Virtual simulation emergency drilling method and system based on deep reinforcement learning under mine fire dynamic catastrophe condition

Abstract

The invention relates to a virtual simulation emergency drilling method and system under a mine fire dynamic catastrophe condition based on deep reinforcement learning, and belongs to the field of mine safety management. According to the method, a mine comprehensive database is constructed, environment, equipment, fire source and ventilation system data of a mine are collected, a mine structure model is generated by using a three-dimensional modeling technology, and a mine fire spreading process is simulated in real time through fire dynamics simulation, deep reinforcement learning optimization and a virtual reality technology. The drill personnel enter the fire disaster simulation environment through the virtual reality equipment to carry out emergency operation, and the system carries out real-time feedback according to the operation of the drill personnel and optimizes the fire disaster spreading simulation model. Parameters such as wind speed and wind quantity of the ventilation system and fire spreading are subjected to coupling calculation, so that the accuracy and dynamic response of fire simulation are ensured. According to the system, through cooperative work of a plurality of modules, intelligence, individuation and high efficiency of mine fire emergency drilling are realized, and the emergency response capability of drilling personnel is effectively improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of mine safety management, and relates to a virtual simulation emergency drill method and system for mine fire under dynamic disaster conditions based on deep reinforcement learning. BACKGROUND

[0002] Mine fire is one of the most common and most threatening disasters in mine safety production, which is usually caused by coal seam spontaneous combustion, gas accumulation, electrical equipment failure, improper operation of miners and other factors. After the occurrence of fire, due to the influence of mine ventilation system, the speed of fire spread and the complex layout of internal roadway, the fire can spread rapidly to multiple areas of the mine. The high temperature, smoke and toxic gases caused by the fire not only cause suffocation and poisoning of miners, but also may cause equipment damage, waste of mineral resources, and bring immeasurable economic losses and social impact.

[0003] Although the frequency of mine fire has decreased, its harmfulness still exists. When a fire occurs, the escape and emergency rescue ability of miners is directly related to life safety. In order to effectively respond to mine fire, fire emergency drill in advance has become an important safety measure. Traditional mine fire emergency drill usually relies on on-site simulation, and the drill personnel perform fire drill through preset scenarios. Although these drill methods can improve the emergency response ability to a certain extent, due to their inherent limitations, they cannot perfectly simulate the complexity and disaster process of mine fire.

[0004] Currently, there are several main problems in mine fire emergency drill:

[0005] (1) Insufficient simulation accuracy: Most of the existing mine fire emergency drills use computer simulation or on-site drill methods, but due to the complexity of fire spread process, the existing simulation methods can only be carried out in a certain area, and it is difficult to consider the influence of various factors such as complex terrain, equipment layout, ventilation system and personnel distribution in the mine on the spread of fire. Therefore, the simulation accuracy of the existing technology is low, and it cannot comprehensively and accurately reflect the disaster process of mine fire.

[0006] (2) Coupling problem of fire and ventilation system: The spread of mine fire is not only affected by factors such as fire source, temperature, smoke, but also closely related to the ventilation system of the mine. The changes of air volume, air speed, fan failure and air door damage of the ventilation system will directly affect the spread speed of the fire and the diffusion path of the smoke. The existing mine fire simulation system often ignores the coupling effect between fire spread and ventilation system, resulting in a large deviation between the simulation results and the actual fire disaster process.

[0007] (3) Lack of dynamic response capability: Traditional mine fire emergency drill systems are mostly static systems. The simulated fire scenarios are usually conducted under fixed conditions and cannot respond in real time to changes in the mine environment and the fire spread process. During the drill, the environmental conditions in the mine (such as wind speed, air pressure, temperature, etc.) will change continuously, which has a significant impact on the fire spread process. However, existing systems often cannot adjust these variables in real time, resulting in a lack of dynamic adaptability in the simulated scenarios.

[0008] (4) Low level of intelligence and lack of personalized adaptability: Existing mine fire emergency drill systems are mostly based on preset scenarios and lack the ability to adapt to different mine environments, fire types, and emergency rescue needs. There are significant differences in mine layout, ventilation systems, and equipment configurations among different mines, making it difficult for traditional systems to flexibly adjust according to the actual conditions of the mine. The system lacks sufficient intelligence and cannot meet the personalized and intelligent needs of emergency drills.

[0009] (5) Limitations of Virtual Reality and Augmented Reality Technologies: Although Virtual Reality (VR) and Augmented Reality (AR) technologies have been applied in mine fire emergency drills, current systems still have certain limitations. Existing VR / AR systems mainly focus on static displays and simulations of some simple dynamic effects, and cannot comprehensively and accurately reproduce the fire spread process and personnel reactions, nor can they update the scene in real time through linkage with the fire simulation system. Existing technologies cannot handle complex environmental changes, nor can they effectively support multi-person collaborative drills and dynamic interactions.

[0010] (6) Lack of comprehensive hazard assessment and feedback mechanisms: Existing mine fire emergency drill systems lack the ability to assess hazards after a fire occurs. For example, real-time information such as temperature, gas concentration, and personnel location after a fire cannot be promptly fed back to the drill participants, making it impossible for them to accurately assess the risk of fire spread and the necessary emergency response measures. Mine fire emergency drill systems should have real-time feedback and dynamic assessment capabilities to help drill participants make better decisions and take action.

[0011] To address the aforementioned issues, existing technologies are no longer sufficient to meet the high-precision, highly intelligent, and diverse requirements of mine fire emergency drills. Therefore, developing a new emergency drill system capable of real-time adjustment of fire spread models, intelligent dynamic drills, and considering the interaction between ventilation systems and fire spread has become an urgent need for current mine fire emergency drill technology.

[0012] This invention proposes a virtual simulation emergency drill method and system for mine fire dynamic disaster conditions based on deep reinforcement learning. It adopts virtual reality (VR) and augmented reality (AR) technologies, combined with fire dynamics models, ventilation system simulation, deep reinforcement learning and other technologies, to simulate the spread process of mine fire, smoke diffusion path, and the impact of ventilation system in real time, providing mines with a more accurate, flexible and safe fire emergency drill plan. Summary of the Invention

[0013] In view of this, the purpose of this invention is to provide a virtual simulation emergency drill method and system based on deep reinforcement learning under dynamic disaster conditions of mine fires.

[0014] To achieve the above objectives, the present invention provides the following technical solution:

[0015] A virtual simulation emergency drill method based on deep reinforcement learning under dynamic disaster conditions of mine fires includes the following steps:

[0016] Step 1: Construct a comprehensive mine database and collect various basic parameter data of the mine, including roadway dimensions, cross-sectional shape, elevation difference, equipment dimensions, layout, ventilation facility dimensions, electromechanical equipment data, mining engineering status data, environmental monitoring data, personnel positioning data, fire prevention and extinguishing material performance data, safety rule data, and emergency plan data;

[0017] Step 2: Using 3D modeling technology, based on the data in the comprehensive mine database, establish a 3D model library of various facilities and structures within the mine, and generate dynamic simulation models of fire, smoke, and flames;

[0018] Step 3: Based on the fire dynamics simulation module, a fire accident proxy model is constructed through BP neural network training, and the three-dimensional model and the fire proxy model are used for real-time fire simulation.

[0019] Step 4: Combine the fire simulation results obtained in Step 3 with environmental variables to dynamically render the fire scene and generate the mine fire disaster environment.

[0020] Step 5: Using virtual reality (VR) or augmented reality (AR) technology, support multi-person collaborative drills. Drill participants perform collaborative operations in a mine fire disaster environment. These collaborative operations affect the simulated fire state and are fed back to the fire dynamics simulation module in real time, forming a closed-loop control.

[0021] Furthermore, the 3D modeling technology includes laser scanning modeling, image modeling, and digital carving modeling.

[0022] Furthermore, the fire dynamics simulation module trains a BP neural network based on the input simulation data, including fire source intensity, tunnel cross-section, initial wind speed, ambient temperature, ambient pressure, monitoring point distance, and fire time, and generates a fire accident proxy model, which is used for real-time fire spread simulation.

[0023] Furthermore, the fire scene includes the shape of the flames, the path of the smoke diffusion, the distortion of the heat wave, the changes in the high-temperature area, and the visual effects of objects burning and deforming.

[0024] A virtual simulation emergency drill system for dynamic disaster conditions of mine fires based on deep reinforcement learning, comprising:

[0025] The mine integrated database module is used to store various basic parameter data of the mine, including roadway dimensions, cross-sectional shape, elevation difference, equipment dimensions, layout, ventilation facility dimensions, electromechanical equipment data, mining engineering status data, environmental monitoring data, personnel positioning data, fire prevention and extinguishing material performance data, safety rule data, and emergency plan data. The database module is connected to the 3D modeling module, fire dynamics simulation module, and drill module.

[0026] The 3D modeling module is used to generate a 3D model library of various facilities and structures in the mine based on the data in the mine integrated database module, and to generate dynamic simulation models of fire, smoke and flame. The 3D modeling module is connected to the fire dynamics simulation module and the high-precision rendering module.

[0027] The fire dynamics simulation module constructs a fire accident proxy model through BP neural network training. The fire dynamics simulation module is connected to the 3D modeling module, receives model data and performs real-time fire simulation, and transmits the simulation results to the high-precision rendering module and the drill module.

[0028] The high-precision rendering module is used to render fire disaster scenarios based on the simulation results transmitted by the fire dynamics simulation module, including dynamic effects such as flame shape, smoke diffusion and heat wave distortion. The high-precision rendering module is connected to the exercise module.

[0029] The exercise module uses VR or AR technology to support multi-person collaborative exercises. The exercise module receives the rendering results from the high-precision rendering module and displays real-time disaster scenarios through virtual reality devices. The operational behavior of the exercise personnel is fed back to the fire dynamics simulation module in real time, forming a closed-loop control.

[0030] Furthermore, the fire dynamics simulation module and the 3D modeling module communicate via a data interface, which transmits parameter data of the mine facilities and the fire model.

[0031] Furthermore, the high-precision rendering module generates dynamic effects of heat waves, smoke, and flames related to the dynamic behavior of a fire through a physically based rendering engine, and the rendering module is connected to the exercise module through a real-time data exchange interface.

[0032] Furthermore, the training module includes a multi-person collaborative operation function, which supports training operations through VR devices. The training module generates dynamic feedback based on the personnel's operating behavior and transmits it to the central control module.

[0033] Furthermore, the fire dynamics simulation module includes a real-time parameter adjustment module, which adjusts the parameters of wind speed, temperature, and smoke concentration in the fire model in real time according to the fire spread, ventilation environment, and equipment status in the mine, so as to achieve a more accurate simulation.

[0034] Furthermore, the real-time parameter adjustment module obtains data on the mine environment, fire source type, and equipment status through communication with the fire dynamics simulation module, and dynamically adjusts simulation parameters such as fire spread rate, smoke concentration, and wind speed based on the data to simulate different emergency rescue scenarios.

[0035] The beneficial effects of this invention are as follows: This invention provides a virtual simulation emergency drill method and system based on deep reinforcement learning under dynamic disaster conditions of mine fires. By combining virtual reality and augmented reality technologies, it can realistically and accurately simulate the spread process of mine fires, smoke diffusion, heat wave distortion, and personnel reactions and emergency responses. Its beneficial effects are mainly reflected in the following aspects:

[0036] (1) By simulating the spread of a mine fire in real time and combining the changes in ventilation system, temperature, wind speed and other factors, this invention can more accurately restore the fire disaster process, provide a more realistic drill scenario and help drill personnel better cope with complex disaster situations.

[0037] (2) This invention can adaptively adjust according to the actual conditions of different mines and the type of fire, so that each exercise can be customized according to actual needs, avoiding the drawbacks of fixed scenarios and lack of flexibility in traditional exercises. The system can dynamically adjust simulation parameters according to different environmental data, fire spread and emergency rescue needs, providing a personalized exercise experience.

[0038] (3) This invention comprehensively considers the coupling effect between fire spread and the mine ventilation system and equipment status, and can adjust the ventilation environment in the simulation in real time to accurately simulate the dynamic changes of fire spread. Compared with traditional simulation methods, this invention can more accurately simulate the interaction between fire spread and the ventilation system, and improve the scientificity and practicality of the simulation results.

[0039] (4) Through deep reinforcement learning technology, this invention can continuously optimize fire simulation strategies, learn and improve fire spread prediction models and emergency drill response strategies. This enables the drill system to have intelligent adjustment capabilities, and can adjust in real time according to changing fire scenarios, equipment failures, wind speed changes and other factors, thereby achieving efficient and automated emergency drills.

[0040] (5) The virtual simulation exercise system of the present invention can monitor and provide feedback on the operational behavior of the exercise personnel in real time, and update the simulation status according to the actual reaction and operation results of the personnel, forming a closed-loop control. This process helps the exercise personnel to better understand the disaster process, improve their decision-making ability, optimize the emergency response process, and thus enhance the mine fire prevention and control capabilities.

[0041] (6) Compared with traditional on-site fire emergency drills, this invention provides a safe training environment through virtual reality and augmented reality technologies, avoiding the risks of actual operations in dangerous environments such as high temperature and high concentration of smoke. In addition, the system's efficient simulation and automated adjustment greatly reduce the time and cost required for traditional drills, improving the economic benefits of the drills.

[0042] Through the above-mentioned beneficial effects, this invention can not only improve the effectiveness and safety of mine fire emergency drills, but also provide strong technical support for mine fire emergency management, effectively improve the mine fire prevention and control capabilities, and reduce the risk of casualties and equipment damage.

[0043] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0044] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0045] Fig. 1 This is a flowchart of the method of the present invention;

[0046] Fig. 2 This is a system structure diagram of the present invention. Detailed Implementation

[0047] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0048] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0049] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0050] Please see Figs. 1-2 The method of the present invention is as follows:

[0051] By utilizing various basic parameters collected on-site from the mine (roadway dimensions, cross-sectional shape, elevation difference, equipment dimensions, layout, dimensions of various equipment, dimensions of ventilation facilities, etc.), hydrological parameters, topographic parameters, ventilation system data, electromechanical equipment and facility data, mining engineering and production status data, environmental monitoring and sensor data, personnel and safety system data, fire prevention and extinguishing material and measure data, safety rules and emergency plan data, etc., a comprehensive mine database module is established, comprising a basic topographic and structure database, a mine geology database, a mine roadway database, a ventilation database, an electromechanical equipment and facility database, a fire prevention and extinguishing material and measure database, a safety rules and emergency plan database, and a mine surveying database. By utilizing the collected parameter database module and employing modeling techniques such as 3D data modeling, laser scanning modeling, and image modeling, combined with 3D Max-assisted modeling data processing methods, 3D models of above-ground and underground structures and objects can be rapidly established. Neural network-assisted modeling methods fill and repair blank areas caused by the scanning modeling process, and automatically search and replace single-sided or multi-sided object models in the model library module after automatic identification, thus rapidly building a static model library. The dynamic model library creates dynamic 3D models by simulating the movement and interaction of a large number of particles, realizing the construction of dynamic models such as smoke, dust, flames, and water flow. For detailed 3D models in the model library module that require operation, digital sculpting modeling is used to reproduce models such as operating tables, complex control panels, and active models. Combined with a scene and task editing system, various mine models and 3D mining spaces can be freely combined and set up to reproduce typical fire-affected mining area scenes and build new high-precision 3D fire scenarios for drills.

[0052] The software's mine scene and task editing system configures and mobilizes the model library module, mine task editing module, mine scene editing module, and configuration file library. Through the aforementioned modeling methods, the mine scene and task editing system is implemented, enabling the rapid construction and reproduction of classic mine fire mining area models and high-precision drill mine fire models. The mine scene and task editing system imports the constructed typical mine fire mining area models into the fire dynamics simulation module. This module primarily performs numerical simulations of fires, achieving numerical simulations at the mining area level. The fire dynamics simulation module, aided by a database of 50 typical fire scenarios and typical mine fires, achieves numerical simulations of typical mine fire reproductions. Simulation data (fire source intensity, roadway cross-sectional area, initial wind speed, inclination angle, ambient temperature, ambient pressure, monitoring point distance, and fire time) is used as input to a BP neural network for training, completing a surrogate model (SM). Using the trained surrogate model (SM), approximate FDS results can be quickly obtained, enabling real-time numerical simulation of mining area-level fires. Simultaneously, a real-time parameter regulator is obtained through training with a BP neural network. The real-time parameter regulator can run a simplified model of FDS. According to the input variable parameters of the intelligent exercise guidance module and the multi-person collaborative exercise module, the parameter regulator will feed back the corrected parameters to FDS and update the simulation state.By combining the proxy model and real-time parameter regulator with the simulated mine fire model, and inputting the data into the central control module, a mine fire simulation exercise at the mining area level is conducted. The central control module, through a hybrid algorithm model, determines the overall spread and damage of the fire disaster model within the scene, and sends the rendering information to the high-performance rendering module. This module renders the fire model's spread and diffusion status and changes in the roadway environment, and then sends the rendering information to the virtual simulation training system and intelligent exercise guidance module. The multi-person collaborative exercise module, intelligent exercise evaluation module, referee evaluation module, and observation module display the situation of the participants in a dynamic mine fire disaster scenario. The multi-person collaborative exercise module enables multiple personnel to conduct collaborative exercises underground using VR virtual equipment, and the data on the impact of the participants' actions on the virtual environment is transmitted along the virtual simulation training system to the central control module. The central control module then sends the exercise data to the fire accident proxy model and real-time parameter regulator. After real-time numerical simulation calculations, the calculation results are sent back to the central control module. The control module, and then the intelligent model control module, determines the extent of the fire's spread and damage in the overall scene. It sends rendering information to the high-performance rendering module and the virtual simulation training system, and displays the situation of the trainees in the dynamic disaster scenario of a mine fire through the intelligent exercise evaluation module and the observation module, providing the trainees with the assessment results. The intelligent exercise guidance module can input variable factors during the exercise, such as sudden decreases in airflow, damage to ventilation equipment, power outages, and partial collapses of roadways, increasing the difficulties faced by the trainees. This information is transmitted along the transmission line to the central control module, which sends the exercise data to the proxy model and real-time parameter adjustment machine. After real-time numerical simulation calculations, the calculation results are sent to the central control module. The central control module then determines the extent of the fire model's spread and damage in the overall scene, sends rendering information to the high-performance rendering module, and displays the situation of the trainees in the dynamic disaster scenario of a mine fire through the intelligent exercise evaluation module, the referee evaluation module, and the observation module.

[0053] The fire disaster simulation model and real-time parameter adjustment machine can input eight types of parameters from mining area nodes in real time (fire source intensity, roadway cross-sectional area, initial wind speed, dip angle, ambient temperature, ambient pressure, monitoring point distance, and fire time). By changing the simulated mining area conditions, the built-in solution control module of the fire disaster simulation model and real-time parameter adjustment machine can transmit dynamic data near the fire to the fire simulation model and real-time parameter adjustment machine in real time, realizing the simulated response changes of the fire area and achieving the purpose of dynamic disaster simulation of mine fires. This method can use VR / AR and MR virtual simulation technology to simulate underground personnel evacuation and post-disaster mine rescue team emergency rescue. It can provide feedback that conforms to the actual disaster environment based on the operation of the drill personnel and the environment, achieving good emergency rescue drills and escape training effects.

[0054] The operation steps are as follows:

[0055] Step 1: Construct a comprehensive mine database and a high-precision 3D model library

[0056] Collect and integrate various basic parameters of the mine (roadway dimensions, cross-sectional shape, elevation difference, equipment size and layout, ventilation facility size and location, electromechanical equipment data, mining engineering status, environmental monitoring data, personnel location data, fire prevention and extinguishing material location and performance, safety rules and emergency plans, etc.). Based on the collected data, construct databases for basic topography and structures, mine geology, mine roadways, ventilation, electromechanical equipment and facilities, fire prevention and extinguishing materials and measures, safety rules and emergency plans, and mine surveying, forming a comprehensive mine model database.

[0057] Utilizing a comprehensive mine model database, combined with 3D data modeling, laser scanning modeling, and image modeling techniques, and supplemented by artificial intelligence (such as neural networks) to assist in modeling data processing, we can rapidly construct 3D models of above-ground and underground structures, equipment, and facilities (especially ventilation facilities and electromechanical equipment). AI is used to assist in repairing blank areas in the scanning modeling. We automatically identify single-sided / multi-sided object models and search and replace them in the model library to quickly build a static model library. We also construct a dynamic model library: through particle systems, we simulate the movement and interaction of a large number of smoke particles, heat particles, and flame particles to create dynamic 3D models of flames, smoke, smoke diffusion, and high-temperature airflow. For models requiring precise operation or activity (such as control panels, fan control panels, air doors, and fire extinguishing equipment), we use digital sculpting modeling to accurately reproduce them. Finally, we create a dynamic rendering model of thermal damage, rendering in real-time the deformation, melting, and burning processes of the model under high fire temperatures, as well as the dynamic effects of the spread, accumulation, and stratification of flames and dense smoke in the roadway, based on the model's material data and the simulated temperature field.

[0058] Step 2: Construct a high-precision 3D mine fire simulation scenario

[0059] Based on the model library module established in step one, and utilizing the software's mine scene editing module and mine task editing module, combined with the configuration file library module, various static and dynamic mine models can be freely combined to quickly build and reproduce typical fire-prone mining area scenarios (such as belt friction fires, electrical fires, gas combustion and explosions, and flammable material fires).

[0060] Step 3: Establish a high-precision numerical simulation and surrogate model for fire accidents

[0061] A fire dynamics simulation module (such as a simulator based on FDS or similar principles) is built to achieve physical simulation of mining area-level fires. A database of typical mine fires (containing 50+ typical scenarios) is used to reproduce the development process of typical fires. High-precision simulation data (fire source intensity, roadway cross-sectional area, initial wind speed, dip angle, ambient temperature, ambient pressure, monitoring point distance, and fire time) are used as input to train a BP neural network to construct a fire accident surrogate model (SM). The surrogate model is used to quickly obtain results with approximate FDS accuracy, achieving real-time numerical simulation of mining area-level fires. Simultaneously, a real-time parameter regulator is trained: simplifying the complex FDS model into a model that can be quickly processed by the regulator. External input variable parameters are received, and the simulation state is quickly corrected.

[0062] Step 4: Simulation Result Processing and Disaster Assessment

[0063] The output of the fire incident proxy model (SM) or real-time parameter regulator (temperature field, smoke concentration field, visibility, toxic gas concentration, fire spread range, etc.) is sent to the central control module. The central control module uses a hybrid algorithm model (which may include a rule engine, a physics-simplified model, etc.) to determine the fire's spread path, damage level (e.g., equipment burnout, support failure, tunnel collapse risk), ventilation system malfunction, and hazardous area extent in the overall scene, and organizes the data required for subsequent rendering.

[0064] Step 5: High-fidelity real-time rendering and scene distribution

[0065] Rendering is performed in a high-performance rendering module using a high-performance virtual simulation computer and high-precision scene rendering algorithms (such as Physically Based Rendering (PBR) and volume rendering for smoke). After the central control module determines the rendering range, it sends the judgment results and scene status information to the high-performance rendering module. The high-performance rendering module renders the dynamic fire process of the entire mining area model (flame morphology, smoke diffusion path, visibility changes, high-temperature areas, thermal damage to objects) and changes in the tunnel environment (such as dense smoke obscuring and heat wave distortion). The rendered dynamic fire disaster scene is then transmitted to the virtual simulation training system and the intelligent drill guidance module. The footage is finally distributed to the multi-person collaborative drill module, the intelligent drill evaluation module, the judge evaluation module, and the observation module.

[0066] Step Six: Multi-person Collaborative Exercises and Real-time Closed-Loop Feedback

[0067] Trainees (rescue / escape personnel) utilize VR virtual equipment in a multi-person collaborative drill module to perform coordinated operations (such as fire extinguishing, ventilation control, escape, search and rescue, and casualty transport) within a rendered fire disaster scenario. Data on the impact of the trainees' actions on the virtual environment (such as closing dampers, starting fans, using fire extinguishers, moving obstacles, and selecting escape routes) is fed back to the central control module in real time through the virtual simulation training system.

[0068] The central control module transmits this impact data (as new boundary conditions or parameter changes) to the Fire Incident Surrogate Model (SM) and the real-time parameter regulator. The Fire Incident Surrogate Model (SM) / real-time parameter regulator performs rapid real-time calculations based on the new parameters. The updated simulation results are transmitted back to the central control module. The central control module performs a new round of disaster spread and damage assessment and sends the updated rendering information to the high-performance rendering module and the virtual simulation training system.

[0069] The high-performance rendering module updates the scene and transmits it back in real time to the multi-person collaborative exercise module, intelligent exercise evaluation module, referee evaluation module, and observer module through the virtual simulation training system, forming a real-time closed loop that allows trainees to immediately see the impact of their operations on the fire situation and the environment.

[0070] Step Seven: Intelligent guidance increases the difficulty of drills and introduces unexpected situations.

[0071] The intelligent drill guidance module can dynamically inject variables and sudden disaster scenarios during the drill to increase the difficulty, complexity, and realism of the exercise. It can randomly set scenarios such as sudden increases / decreases in airflow, main ventilator malfunctions, and local ventilator shutdowns. It can also simulate damage to critical ventilation facilities (door failure, air bridge collapse) and power outages (affecting ventilation, drainage, and lighting). Simulations include abnormal gas outbursts / accumulation, coal dust explosions, and partial roof falls / collapses in tunnels blocking escape or ventilation paths. The intelligent drill guidance module transmits the set emergency parameters to the central control module. The central control module then transmits the information to the Fire Accident Surveillance Model (SM) and the real-time parameter regulator. The SM / real-time parameter regulator performs rapid real-time calculations based on the new parameters. The updated simulation results are transmitted back to the central control module. The central control module performs a new round of disaster spread and damage assessment and sends the updated rendering information to the high-performance rendering module.

[0072] The high-performance rendering module updates the scene and transmits it back in real time to the multi-person collaborative exercise module, intelligent exercise evaluation module, referee evaluation module, and observer module through the virtual simulation training system, presenting new disaster scenarios under emergencies.

[0073] Step 8: Intelligent Evaluation and Exercise Report Generation

[0074] Through the intelligent exercise evaluation module and the referee evaluation module, combined with observations from the observation module:

[0075] Real-time and post-event comprehensive evaluations are conducted on the decision-making, operational procedures, and execution of key tasks (such as wind control, fire extinguishing, search and rescue, self-rescue and mutual rescue, and escape route selection) of trainees (rescue teams / escapees) in dynamic disaster environments.

[0076] Assess teamwork, resource utilization, emergency response speed, and safety.

[0077] The intelligent drill evaluation module automatically analyzes drill data and generates detailed rescue or escape assessment reports in the form of text reports, charts, and screenshots / replay clips of key scenes. The report content facilitates the evaluation and analysis of drill effectiveness, identifies strengths and weaknesses, and provides a basis for subsequent training, plan revision, and improving the efficiency of mine emergency rescue.

[0078] The innovation of this invention lies in realizing a multi-person collaborative drill and emergency rescue method for dynamic disaster evolution of underground mine fires, based on the rapid construction of intelligent mine models, the deep integration of fire dynamics proxy models and real-time adjustment technology of semi-physical parameters, and computer virtual simulation technology. This method possesses the following core functions:

[0079] Simulation of dynamic evolution of fire under the influence of mine ventilation: Accurately simulate the development, spread and smoke diffusion process of fire under the influence of complex mine ventilation networks (including normal ventilation, wind adjustment, reverse ventilation, airflow short circuit, ventilation facility failure and other conditions).

[0080] Simulation of fire source characteristics and disaster spread: Simulate the changes in combustion intensity and heat release rate of different types of fire sources (solid, liquid, gas), as well as the chain reactions such as flame propagation, high-temperature heat flow, and ignition of combustibles caused by these changes.

[0081] Multiphysics Coupled Disaster Simulation:

[0082] Simulation of the generation and transport of toxic and harmful gases: Dynamic simulation of the generation, diffusion, accumulation and concentration distribution changes of toxic and harmful gases such as carbon monoxide (CO), carbon dioxide (CO2), hydrogen sulfide (H2S), and nitrogen oxides (NOx) generated by fire under the action of airflow.

[0083] Simulation of secondary disasters induced by high temperature: Simulates secondary disasters caused by high temperature during fire, such as support failure, equipment damage, cable burning, gas accumulation and explosion risk, and coal dust explosion risk.

[0084] Visibility dynamic decay simulation: Based on smoke concentration and particulate matter distribution, calculate and render the changes in visibility in the disaster environment in real time, affecting escape and rescue visibility.

[0085] Dynamic feedback simulation of fire and ventilation systems:

[0086] The fire incident proxy model (SM) and real-time parameter regulator allow for the input of key node parameters (fire source intensity, ventilation resistance, air volume, temperature, pressure, etc.) in real time.

[0087] The ventilation calculation and control module senses and calculates the impact of fire on the ventilation network in real time (such as throttling effect and fire-wind pressure generation), and feeds back the adjusted ventilation status to the agent model / regulator.

[0088] It enables rapid and high-precision simulation of the effects of air control and fire extinguishing measures such as air adjustment, air reversal, fan start-up and shutdown, opening and closing of air doors / windows, inert (nitrogen / carbon dioxide) fire extinguishing, and construction of temporary airtight enclosures, and dynamically reflects changes in the state of the fire zone (such as fire being under control, changes in smoke flow direction, and transfer of dangerous areas).

[0089] Physically based automatic scene rendering: Based on the results of fire dynamics simulation (temperature field, smoke concentration field, gas concentration field, visibility field) and damage judgment results, a high-fidelity dynamic disaster scene of a mine fire is automatically generated through a high-performance rendering engine, which conforms to the laws of physics. This includes realistic visual effects such as flames, smoke diffusion, high temperature heat wave distortion, object burning and deformation, and ambient light occlusion.

[0090] The core of this invention lies in:

[0091] Fire Accident Surrogate Model (SM) and Real-time Parameter Regulator: As the core engine, it receives initial parameters from the central control module and dynamic variables (such as changes in fire source and ventilation operation) from the exercise interaction / guidance input, quickly calculates the fire development and the feedback effect with the ventilation system, and outputs key disaster parameters.

[0092] Ventilation-fire dynamic coupling mechanism: The system's built-in ventilation calculation and control module is tightly integrated with the fire proxy model / real-time parameter regulator to ensure that changes in airflow status (active regulation or passive disruption) can dynamically and in real time affect the fire and smoke spread, and vice versa, changes in the fire situation can also affect the ventilation network in real time. This constitutes the core of "semi-physical parametric simulation", that is, based on physical mechanisms, it realizes real-time simulation of complex system feedback through efficient models.

[0093] Closed-loop interactive drills: Combining VR / AR / MR virtual simulation technologies, a real-time closed loop is constructed, encompassing "personnel operation → environmental impact → model calculation → scene update → personnel perception." The actions of drill participants (escapees or rescue team members) in the virtual disaster environment (such as ventilation control, fire extinguishing, and escape route selection) instantly affect simulation parameters and receive physically accurate feedback through updated high-fidelity rendered scenes, greatly enhancing the realism of the drills and the training effectiveness. This achieves the advanced goal of conducting multi-person collaborative escape drills and emergency rescue training in a highly realistic and physically reliable dynamic disaster environment of a mine fire, significantly improving the level of mine safety training and emergency preparedness.

[0094] Example 1: A Virtual Simulation Emergency Drill Method Based on Deep Reinforcement Learning for Dynamic Disaster Conditions in Mine Fires

[0095] This embodiment provides a virtual simulation emergency drill method for mine fires under dynamic disaster conditions based on deep reinforcement learning. It combines virtual reality (VR), augmented reality (AR) technologies with deep reinforcement learning algorithms. This method achieves intelligent and personalized mine fire emergency drills by simulating the dynamic processes of mine fire spread, smoke diffusion, and fire intensity in real time.

[0096] The workflow is as follows:

[0097] S1: Mine Data Acquisition and Processing

[0098] Basic data for the mine is collected through sensors, monitoring equipment, and manual data entry. The specific data collected are as follows:

[0099] Mine Geographic and Structural Data: Tunnel Length: 200 meters; Tunnel Width: 6 meters; Tunnel Height: 4 meters; Equipment Layout: The mine has 2 ventilation fans, each with a rated air volume of 1000 m³ / h. 3 / h, with an operating wind speed of 5m / s;

[0100] Environmental monitoring data: Initial temperature: 25℃; Initial humidity: 50%; Initial air pressure: 1013hPa;

[0101] Fire-related data: Location of the fire source: 150 meters deep in the mine, in the equipment area; Fire intensity: 5000 W / m 2 Coal seam temperature: 600℃ (initial state of ignition source);

[0102] Ventilation system data: Air velocity: 5 m / s (initial ventilation rate); Air volume: 1000 m³ / s 3 / h (air volume provided by each fan); Fan operating status: 2 fans fully open, air damper open; Through these data, the mine's comprehensive database provides the necessary data support for subsequent fire simulation.

[0103] S2: 3D Modeling and Fire Simulation Model Generation

[0104] Using data from a comprehensive mine database, a 3D structural model of the mine, including roadways, equipment, and ventilation systems, was created using 3D modeling technology. The resulting fire simulation model incorporates factors such as smoke diffusion and flame spread, enabling the simulation of fire propagation within the mine.

[0105] S3: Combination of Fire Dynamics Simulation and Deep Reinforcement Learning

[0106] The fire spread process is modeled using fire dynamics simulation equations. Based on the physical laws of heat conduction and convection, and combined with data on the temperature field, airflow, and ignition source intensity within the mine, the fire dynamics simulation model simulates the fire spread process in real time.

[0107] The formula for simulating fire spread is as follows:

[0108]

[0109] Where T is temperature in Kelvin (K), representing the temperature at each point; t is time in seconds, representing the time progression in the simulation; and α is thermal diffusivity in m² / s, representing the ability of a substance to conduct heat, set at 1.2 × 10⁻⁶. -5 m 2 / s; It represents the degree of diffusion of the temperature field, measured in units of 1 / m. 2 Q is the heat release intensity per unit volume, expressed in W / m³. 3 The heat source intensity for spontaneous combustion of coal seams is 5000 W / m. 3 .

[0110] Deep reinforcement learning technology optimizes the simulation strategy of fire spread through the training process. The system can adjust the speed and path of fire spread based on real-time data (such as equipment failure, changes in fire source, changes in environment, etc.).

[0111] S4: Dynamic Adjustment Calculation Process for Fire Spread and Ventilation System Coupling

[0112] The interaction between the fire spread process and the mine ventilation system is dynamically calculated using the following coupling equations. Changes in the ventilation system (such as wind speed and air volume) have a significant impact on the speed and direction of fire spread.

[0113] The specific calculation process is as follows:

[0114] S41 coupling equation:

[0115] Based on the aforementioned formula:

[0116]

[0117] Where T is temperature, in Kelvin (K); α is thermal diffusivity, in m³. 2 / s; β is the influence coefficient of airflow on the temperature field, with units of m. 2 / s, set to 0.5; v is the airflow velocity, in m / s, set to 5 m / s, i.e., the ventilation velocity; Q is the heat release intensity per unit volume, in W / m³. 3 Set to 5000W / m 3 .

[0118] S42 Calculation Example:

[0119] Suppose a mine tunnel is 200 meters long, 6 meters wide, and 4 meters high, with an ignition source intensity of 5000 W / m. 2 The initial temperature is 300K, the airflow velocity is 5m / s, and the thermal diffusivity is 1.2×10⁻⁶. -5 m2 / s, the influence coefficient of airflow on temperature is 0.5.

[0120] Numerical calculations are performed using the finite difference method, assuming each time step is 1 second, to calculate the temperature change during the fire spread process. and A certain value is calculated, and the final result is:

[0121] T(t+1)≈305K

[0122] The temperature result will be passed to the next time step (i.e., the next second), and the temperature will continue to change as wind speed changes and the ventilation system is adjusted.

[0123] S5: Demonstration Process of Virtual Reality and Augmented Reality Technologies

[0124] In this embodiment, participants conduct a mine fire emergency drill using virtual reality (VR) or augmented reality (AR) devices. The drill procedure is as follows:

[0125] S51: Enter the virtual training environment:

[0126] Participants wore VR headsets and handheld control devices, entering a virtual fire drill scenario within a mine. The system loaded a 3D modeling module to create a mine environment, including elements such as roadways, equipment, ventilation systems, and fire sources.

[0127] S52: Simulated fire spread:

[0128] At the start of the drill, the fire spread model was initialized based on real-time data from within the mine (such as fire intensity, temperature, and wind speed). The system began simulating the fire spread process, and participants could observe smoke, flame expansion, and temperature changes around the fire source. In the virtual scenario, the fire spread rate and smoke concentration were adjusted according to changes in wind speed and the fire source.

[0129] S53: Operations of the trainees:

[0130] Participants used virtual equipment to simulate firefighting and evacuation operations. For example:

[0131] Activate the fire extinguishing equipment and control the fire extinguisher through virtual devices to simulate fire extinguishing operations.

[0132] Turn on the ventilation equipment and adjust the wind speed through simulated operation to control the spread of the fire.

[0133] The participants can make evacuation decisions based on the spread of the fire and use virtual reality equipment to navigate to the mine's safety exits.

[0134] S54: System Feedback and Dynamic Adjustment

[0135] The system monitors the behavior of the participants in real time. For example, when participants activate fire extinguishing equipment, the system automatically adjusts the fire spread rate and reduces the fire intensity based on the equipment's effectiveness. When participants take incorrect actions (such as failing to evacuate or extinguish the fire in time), the system provides negative feedback and simulates the corresponding fire spread consequences (e.g., increased fire spread rate, increased temperature).

[0136] S55: Exercise Summary and Optimization:

[0137] After the drill, the system generates a report to evaluate the performance of the participants and summarizes the results based on drill data (such as fire extinguishing effectiveness and evacuation time). Through reinforcement learning algorithms, the system continuously optimizes the simulation process to improve the effectiveness of future drills.

[0138] Through the above steps, participants can conduct emergency operations in a fully simulated mine fire environment. The system dynamically adjusts the simulated scenario based on real-time data (such as wind speed and fire intensity) and provides an immersive training experience through virtual reality and augmented reality technologies. Furthermore, by applying reinforcement learning algorithms, the system can optimize the fire spread model and emergency response strategies based on real-time feedback during the exercise, ensuring the efficiency and safety of the drills.

[0139] Example 2: Mine Fire Emergency Drill System Based on Deep Reinforcement Learning

[0140] This embodiment provides a virtual simulation emergency drill system for dynamic disaster conditions of mine fires based on deep reinforcement learning. The system adopts a modular structure and can provide personalized fire emergency drill plans according to the actual mine environment and fire simulation data. The core functions of the system include fire dynamics simulation, dynamic adjustment of the ventilation system, virtual reality drills, and deep reinforcement learning optimization.

[0141] System composition and functions of each module

[0142] The system consists of multiple functional modules, each undertaking a specific task, and forming the overall functionality through data flow and interaction.

[0143] The following are the main modules of the system and the connections between them:

[0144] 1. Mine Integrated Database Module

[0145] Function: Used to store and manage basic mine data, including mine geographical structure, ventilation system, equipment layout, fire-related data, etc. Mine data is updated in real time through sensors and manual input.

[0146] Data input: Real-time data is obtained from mine environmental monitoring equipment, fire sensors, personnel positioning systems, etc.

[0147] Data output: The mine data is transmitted to the 3D modeling module, the fire dynamics simulation module, and other downstream modules to ensure data consistency and real-time performance.

[0148] 2. 3D Modeling Module

[0149] Function: Based on data from the mine's comprehensive database, this module generates a 3D structural model of the mine, mainly including roadways, equipment, and ventilation systems. It also generates fire simulation models, including fire sources, smoke, and heat waves.

[0150] Data input: Receives data from the mine integrated database module (such as roadway structure, equipment layout, ignition source data, etc.) and generates a 3D model.

[0151] Data output: The generated 3D model data is transmitted to the fire dynamics simulation module, virtual reality exercise module, etc.

[0152] 3. Fire Dynamics Simulation Module

[0153] Function: This module simulates the spread of a mine fire based on fire dynamics formulas, including temperature changes, smoke diffusion, and changes in the ignition source. It incorporates factors such as the temperature field, airflow velocity, and ignition source intensity within the mine to simulate fire spread.

[0154] Data input: Receives mine structure data from the 3D modeling module and real-time environmental data (such as wind speed, equipment failure, etc.) from the mine integrated database module, and optimizes the simulation process based on deep reinforcement learning algorithms.

[0155] Data output: The fire simulation results are transmitted to the ventilation system dynamic adjustment module, high-precision rendering module, virtual reality simulation module, etc., for real-time feedback of the simulation results.

[0156] 4. Dynamic adjustment module for ventilation system

[0157] Function: Simulates the impact of mine ventilation systems on fire spread. This module can dynamically adjust parameters such as wind speed and air volume of the ventilation system to simulate the fire spread process under different ventilation conditions.

[0158] Data input: Receives fire spread results (such as fire spread rate, temperature changes, etc.) from the fire dynamics simulation module, and adjusts wind speed, air volume, etc. based on real-time mine monitoring data.

[0159] Data output: The data after ventilation adjustment is transmitted back to the fire dynamics simulation module to ensure the accuracy and dynamic response of the fire simulation.

[0160] 5. Deep Reinforcement Learning Optimization Module

[0161] Function: This module uses deep reinforcement learning algorithms to optimize the simulated fire spread process and adjusts emergency response strategies based on feedback from drill participants and the results of the fire spread. It can learn and optimize the impact of different operations on fire spread during the mine fire process.

[0162] Data input: The module receives fire spread data from the fire dynamics simulation module, feedback data from the drill module, and ventilation system adjustment data. The deep reinforcement learning optimization module receives input data including fire spread data from the fire dynamics simulation module, feedback data from the drill module, and real-time mine monitoring data. Output data consists of optimized fire spread model parameters and emergency response strategies, which are fed back to the fire dynamics simulation module and the drill module to improve the accuracy of the simulation and the effectiveness of the drills.

[0163] Data output: Optimize the fire spread model and emergency response strategy based on feedback data, adjust fire spread simulation parameters (such as fire spread rate, temperature, smoke concentration, etc.), and transmit the optimized results to the fire dynamics simulation module and the drill module.

[0164] 6. High-precision rendering module

[0165] Function: Based on the output data of the fire dynamics simulation module and the ventilation system adjustment module, render the dynamic scene of the mine fire in real time, including effects such as flames, smoke, and heat wave diffusion.

[0166] Data Input: The high-precision rendering module receives simulation results from the fire dynamics simulation module and the ventilation system adjustment module, generating dynamic visual effects of the fire. Input data includes fire spread data (such as flame temperature, smoke concentration, and fire source location) from the fire dynamics simulation module, wind speed and airflow change data from the ventilation system adjustment module, and mine model data from the 3D modeling module. Output data is rendered graphical information, including dynamic effects of the fire scene (such as visualizations of smoke, flames, and temperature changes). This data is then transmitted to the exercise module to generate a virtual exercise environment.

[0167] Data output: The rendering results are passed to the virtual reality training module to provide trainees with a visualized fire environment.

[0168] 7. Virtual Reality Training Module

[0169] Function: Provides participants with a virtual mine fire scenario, supporting multi-person collaborative drills. Participants perform emergency response operations in the virtual environment, such as firefighting, evacuation, and rescue. The system provides real-time feedback and adjustments based on the participants' actions. The virtual reality drill module presents the dynamic fire scene data received from the high-precision rendering module in real time, providing interactive drills in both virtual reality and augmented reality environments. The high-precision rendering module transmits the visual effects of fire spread (such as flames, smoke diffusion, and temperature fields) to the drill module in real time, ensuring that participants see a realistic and dynamically changing disaster scenario. The participants' actions (such as firefighting and evacuation) are fed back to the virtual reality drill module, thus affecting the rendering results and forming a closed-loop control mechanism.

[0170] Data Input: The virtual reality (VR) training module receives fire simulation scenarios from the high-precision rendering module, feedback data from the deep reinforcement learning optimization module, and real-time mine monitoring data. Input data includes fire simulation scenario data from the high-precision rendering module, mine environment data, and operational data from the trainees. The VR training module presents dynamic fire scenarios through virtual reality devices and provides real-time feedback based on the trainees' actions (such as firefighting, evacuation, and equipment operation). This feedback data is then transmitted to the deep reinforcement learning optimization module to optimize the fire spread model and emergency response strategies.

[0171] Data output: The operational data of the drill participants (such as fire extinguishing, evacuation, equipment operation, etc.) are fed back to the deep reinforcement learning optimization module to optimize emergency response strategies and simulation results.

[0172] 8. Data flow relationships:

[0173] The mine integrated database module provides basic data for the entire system and provides input to the 3D modeling module, fire dynamics simulation module, virtual reality drill module, etc.

[0174] The 3D modeling module generates structural and fire simulation models of the mine and transmits the data to the fire dynamics simulation module.

[0175] The fire dynamics simulation module simulates the fire spread process and transmits the results to the ventilation system dynamic adjustment module, high-precision rendering module, virtual reality simulation module, etc., to ensure that the simulation results are consistent with the actual situation in the mine.

[0176] The ventilation system dynamic adjustment module adjusts the ventilation system based on real-time monitoring data, influencing the fire spread process, and feeds back the adjusted data to the fire dynamics simulation module.

[0177] The deep reinforcement learning optimization module optimizes the fire spread model and emergency response strategy based on the exercise feedback, and transmits the optimized results to the fire dynamics simulation module and the virtual reality exercise module to optimize the exercise effect.

[0178] The high-precision rendering module is responsible for rendering the fire scene in real time and providing the rendering results to the virtual reality training module to ensure that trainees can see the realistic fire simulation effect in real time.

[0179] There is data sharing and interaction between the mine integrated database module and the 3D modeling module, fire dynamics simulation module, and virtual reality exercise module.

[0180] The 3D modeling module and the fire dynamics simulation module are closely connected. The 3D model provides environmental support for fire simulation, and the fire simulation results are fed back to the virtual reality exercise module.

[0181] The ventilation system dynamic adjustment module and the fire dynamics simulation module communicate via real-time data streams, influencing the fire spread simulation.

[0182] Dynamic feedback is established between the deep reinforcement learning optimization module and the fire dynamics simulation module, virtual reality exercise module, and ventilation system adjustment module to optimize the exercise effect and simulation results.

[0183] 9. System Workflow

[0184] Data Acquisition: The system collects various basic data of the mine (such as ventilation system, equipment layout, fire source data, etc.) through the mine comprehensive database module.

[0185] 3D Modeling: The 3D modeling module generates mine environment models and fire simulation models based on mine data.

[0186] Fire Simulation: The fire dynamics simulation module simulates the fire spread process based on the fire dynamics model and performs coupled calculations with the mine's ventilation system.

[0187] Ventilation regulation: The dynamic regulation module of the ventilation system adjusts the mine's ventilation system based on real-time monitoring data, simulating the impact of wind speed and air volume on the spread of fire.

[0188] Deep reinforcement learning optimization: The deep reinforcement learning module adjusts fire simulation parameters and emergency response strategies based on feedback from drill participants and the results of fire spread.

[0189] High-precision rendering and drills: The high-precision rendering module converts the fire simulation results into a dynamic fire scene in a virtual environment, allowing drill participants to perform emergency operations in a virtual reality environment.

[0190] Exercise Feedback and Optimization: After the exercise, the data is fed back to the deep reinforcement learning module to optimize the fire spread simulation and emergency response strategies during the exercise.

[0191] Through the above embodiments, the mine fire emergency drill system of the present invention can realize dynamic fire spread simulation and personalized emergency response drills based on deep reinforcement learning. The system, through the collaborative work of multiple modules, ensures the realism, intelligence, and efficiency of the drill process, improves the mine fire emergency response capability, reduces safety risks in actual drills, and optimizes mine fire prevention and control strategies.

[0192] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A virtual simulation emergency drill method for mine fire dynamic disaster conditions based on deep reinforcement learning, characterized in that: Includes the following steps: Step 1: Construct a comprehensive mine database and collect various basic parameter data of the mine, including roadway dimensions, cross-sectional shape, elevation difference, equipment dimensions, layout, ventilation facility dimensions, electromechanical equipment data, mining engineering status data, environmental monitoring data, personnel positioning data, fire prevention and extinguishing material performance data, safety rule data, and emergency plan data; Step 2: Using 3D modeling technology, based on the data in the comprehensive mine database, establish a 3D model library of various facilities and structures within the mine, and generate dynamic simulation models of fire, smoke, and flames; Step 3: Based on the fire dynamics simulation module, a fire accident proxy model is constructed through BP neural network training, and the three-dimensional model and the fire proxy model are used for real-time fire simulation. Step 4: Combine the fire simulation results obtained in Step 3 with environmental variables to dynamically render the fire scene and generate the mine fire disaster environment. Step 5: Using virtual reality (VR) or augmented reality (AR) technology, support multi-person collaborative drills. Drill participants perform collaborative operations in a mine fire disaster environment. These collaborative operations affect the simulated fire state and are fed back to the fire dynamics simulation module in real time, forming a closed-loop control.

2. The virtual simulation emergency drill method for mine fire dynamic disaster conditions based on deep reinforcement learning according to claim 1, characterized in that: The 3D modeling technology includes laser scanning modeling, image modeling, and digital carving modeling.

3. The virtual simulation emergency drill method for mine fire dynamic disaster conditions based on deep reinforcement learning according to claim 1, characterized in that: The fire dynamics simulation module trains a backpropagation neural network based on the input simulation data, including fire source intensity, tunnel cross-section, initial wind speed, ambient temperature, ambient pressure, monitoring point distance, and fire time, and generates a fire accident proxy model. The proxy model is used for real-time fire spread simulation.

4. The virtual simulation emergency drill method for dynamic disaster conditions of mine fire based on deep reinforcement learning according to claim 1, characterized in that: The fire scene includes the shape of the flames, the path of the smoke diffusion, the distortion of the heat wave, the changes in the high-temperature area, and the visual effects of objects burning and deforming.

5. A virtual simulation emergency drill system for dynamic disaster conditions of mine fires based on deep reinforcement learning, characterized in that: include: The mine integrated database module is used to store various basic parameter data of the mine, including roadway dimensions, cross-sectional shape, elevation difference, equipment dimensions, layout, ventilation facility dimensions, electromechanical equipment data, mining engineering status data, environmental monitoring data, personnel positioning data, fire prevention and extinguishing material performance data, safety rule data, and emergency plan data. The database module is connected to the 3D modeling module, fire dynamics simulation module, and drill module. The 3D modeling module is used to generate a 3D model library of various facilities and structures in the mine based on the data in the mine integrated database module, and to generate dynamic simulation models of fire, smoke and flame. The 3D modeling module is connected to the fire dynamics simulation module and the high-precision rendering module. The fire dynamics simulation module constructs a fire accident proxy model through BP neural network training. The fire dynamics simulation module is connected to the 3D modeling module, receives model data and performs real-time fire simulation, and transmits the simulation results to the high-precision rendering module and the drill module. The high-precision rendering module is used to render fire disaster scenarios based on the simulation results transmitted by the fire dynamics simulation module, including dynamic effects such as flame shape, smoke diffusion and heat wave distortion. The high-precision rendering module is connected to the exercise module. The exercise module uses VR or AR technology to support multi-person collaborative exercises. The exercise module receives the rendering results from the high-precision rendering module and displays real-time disaster scenarios through virtual reality equipment. The operational behavior of the exercise personnel is fed back to the fire dynamics simulation module in real time, forming a closed-loop control.

6. The virtual simulation emergency drill system for dynamic disaster conditions of mine fire based on deep reinforcement learning according to claim 5, characterized in that: The fire dynamics simulation module and the 3D modeling module communicate via a data interface, which transmits parameter data of the mine facilities and the fire model.

7. The virtual simulation emergency drill system for dynamic disaster conditions of mine fire based on deep reinforcement learning according to claim 5, characterized in that: The high-precision rendering module generates dynamic effects of heat waves, smoke, and flames related to the dynamic behavior of a fire through a physical rendering engine. The rendering module is connected to the exercise module through a real-time data exchange interface.

8. The virtual simulation emergency drill system for dynamic disaster conditions of mine fire based on deep reinforcement learning according to claim 5, characterized in that: The training module includes a multi-person collaborative operation function, which supports training operations through VR devices. The training module generates dynamic feedback based on the personnel's operation behavior and transmits it to the central control module.

9. The virtual simulation emergency drill system for dynamic disaster conditions of mine fire based on deep reinforcement learning according to claim 5, characterized in that: The fire dynamics simulation module includes a real-time parameter adjustment module, which adjusts the parameters of wind speed, temperature and smoke concentration in the fire model in real time according to the fire spread, ventilation environment and equipment status in the mine, so as to achieve a more accurate simulation.

10. The virtual simulation emergency drill system for dynamic disaster conditions of mine fire based on deep reinforcement learning according to claim 9, characterized in that: The real-time parameter adjustment module obtains data on the mine environment, fire source type, and equipment status through communication with the fire dynamics simulation module, and dynamically adjusts simulation parameters such as fire spread rate, smoke concentration, and wind speed based on the data to simulate different emergency rescue scenarios.

Images