An underground space evacuation risk assessment method

Abstract

This invention proposes a method for assessing evacuation risks in underground spaces. The method includes: step 1) constructing a physical field model for toxic gas diffusion; step 2) constructing a pedestrian evacuation model; step 3) achieving dynamic data coupling of multiple physics fields; step 4) setting simulation conditions and executing coupled simulation; and step 5) analyzing the results and extracting risk thresholds. This invention can couple the physical diffusion of toxic gas with the dynamic response of personnel behavior in real time, overcoming the limitation of traditional methods that separate the two. It can realistically simulate the complex dynamic interaction process between toxic gas diffusion and personnel evacuation, significantly improving the realism and reliability of dynamic risk assessment for personnel evacuation.

Description

Technical Field

[0001] This invention relates to a method for assessing evacuation risks in underground spaces, belonging to the field of public safety and emergency management technology. Background Technology

[0002] Underground spaces (such as subway stations, utility tunnels, and underground shopping malls) face severe evacuation challenges in the event of toxic gas leaks due to their enclosed nature, high population density, and limited evacuation routes. Accurately assessing the evacuation risks under such accidents is crucial for developing effective emergency strategies and ensuring public safety.

[0003] Currently, research methods for emergency response to toxic gas leaks mainly fall into two categories: one is to use CFD (Computational Fluid Dynamics) software to simulate the diffusion process of toxic gas and obtain the spatial concentration distribution, but such studies are usually not coupled with dynamic personnel evacuation behavior in real time; the other is to use evacuation simulation software to simulate personnel movement, but often presupposes the toxic gas hazard as a static or simplified condition, which cannot reflect the dynamic interaction between leak diffusion and pedestrian behavior. Existing methods are difficult to quantify the real-time evolution of personnel physiological state in a dynamically changing toxic environment and its feedback impact on the efficiency of group evacuation, resulting in insufficient reliability of risk assessment and scientific rigor of system safety design. Summary of the Invention

[0004] This invention proposes a method for assessing evacuation risks in underground spaces, aiming to improve the accuracy and reliability of evacuation risk prediction and the scientific nature of emergency strategies.

[0005] The technical solution of the present invention: a method for assessing evacuation risks in underground spaces, the method comprising:

[0006] Step 1) Construct a physical field model for the diffusion of toxic gas;

[0007] Step 2) Construct a pedestrian evacuation model;

[0008] Step 3) Achieve dynamic data coupling for multiphysics;

[0009] Step 4) Set the simulation conditions and perform the coupled simulation;

[0010] Step 5) Analyze the results and extract the risk threshold.

[0011] Furthermore, the construction of the physical field model for toxic gas diffusion specifically includes: establishing a CFD calculation model based on the actual three-dimensional geometric structure of the target underground space, setting different leakage source locations, leakage outlet directions and gas types, simulating the dynamic diffusion process of toxic gas in a closed space, and extracting spatiotemporal concentration field data of a preset monitoring surface.

[0012] Furthermore, the construction of the pedestrian evacuation model specifically includes: establishing a pedestrian evacuation model in an intelligent agent simulation platform, and defining the initial state, movement logic, and decision rules of pedestrians.

[0013] Furthermore, the intelligent agent simulation platform is the AnyLogic simulation platform; the AnyLogic simulation platform is equipped with a response mechanism, that is, after the warning stage is triggered, the AnyLogic simulation platform will activate a preset response time T. 响应 When the response time T 响应 After the event concluded, pedestrians officially began to disperse.

[0014] Furthermore, the pedestrian evacuation model introduces a pedestrian status determination mechanism based on real-time toxic gas exposure dose. By acquiring the toxic gas concentration at the pedestrian's location in real time, calculating the cumulative amount of toxic gas inhaled, and dynamically determining the pedestrian's status as normal, injured, or dead according to preset dose-state transition rules, the model adjusts the pedestrian's movement speed accordingly.

[0015] Furthermore, the dose-state transition rule includes the AEGL standard and the mortality probability curve, and the AEGL standard is curve-fitted using MATLAB.

[0016] Furthermore, the realization of multi-physics dynamic data coupling specifically includes: to solve the problem of data interaction and time synchronization between the CFD physical field and the ABM behavioral field, the spatiotemporal concentration field data output in step 1) is interpolated and format converted by data processing software, and the spatiotemporal concentration field data is processed into a concentration function that can be called by pedestrians in real time in step 2), thereby constructing a second-level interaction interface between the CFD physical field and the ABM behavioral field, and realizing closed-loop feedback simulation between the CFD physical field and the ABM behavioral field.

[0017] Furthermore, the process of processing the spatiotemporal concentration field data into a concentration function that can be called by pedestrians in real time in step 2) to construct a second-level interactive interface between the CFD physical field and the ABM behavioral field specifically includes using the data fitting toolbox cftool (Curve Fitting Tool) provided by MATLAB software to establish a spatiotemporally aligned toxic gas concentration query function for the ABM behavioral field to call in real time within the simulation step.

[0018] Furthermore, the setting of simulation conditions and execution of coupled simulation specifically includes: comprehensively considering the leakage scenario, evacuation scale and emergency response time, designing several sets of comparative simulation conditions; and sequentially running the CFD physical field and ABM behavioral field to simulate the entire process from toxic gas leakage, monitoring alarm, personnel response to completion of evacuation.

[0019] Furthermore, the analysis results and risk threshold extraction specifically include: collecting and statistically analyzing personnel casualty data and evacuation time distribution information under various working conditions; quantifying and revealing key risk influencing factors through data fitting and comparison, identifying the critical emergency response time threshold that determines the system's safety status, clarifying its nonlinear effect, and proposing dynamic and targeted emergency evacuation optimization strategies accordingly.

[0020] The beneficial effects of this invention are:

[0021] 1) This invention can couple the physical diffusion of toxic gas with the dynamic response of personnel behavior in real time, and establishes a simulation framework for the second-level dynamic coupling of the physical field and behavioral field of CFD (Computational Fluid Dynamics) and ABM (agent-based modeling). It overcomes the limitation of traditional methods that separate the two, and can realistically simulate the complex dynamic interaction process between toxic gas diffusion and personnel evacuation, significantly improving the realism and reliability of dynamic risk assessment of personnel evacuation.

[0022] 2) The pedestrian status dynamic determination mechanism based on real-time cumulative inhalation of toxic gas proposed in this invention combines the toxic mechanism of the toxin with individual behavioral feedback, and realizes a refined modeling of the physiological and behavioral evolution of people in hazardous environments, providing a new method for quantifying individual and group risks.

[0023] 3) Through systematic simulation, this invention can quantitatively identify the key control variables and their critical thresholds (such as emergency response time) that affect system safety. The conclusions can be directly used to guide the optimization of emergency plans (such as setting alarm response time red lines) and the formulation of dynamic evacuation strategies (such as dynamically adjusting the guidance scheme according to the direction of leakage and gas density). It has important practical application value and helps to improve the safety resilience of underground spaces. Attached Figure Description

[0024] Appendix Figure 1 This is a schematic diagram of the overall process of the method of the present invention.

[0025] Appendix Figure 2 This is a schematic diagram of the concentration field of toxic gas diffusion simulated by CFD in Embodiment 1 of the present invention (taking chlorine as an example).

[0026] Appendix Figure 3 This is a schematic diagram of the response mechanism logic in Embodiment 1 of the present invention. Detailed Implementation

[0027] A method for assessing evacuation risks in underground spaces, comprising the following steps:

[0028] Step 1) Construct a physical field model for the diffusion of toxic gas;

[0029] Step 2) Construct a pedestrian evacuation model;

[0030] Step 3) Achieve dynamic data coupling for multiphysics;

[0031] Step 4) Set the simulation conditions and perform the coupled simulation;

[0032] Step 5) Analyze the results and extract the risk threshold.

[0033] The construction of the physical field model for toxic gas diffusion specifically includes: establishing a CFD calculation model based on the actual three-dimensional geometric structure of the target underground space, setting different leakage source locations, leakage outlet directions (X, Y, Z) and gas types (such as chlorine Cl2, carbon monoxide CO), simulating the dynamic diffusion process of toxic gas in a closed space, and extracting spatiotemporal concentration field data of a preset monitoring surface; preferably, the RNG k-ε turbulence model is used to simulate the dynamic diffusion process of toxic gas in a closed space.

[0034] The construction of the pedestrian evacuation model specifically includes: establishing a pedestrian evacuation model in an intelligent agent simulation platform, defining the initial state, movement logic, and decision rules of pedestrians; the intelligent agent simulation platform is preferably AnyLogic simulation platform.

[0035] The core of the pedestrian evacuation model lies in introducing a pedestrian status determination mechanism based on real-time toxic gas exposure dose. By acquiring the toxic gas concentration at the pedestrian's location in real time, calculating the cumulative amount of toxic gas inhaled, and dynamically determining the pedestrian's status as normal, injured, or dead according to a preset dose-state transition rule (referencing the AEGL standard for function fitting and death probability curves), the model adjusts the pedestrian's movement speed accordingly. The injury includes minor, moderate, and serious injuries.

[0036] The dose-state transition rules include: existing AEGL (Acute Exposure Guideline Levels) standards and existing mortality probability curves, which are used as one of the criteria for determining pedestrian mortality.

[0037] Since the shortest time standard of the existing AEGL standard is 10 minutes, it is not suitable for simulation scenarios of short-term high-concentration exposure; therefore, MATLAB is used to perform curve fitting on the existing AEGL standard to obtain the AEGL standard for short time.

[0038] The implementation of multi-physics dynamic data coupling specifically includes: to solve the problem of data interaction and time synchronization between the CFD physical field and the ABM behavioral field, the spatiotemporal concentration field data output in step 1) is interpolated and format converted by data processing software, and the spatiotemporal concentration field data is processed into a concentration function that can be called by pedestrians in real time in step 2), thereby constructing a second-level interaction interface between the CFD physical field and the ABM behavioral field, and realizing closed-loop feedback simulation between the CFD physical field and the ABM behavioral field.

[0039] The process of processing the spatiotemporal concentration field data into a concentration function that can be called in real time by pedestrians in step 2), thereby constructing a second-level interactive interface between the CFD physical field and the ABM behavioral field, specifically includes: using MATLAB software, the core of which is to use the data fitting toolbox cftool of MATLAB software to establish a spatiotemporally aligned toxic gas concentration query function, which can be called in real time by the ABM behavioral field within the simulation step.

[0040] The AnyLogic simulation platform is equipped with a response mechanism, meaning that after the warning phase is triggered, the AnyLogic simulation platform will activate a preset response time T. 响应 T 响应 The response time T represents the total time from when pedestrians and staff discover an anomaly and evacuation conditions are met, report it to the command center for assessment and alert issuance, to when pedestrians receive the alert and begin evacuation procedures. This timeframe specifically includes key delays such as perception delay, decision-making delay, and information transmission delay. 响应 After the event concludes, pedestrians officially enter the evacuation phase; when all surviving pedestrians have successfully evacuated to a safe location, the ABM behavioral field simulation of pedestrian status ends.

[0041] The setting of simulation conditions and execution of coupled simulations specifically includes: comprehensively considering the leakage scenario (direction, gas type), evacuation scale and emergency response time, designing several sets of control simulation conditions; sequentially running the CFD physical field and ABM behavioral field to simulate the entire process from toxic gas leakage, monitoring alarm, personnel response to completion of evacuation; to eliminate random errors, the simulation conditions of each parameter combination in the ABM behavioral field simulation are repeated more than 10 times, preferably 10 times, and the average value of the results is taken.

[0042] The analysis results and risk threshold extraction specifically include: collecting and statistically analyzing personnel casualty data and evacuation time distribution information under various working conditions; quantifying and revealing key risk influencing factors through data fitting and comparison, especially identifying the critical emergency response time threshold that determines the system's safety status, clarifying its nonlinear effect, and proposing dynamic and targeted emergency evacuation optimization strategies accordingly.

[0043] Example 1

[0044] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0045] A dynamic evacuation risk assessment under a sudden toxic gas leak scenario at a subway platform level, specifically including:

[0046] S1: Physical field modeling and simulation calculation of toxic gas diffusion;

[0047] S2: Pedestrian evacuation behavior field modeling and dynamic state determination;

[0048] S3: Implementation of the multiphysics dynamic coupling interface;

[0049] S4: Simulation condition execution;

[0050] S5: Results Analysis and Strategy Extraction.

[0051] S1: Physical field modeling and simulation calculation of toxic gas diffusion specifically includes:

[0052] A three-dimensional geometric model with a length of 115 m, a width of 11.6 m, and a height of 4 m was established using a standard subway platform. The model was then imported into ANSYS Fluent 2023 R1 for fluid computation mesh generation. The total number of meshes was approximately 2.14 million, and mesh independence was verified.

[0053] The toxic gas leak source was located near the staircase on one side of the platform. Leakage outlets were set along three directions in a Cartesian coordinate system: X (horizontal), Y (vertical upward), and Z (horizontal and vertical). The leakage and diffusion of typical toxic gases, chlorine (Cl2) and carbon monoxide (CO), were simulated. An RNG k-ε turbulence model was used to simulate the unsteady, buoyancy-affected complex flow within the platform. The governing equations are as follows:

[0054] (1); (2);

[0055] in, This indicates the generation of average turbulent kinetic energy; This indicates the generation of turbulent kinetic energy caused by buoyancy; It is the contribution of fluctuating expansion in compressible turbulence to the total dissipation rate; and They are and The inverse effective Prandtl number; and It is a user-defined source item; For fluid density; For time; , For Cartesian coordinate components; For velocity components, corresponding direction; It is turbulent kinetic energy; The turbulent dissipation rate; Effective dynamic viscosity; , and These are model constants; This is an additional feature unique to the RNG model.

[0056] The flow solver used in Fluent calculations employs a semi-implicit method (SIMPLE) algorithm that couples pressure and velocity, using a second-order upwind scheme. Temperature affects the thermal plume around the human body and the buoyancy of particles, and may generate turbulence around the human body; therefore, the Boussinesq model is used to consider the buoyancy effect.

[0057] S2: Pedestrian evacuation behavior field modeling and dynamic state determination specifically includes:

[0058] A pedestrian evacuation model was built in the AnyLogic simulation platform, defining four types of pedestrians with different behavioral characteristics. The model simulated daily scenarios of "waiting for a bus - boarding the bus" and "going upstairs - exiting the station" within 60 seconds before the leak occurred. Pedestrian movement followed a social force model.

[0059]

[0060] In the formula, It is the quality of pedestrian i. It is the actual speed of pedestrian i. (t) represents the direction of the actual velocity. This represents the expected speed of pedestrian i. Indicates the direction of the desired velocity. Corresponding to the adaptation time of pedestrian i, This represents the repulsive force from pedestrian j, while This represents the repulsive force from fixed obstacles such as walls. For time.

[0061]

[0062] In the formula, Let t be the position of the pedestrian. The concentration of toxic gas below; Let t be the pedestrian's temporal velocity. This represents the cumulative amount of toxic gas inhaled by pedestrians.

[0063] in accordance with The values ​​are used to dynamically determine the pedestrian's state based on a pre-set dose-state transition rule (fitted with reference to the AEGL standard), and adjust their mobility accordingly; the velocity influence coefficient for each state... As shown in the table below:

[0064] ;

[0065] In the formula, The pedestrian's expected speed of movement; Theoretical evacuation speed for pedestrians; represents the influence coefficient under each state.

[0066] Meanwhile, to more scientifically reflect the randomness of individual risk, a probability-based mortality determination mechanism is introduced into the pedestrian status determination mechanism, including the mortality probability. Calculated using the following formula:

[0067]

[0068]

[0069] In the formula, The cumulative inhaled toxic gas dose calculated above, in mg; and The constant is related to the type of poison gas; F is used to calculate the probability of death. Intermediate variables.

[0070] The implementation of S3, the multiphysics dynamic coupling interface, specifically includes:

[0071] To bridge the data barrier between CFD physical fields and ABM behavioral fields, a data conversion script was written using MATLAB R2021a. This script performs three-dimensional interpolation on the discrete spatiotemporal concentration data exported from Fluent, generating a continuous query function f(t) with respect to spatial coordinates and simulation time. In the AnyLogic simulation platform, within each simulation step, the pedestrian, based on their real-time coordinates (taking breathing height),... The function is called at the current simulation time to obtain the accurate local toxic gas concentration value, thereby achieving second-level data synchronization and closed-loop interaction between the two models;

[0072] In the AnyLogic simulation platform, set the corresponding variable values, pedestrian state parameters, and functions used for numerical calculation, as shown in Table 2.

[0073]

[0074] This embodiment utilizes the logic and event-driven mechanism built into the AnyLogic simulation platform to achieve real-time response and automated management of pedestrian evacuation processes in dynamic toxic gas environments. The specific implementation process of its technical solution is as follows:

[0075] X1: Pedestrian environmental perception and data input;

[0076] The ABM behavior field endows each pedestrian with real-time spatial positioning capabilities. By continuously acquiring their planar coordinates (x, y) and matching them with predefined monitoring area boundaries, it automatically determines the toxic gas monitoring zone to which the pedestrian belongs. Subsequently, the system calls the toxic gas function f(t), which is generated by MATLAB and bound to that zone, to read the toxic gas concentration value at the pedestrian's location in real time. ;

[0077] X2: Dynamic calculation of individual cumulative exposure dose;

[0078] Based on the cumulative inhaled toxic gas data acquired in real time, the ABM behavior field continuously calculates and updates the cumulative inhaled toxic gas volume of each pedestrian at a fixed time step (every 1 second) according to the integral algorithm shown in formula (4). The calculation process is completed automatically during simulation, ensuring that the dose value is strictly synchronized with the changes in the spatiotemporal toxic gas field.

[0079] X3: Pedestrian status determination and behavior parameter adjustment;

[0080] The ABM behavioral field is pre-set based on the cumulative amount of toxic gas inhaled. The system automatically changes the pedestrian's current state transition rules in each calculation cycle. The value is compared and calculated with the rules to determine the pedestrian's physiological state (such as normal, minor injury, moderate injury, serious injury, or death). Based on the rules defined in Table 1, the pedestrian's moving speed is dynamically adjusted (by multiplying by the speed influence coefficient α).

[0081] X4: Group status monitoring and emergency response triggering;

[0082] During the simulation, the ABM behavior field automatically performs real-time statistics on the status of all pedestrians and monitors the changes in the number of people at each status level (especially the level representing casualties). When the number of people in a specific status reaches the preset emergency response trigger condition (e.g., the number of seriously injured people exceeds the threshold), the system will activate a delay mechanism. After a preset, configurable emergency response time (e.g., 5 seconds, 15 seconds, 30 seconds), the model automatically switches from "normal state" to "evacuation mode".

[0083] X5: Evacuation mode execution and simulation result generation;

[0084] Once evacuation mode is entered, all pedestrians not determined to be dead will have their movement target changed to the designated exit, and their speed will be adjusted accordingly based on the status determined by X3. The simulation will continue to run until all surviving pedestrians have left the simulation area. The ABM behavior field will automatically record and output key result data throughout the process, including but not limited to: the number of pedestrians in different states at each time, the total evacuation time for each pedestrian, and the final number of casualties, thus providing a complete data foundation for subsequent risk quantification analysis.

[0085] Through the aforementioned series of automated data flow, logical judgment, and state control, this invention achieves seamless and dynamic coupling between the CFD physical field and the ABM behavioral field, fully embodying the core innovation of the integrated simulation of "perception-computation-determination-response" in this invention.

[0086] S4: Simulation condition execution, specifically includes:

[0087] The design incorporates a combination of evacuation scale, response time, and leakage conditions (X-direction Cl2 leakage) to form 6 sets of toxic gas simulations and 24 sets of evacuation simulation conditions. In each evacuation simulation, the coupled model is run 10 times to reduce randomness, and data such as the number of casualties and individual evacuation time are recorded in each simulation.

[0088] S5: Result Analysis and Strategy Extraction specifically includes:

[0089] Statistical analysis revealed that when the response time was ≤15 seconds, the casualty rate was kept at a low level (e.g., <15%). Once the response time exceeded 15 seconds, the casualty rate increased rapidly and non-linearly with the increase in the number of evacuees. Reducing the response time from 30 seconds to 5 seconds increased the average survival rate by more than 60%. Based on this, the core emergency strategy is proposed: the response time from system detection to issuing the evacuation order must be controlled within 15 seconds. At the same time, differentiated dynamic path guidance should be implemented according to the direction of leakage (e.g., the influence range of X-direction jet is large) and the properties of the gas (Cl2 settles, CO floats).

Claims

1. A method for assessing evacuation risks in underground spaces, characterized by: include: Step 1) Construct a physical field model for the diffusion of poison gas; Step 2) Construct a pedestrian evacuation model; Step 3) Achieve dynamic data coupling for multiphysics; Step 4) Set the simulation conditions and perform the coupled simulation; Step 5) Analyze the results and extract the risk threshold.

2. The method for assessing evacuation risks in underground spaces according to claim 1, characterized in that: The construction of the physical field model for toxic gas diffusion specifically includes: establishing a CFD calculation model based on the actual three-dimensional geometric structure of the target underground space, setting different leakage source locations, leakage outlet directions and gas types, simulating the dynamic diffusion process of toxic gas in a closed space, and extracting spatiotemporal concentration field data of a preset monitoring surface.

3. The method for assessing evacuation risks in underground spaces according to claim 1, characterized in that: The construction of the pedestrian evacuation model specifically includes: establishing a pedestrian evacuation model in an intelligent agent simulation platform, and defining the initial state, movement logic, and decision rules of pedestrians.

4. The method for assessing evacuation risks in underground spaces according to claim 3, characterized in that: The intelligent agent simulation platform is the AnyLogic simulation platform; the AnyLogic simulation platform is equipped with a response mechanism, that is, after the warning stage is triggered, the AnyLogic simulation platform will activate a preset response time T. 响应 When the response time T 响应 After the event concluded, pedestrians officially began to disperse.

5. The method for assessing evacuation risks in underground spaces according to claim 1, characterized in that: The pedestrian evacuation model introduces a pedestrian status determination mechanism based on real-time toxic gas exposure dose. By acquiring the toxic gas concentration at the pedestrian's location in real time, calculating the cumulative amount of toxic gas inhaled, and dynamically determining the pedestrian's status as normal, injured, or dead according to preset dose-state transition rules, the model adjusts the pedestrian's movement speed accordingly.

6. The method for assessing evacuation risks in underground spaces according to claim 5, characterized in that: The dose-state transition rule includes the AEGL standard and the mortality probability curve, and the AEGL standard is curve-fitted using MATLAB.

7. The method for assessing evacuation risks in underground spaces according to claim 2, characterized in that: The implementation of multi-physics dynamic data coupling specifically includes: to solve the problem of data interaction and time synchronization between the CFD physical field and the ABM behavioral field, the spatiotemporal concentration field data output in step 1) is interpolated and format converted by data processing software, and the spatiotemporal concentration field data is processed into a concentration function that can be called by pedestrians in real time in step 2), thereby constructing a second-level interaction interface between the CFD physical field and the ABM behavioral field, and realizing closed-loop feedback simulation between the CFD physical field and the ABM behavioral field.

8. The method for assessing evacuation risks in underground spaces according to claim 7, characterized in that: The process of processing the spatiotemporal concentration field data into a concentration function that can be called in real time by pedestrians in step 2) to construct a second-level interactive interface between the CFD physical field and the ABM behavioral field specifically includes using the data fitting toolbox cftool in MATLAB software to establish a spatiotemporally aligned toxic gas concentration query function, which can be called in real time by the ABM behavioral field within the simulation step.

9. The method for assessing evacuation risks in underground spaces according to claim 1, characterized in that: The setting of simulation conditions and execution of coupled simulation specifically includes: comprehensively considering the leakage scenario, evacuation scale and emergency response time, designing several sets of comparative simulation conditions; and sequentially running the CFD physical field and ABM behavioral field to simulate the entire process from toxic gas leakage, monitoring alarm, personnel response to completion of evacuation.

10. The method for assessing evacuation risks in underground spaces according to claim 1, characterized in that: The analysis results and risk threshold extraction specifically include: collecting and statistically analyzing personnel casualty data and evacuation time distribution information under various working conditions; quantifying and revealing key risk influencing factors through data fitting and comparison, identifying the critical emergency response time threshold that determines the system's safety status, clarifying its nonlinear effect, and proposing dynamic and targeted emergency evacuation optimization strategies accordingly.

Images