ignition locations one by one. The algorithm comprehensively analyzes key fire development parameters such as fire spread rate, smoke diffusion range, and temperature conduction efficiency under actual ventilation conditions at different ignition locations. At the same time, it integrates information such as the spatial layout characteristics of each ignition location and the distribution of surrounding combustibles, and finally constructs fire scenario parameters that cover multiple scenarios, multiple dimensions, and fit the actual working conditions of the site.

[0043] Based on the constructed 3D spatial structure data, the accessibility of areas such as passages, stairs, and entrances / exits within the target rail transit station is first analyzed. Impassable areas such as walls and obstacles are eliminated, and the access boundaries and connectivity of each area are clarified to construct an accessible space that conforms to the actual access logic of the station. Subsequently, historical passenger flow data of the target station is introduced, including key information such as passenger flow density and direction in different time periods and areas. Combined with the characteristics of the accessible space area and access capacity, the historical passenger flow data is reasonably allocated to various areas of the accessible space through a data mapping algorithm to simulate the natural distribution of passengers in the station during actual operation. Finally, initial passenger distribution data that accurately reflects the passenger flow distribution pattern of the station and is highly consistent with the actual scenario is generated.

[0044] In one possible implementation, step S200 further includes:

[0045] Step S210: Discretize the three-dimensional spatial structure data into a raster and construct a sparse network topology map.

[0046] Step S220: Based on the evacuation network topology map, initialize the environment of the target rail transit station and construct basic environmental space units.

[0047] Step S230: Map the initial passenger distribution data to the evacuation network topology map one by one according to the basic environmental space unit to construct multiple intelligent passenger agents.

[0048] Step S240: Traverse the multiple intelligent agent passengers to configure attributes and generate multiple sets of individual attribute parameters, wherein the multiple sets of individual attribute parameters correspond to the multiple intelligent agent passengers.

[0049] Step S250: Set multiple simulation time steps, and calculate the spatial proximity relationship based on the multiple intelligent passenger agents and the multiple sets of individual attribute parameters according to the multiple simulation time steps.

[0050] Step S260: Based on the spatial proximity relationship, map multiple intelligent passenger agents to construct a neighborhood list.

[0051] Step S270: Based on the neighborhood list, perform dynamic interaction analysis according to the multiple simulation time steps to construct a dynamic interaction network.

[0052] Step S280: Add the dynamic interactive network to the evacuation simulation model of the intelligent passenger agent.

[0053] Specifically, the predefined grid size standard is first established. Considering the site's spatial accuracy requirements and simulation efficiency, a reasonable grid edge length is determined. The constructed 3D spatial structure data is then discretized into grids, uniformly dividing the overall site space into several regular 3D grid units. Each grid unit is assigned a unique spatial coordinate identifier. Subsequently, based on key structural elements in the 3D spatial structure data, such as passable areas like passageways, stairs, and entrances / exits, and impassable areas like walls and obstacles, each grid unit is labeled with accessibility attributes, indicating whether it is passable, impassable, or has restricted access. The connectivity relationships between passable grid units are also clarified, such as the direction of travel between adjacent grids and the path connection status. Based on this, using grid units as nodes and connectivity relationships as edges, an evacuation network topology is constructed using graph theory modeling. This topology not only fully preserves the site space's accessibility logic and layout characteristics but also clearly presents the associated paths between each passable area.

[0054] Using the constructed evacuation network topology as a spatial framework, the simulation environment of the target rail transit station is fully initialized and configured: First, core information such as spatial coordinates, access attributes, and connectivity relationships of each grid cell in the topology is extracted. Combined with the actual building dimensions and facility layout in the 3D spatial structure data, each grid cell is assigned realistic physical spatial attributes, such as area, height, and whether it is adjacent to stairs / exits. Second, key facilities within the station, such as safety exits, emergency exits, and fire equipment storage points, are marked with coordinates and functional attributes to clarify their role and priority in the simulation environment. Finally, the grid cells that integrate spatial attributes, facility information, and access rules are defined as basic environmental spatial units. Each unit contains complete attribute information such as an independent identifier ID, spatial coordinate range, functional type, and access capacity threshold. A structured simulation environment system is formed through the association mapping between units.

[0055] First, the core information contained in the initial passenger distribution data, such as the number of passengers and the passenger flow density in each area, is identified. Using the constructed basic environmental spatial units as the mapping benchmark, the initial passenger distribution data is decomposed into the corresponding basic environmental spatial units of the evacuation network topology according to the principle of corresponding passenger flow to each unit. For the passenger data in each unit, an independent identity is assigned to it through intelligent agent modeling technology. At the same time, spatial information such as unit coordinates and passage attributes are associated to transform the discrete passenger data into intelligent agent entities with spatial perception capabilities. Each intelligent agent passenger is precisely matched with the spatial characteristics and passenger flow distribution characteristics of the corresponding unit. Finally, multiple intelligent agent passengers are constructed with a number and distribution that are highly consistent with the actual initial passenger situation and can respond to changes in the simulation environment.

[0056] A traversal algorithm is used to access all the constructed intelligent passenger agents one by one. Combining the actual population characteristics of passengers at rail transit stations with the needs of evacuation scenarios, multi-dimensional individual attribute parameters are configured for each intelligent passenger agent. These include basic attributes such as age group, gender, and physical condition; behavioral attributes such as movement speed, reaction sensitivity, and evacuation experience; and decision-making attributes such as evacuation priority, risk preference, and familiarity with the environment. At the same time, each set of attribute parameters is assigned a quantitative value or classification label to ensure that each set of individual attribute parameters establishes a unique mapping relationship with the corresponding intelligent passenger agent. This not only gives each intelligent agent differentiated behavioral logic and decision-making basis, but also provides accurate data support for subsequent calculation of spatial proximity relationships and simulation of dynamic interaction behavior based on attribute parameters, making the behavior of intelligent passenger agents more consistent with the evacuation characteristics of real populations.

[0057] First, based on the balance between accuracy requirements and computational efficiency in the simulation of evacuation at rail transit stations, multiple continuous and equal simulation time steps are set, such as 0.5 seconds or 1 second. Then, a traversal matching algorithm is used to associate each agent passenger with each simulation time step, generating a unique matching identifier. Based on this matching identifier, combined with the individual attribute parameters corresponding to each group of agent passengers, such as movement speed and mobility, the agent passengers are accurately mapped to the corresponding basic environmental spatial units, determining their specific unit coordinates at each time step. Using the Euclidean distance calculation formula, pairwise distance calculations are performed on the unit coordinates of all agent passengers at the same time step, obtaining multiple coordinate Euclidean distance data. The agent passengers are sorted and marked according to the distance values, constructing a global distance matrix covering the spatial location associations of all agents. Finally, based on this global distance matrix, other agents within a preset distance range around each agent are selected, completing the proximity analysis and obtaining spatial proximity relationships that clearly reflect the spatial location relationships between agents.

[0058] Based on the obtained spatial proximity relationships, a unique identity is first assigned to each intelligent agent passenger. Then, according to the mapping logic of individual association with the surrounding environment, all intelligent agent passengers are traversed, and other intelligent agents within a preset distance threshold around each intelligent agent, that is, intelligent agents with spatial proximity relationships, are associated and matched. The identity, relative spatial position, and individual attribute parameters of the associated intelligent agents, such as movement speed and action status, are recorded. Through a structured data organization method, a unique neighborhood information entry is generated for each intelligent agent. The entry clearly contains the complete information of the surrounding associated intelligent agents and their mutual spatial relationships. Then, the neighborhood information entries of all intelligent agents are integrated and sorted to construct a neighborhood list indexed by the intelligent agent identity, which can quickly query the surrounding associated objects. This list provides clear and accurate association data support for subsequent time-step-based dynamic interaction analysis of intelligent agents.

[0059] This study employs a combination of multi-time-step iterative analysis and agent interaction rule modeling. Using a neighborhood list as the foundation for interaction data, Agent-Based Modeling decomposes the behavioral logic of agents within the neighborhood at each simulation time step. It pre-defines avoidance priority algorithms, congestion feedback mechanisms, and path selection coordination rules to quantitatively analyze dynamic interaction relationships among agents, such as speed adaptation, space occupancy conflicts, and directional coordination. Simultaneously, temporal correlation analysis is introduced to track changes in the neighborhood list at different time steps, such as the addition or removal of interaction objects and distance changes, recording the temporal correlation and evolution patterns of interaction behaviors. Combining spatial constraints of the evacuation network topology, such as grid access restrictions and channel width thresholds, a graph neural network (GNN) is used to model agents as nodes and interaction strength as edge weights, constructing a dynamically weighted interaction network. This network can adjust node association states in real time according to the interaction data updates at each time step, ultimately forming a dynamic interaction network that accurately replicates the dynamic interaction characteristics between agents and supports subsequent path decisions.

[0060] The constructed dynamic interaction network is deeply integrated with the previously established intelligent agent passenger basic simulation framework. First, the core data structures in the dynamic interaction network, such as intelligent agent nodes, interaction edge weights, and temporal interaction rules, are defined, and a one-to-one correspondence is established with modules such as intelligent agent identity identifiers, basic environmental spatial units, and individual attribute parameters in the evacuation simulation model. Then, through interface adaptation and data association algorithms, dynamic interaction logic such as avoidance, following, and crowding conflicts between intelligent agents is embedded into the behavioral decision layer of the model, enabling intelligent agent passengers to call neighborhood list information in real time during the simulation and respond to changes in the behavior of surrounding intelligent agents. At the same time, the model's calculation process is optimized to ensure that the updates of the dynamic interaction network with the simulation time step are coordinated with the intelligent agent's path decision and position movement. Finally, a complete intelligent agent passenger evacuation simulation model with individual differentiated attributes, spatial perception capabilities, and dynamic interaction characteristics of real crowds is formed, providing reliable model support for subsequent coupling with dynamic fire risk field data and achieving accurate evacuation simulation.

[0061] In one possible implementation, step S250 further includes:

[0062] Step S251: Iterate through the multiple intelligent passenger agents and match them with the multiple simulation time steps to obtain multiple matching identifiers.

[0063] Step S252: Based on the multiple matching identifiers and the multiple sets of individual attribute parameters, map them to the basic environmental space unit to determine the unit coordinates of multiple intelligent passenger agents.

[0064] Step S253: Calculate the distance between the unit coordinates of the multiple intelligent passenger agents to obtain the Euclidean distance between the multiple coordinates.

[0065] Step S254: Arrange the distance identifiers of the multiple intelligent passenger agents according to the multiple coordinate Euclidean distances to construct a global distance matrix.

[0066] Step S255: Perform proximity analysis on multiple intelligent passenger agents based on the global distance matrix to obtain spatial proximity relationships.

[0067] Specifically, an iterative matching algorithm is adopted. First, the unique identity identifiers of all constructed intelligent agent passengers are identified, such as ID codes and multiple preset continuous simulation time steps, including parameters such as time sequence number and duration. Then, according to the principle of one intelligent agent and one time step, each intelligent agent passenger is matched with each simulation time step in turn. A unique matching identifier is assigned to each combination of intelligent agent identity identifier and simulation time step sequence number. The structured encoding form of intelligent agent ID and time step sequence number is adopted to ensure that each matching identifier can accurately associate the behavior calculation dimension of a specific intelligent agent at a specific time step, and finally generate multiple unique and traceable matching identifiers.

[0068] First, the agent's identity and corresponding simulation time step information contained in each matching identifier are analyzed. This identifier is then linked to multiple sets of individual attribute parameters, such as movement speed, mobility, and reaction sensitivity. Subsequently, using the spatial coordinate range and access attributes of the basic environmental spatial unit as a mapping benchmark, the agent's movement state and position preference at the current time step are simulated in combination with the individual attribute parameters. For example, the spatial range that can be covered within the time step is calculated based on the movement speed, and whether the agent can pass through the restricted access unit is determined based on the mobility. Through the spatial mapping algorithm, each agent is accurately matched to the corresponding basic environmental spatial unit in the evacuation network topology diagram, and finally, the unique unit coordinates of each agent passenger at a specific time step are determined.

[0069] Based on the defined unit coordinates of multiple intelligent passenger agents at corresponding simulation time steps, the three-dimensional spatial values ​​of each agent's unit coordinates are first determined, including x, y, and z axis coordinate parameters. Then, a batch Euclidean distance calculation algorithm is used to perform pairwise combination calculations on the unit coordinates of all intelligent passenger agents within the same simulation time step. The spatial straight-line distance between each pair of agents is accurately solved using the Euclidean distance formula d=√[(x2-x1)²+(y2-y1)²+(z2-z1)²], ultimately generating multiple coordinate Euclidean distance data representing the spatial distance relationship between agents. Each distance data is associated with the identity identifier and time step information of the corresponding agent, providing quantitative distance support for subsequent global distance matrix construction and spatial proximity analysis.

[0070] The obtained Euclidean distance data of multiple coordinates are preprocessed, and the agent identity and simulation time step information corresponding to each pair of distance data are associated. Then, the distance relationship between each agent and all other agents is sorted and labeled according to the Euclidean distance values ​​in ascending or descending order, such as "1" indicating the closest and "n" indicating the farthest, where n is the total number of agents. Using the agent identity as the row and column index, the sorted distance values ​​and corresponding distance order labels are filled into the corresponding positions of the matrix to construct a global distance matrix. Each element in the matrix contains both the specific Euclidean distance value between two specific agents and the label of the relative distance order, fully presenting the spatial distance relationship between all agents under the same simulation time step.

[0071] Using the constructed global distance matrix as the core analytical basis, a reasonable spatial distance threshold is first set by combining the accuracy requirements of evacuation simulation with the spatial characteristics of the site. This threshold needs to balance the realism of agent interaction and computational efficiency, such as being determined based on the width of the site passage and the safe distance during crowd evacuation. Then, the row / column data corresponding to each agent in the global distance matrix is ​​traversed to filter out the identifiers and corresponding distance information of other agents whose distance values ​​are less than or equal to the preset threshold, thus identifying the neighboring objects of each agent at the current simulation time step. The filtering results are optimized and verified through proximity relationship determination rules, such as prioritizing the retention of the N closest agents and excluding agents in non-connected areas, to eliminate invalid proximity associations. Finally, the effective neighboring object information of each agent is integrated to form a spatial proximity relationship that can clearly represent the spatial location association of each agent at a specific time step.

[0072] In one possible implementation, step S300 further includes:

[0073] Step S310: Perform physical field simulation based on the fire scene parameters to construct a fire physical field engine.

[0074] Step S320: Based on the fire physics engine, perform spatial matching of the target rail transit stations to generate a computational domain network.

[0075] Step S330: Map the multiple potential ignition locations to a computing domain network for driving simulation to obtain spatiotemporal distribution data of fire source disasters.

[0076] Step S340: Conduct a risk assessment based on the spatiotemporal distribution data of the fire source disaster and construct a dynamic fire risk factor.

[0077] Step S350: Map the dynamic fire risk factors to the evacuation network topology map for data association to generate dynamic fire risk field data.

[0078] Specifically, taking fire scene parameters that integrate the characteristics of multiple potential ignition locations, site environment ventilation conditions, fire source power growth law, and spatial environment attributes as core inputs, the system employs computational fluid dynamics (CFD) and fire dynamics coupled simulation technology. It solves the mass conservation, momentum conservation, energy conservation, and component transport control equations during the fire spread process using numerical discretization methods. It integrates key physical models such as the fire source combustion rate model, smoke diffusion turbulence model, temperature field conduction model, and toxic and harmful gas generation and diffusion models, such as CO and CO2. Simultaneously, it embeds boundary constraints from the fire scene parameters, such as ventilation volume, building fire resistance rating, and combustible material distribution density. Through multi-physics coupling algorithms, it achieves accurate replication of physical phenomena such as thermal radiation, thermal convection, smoke flow, and toxic gas diffusion during fire development, constructing a fire physics engine with dynamic calculation and real-time output capabilities.

[0079] First, the computational dimension parameters of the fire physics engine are extracted, including mesh accuracy, physics calculation boundary conditions, and coordinate system. Simultaneously, core spatial information such as building outlines, passageway dimensions, wall locations, and facility distribution from the target rail transit station's 3D spatial structure data is retrieved. A spatial coordinate calibration algorithm is used to uniformly adapt the station's spatial coordinates with the physics engine's computational coordinates, ensuring consistency in spatial dimensions. Then, a structured mesh generation technique is employed, combining station spatial characteristics with fire simulation accuracy requirements, to divide the overall station space into regular mesh units matching the physics engine's computational capabilities. The spatial coordinates and boundary attributes of each mesh are clearly defined, such as building boundaries, open space boundaries, and connectivity. Boundary constraint parameters required by the fire physics engine are embedded, such as wall fire resistance, ventilation opening locations and ventilation volumes, and combustible material distribution areas. Attribute annotations and parameter assignments are performed on the mesh units, ultimately generating a computational domain network containing structured meshes, spatial topological relationships, boundary condition parameters, and attributes required for physics calculations, achieving precise coupling between the station space and the fire physics engine.

[0080] Through a spatial coordinate mapping algorithm, multiple potential ignition locations are sequentially associated with corresponding structured grid cells in the computational domain network. This clarifies the precise coordinates of each ignition location within the grid and its surrounding environmental attributes, such as combustible material distribution and ventilation conditions. Subsequently, driven by a fire physics engine, corresponding fire scene parameters are loaded for each mapped ignition location, including key parameters such as initial fire power, power growth curve, and combustion type. Boundary constraints in the computational domain network are also applied, such as the fire resistance of building walls and airflow parameters at ventilation openings. Through multi-time-step iterative simulation, the fire evolution process at different ignition locations at different times is simulated, and dynamic data such as the spatial expansion trajectory of the fire spread range, the distribution of smoke concentration in each grid cell, the spatial gradient change of the temperature field, and the generation and diffusion paths of toxic and harmful gases are output in real time. Finally, these fire-related data spanning time and space are integrated and archived to form spatiotemporal distribution data of fire source disasters covering three-dimensional characteristics of time, space, and disaster intensity, comprehensively presenting the dynamic evolution law of fire development.

[0081] Using spatiotemporal distribution data of fire hazards as the core, this study first screens key risk indicators such as temperature exceedance coefficient, smoke concentration exceedance multiple, toxic gas volume fraction, fire spread rate, and remaining safe evacuation time through an indicator system construction technique. The Analytic Hierarchy Process (AHP) combined with entropy weighting is then used to determine the comprehensive weight of each indicator. Next, a fuzzy comprehensive evaluation model is used to transform the raw disaster data at each spatiotemporal node into a quantitative risk level of 0-10. Simultaneously, a Long Short-Term Memory (LSTM) time-series prediction model is introduced, and a risk trend prediction model is trained based on historical disaster evolution data, outputting the probability of risk level changes and diffusion paths for each region over multiple future time steps. Finally, the quantitative risk level, indicator weight ratio, temporal evolution trend, and risk impact boundary data are integrated, and a dynamic fire risk factor is constructed using structured data encapsulation technology. This factor contains five core dimensions of information: spatial coordinates, time node, risk level, evolution trend, and indicator contribution, and can be updated in real time with the simulation process, providing accurate risk quantification basis for subsequent evacuation route decisions.

[0082] By employing spatial coordinate calibration and correlation algorithms, a precise mapping relationship is established between the computational domain network of the fire physical field simulation and the evacuation network topology, clarifying the one-to-one correspondence between grid cells and raster cells in both types of networks. Subsequently, Kriging spatial interpolation is used to allocate dynamic fire risk factors, including risk level, evolution trend, and indicator contribution, to each basic environmental spatial cell of the evacuation network topology according to their corresponding spatial coordinates, ensuring that risk factors are continuously and accurately distributed in the topology. Simultaneously, the original initial accessibility attribute data in the evacuation network topology, such as cell access width and maximum capacity, is retrieved. Through data fusion algorithms, the dynamic fire risk factors are correlated and coupled with the initial accessibility data. For example, the accessibility weight of cells is dynamically adjusted according to the risk level, with reduced accessibility weight in high-risk areas and maintained or optimized weight in low-risk areas. Finally, dynamic fire risk field data that can reflect the risk level and accessibility of each evacuation cell in real time and is consistent with the data format of the intelligent agent evacuation simulation model is generated.

[0083] In one possible implementation, step S350 further includes:

[0084] Step S351: Perform spatial coordinate correlation analysis based on the computing domain network and the evacuation network topology to construct a spatial coordinate mapping relationship.

[0085] Step S352: Based on the spatial coordinate mapping relationship, the dynamic fire risk factor is allocated to the evacuation network topology map by spatial interpolation to obtain the spatial interpolation result.

[0086] Step S353: Based on the evacuation network topology map, conduct a traffic assessment and construct initial traffic capacity attribute data.

[0087] Step S354: Based on the spatial interpolation results, associate the dynamic fire risk factor with the initial accessibility attribute data to generate the dynamic fire risk field data.

[0088] Specifically, the three-dimensional coordinate system of the computational domain network is first extracted, including core parameters such as the coordinate range, origin, axis orientation, and grid accuracy of the grid cells, and the grid cell coordinate information of the evacuation network topology, including grid number, spatial coordinate boundary, and topological connection relationship. A coordinate unification algorithm is then used to align the coordinate references of the two types of networks, eliminating spatial deviations caused by differences in modeling software and coordinate system settings. Subsequently, feature point matching technology is employed to select landmark elements with fixed spatial locations within the target rail transit station, such as the center point of a column, corners of evacuation passages, and vertices of equipment foundation outlines, as... Coplanar feature points are identified, and their coordinate data in the computational domain network and the evacuation network topology are obtained. The coordinates of the two sets of feature points are then registered and optimized using the Iterative Closest Point (ICP) algorithm. The optimal coordinate transformation matrix, including translation, rotation, and scaling parameters, is solved to verify whether the spatial error of the registered feature points meets a preset threshold, such as an error ≤ 0.1m. Finally, this coordinate transformation matrix is ​​used to establish a one-to-one correspondence between each grid cell in the computational domain network and the corresponding grid cell in the evacuation network topology, thus constructing a spatial coordinate mapping relationship that enables accurate mapping of spatial locations between the two types of networks and smooth data association across networks.

[0089] Based on the constructed spatial coordinate mapping relationship, the dynamic fire risk factors are first identified, encompassing quantitative indicators such as risk level, temperature, smoke concentration, and toxic gas content. The specific numerical values ​​and spatial distribution characteristics of these indicators within each grid cell of the computational domain network are then determined. Next, combined with the grid cell division rules of the evacuation network topology, either inverse distance weighted interpolation (IRW) or kriging interpolation algorithms are selected. IWT prioritizes the rapid allocation of discretized risk data, while kriging interpolation is suitable for high-precision scenarios requiring consideration of spatial correlation. Subsequently, according to the coordinate mapping relationship, each grid cell in the computational domain network is... The risk factor data is accurately mapped to the corresponding grid cells in the evacuation network topology map. For blank grid cells in the topology map that are not directly associated with the computational domain grid, risk estimation is performed based on the surrounding grid cells with assigned risk values ​​using an interpolation algorithm. At the same time, the interpolation error is controlled within a preset threshold, such as risk level error ≤ 0.5 and quantification index error ≤ 5%. This ensures that the risk factors form a continuous, smooth spatial distribution in the evacuation network topology map that is consistent with the actual fire risk evolution pattern. Finally, a spatial interpolation result is obtained that fully covers all grid cells in the topology map and contains accurate risk quantification data for each cell.

[0090] Using the discretized grid-processed evacuation network topology as the core analysis vehicle, key spatial characteristic parameters of each grid cell in the topology are first extracted, including cell passage width, channel clearance height, ground material (such as the friction coefficient of floor tiles or anti-slip mats), the presence and area occupied by fixed obstacles such as columns / equipment, and connectivity with evacuation exits. Subsequently, combining the evacuation design specifications and personnel evacuation safety standards of rail transit stations, a personnel density-passage speed correlation model and a formula for calculating the passage volume per unit width per unit time, such as the Petrusek formula, are introduced to quantify the core indicators of each grid cell under normal conditions without fire impact, such as the maximum number of people it can carry, theoretical passage speed, and maximum passage volume per unit time. At the same time, abnormal data, such as invalid cells with zero passage capacity due to excessive obstruction, are eliminated through a verification algorithm. The passage potential of key grid cells adjacent to evacuation exits is enhanced and evaluated. Finally, initial passage capacity attribute data containing multi-dimensional information such as grid cell number, spatial characteristic parameters, maximum passage capacity, passage speed threshold, and connectivity priority is constructed.

[0091] Using spatial interpolation results as the correlation benchmark, the correspondence between quantitative indicators such as risk level, temperature exceedance coefficient, and smoke concentration in dynamic fire risk factors and grid cells in the evacuation network topology is first analyzed. Then, core parameters such as maximum throughput, throughput speed threshold, and connectivity priority in the initial accessibility attribute data are retrieved. A correlation mapping rule between the two types of data is established through a data coupling algorithm. That is, a dynamic correction coefficient is set according to the quantitative value of the risk factor. For high-risk cells, such as risk level ≥ 8 and smoke concentration exceeding the standard by more than 2 times, a correction coefficient of 0.3~0.6 is used to reduce the initial accessibility. For medium-risk units (risk levels 4-7), the accessibility is fine-tuned using a correction coefficient of 0.7-0.9. For low-risk units (risk levels ≤3), the initial accessibility coefficient is maintained at 1.0. Subsequently, the corrected accessibility data is structurally bound to the risk factor data of the corresponding grid units, and the unit coordinates and time step correlation information are supplemented. Finally, dynamic fire risk field data is generated, which includes four-dimensional core information such as grid number, risk quantification index, dynamic accessibility, and connectivity status, and is adapted to the data format of the evacuation simulation model, thus achieving precise integration of fire risk and access constraints.

[0092] In one possible implementation, step S400 further includes:

[0093] Step S410: Calculate the passenger density based on the initial passenger distribution data and construct the passenger density influence factor.

[0094] Step S420: Based on multiple time simulation steps, map the dynamic fire risk field data to the evacuation network topology map to perform access calculation and obtain the initial access cost.

[0095] Step S430: Based on the personnel density influence factor, the initial passage cost, and the dynamic fire risk factor, perform a weighted summation to obtain the comprehensive passage cost.

[0096] Step S440: Based on the comprehensive passage cost, drive the intelligent agent passenger to perform simulation and generate multiple movement intentions.

[0097] Step S450: Perform motion conflict analysis based on the multiple motion intentions and extract discrete conflict events.

[0098] Step S460: Map the discrete conflict events to the evacuation network topology graph for path decision-making, and construct the evacuation simulation results.

[0099] Specifically, based on initial passenger distribution data, including passenger numbers and spatial location relationships in each basic environmental spatial unit, spatial parameters such as grid area and effective passage area ratio of each basic environmental spatial unit in the evacuation network topology are first extracted. Then, using a standardized personnel density calculation formula (personnel density = actual number of passengers in the unit ÷ effective passage area of ​​the unit), the personnel density value of each unit in its initial state is accurately calculated. Subsequently, referring to the personnel evacuation safety regulations for rail transit stations, four threshold standards—low density, medium density, high density, and congestion density—are set, such as low density ≤ 1. The density of each unit is classified into four levels: people / m², medium density (1-2 people / m²), high density (2-3 people / m²), and congestion density (>3 people / m²). The density values ​​of each unit are compared with the thresholds to complete the classification. Through normalization, the density level is converted into a quantitative coefficient in the range of 0 to 1, such as a coefficient of 1.0 for congestion density and a coefficient of 0.2 for low density. At the same time, information such as unit number, spatial coordinates, and density value are associated to finally construct a personnel density influence factor containing four-dimensional information including unit identifier, density value, density level, and quantitative influence coefficient. This factor comprehensively represents the constraint effect of the degree of personnel congestion in different areas on evacuation and traffic efficiency.

[0100] First, multiple preset simulation steps are defined, including parameters such as step time interval and total simulation duration. Following the principle of iterative step-by-step approach, at each time step, the core information such as real-time risk level, dynamic accessibility, and risk evolution trend of each grid cell in the dynamic fire risk field data is accurately mapped to the basic environmental spatial cell corresponding to the evacuation network topology using spatial coordinate association technology. Then, based on the connectivity relationships of each cell in the topology, dynamic accessibility constraints, and the evacuation goals of the agent's passengers, such as the nearest safe exit and low-risk areas, an improved Dijkstra algorithm is used to calculate the travel distance cost, travel time cost, and risk exposure cost of the agent from its current cell to each potential target cell. Through weighted integration, such as a distance weight of 0.3, a time weight of 0.4, and a risk weight of 0.3, the three types of costs are transformed into a unified quantitative index. Finally, the initial access cost value corresponding to each cell in the evacuation network topology at each time step is obtained, forming an initial access cost matrix that is dynamically updated over time and covers the entire topology.

[0101] First, considering the actual needs and safety priorities of fire evacuation in rail transit, weight coefficients are determined by coupling the Analytic Hierarchy Process (AHP) with the entropy weight method. The weight of the personnel density factor focuses on the constraint of congestion on traffic efficiency, such as setting it to 0.3; the weight of the initial traffic cost factor emphasizes basic traffic costs, such as setting it to 0.4; and the weight of the dynamic fire risk factor highlights the priority of safety risks, such as setting it to 0.3, ensuring that the weight allocation is both subjective and objective. Subsequently, the three types of factors are normalized to eliminate calculation biases caused by differences in data dimensions and units. The quantitative coefficients of personnel density, initial passage cost, and fire risk level are uniformly mapped to the range of 0 to 1. The comprehensive passage cost is calculated unit by unit according to the weighted summation formula: comprehensive passage cost = personnel density influence factor × corresponding weight + initial passage cost × corresponding weight + dynamic fire risk factor × corresponding weight. The three types of data in each unit of the evacuation network topology at each time simulation step are calculated to obtain a unified quantitative index that can comprehensively reflect the degree of congestion, basic passage cost, and real-time fire risk. This provides a core basis for intelligent passenger route decision-making that balances efficiency and safety.

[0102] Based on the comprehensive passage cost value as the core decision-making basis, each intelligent agent passenger is first bound to a preset evacuation target, such as the nearest safety exit or the evacuation route with the lowest real-time risk. Combining the coordinates of the basic environmental spatial unit where the intelligent agent is currently located and individual attribute parameters, such as movement speed and emergency response efficiency, a heuristic search algorithm, such as the improved A* algorithm, is used to traverse the comprehensive passage cost value of adjacent units in the evacuation network topology. The 3 to 5 adjacent units with the optimal cost value, that is, the lowest comprehensive congestion level, passage cost and fire risk, are selected as potential movement directions. At the same time, the movement distance, expected passage time and risk exposure level of each potential direction are calculated to clarify the target unit, movement priority and alternative solutions of the intelligent agent to move to. Finally, multiple movement intentions are generated for each intelligent agent passenger, including target unit coordinates, movement priority, expected cost and alternative directions, to comprehensively cover the passage possibilities in different scenarios.

[0103] First, a multi-agent movement intention conflict detection framework is built. This framework iterates through multiple movement intentions of all agents, extracting the target unit coordinates and movement time window corresponding to each intention, including core information such as start and end times and movement path trajectory. A spatial overlap detection algorithm is used to determine whether different agents intend to enter the same target unit within the same time window or whether their movement paths overlap within the same grid unit. Simultaneously, constraints such as the maximum capacity and passage width of each unit in the evacuation network topology are considered to filter out concurrent movement intentions exceeding the unit's passage capacity limit. Detected conflicts are categorized and labeled, such as space occupancy conflicts, path intersection conflicts, and passage capacity exceeding limits. The conflict-related agent ID, conflict unit coordinates, conflict time step, conflict type, and details of the involved movement intentions are extracted. This information is encapsulated into structured discrete conflict events, ensuring that the key elements of each conflict event are complete and traceable.

[0104] A priority-based conflict arbitrator is constructed, which combines the distance between passengers and evacuation exits, individual movement speed attributes, dynamic accessibility of conflicting units, and fire risk level to adjudicate discrete conflict events, generating movement intention modification instructions that include adjusting movement order, changing alternative target units, and delaying movement time steps. Subsequently, based on these instructions, the evacuation network topology is traversed to select multiple alternative target units with the optimal comprehensive accessibility cost and within the unit's capacity limit for the conflicting agents. Combining unit connectivity and real-time dynamic fire risk field data, the original movement intention is updated and optimized to generate a passenger target movement intention that accurately matches the current scenario. A feasibility verification algorithm is used to verify the target movement intention, such as verifying the accessibility status of target units and the existence of secondary conflicts. Once verified, the agent is driven to move, and the temporal position, movement path trajectory, and conflict resolution results of all agents at each time step are recorded. Finally, these data are correlated and integrated with the evacuation network topology and dynamic fire risk field data to construct a complete evacuation simulation result that includes individual agent trajectories, overall evacuation efficiency, conflict distribution characteristics, and accessibility in risk areas, comprehensively reflecting the dynamic process of personnel evacuation in a fire scenario.

[0105] In one possible implementation, step S460 further includes:

[0106] Step S461: Construct a conflict arbitrator, and use the conflict arbitrator to calculate and adjudicate discrete conflict events to generate a movement intention modification instruction.

[0107] Step S462: Perform motion analysis on multiple intelligent passenger agents using the motion intent modification instruction to generate multiple candidate target units.

[0108] Step S463: Update multiple movement intentions based on the multiple candidate target units to generate passenger target movement intentions.

[0109] Step S464: Verify the passenger's target movement intention. If the verification is successful, execute the passenger's target movement intention to perform path simulation and construct the temporal positions of multiple intelligent passenger agents.

[0110] Step S465: Add the temporal locations of the multiple intelligent passenger agents to the evacuation simulation results.

[0111] Specifically, a conflict arbitrator is constructed using a combination of multi-dimensional priority quantification and conflict type-adaptive adjudication. This arbitrator integrates individual attribute parameters of the intelligent passenger agent, including movement speed and emergency response level; spatial location characteristics, including straight-line distance to evacuation exits and fire risk level of the area; and conflict scenario constraints, including dynamic passage capacity of conflict units and remaining evacuation time. A comprehensive conflict adjudication evaluation system is established, and the weights of each evaluation indicator are determined using the analytic hierarchy process (AHP) and entropy weighting method. Distance weight is set to 0.3, risk level weight to 0.4, and movement speed weight to 0.3. The arbitrator accurately identifies spatial occupancy conflicts and path intersections through a conflict type identification algorithm. Different conflict types, such as conflicts and capacity overrun conflicts, are addressed by quantifying and calculating data such as conflict type, associated agent information, and conflict unit status in discrete conflict events. The movement priority of each agent is determined according to the principles of safety first and efficiency adaptation. Based on the preset conflict handling rule base, i.e., high-priority agents pass first and low-priority agents adjust paths, a targeted movement intention modification instruction is generated for each conflict event through a rule matching algorithm. The instruction clearly includes specific executable content such as the timing parameters for adjusting the movement order, the coordinate range of the alternative target units, the number of steps for delaying movement, and the node positions of the split path, ensuring that the instruction can accurately adapt to the conflict scenario and agent characteristics.

[0112] Based on the movement intention modification command as the core basis, the system retrieves the information of adjacent grid cells of the basic environmental spatial unit where each conflicting agent is currently located in the evacuation network topology map. Simultaneously, it acquires the risk level, dynamic passage capacity, and comprehensive passage cost from the real-time dynamic fire risk field data. Through a heuristic search algorithm, adjacent grid cells are screened one by one, prioritizing the retention of grid cells with low risk level, sufficient passage capacity, and not exceeding the maximum capacity of the unit. At the same time, combined with the agent's preset evacuation target direction, the system focuses on selecting units that are highly consistent with the original movement path or closer to the safety exit. After eliminating grid cells with secondary conflict risks, multiple candidate target units that meet the requirements of safety and efficiency are generated. Each candidate target unit is clearly marked with the corresponding comprehensive passage cost, passage distance, expected movement time, and conflict avoidance probability, forming a candidate target unit list with a clear structure and complete parameters.

[0113] Based on multiple candidate target units as the core update basis, the original movement intentions of each intelligent agent passenger are matched one by one. Combining the priority requirements of movement intention modification instructions and the individual attribute parameters of the intelligent agent, the comprehensive passage cost, passage distance, expected movement time, and conflict avoidance probability of the candidate target units are compared from multiple dimensions. The candidate target units with the highest fit with the intelligent agent's evacuation goal and the best safety and efficiency are selected as the core update direction. Targeted adjustments are made to key information such as target units, movement path planning, and passage timing in the original movement intention. If there is a conflict risk in the original path, it is replaced with the optimal path corresponding to the candidate target unit. If only the timing needs to be optimized, the movement step size is adjusted based on the passage status of the candidate unit. At the same time, the real-time risk level, dynamic passage capability, and other related parameters of the target unit are supplemented. Finally, a passenger target movement intention that is both targeted and feasible is generated to ensure that the movement plan of each intelligent agent not only meets the requirements of the overall evacuation scenario, but also adapts to individual characteristics and real-time environmental changes.

[0114] A multi-dimensional passenger target movement intention verification system was established, and verification was carried out from three core dimensions: spatial feasibility, risk controllability, and conflict-free nature. Spatial feasibility verification focused on the target unit and planned path that the intelligent agent passenger intended to move to, confirming that the target unit had sufficient passage space and an effective connection path with the current unit, and that there were no physical obstacles blocking the way. Risk controllability verification combined dynamic fire risk field data to check that the real-time fire risk level, smoke concentration, temperature, and other indicators of the target unit and movement path did not exceed the safety threshold, ensuring that the risk exposure during the intelligent agent's passage was within a controllable range. Conflict-free nature verification, by traversing the target movement intentions of all intelligent agent passengers, checked whether the current intelligent agent's movement plan had new conflicts such as overlapping space occupation or path intersections with other intelligent agents at the same time step. When all three verifications meet the preset standards and there are no abnormalities, the verification is deemed successful. Then, the intelligent agent passenger is driven to perform path simulation according to the target movement intention. Based on the set multiple simulation time steps, the spatial coordinates, movement status and basic environmental spatial unit information of each intelligent agent at different time nodes are recorded in real time. Finally, a dataset of multiple intelligent agent passenger time sequence locations containing time node, intelligent agent number, spatial coordinates and movement status is constructed.

[0115] The temporal location datasets of multiple intelligent agents are structured to clarify the agent number, time node, spatial coordinates, movement status, and associated conflict resolution records corresponding to each temporal location, ensuring data integrity and uniform format. This structured dataset is then adapted to the pre-defined framework of the evacuation simulation results. Following the time simulation step order, the temporal locations of each agent are embedded into the result system one by one, synchronously linking grid cell information of the evacuation network topology, real-time data of the dynamic fire risk field, and changes in the comprehensive passage cost. Through data integration algorithms, the temporal location data is deeply integrated with existing simulation results, such as conflict distribution characteristics and overall evacuation progress. Statistical information such as the overall distribution density of agents and regional passage efficiency at each time node is supplemented, ultimately improving the spatiotemporal dimension data of the evacuation simulation results and forming a complete result system that comprehensively reflects the dynamic process of personnel evacuation in a fire scenario.

[0116] In one possible implementation, step S500 further includes:

[0117] Step S510: Perform state analysis based on the temporal location of the multiple intelligent passenger agents to obtain spatiotemporal state record data of the multiple intelligent passenger agents.

[0118] Step S520: Based on the time sequence location and the spatiotemporal state recording data, perform trajectory structuring to construct multiple trajectory chains.

[0119] Step S530: Perform data augmentation on the multiple trajectory chains to construct a passenger individual trajectory dataset.

[0120] Step S540: Perform spatiotemporal clustering based on the passenger individual trajectory dataset to generate multiple spatiotemporal clusters.

[0121] Step S550: Dynamically mine according to the multiple spatiotemporal clusters to identify multiple fire scene evacuation path patterns, including mainstream evacuation path patterns, secondary path patterns, and abnormal path patterns.

[0122] Step S560: Based on the evacuation network topology map and the dynamic fire risk field data, perform a safety assessment on the mainstream evacuation path pattern, the secondary path pattern, and the abnormal path pattern, and generate a safety diagnosis report.

[0123] Step S570: Based on the safety diagnosis report and in conjunction with the mainstream evacuation route mode, the secondary route mode, and the abnormal route mode, evacuation guidance is provided to the target rail transit station, and the evacuation route optimization suggestions are constructed.

[0124] Specifically, the analysis focuses on the temporal location of multiple intelligent passenger agents. It extracts the spatial coordinates, basic environmental spatial units, movement speed, dwell status, and associated conflict resolution records of each agent at different time points. Combining the real-time risk level in the dynamic fire risk field data with the unit attributes of the evacuation network topology, it conducts a multi-dimensional analysis of the agents' passage status, risk exposure, and path adaptability. It clarifies the key state changes of each agent during the evacuation process, such as acceleration, deceleration, avoidance, and stagnation, and finally forms spatiotemporal state record data of multiple intelligent passenger agents, ensuring comprehensive state description and complete data dimensions.

[0125] Based on the temporal location and spatiotemporal state record data, the data is grouped and associated according to the agent number. The spatial coordinates of each agent are continuously linked in order of time simulation step size. The position data of adjacent time nodes are bound with the corresponding state record data. The starting point, ending point, movement direction and state change reason of the trajectory segment are clarified. After removing invalid and redundant data, multiple trajectory chains are constructed, each containing a temporal sequence, position chain, state change chain and association parameters, so as to realize the structured presentation of the agent evacuation trajectory.

[0126] Data augmentation and quality optimization were combined to enhance multiple trajectory chains. Missing time node location data in the trajectory chains were supplemented by interpolation, and abnormal state identifiers were corrected based on adjacent state record data. At the same time, individual attribute parameters of the agents, real-time change data of dynamic fire risk field, and unit features of evacuation network topology were integrated to supplement the risk level, traffic capacity, comprehensive traffic cost and other related information of each node in the trajectory chain. All enhanced trajectory chains were integrated in a unified format to construct a passenger individual trajectory dataset that covers all agents, has rich data dimensions, and meets the accuracy standards.

[0127] Based on passenger individual trajectory datasets, density peak clustering algorithm combined with spatiotemporal distance metric model is adopted. Reasonable time thresholds and spatial distance thresholds are set to perform spatiotemporal clustering analysis on the trajectory chains of all agents. Trajectory chains with similar time series and high spatial path overlap are grouped into the same cluster. At the same time, isolated abnormal trajectory segments are removed to generate multiple compact and distinctive spatiotemporal clusters. Each cluster clearly includes core trajectory features, the proportion of agents, the time distribution range and the spatial coverage area, intuitively presenting the evacuation trajectory clustering characteristics of different groups.

[0128] For multiple spatiotemporal clusters, a trajectory pattern mining algorithm is used to extract key indicators such as core path features, traffic efficiency, risk exposure level, and conflict frequency for each cluster. The clusters are then classified and identified based on their size, coverage, and frequency of occurrence: the trajectory patterns of the largest, most extensive, and most frequent clusters are identified as the mainstream evacuation path patterns; the trajectory patterns of medium-sized clusters with specific regional adaptability are identified as secondary path patterns; and trajectory patterns that deviate from the conventional path, exhibit obvious detours, stagnation, or high-risk exposure are identified as abnormal path patterns. This comprehensive analysis of the characteristics of various evacuation paths in fire scenarios is used to identify these patterns.

[0129] Using the evacuation network topology map as a spatial reference, the trajectory data of mainstream evacuation path patterns, secondary path patterns, and abnormal path patterns are accurately mapped to the grid cells of the topology map through spatial coordinate association technology, clarifying the basic environmental spatial units, connectivity relationships, and key nodes covered by each path pattern. Simultaneously, dynamic fire risk field data is loaded, and core risk parameters such as real-time risk level, smoke concentration, temperature distribution, and risk diffusion rate are extracted from the corresponding grid cells of each path pattern. A risk quantification model is used to transform multi-dimensional risk indicators into unified risk quantification values. Finally, a safety assessment system is constructed, including risk exposure assessment, traffic efficiency assessment, and bottleneck identification assessment. The comprehensive assessment system includes risk exposure assessment, which calculates the risk exposure duration, high-risk area traversal ratio, and risk accumulation value for each route mode; traffic efficiency assessment, which analyzes the average traffic speed, congestion node distribution, and evacuation time; and bottleneck identification assessment, which locates key road sections with insufficient capacity, prone to congestion, or concentrated risks. Based on the assessment results, safety ratings are given to the three types of route modes, and a list of risk points, efficiency shortcomings, and optimization directions are compiled. Finally, through structured report generation technology, basic route information, safety ratings, quantitative assessment data, risk point details, and improvement suggestions are integrated to generate a complete and data-supported safety diagnostic report.

[0130] Based on safety diagnostic reports and considering the characteristics of three evacuation route patterns and the actual spatial structure of target rail transit stations, a multi-dimensional evacuation guidance optimization system is constructed. For mainstream evacuation route patterns, route signage is optimized, bottleneck sections are widened to improve traffic efficiency, and protection and guidance in high-risk areas are strengthened. For secondary route patterns, connectivity design is improved, emergency signage is added, and their reliability and accessibility as alternative routes are enhanced. For abnormal route patterns, the causes are analyzed, and targeted avoidance strategies are developed, guiding passengers to avoid risky routes through broadcast guidance and ground signage. Simultaneously, the optimization schemes for the three route patterns are integrated, and in conjunction with the station's emergency management needs, hierarchical and implementable evacuation route optimization recommendations are formulated, covering hardware facility upgrades, guidance mechanism improvements, and key points of emergency drills, providing comprehensive support for improving the safety and efficiency of fire emergency evacuation at rail transit stations.

[0131] Example 2, based on the same inventive concept as the simulation analysis method for emergency evacuation in rail transit fires described in the previous examples, such as... Figure 2 As shown, this application provides a simulation analysis system for emergency evacuation in rail transit fires. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0132] The data acquisition module 10 is used to collect three-dimensional spatial structure data of the target rail transit station, fire scene parameters, and initial passenger distribution data.

[0133] The evacuation simulation model construction module 20 is used to construct an evacuation simulation model of the intelligent passenger based on the three-dimensional spatial structure data and the initial passenger distribution data.

[0134] The dynamic fire risk field data generation module 30 is used to load the fire scene parameters, dynamically simulate the fire development process, and generate dynamic fire risk field data.

[0135] The evacuation simulation result construction module 40 is used to couple the dynamic fire risk field data with the evacuation simulation model, drive the intelligent passenger to make path decisions, and construct the evacuation simulation result.

[0136] The evacuation route optimization suggestion construction module 50 is used to identify the trajectory of intelligent passenger based on the evacuation simulation results, generate a passenger individual trajectory dataset for dynamic path analysis, and construct evacuation route optimization suggestions.

[0137] Furthermore, the system also includes:

[0138] A panoramic scan of the target rail transit station is performed using a 3D laser scanner to acquire 3D point cloud data. Noise reduction and registration are then performed on the 3D point cloud data to construct a 3D point cloud model. Building identification is performed throughout the target rail transit station to extract key structural elements, which are then mapped onto the 3D point cloud model to construct 3D spatial structure data. A fire risk assessment is conducted based on the 3D point cloud model, and fire source power analysis is performed on multiple potential ignition locations to generate fire source power growth curves. Ventilation and smoke extraction markers are then applied to the 3D point cloud model to determine the station's environmental ventilation conditions. Based on the station's environmental ventilation conditions and the fire source power growth curves, multiple potential ignition locations are mapped to construct fire scene parameters. Finally, accessibility analysis is performed on the 3D spatial structure data to construct accessible spaces. Historical passenger flow data from the target rail transit station is then mapped onto the accessible spaces for spatial distribution, generating the initial passenger distribution data.

[0139] Furthermore, the system also includes:

[0140] The three-dimensional spatial structure data is discretized into a raster to construct an evacuation network topology map. Based on this topology map, the target rail transit station is initialized to construct basic environmental spatial units. The initial passenger distribution data is mapped one by one to the evacuation network topology map according to these basic environmental spatial units, constructing multiple intelligent passenger agents. The attributes of these intelligent passenger agents are configured, generating multiple sets of individual attribute parameters, which correspond to the individual passenger agents. Multiple simulation time steps are set, and spatial proximity relationships are obtained based on the individual passenger agents and their attribute parameters at each time step. These spatial proximity relationships are mapped to the multiple intelligent passenger agents to construct a neighborhood list. Dynamic interaction analysis is performed at each time step based on the neighborhood list to construct a dynamic interaction network. The dynamic interaction network is then added to the evacuation simulation model of the intelligent passenger agents.

[0141] Furthermore, the system also includes:

[0142] The process involves iterating through multiple intelligent passenger agents and matching them with multiple simulation time steps to obtain multiple matching identifiers. Based on these matching identifiers and multiple sets of individual attribute parameters, the process maps them to basic environmental spatial units to determine the unit coordinates of the multiple intelligent passenger agents. Distance calculations are performed on the unit coordinates of the multiple intelligent passenger agents to obtain multiple Euclidean distances. The multiple intelligent passenger agents are then arranged according to their distance identifiers based on these Euclidean distances to construct a global distance matrix. Finally, proximity analysis is performed on the multiple intelligent passenger agents based on the global distance matrix to obtain spatial proximity relationships.

[0143] Furthermore, the system also includes:

[0144] Based on the fire scenario parameters, a physical field simulation is performed to construct a fire physical field engine; based on the fire physical field engine, spatial matching of target rail transit stations is performed to generate a computational domain network; the multiple potential ignition locations are mapped to the computational domain network for driving simulation to obtain spatiotemporal distribution data of fire source disasters; based on the spatiotemporal distribution data of fire source disasters, risk assessment is performed to construct dynamic fire risk factors; the dynamic fire risk factors are mapped to the evacuation network topology map for data association to generate dynamic fire risk field data.

[0145] Furthermore, the system also includes:

[0146] Based on the computational domain network and the evacuation network topology, a spatial coordinate correlation analysis is performed to construct a spatial coordinate mapping relationship. According to the spatial coordinate mapping relationship, the dynamic fire risk factors are spatially interpolated and assigned to the evacuation network topology to obtain spatial interpolation results. Based on the evacuation network topology, a traffic assessment is performed to construct initial traffic capacity attribute data. According to the spatial interpolation results, the dynamic fire risk factors are correlated with the initial traffic capacity attribute data to generate the dynamic fire risk field data.

[0147] Furthermore, the system also includes:

[0148] Based on the initial passenger distribution data, personnel density is calculated to construct a personnel density influencing factor; based on multiple time simulation steps, the dynamic fire risk field data is mapped to the evacuation network topology for passage calculation to obtain an initial passage cost; based on the personnel density influencing factor, the initial passage cost, and the dynamic fire risk factor, a weighted sum is performed to obtain a comprehensive passage cost; based on the comprehensive passage cost, the intelligent agent passenger is driven to perform simulation to generate multiple movement intentions; based on the multiple movement intentions, movement conflict analysis is performed to extract discrete conflict events; the discrete conflict events are mapped to the evacuation network topology for path decision-making to construct the evacuation simulation results.

[0149] Furthermore, the system also includes:

[0150] A conflict arbitrator is constructed to calculate and adjudicate discrete conflict events, generating a movement intention modification instruction. This instruction is then used to analyze the movement of multiple intelligent passenger agents, generating multiple candidate target units. The movement intentions are updated based on these candidate target units to generate a passenger target movement intention. The passenger target movement intention is verified; if verification is successful, the passenger target movement intention is executed to perform path simulation, constructing the temporal positions of the multiple intelligent passenger agents. Finally, the temporal positions of the multiple intelligent passenger agents are added to the evacuation simulation results.

[0151] Furthermore, the system also includes:

[0152] State analysis is performed based on the temporal locations of multiple intelligent passenger agents to obtain spatiotemporal state record data of multiple intelligent passenger agents; trajectory structuring is performed based on the temporal locations and the spatiotemporal state record data to construct multiple trajectory chains; data augmentation is performed on the multiple trajectory chains to construct a passenger individual trajectory dataset; spatiotemporal clustering is performed based on the passenger individual trajectory dataset to generate multiple spatiotemporal clusters; dynamic mining is performed according to the multiple spatiotemporal clusters to identify multiple fire scenario evacuation path patterns, which include mainstream evacuation path patterns, secondary path patterns, and abnormal path patterns; safety assessments are performed on the mainstream evacuation path patterns, secondary path patterns, and abnormal path patterns based on the evacuation network topology map and the dynamic fire risk field data to generate a safety diagnosis report; based on the safety diagnosis report and the mainstream evacuation path patterns, secondary path patterns, and abnormal path patterns, evacuation guidance is provided to the target rail transit station to construct the evacuation path optimization suggestions.

[0153] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0154] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0155] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A simulation analysis method for emergency evacuation in rail transit fires, characterized in that, The method includes: Collect three-dimensional spatial structure data, fire scene parameters, and initial passenger distribution data of the target rail transit station; Based on the three-dimensional spatial structure data and the initial passenger distribution data, an evacuation simulation model of the intelligent passenger is constructed. The fire scenario parameters are loaded to dynamically simulate the fire development process and generate dynamic fire risk field data. The dynamic fire risk field data is coupled with the evacuation simulation model to drive the intelligent passenger to make path decisions and construct the evacuation simulation results. Based on the evacuation simulation results, the trajectory of passengers is identified by the intelligent agent, a dataset of individual passenger trajectories is generated for dynamic path analysis, and evacuation path optimization suggestions are constructed.

2. The simulation analysis method for emergency evacuation in rail transit fires as described in claim 1, characterized in that, The methods for collecting three-dimensional spatial structure data, fire scene parameters, and initial passenger distribution data of the target rail transit station include: A panoramic scan of the target rail transit station is performed using a 3D laser scanner to obtain 3D point cloud data; Denoising and registration are performed on the aforementioned 3D point cloud data to construct a 3D point cloud model. The target rail transit stations are traversed to identify buildings, extract key structural elements, and map the key structural elements to the three-dimensional point cloud model to construct three-dimensional spatial structure data. Fire risk assessment is performed based on the three-dimensional point cloud model. Multiple potential ignition locations are preset for fire source power analysis, and fire source power growth curves are plotted. Based on the three-dimensional point cloud model, ventilation and smoke extraction markings are made to determine the site's environmental ventilation conditions. Based on the site's environmental ventilation conditions and the fire source power growth curve, multiple potential ignition locations are mapped to construct fire scenario parameters. Based on the three-dimensional spatial structure data, a passability analysis is performed to construct a passable space. Historical passenger flow data of the target rail transit station is then mapped to the passable space for spatial distribution, generating the initial passenger distribution data.

3. The simulation analysis method for emergency evacuation in rail transit fires as described in claim 2, characterized in that, Based on the aforementioned three-dimensional spatial structure data and the initial passenger distribution data, an evacuation simulation model for intelligent passenger agents is constructed, the method comprising: The three-dimensional spatial structure data is discretized into a raster to construct an evacuation network topology map; Based on the evacuation network topology, the target rail transit station is initialized environmentally to construct basic environmental spatial units. The initial passenger distribution data is mapped one by one to the evacuation network topology according to the basic environmental spatial unit, and multiple intelligent passenger agents are constructed. The multiple intelligent agent passengers are traversed and their attributes are configured to generate multiple sets of individual attribute parameters, and the multiple sets of individual attribute parameters have a corresponding relationship with the multiple intelligent agent passengers. Multiple simulation time steps are set, and spatial proximity relationships are obtained by calculating based on the multiple intelligent passenger agents and the multiple sets of individual attribute parameters according to the multiple simulation time steps. Based on the spatial proximity relationship, a neighborhood list is constructed by mapping multiple intelligent agent passengers; Based on the neighborhood list, a dynamic interaction network is constructed by performing dynamic interaction analysis according to the multiple simulation time steps. The dynamic interactive network is added to the evacuation simulation model of the intelligent passenger agent.

4. The simulation analysis method for emergency evacuation in rail transit fires as described in claim 3, characterized in that, Multiple simulation time steps are set, and spatial proximity relationships are obtained by calculating based on the multiple intelligent passenger agents and the multiple sets of individual attribute parameters according to the multiple simulation time steps. The method includes: The multiple intelligent passenger agents are matched with the multiple simulation time steps to obtain multiple matching identifiers; Based on the multiple matching identifiers and the multiple sets of individual attribute parameters, the unit coordinates of multiple intelligent passenger agents are determined by mapping them to basic environmental space units. Distance calculations are performed on the unit coordinates of the multiple intelligent passenger agents to obtain multiple coordinate Euclidean distances; The multiple intelligent passenger agents are arranged according to the Euclidean distance of the multiple coordinates to construct a global distance matrix; Based on the global distance matrix, a proximity analysis is performed on multiple intelligent passenger agents to obtain spatial proximity relationships.

5. The simulation analysis method for emergency evacuation in rail transit fires as described in claim 3, characterized in that, The method for dynamically simulating the fire development process by loading the fire scene parameters and generating dynamic fire risk field data includes: Based on the fire scene parameters, a physical field simulation is performed to construct a fire physical field engine. Based on the fire physics engine, spatial matching of target rail transit stations is performed to generate a computational domain network; The multiple potential fire locations are mapped to a computational domain network for driving simulation to obtain spatiotemporal distribution data of fire source disasters; Risk assessment is conducted based on the spatiotemporal distribution data of fire sources to construct dynamic fire risk factors. The dynamic fire risk factors are mapped to the evacuation network topology for data association, generating dynamic fire risk field data.

6. The simulation analysis method for emergency evacuation in rail transit fires as described in claim 5, characterized in that, The method involves mapping the dynamic fire risk factors to the evacuation network topology for data association to generate dynamic fire risk field data, including: Based on the computational domain network and the evacuation network topology, spatial coordinate correlation analysis is performed to construct a spatial coordinate mapping relationship; Based on the spatial coordinate mapping relationship, the dynamic fire risk factors are spatially interpolated and assigned to the evacuation network topology to obtain spatial interpolation results; Based on the evacuation network topology map, a traffic assessment is performed to construct initial traffic capacity attribute data; Based on the spatial interpolation results, the dynamic fire risk factor is correlated with the initial accessibility attribute data to generate the dynamic fire risk field data.

7. The simulation analysis method for emergency evacuation in rail transit fires as described in claim 5, characterized in that, The method involves coupling the dynamic fire risk field data with the evacuation simulation model to drive the intelligent passenger agent to make path decisions and construct the evacuation simulation results. Based on the initial passenger distribution data, the passenger density is calculated, and a passenger density influencing factor is constructed. Based on multiple time simulation steps, the dynamic fire risk field data is mapped to the evacuation network topology map for passage calculation to obtain the initial passage cost. The comprehensive passage value is obtained by weighted summation of the personnel density influence factor, the initial passage value, and the dynamic fire risk factor. Based on the comprehensive passage cost, the intelligent agent passenger is driven to simulate and generate multiple movement intentions; Based on the multiple movement intentions, perform movement conflict analysis to extract discrete conflict events; The discrete conflict events are mapped to the evacuation network topology for path decision-making, and the evacuation simulation results are constructed.

8. The simulation analysis method for emergency evacuation in rail transit fires as described in claim 7, characterized in that, The method involves mapping the discrete conflict events to the evacuation network topology for path decision-making and constructing the evacuation simulation results. Construct a conflict arbitrator, and use the conflict arbitrator to calculate and adjudicate discrete conflict events, generating a movement intention modification instruction; The movement analysis of multiple intelligent passenger agents is performed using the movement intent modification command to generate multiple candidate target units. Multiple movement intentions are updated based on the multiple candidate target units to generate passenger target movement intentions; Verify the passenger's target movement intention. If the verification is successful, execute the passenger's target movement intention to perform path simulation and construct the temporal positions of multiple intelligent agent passengers. The temporal locations of the multiple intelligent passenger agents are added to the evacuation simulation results.

9. The simulation analysis method for emergency evacuation in rail transit fires as described in claim 8, characterized in that, Based on the evacuation simulation results, the trajectory of passengers is identified by the intelligent agent, a dataset of individual passenger trajectories is generated for dynamic path analysis, and evacuation path optimization suggestions are constructed. The method includes: Based on the temporal location of the multiple intelligent passenger agents, state analysis is performed to obtain spatiotemporal state record data of the multiple intelligent passenger agents. Based on the temporal location and the spatiotemporal state recording data, the trajectory is structured to construct multiple trajectory chains; The multiple trajectory chains are augmented to construct a dataset of individual passenger trajectories; Based on the passenger individual trajectory dataset, spatiotemporal clustering is performed to generate multiple spatiotemporal clusters; Dynamic mining is performed based on the multiple spatiotemporal clusters to identify multiple fire scene evacuation path patterns, which include mainstream evacuation path patterns, secondary path patterns, and abnormal path patterns. Based on the evacuation network topology and the dynamic fire risk field data, a safety assessment is performed on the mainstream evacuation path pattern, the secondary path pattern, and the abnormal path pattern, and a safety diagnosis report is generated. Based on the safety diagnosis report, combined with the mainstream evacuation route mode, the secondary route mode, and the abnormal route mode, evacuation guidance is provided for the target rail transit station, and the evacuation route optimization suggestions are constructed.

10. A simulation analysis system for emergency evacuation in rail transit fires, characterized in that, The system is used to implement the simulation analysis method for emergency evacuation in rail transit fires as described in any one of claims 1-9, and the system includes: The data acquisition module is used to collect three-dimensional spatial structure data, fire scene parameters, and initial passenger distribution data of the target rail transit station. An evacuation simulation model construction module is used to construct an evacuation simulation model of intelligent passenger based on the three-dimensional spatial structure data and the initial passenger distribution data. The dynamic fire risk field data generation module is used to load the fire scene parameters, dynamically simulate the fire development process, and generate dynamic fire risk field data. The evacuation simulation result construction module is used to couple the dynamic fire risk field data with the evacuation simulation model, drive the intelligent passenger to make path decisions, and construct the evacuation simulation result. The evacuation route optimization suggestion construction module is used to identify the trajectory of intelligent passenger based on the evacuation simulation results, generate individual passenger trajectory datasets for dynamic path analysis, and construct evacuation route optimization suggestions.

Images