Fire safety data management method and system based on edge computing

Abstract

The present application relates to a kind of fire safety data management methods based on edge computing, it is related to data management technical field, the present application is based on edge computing architecture fusion multi-modal sensor data and the accurate three-dimensional coordinate of local BIM model output fire source, substantially shorten the search time of fire rescue personnel, utilize edge node to carry out local anomaly identification and positioning calculation, effectively reduce the cloud communication time delay and network congestion risk, by calling global digital twin model, and depth coupling real-time wind speed, heating and air conditioning operating condition and material combustion characteristic etc. Physical environment parameters carry out CFD simulation calculation, can high-precision, forward-lookingly evolve fire boundary and the spread track of toxic smoke, and the regional function priority and dynamic fire spread prediction model are comprehensively considered, can calculate the safety weight of each passage in real time, actively avoid fire and smoke sealing area, generate dynamic optimal path, maximum degree guarantee personnel life safety.

Description

Technical Field

[0001] This invention relates to the field of data management technology, and in particular to a fire safety data management method and system based on edge computing. Background Technology

[0002] With the development of IoT technology, enterprise fire management systems are gradually evolving from independent operation to networked monitoring. However, existing fire management systems still have the following problems in complex industrial buildings or high-rise buildings: Cross-system coordination suffers from high latency and fragmentation: existing systems mostly rely on cloud servers for data processing and command issuance. When a fire occurs, due to network latency or protocol barriers, it is impossible to complete the seamless coordination of "alarm - cut off non-fire-fighting power - start smoke exhaust - open access control - video pop-up" in sub-second time, which can easily delay the best rescue opportunity.

[0003] Inaccurate fire location: Traditional alarms can only provide two-dimensional zone or loop addresses. In complex buildings, they cannot provide accurate three-dimensional spatial coordinates, which makes it time-consuming for rescuers to search for the source of the fire.

[0004] Lack of high-precision simulation and dynamic evacuation capabilities for fire spread: Existing systems mostly rely on static preset evacuation routes, lacking the ability to extrapolate from real-time fire environment conditions (such as temperature field and smoke concentration field). If the ignition point or smoke spread direction happens to be located on the preset evacuation route, static signs can easily lead people into dangerous areas; To address the aforementioned technical deficiencies, a solution is proposed. Summary of the Invention

[0005] The purpose of this invention is to utilize edge nodes for local anomaly identification and location calculation, effectively reducing cloud communication latency and network congestion risks. By calling a global digital twin model and deeply coupling real-time wind speed, HVAC operating conditions, and material combustion characteristics and other physical environment parameters for CFD simulation calculation, it can accurately and proactively evolve the fire boundary and the spread trajectory of toxic smoke. It also comprehensively considers regional functional priorities and dynamic fire spread prediction models, and can calculate the safety weight of each channel in real time, actively avoid fire and smoke blockage areas, generate dynamic optimal paths, and maximize the protection of personnel safety.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a fire safety data management method based on edge computing, comprising the following steps: S1: By deploying edge computing nodes in various regions of the enterprise, multimodal fire environment data within the node's radiation area is collected in real time. When abnormal features are identified, the precise three-dimensional coordinates of the fire occurrence point are calculated based on the local building BIM model built into the edge computing node and the spatial distribution characteristics of the multimodal fire environment data. The multimodal fire environment data includes temperature sensing data, smoke concentration data, infrared thermal imaging video stream, and visible light video stream. S2: The precise three-dimensional coordinates of the fire location and real-time multimodal fire environment data are uploaded to the cloud server through edge computing nodes. The cloud server calls the global building digital twin model and combines real-time wind speed, HVAC status and building material combustion characteristics. Computational fluid dynamics algorithm is used to perform high-precision three-dimensional simulation of fire spread and smoke flow to generate a dynamic fire spread prediction model. S3: Based on the functional division of each area of ​​the enterprise, the area types are obtained, including densely populated areas, high-value areas, ordinary areas and uninhabited areas. Priority evacuation weights are assigned to the area types, and the safety weights of each evacuation route are further optimized based on the distance between the area distribution location and the precise three-dimensional coordinates of the fire point. S4: Calculate the safety weight of each evacuation route in real time based on the dynamic fire spread prediction model, generate the dynamic optimal evacuation route that avoids the spread of fire and toxic smoke, and send control commands to the edge computing node, which drives the intelligent evacuation signs in the building to dynamically change their direction.

[0007] Furthermore, when abnormal features are detected, a locally preset hardware-level linkage strategy is triggered, and multiple control signals are output synchronously within a sub-second delay. These multiple control signals include a fire alarm signal sent to the fire control panel, cutting off the non-fire-fighting power supply to the fire area, turning on the smoke exhaust fan of the corresponding defense zone, unlocking the access control gate of the escape route, and pushing a real-time video pop-up of the fire point to the security monitoring center.

[0008] Furthermore, the specific process for obtaining the precise three-dimensional coordinates of the fire location is as follows: S11. The edge computing node acquires the visible light video stream and infrared thermal imaging video stream within the radiation area of ​​the node it is located. Based on the preset camera intrinsic and extrinsic parameter matrix, it performs spatiotemporal synchronization and affine registration on the visible light video stream and infrared thermal imaging video stream to obtain a dual-modal fusion image. S12. Input the dual-modal fused image into the pre-trained fire visual detection model, extract the two-dimensional pixel bounding box of the abnormal area, and extract the coordinates of the highest temperature pixel in the infrared thermal imaging video stream of the abnormal area. S13. Using the coordinates of the highest temperature pixel as the target feature point, call the local building BIM model built into the edge computing node to obtain the spatial absolute coordinates and monitoring field of view of the camera within the node's radiation area, and calculate the relative depth distance of the target feature point relative to the camera based on the principle of binocular visual parallax. S14. Using a coordinate system transformation matrix, the relative depth distance and the coordinates of the highest temperature pixel are mapped to the geodetic coordinate system of the local building BIM model to obtain the initial visual three-dimensional coordinates. S15. Obtain the topological location of each temperature sensor and smoke sensor deployed in the node radiation area, as well as the currently collected temperature sensing data and smoke concentration data. Using the spatial coordinates of the topological location as the reference point, construct a three-dimensional spatial gradient field of temperature and smoke concentration in the spatial grid of the local building BIM model using a spatial interpolation algorithm. Use the gradient descent method to perform source tracing iteration along the direction of increase in concentration and temperature in the three-dimensional spatial gradient field, and calculate the highest aggregation center point of temperature and smoke concentration as the coordinates of the sensing spatial source point. S16. The initial visual 3D coordinates and the coordinates of the sensing spatial source point are weighted and fused using the Kalman filter algorithm to obtain fused 3D coordinates. The fused 3D coordinates are then projected onto the local building BIM model for physical constraint verification. If the fused 3D coordinates fall in the suspended area of ​​the local building BIM model, they are projected in reverse along the direction of gravity to the surface of the nearest physical component. The corrected fused 3D coordinates are the final output accurate 3D coordinates of the fire occurrence point.

[0009] Furthermore, the specific process for generating a dynamic fire spread prediction model is as follows: S21. Obtain the precise three-dimensional coordinates of the fire occurrence point uploaded by the edge computing node, locate the initial fire space in the global building digital twin model, perform adaptive three-dimensional mesh division on the initial fire space and adjacent connected spaces, generate a high-density mesh in the location near the fire occurrence point and ventilation opening, and generate a low-density mesh in the flat area away from the fire source. S22. Based on the temperature sensing data and infrared thermal imaging image uploaded in real time by the edge computing node at the fire point, the heat release rate of the fire point is estimated by inversion, and the heat release rate is used as the initial boundary condition for the three-dimensional mesh where the fire point is located. S23. Obtain real-time external wind speed and direction data at the fire location, as well as the real-time operating conditions of the building's HVAC system. The real-time operating conditions include the opening and closing status and air volume parameters of each air supply outlet, return air outlet, and smoke exhaust outlet. Map the real-time external wind speed and direction data and the real-time operating conditions to the three-dimensional mesh boundary of the global building digital twin model to construct fluid dynamic boundary conditions. S24. Obtain the BIM attribute labels of building materials around the fire point in the global building digital twin model, match the preset material combustion database according to the BIM attribute labels, obtain the ignition temperature, combustion heat value and smoke generation rate of the building materials, and construct the pyrolysis boundary conditions on the surface of solid materials. S25. Based on the set initial boundary conditions, fluid dynamics boundary conditions and pyrolysis boundary conditions, the computational fluid dynamics solver of the cloud server is used to solve the mass conservation equation, momentum conservation equation, energy conservation equation and component transport equation simultaneously. Then, iterative calculation is performed according to the preset time step to output the spatiotemporal evolution data of temperature field, smoke concentration field and visibility field of each three-dimensional grid node in the future continuous time segment, and integrated into a simulation dataset. S26. Using a data assimilation algorithm, the continuously obtained multimodal fire environment data is compared with the simulation dataset at the same time. If the error between the temperature field and smoke concentration field in the simulation dataset and the multimodal fire environment data exceeds a preset threshold, the Kalman filter algorithm is used to dynamically adjust the heat release rate of the fire source and the fluid boundary parameters, and the iterative calculation of the next time window is restarted to obtain the simulation dataset after continuous dynamic correction, which constitutes the dynamic fire spread prediction model.

[0010] Furthermore, the specific process for obtaining the safety weights of each evacuation route is as follows: S31. Extract the physical features of passageways from the global building digital twin model. Abstract the intersections of rooms, lobbies, and corridors within the building into a node set V. Abstract the physical evacuation channels connecting the nodes in node set V into an edge set E. Construct a three-dimensional evacuation topology graph G=(V, E) of the building space. Extract each edge from edge set E. The physical properties, including channel physical length and the effective passage width ; S32. Based on the functional use of each area within the enterprise, assign an area type label to each node in the 3D evacuation topology map, and, adhering to the principle of prioritizing life safety while also protecting property, preset priority evacuation weight coefficients, where: When the area type is a densely populated area, the highest first weight threshold is assigned; When the region type is a high-value region, a secondary weight threshold is assigned; When the region type is a normal region, a third weight threshold is assigned; When the region type is an uninhabited region, the lowest fourth weight threshold is assigned; S33. For any evacuation route Extract the priority evacuation weight coefficients of its two end nodes. and Calculate evacuation routes Basic evacuation demand :

[0011] The basic evacuation demand Used to reflect the natural carrying capacity of evacuation routes for the evacuation of people; S34. Obtain the precise three-dimensional coordinates of the fire location for each evacuation route in the three-dimensional evacuation topology map. The shortest Euclidean distance Dij between the three-dimensional line segment containing the evacuation route and the fire location is calculated using a spatial geometry algorithm. Then, a spatial distance attenuation function is introduced to calculate the distance penalty factor of the evacuation route affected by the location of the fire location. : ; in A preset attenuation constant related to the degree of enclosure of the building space; The smaller the shortest Euclidean distance Dij, the smaller the distance penalty factor. The closer it gets to 0; The larger the shortest Euclidean distance Dij, the greater the distance penalty factor. The closer it is to 1; S35, The basic evacuation demand level Distance penalty factor The initial safety weights of the evacuation routes are calculated by performing weighted fusion. : ; in, This is a dimensional unification adjustment coefficient; The initial security weight The higher the value, the higher the relative safety and evacuation priority of the evacuation route at the static physical level and in the initial stage of fire.

[0012] Furthermore, the specific process for generating the dynamically optimal evacuation path is as follows: S41. Obtain the evolution data within the future time window t in the dynamic fire spread prediction model, and map the evolution data onto the three-dimensional evacuation topology map. For each evacuation channel in the three-dimensional evacuation topology map, obtain the predicted average temperature value within the three-dimensional grid area where the evacuation channel is located. Predicted values ​​of toxic smoke concentration and visibility forecast values It also presets critical safety thresholds for the limits of human tolerance, including extreme temperature thresholds. Limiting flue gas concentration threshold and minimum safe visibility threshold ; S42. Construct a dynamic penalty coefficient based on multidimensional disaster characteristics. The specific calculation formula is as follows: ; in, , , These correspond to the average temperature prediction values. Predicted values ​​of toxic smoke concentration and visibility forecast values The disaster sensitivity weight factor, and satisfying + + =1; S43. Obtain initial security weights Calculate the comprehensive dynamic safety weight of the evacuation route at time t. : ; in, This is the preset dynamic adjustment coefficient; Simultaneously, a circuit breaker mechanism is set: if at time t, there is a circuit breaker within the channel... ≥ , ≥ , In any of the following situations, it is determined that the evacuation route is blocked by fire and thick smoke, and the comprehensive dynamic safety weight DSij(t) of the evacuation route is forcibly assigned to a minimum value, and the evacuation route is cut off in the three-dimensional evacuation topology map. S44. Using all densely populated areas, high-value areas, and ordinary areas within the fire spread area in the dynamic fire spread prediction model as the pathfinding starting point set, and the nodes of the safety exits on the ground floor and the refuge floors on the roof as the target endpoint set, the improved Dijkstra algorithm is used to search for the optimal path in the three-dimensional evacuation topology map with comprehensive dynamic safety weights. During the search process, the dynamic optimal evacuation path from each pathfinding starting point to the target endpoint is calculated with the objective function of maximizing the overall safety weight product of the path.

[0013] This invention also provides a fire safety data management system based on edge computing, including a fire judgment unit, a fire spread prediction unit, an evacuation planning unit, and a path generation unit, wherein: The fire detection unit is used to collect multimodal fire environment data in real time within the radiation area of ​​edge computing nodes deployed in various areas of the enterprise. When abnormal features are identified, the precise three-dimensional coordinates of the fire occurrence point are calculated based on the local building BIM model built into the edge computing node and the spatial distribution characteristics of the multimodal fire environment data. The fire spread prediction unit is used to upload the precise three-dimensional coordinates of the fire occurrence point and real-time multimodal fire environment data to the cloud server through edge computing nodes. The cloud server calls the global building digital twin model, and combines real-time wind speed, HVAC status and building material combustion characteristics to use computational fluid dynamics algorithms to perform high-precision three-dimensional simulation of fire spread and smoke flow, generating a dynamic fire spread prediction model. The evacuation planning unit is used to divide the area into different types based on the functional use of each area of ​​the enterprise. The area types include densely populated areas, high-value areas, ordinary areas, and uninhabited areas. Priority evacuation weights are assigned to the area types, and the safety weights of each evacuation route are further optimized based on the distance between the area distribution location and the precise three-dimensional coordinates of the fire point. The path generation unit is used to calculate the safety weight of each evacuation route in real time according to the dynamic fire spread prediction model, generate a dynamic optimal evacuation route that avoids the spread of fire and toxic smoke, and send control commands to the edge computing node, which drives the intelligent evacuation signs in the building to dynamically change their direction. In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This edge computing-based fire safety data management method and system leverages an edge computing architecture to integrate multimodal sensor data and local BIM models to output precise 3D coordinates of the fire source, significantly reducing search time for firefighters. Utilizing edge nodes for local anomaly identification and location calculations effectively reduces cloud communication latency and network congestion risks. By calling a global digital twin model and deeply coupling real-time wind speed, HVAC operating conditions, and material combustion characteristics for CFD simulation calculations, it can accurately and proactively evolve fire boundaries and the spread trajectory of toxic smoke. Furthermore, by comprehensively considering regional functional priorities and dynamic fire spread prediction models, it can calculate the safety weights of each passage in real time, proactively avoid fire and smoke blockage areas, and generate dynamically optimal paths, maximizing the protection of personnel safety. Attached Figure Description

[0014] Figure 1 A schematic diagram of the overall external structure of the present invention is shown; Figure 2 A schematic diagram of the overall method flow of the present invention is shown. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Example 1:

[0017] like Figure 1 As shown, a fire safety data management method based on edge computing is characterized by the following steps: S1: By deploying edge computing nodes in various regions of the enterprise, multimodal fire environment data within the node's radiation area is collected in real time. When abnormal features are identified, the precise three-dimensional coordinates of the fire occurrence point are calculated based on the local building BIM model built into the edge computing node and the spatial distribution characteristics of the multimodal fire environment data. The multimodal fire environment data includes temperature sensing data, smoke concentration data, infrared thermal imaging video stream, and visible light video stream. The specific process for obtaining the precise three-dimensional coordinates of the fire location is as follows: S11. The edge computing node acquires the visible light video stream and infrared thermal imaging video stream within the radiation area of ​​the node it is located. Based on the preset camera intrinsic and extrinsic parameter matrix, it performs spatiotemporal synchronization and affine registration on the visible light video stream and infrared thermal imaging video stream to obtain a dual-modal fusion image. S12. Input the dual-modal fused image into the pre-trained fire visual detection model, extract the two-dimensional pixel bounding box of the abnormal area, and extract the coordinates of the highest temperature pixel in the infrared thermal imaging video stream of the abnormal area. S13. Using the coordinates of the highest temperature pixel as the target feature point, call the local building BIM model built into the edge computing node to obtain the spatial absolute coordinates and monitoring field of view of the camera within the node's radiation area, and calculate the relative depth distance of the target feature point relative to the camera based on the principle of binocular visual parallax. S14. Using a coordinate system transformation matrix, the relative depth distance and the coordinates of the highest temperature pixel are mapped to the geodetic coordinate system of the local building BIM model to obtain the initial visual three-dimensional coordinates. S15. Obtain the topological location of each temperature sensor and smoke sensor deployed in the node radiation area, as well as the currently collected temperature sensing data and smoke concentration data. Using the spatial coordinates of the topological location as the reference point, construct a three-dimensional spatial gradient field of temperature and smoke concentration in the spatial grid of the local building BIM model using a spatial interpolation algorithm. Use the gradient descent method to perform source tracing iteration along the direction of increase in concentration and temperature in the three-dimensional spatial gradient field, and calculate the highest aggregation center point of temperature and smoke concentration as the coordinates of the sensing spatial source point. S16. The initial visual 3D coordinates and the coordinates of the sensing spatial source point are weighted and fused using the Kalman filter algorithm to obtain fused 3D coordinates. The fused 3D coordinates are then projected onto the local building BIM model for physical constraint verification. If the fused 3D coordinates fall in the suspended area (i.e., an air grid without physical support) in the local building BIM model, they are projected in the opposite direction of gravity to the surface of the nearest physical component (such as the surface of an equipment casing, floor, or wall). The corrected fused 3D coordinates are the final output accurate 3D coordinates of the fire occurrence point.

[0018] Specifically: When the infrared thermal imaging video stream is not obstructed by obstacles and the line of sight is clear, a higher fusion weight is given to the initial three-dimensional visual coordinates. When dense smoke is detected, causing a decrease in visibility and infrared transmittance, the coordinates of the source point in the sensing space are assigned a higher fusion weight.

[0019] S2: The precise three-dimensional coordinates of the fire location and real-time multimodal fire environment data are uploaded to the cloud server through edge computing nodes. The cloud server calls the global building digital twin model and combines real-time wind speed, HVAC status and building material combustion characteristics. Computational fluid dynamics algorithm is used to perform high-precision three-dimensional simulation of fire spread and smoke flow to generate a dynamic fire spread prediction model. The specific process for generating a dynamic fire spread prediction model is as follows: S21. Obtain the precise three-dimensional coordinates of the fire occurrence point uploaded by the edge computing node, locate the initial fire space in the global building digital twin model, perform adaptive three-dimensional mesh division on the initial fire space and adjacent connected spaces, generate a high-density mesh in the location near the fire occurrence point and ventilation opening, and generate a low-density mesh in the flat area away from the fire source. S22. Based on the temperature sensing data and infrared thermal imaging image uploaded in real time by the edge computing node at the fire point, the heat release rate of the fire point is estimated by inversion, and the heat release rate is used as the initial boundary condition for the three-dimensional mesh where the fire point is located. S23. Obtain real-time external wind speed and direction data at the fire location, as well as the real-time operating conditions of the building's HVAC system. The real-time operating conditions include the opening and closing status and air volume parameters of each air supply outlet, return air outlet, and smoke exhaust outlet. Map the real-time external wind speed and direction data and the real-time operating conditions to the three-dimensional mesh boundary of the global building digital twin model to construct fluid dynamic boundary conditions. S24. Obtain the BIM attribute labels of building materials around the fire point in the global building digital twin model, match the preset material combustion database according to the BIM attribute labels, obtain the ignition temperature, combustion heat value and smoke generation rate of the building materials, and construct the pyrolysis boundary conditions on the surface of solid materials. S25. Based on the set initial boundary conditions, fluid dynamics boundary conditions and pyrolysis boundary conditions, the computational fluid dynamics solver of the cloud server is used to solve the mass conservation equation, momentum conservation equation, energy conservation equation and component transport equation simultaneously. Then, iterative calculation is performed according to the preset time step to output the spatiotemporal evolution data of temperature field, smoke concentration field and visibility field of each three-dimensional grid node in the future continuous time segment, and integrated into a simulation dataset. S26. Using a data assimilation algorithm, the continuously obtained multimodal fire environment data is compared with the simulation dataset at the same time. If the error between the temperature field and smoke concentration field in the simulation dataset and the multimodal fire environment data exceeds a preset threshold, the Kalman filter algorithm is used to dynamically adjust the heat release rate of the fire source and the fluid boundary parameters, and the iterative calculation of the next time window is restarted to obtain the simulation dataset after continuous dynamic correction, which constitutes the dynamic fire spread prediction model.

[0020] S3: Based on the functional division of each area of ​​the enterprise, the area types are obtained, including densely populated areas, high-value areas, ordinary areas and uninhabited areas. Priority evacuation weights are assigned to the area types, and the safety weights of each evacuation route are further optimized based on the distance between the area distribution location and the precise three-dimensional coordinates of the fire point. The specific process for obtaining the safety weights of each evacuation route is as follows: S31. Extract the physical features of passageways from the global building digital twin model. Abstract the intersections of rooms, lobbies, and corridors within the building into a node set V. Abstract the physical evacuation routes (including corridors, staircases, and safety exits) connecting the nodes in node set V into an edge set E. Construct a three-dimensional evacuation topology graph G=(V, E) of the building space. Extract each edge from edge set E. The physical properties of the evacuation route (i.e., the evacuation path connecting node i and node j), including the physical length of the route. and the effective passage width ; S32. Based on the functional use of each area within the enterprise, assign an area type label to each node in the 3D evacuation topology map, and, adhering to the principle of prioritizing life safety while also protecting property, preset priority evacuation weight coefficients, where: When the area type is a densely populated area (such as a centralized office area or a meeting room), the highest first weight threshold is assigned; When the area type is a high-value area (such as a precision instrument room or a core computer room), a secondary weight threshold is assigned. When the region type is a normal region, a third weight threshold is assigned; When the area type is an uninhabited area (such as an abandoned warehouse or equipment mezzanine), the lowest fourth weight threshold is assigned. S33. For any evacuation route Extract the priority evacuation weight coefficients of its two end nodes. and Calculate evacuation routes Basic evacuation demand :

[0021] The basic evacuation demand Used to reflect the natural carrying capacity of evacuation routes for the evacuation of people; S34. Obtain the precise three-dimensional coordinates of the fire location for each evacuation route in the three-dimensional evacuation topology map. The shortest Euclidean distance Dij between the three-dimensional line segment containing the evacuation route and the fire location is calculated using a spatial geometry algorithm. Then, a spatial distance attenuation function is introduced to calculate the distance penalty factor of the evacuation route affected by the location of the fire location. : ; in A preset attenuation constant related to the degree of enclosure of the building space; The smaller the shortest Euclidean distance Dij (i.e., the closer the passage is to the ignition point), the smaller the distance penalty factor. The closer it gets to 0; The larger the shortest Euclidean distance Dij, the greater the distance penalty factor. The closer it is to 1; S35, The basic evacuation demand level Distance penalty factor The initial safety weights of the evacuation routes are calculated by performing weighted fusion. : ; in, This is a dimensional unification adjustment coefficient; The initial security weight The higher the value, the higher the relative safety and evacuation priority of the evacuation route at the static physical level and in the initial stage of fire.

[0022] S4: Calculate the safety weight of each evacuation route in real time based on the dynamic fire spread prediction model, generate the dynamic optimal evacuation route that avoids the spread of fire and toxic smoke, and send control commands to the edge computing node, which drives the intelligent evacuation signs in the building to dynamically change their direction.

[0023] The specific process of generating the dynamic optimal evacuation path is as follows: S41. Obtain the evolution data within the future time window t in the dynamic fire spread prediction model, and map the evolution data onto the three-dimensional evacuation topology map. For each evacuation channel in the three-dimensional evacuation topology map, obtain the predicted average temperature value within the three-dimensional grid area where the evacuation channel is located. Predicted values ​​of toxic smoke concentration and visibility forecast values It also presets critical safety thresholds for the limits of human tolerance, including extreme temperature thresholds. Limiting flue gas concentration threshold and minimum safe visibility threshold ; S42. Construct a dynamic penalty coefficient based on multidimensional disaster characteristics. The specific calculation formula is as follows: ; in, , , These correspond to the average temperature prediction values. Predicted values ​​of toxic smoke concentration and visibility forecast values The disaster sensitivity weight factor, and satisfying + + =1; S43. Obtain initial security weights Calculate the comprehensive dynamic safety weight of the evacuation route at time t. : ; in, This is the preset dynamic adjustment coefficient; Simultaneously, a circuit breaker mechanism is set: if at time t, there is a circuit breaker within the channel... ≥ , ≥ , In any of the following situations, it is determined that the evacuation route is blocked by fire and thick smoke, and the comprehensive dynamic safety weight DSij(t) of the evacuation route is forcibly assigned to a minimum value, and the evacuation route is cut off in the three-dimensional evacuation topology map. S43. Using all densely populated areas, high-value areas, and ordinary areas within the fire spread area in the dynamic fire spread prediction model as the set of pathfinding starting points, and the nodes of the safety exits on the ground floor and the refuge floors on the roof as the set of target endpoints, the improved Dijkstra algorithm is used to search for the optimal path in a three-dimensional evacuation topology map with comprehensive dynamic safety weights. During the search process, the dynamic optimal evacuation path from each pathfinding starting point to the target endpoint is calculated with the objective function of maximizing the overall safety weight product of the path.

[0024] The generated dynamic optimal evacuation path is distributed to the edge computing node corresponding to the fire spread area via the cloud server. Edge computing nodes parse the dynamic optimal evacuation path and convert it into underlying hardware control messages (such as control commands based on RS485 bus or LoRa wireless protocol), and send the hardware control messages point-to-point to the smart evacuation signs deployed in the building. After receiving instructions, the microcontroller inside the intelligent evacuation sign drives the LED light array to dynamically change the direction of the arrow, or activates flashing and voice warnings when it determines that there is a dead end ahead, thus completing the closed-loop guidance control from computing power to physical entity.

[0025] When abnormal features are detected, a locally preset hardware-level linkage strategy is triggered, and multiple control signals are output synchronously within a sub-second delay. These multiple control signals include a fire alarm signal sent to the fire control panel, cutting off the non-fire-fighting power supply to the fire area, turning on the smoke exhaust fan of the corresponding defense zone, unlocking the access control gate of the escape route, and pushing a real-time video pop-up of the fire point to the security monitoring center.

[0026] Example 2:

[0027] like Figure 2 As shown, a fire safety data management system based on edge computing is characterized by comprising a fire judgment unit, a fire spread prediction unit, an evacuation planning unit, and a path generation unit, wherein: The fire detection unit is used to collect multimodal fire environment data in real time within the radiation area of ​​edge computing nodes deployed in various areas of the enterprise. When abnormal features are identified, the precise three-dimensional coordinates of the fire occurrence point are calculated based on the local building BIM model built into the edge computing node and the spatial distribution characteristics of the multimodal fire environment data. The fire spread prediction unit is used to upload the precise three-dimensional coordinates of the fire occurrence point and real-time multimodal fire environment data to the cloud server through edge computing nodes. The cloud server calls the global building digital twin model, and combines real-time wind speed, HVAC status and building material combustion characteristics to use computational fluid dynamics algorithms to perform high-precision three-dimensional simulation of fire spread and smoke flow, generating a dynamic fire spread prediction model. The evacuation planning unit is used to divide the area into different types based on the functional use of each area of ​​the enterprise. The area types include densely populated areas, high-value areas, ordinary areas, and uninhabited areas. Priority evacuation weights are assigned to the area types, and the safety weights of each evacuation route are further optimized based on the distance between the area distribution location and the precise three-dimensional coordinates of the fire point. The path generation unit is used to calculate the safety weight of each evacuation route in real time according to the dynamic fire spread prediction model, generate a dynamic optimal evacuation route that avoids the spread of fire and toxic smoke, and send control commands to the edge computing node, which drives the intelligent evacuation signs in the building to dynamically change their direction.

[0028] The size of the interval and threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value.

[0029] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation. In the two embodiments provided in this application, it should be understood that the disclosed methods and systems can be implemented in other ways; for example, the division of modules is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed; another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices or modules, and may be electrical, mechanical or other forms. The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A fire safety data management method based on edge computing, characterized in that, Includes the following steps: S1: By deploying edge computing nodes in various regions of the enterprise, multimodal fire environment data within the node's radiation area is collected in real time. When abnormal features are identified, the precise three-dimensional coordinates of the fire occurrence point are calculated based on the local building BIM model built into the edge computing node and the spatial distribution characteristics of the multimodal fire environment data. The multimodal fire environment data includes temperature sensing data, smoke concentration data, infrared thermal imaging video stream, and visible light video stream. S2: The precise three-dimensional coordinates of the fire location and real-time multimodal fire environment data are uploaded to the cloud server through edge computing nodes. The cloud server calls the global building digital twin model and combines real-time wind speed, HVAC status and building material combustion characteristics. Computational fluid dynamics algorithm is used to perform high-precision three-dimensional simulation of fire spread and smoke flow to generate a dynamic fire spread prediction model. S3: Based on the functional division of each area of ​​the enterprise, the area types are obtained, including densely populated areas, high-value areas, ordinary areas and uninhabited areas. Priority evacuation weights are assigned to the area types, and the safety weights of each evacuation route are further optimized based on the distance between the area distribution location and the precise three-dimensional coordinates of the fire point. S4: Calculate the safety weight of each evacuation route in real time based on the dynamic fire spread prediction model, generate the dynamic optimal evacuation route that avoids the spread of fire and toxic smoke, and send control commands to the edge computing node, which drives the intelligent evacuation signs in the building to dynamically change their direction.

2. The fire safety data management method based on edge computing according to claim 1, characterized in that, When abnormal features are detected, a locally preset hardware-level linkage strategy is triggered, and multiple control signals are output synchronously within a sub-second delay. These multiple control signals include a fire alarm signal sent to the fire control panel, cutting off the non-fire-fighting power supply to the fire area, turning on the smoke exhaust fan of the corresponding defense zone, unlocking the access control gate of the escape route, and pushing a real-time video pop-up of the fire point to the security monitoring center.

3. The fire safety data management method based on edge computing according to claim 1, characterized in that, The specific process for obtaining the precise three-dimensional coordinates of the fire location is as follows: S11. The edge computing node acquires the visible light video stream and infrared thermal imaging video stream within the radiation area of ​​the node it is located. Based on the preset camera intrinsic and extrinsic parameter matrix, it performs spatiotemporal synchronization and affine registration on the visible light video stream and infrared thermal imaging video stream to obtain a dual-modal fusion image. S12. Input the dual-modal fused image into the pre-trained fire visual detection model, extract the two-dimensional pixel bounding box of the abnormal area, and extract the coordinates of the highest temperature pixel in the infrared thermal imaging video stream of the abnormal area. S13. Using the coordinates of the highest temperature pixel as the target feature point, call the local building BIM model built into the edge computing node to obtain the spatial absolute coordinates and monitoring field of view of the camera within the node's radiation area, and calculate the relative depth distance of the target feature point relative to the camera based on the principle of binocular visual parallax. S14. Using a coordinate system transformation matrix, the relative depth distance and the coordinates of the highest temperature pixel are mapped to the geodetic coordinate system of the local building BIM model to obtain the initial visual three-dimensional coordinates. S15. Obtain the topological location of each temperature sensor and smoke sensor deployed in the node radiation area, as well as the currently collected temperature sensing data and smoke concentration data. Using the spatial coordinates of the topological location as the reference point, construct a three-dimensional spatial gradient field of temperature and smoke concentration in the spatial grid of the local building BIM model using a spatial interpolation algorithm. Use the gradient descent method to perform source tracing iteration along the direction of increase in concentration and temperature in the three-dimensional spatial gradient field, and calculate the highest aggregation center point of temperature and smoke concentration as the coordinates of the sensing spatial source point. S16. The initial visual 3D coordinates and the coordinates of the sensing spatial source point are weighted and fused using the Kalman filter algorithm to obtain fused 3D coordinates. The fused 3D coordinates are then projected onto the local building BIM model for physical constraint verification. If the fused 3D coordinates fall in the suspended area of ​​the local building BIM model, they are projected in reverse along the direction of gravity to the surface of the nearest physical component. The corrected fused 3D coordinates are the final output accurate 3D coordinates of the fire occurrence point.

4. The fire safety data management method based on edge computing according to claim 1, characterized in that, The specific process for generating a dynamic fire spread prediction model is as follows: S21. Obtain the precise three-dimensional coordinates of the fire occurrence point uploaded by the edge computing node, locate the initial fire space in the global building digital twin model, perform adaptive three-dimensional mesh division on the initial fire space and adjacent connected spaces, generate a high-density mesh in the location near the fire occurrence point and ventilation opening, and generate a low-density mesh in the flat area away from the fire source. S22. Based on the temperature sensing data and infrared thermal imaging image uploaded in real time by the edge computing node at the fire point, the heat release rate of the fire point is estimated by inversion, and the heat release rate is used as the initial boundary condition for the three-dimensional mesh where the fire point is located. S23. Obtain real-time external wind speed and direction data at the fire location, as well as the real-time operating conditions of the building's HVAC system. The real-time operating conditions include the opening and closing status and air volume parameters of each air supply outlet, return air outlet, and smoke exhaust outlet. Map the real-time external wind speed and direction data and the real-time operating conditions to the three-dimensional mesh boundary of the global building digital twin model to construct fluid dynamic boundary conditions. S24. Obtain the BIM attribute labels of building materials around the fire point in the global building digital twin model, match the preset material combustion database according to the BIM attribute labels, obtain the ignition temperature, combustion heat value and smoke generation rate of the building materials, and construct the pyrolysis boundary conditions on the surface of solid materials. S25. Based on the set initial boundary conditions, fluid dynamics boundary conditions and pyrolysis boundary conditions, the computational fluid dynamics solver of the cloud server is used to solve the mass conservation equation, momentum conservation equation, energy conservation equation and component transport equation simultaneously. Then, iterative calculation is performed according to the preset time step to output the spatiotemporal evolution data of temperature field, smoke concentration field and visibility field of each three-dimensional grid node in the future continuous time segment, and integrated into a simulation dataset. S26. Using a data assimilation algorithm, the continuously obtained multimodal fire environment data is compared with the simulation dataset at the same time. If the error between the temperature field and smoke concentration field in the simulation dataset and the multimodal fire environment data exceeds a preset threshold, the Kalman filter algorithm is used to dynamically adjust the heat release rate of the fire source and the fluid boundary parameters, and the iterative calculation of the next time window is restarted to obtain the simulation dataset after continuous dynamic correction, which constitutes the dynamic fire spread prediction model.

5. The fire safety data management method based on edge computing according to claim 1, characterized in that, The specific process for obtaining the safety weights of each evacuation route is as follows: S31. Extract the physical features of passageways from the global building digital twin model. Abstract the intersections of rooms, lobbies, and corridors within the building into a node set V. Abstract the physical evacuation channels connecting the nodes in node set V into an edge set E. Construct a three-dimensional evacuation topology graph G=(V, E) of the building space. Extract each edge from edge set E. The physical properties, including channel physical length and the effective passage width ; S32. Based on the functional use of each area within the enterprise, assign an area type label to each node in the 3D evacuation topology map, and, adhering to the principle of prioritizing life safety while also protecting property, preset priority evacuation weight coefficients, where: When the area type is a densely populated area, the highest first weight threshold is assigned; When the region type is a high-value region, a secondary weight threshold is assigned; When the region type is a normal region, a third weight threshold is assigned; When the region type is an uninhabited region, the lowest fourth weight threshold is assigned; S33. For any evacuation route Extract the priority evacuation weight coefficients of its two end nodes. and Calculate evacuation routes Basic evacuation demand : The basic evacuation demand Used to reflect the natural carrying capacity of evacuation routes for the evacuation of people; S34. Obtain the precise three-dimensional coordinates of the fire location for each evacuation route in the three-dimensional evacuation topology map. The shortest Euclidean distance Dij between the three-dimensional line segment containing the evacuation route and the fire location is calculated using a spatial geometry algorithm. Then, a spatial distance attenuation function is introduced to calculate the distance penalty factor of the evacuation route affected by the location of the fire location. : ； in A preset attenuation constant related to the degree of enclosure of the building space; The smaller the shortest Euclidean distance Dij, the smaller the distance penalty factor. The closer it gets to 0; The larger the shortest Euclidean distance Dij, the greater the distance penalty factor. The closer it is to 1; S35, The basic evacuation demand level Distance penalty factor The initial safety weights of the evacuation routes are calculated by performing weighted fusion. : ； in, This is a dimensional unification adjustment coefficient; The initial security weight The higher the value, the higher the relative safety and evacuation priority of the evacuation route at the static physical level and in the initial stage of fire.

6. The fire safety data management method based on edge computing according to claim 1, characterized in that, The specific process of generating the dynamic optimal evacuation path is as follows: S41. Obtain the evolution data within the future time window t in the dynamic fire spread prediction model, and map the evolution data onto the three-dimensional evacuation topology map. For each evacuation channel in the three-dimensional evacuation topology map, obtain the predicted average temperature value within the three-dimensional grid area where the evacuation channel is located. Predicted values ​​of toxic smoke concentration and visibility forecast values It also presets critical safety thresholds for the limits of human tolerance, including extreme temperature thresholds. Limiting flue gas concentration threshold and minimum safe visibility threshold ; S42. Construct a dynamic penalty coefficient based on multidimensional disaster characteristics. The specific calculation formula is as follows: ； in, , , These correspond to the average temperature prediction values. Predicted values ​​of toxic smoke concentration and visibility forecast values The disaster sensitivity weight factor, and satisfying + + =1; S43. Obtain initial security weights Calculate the comprehensive dynamic safety weight of the evacuation route at time t. : ； in, This is the preset dynamic adjustment coefficient; Simultaneously, a circuit breaker mechanism is set: if at time t, there is a circuit breaker within the channel... ≥ , ≥ , In any of the following situations, it is determined that the evacuation route is blocked by fire and thick smoke, and the comprehensive dynamic safety weight DSij(t) of the evacuation route is forcibly assigned to a minimum value, and the evacuation route is cut off in the three-dimensional evacuation topology map. S44. Using all densely populated areas, high-value areas, and ordinary areas within the fire spread area in the dynamic fire spread prediction model as the pathfinding starting point set, and the nodes of the safety exits on the ground floor and the refuge floors on the roof as the target endpoint set, the improved Dijkstra algorithm is used to search for the optimal path in the three-dimensional evacuation topology map with comprehensive dynamic safety weights. During the search process, the dynamic optimal evacuation path from each pathfinding starting point to the target endpoint is calculated with the objective function of maximizing the overall safety weight product of the path.

7. A fire safety data management system based on edge computing, characterized in that, It includes a fire assessment unit, a fire spread prediction unit, an evacuation planning unit, and a path generation unit, among which: The fire detection unit is used to collect multimodal fire environment data in real time within the radiation area of ​​edge computing nodes deployed in various areas of the enterprise. When abnormal features are identified, the precise three-dimensional coordinates of the fire occurrence point are calculated based on the local building BIM model built into the edge computing node and the spatial distribution characteristics of the multimodal fire environment data. The fire spread prediction unit is used to upload the precise three-dimensional coordinates of the fire occurrence point and real-time multimodal fire environment data to the cloud server through edge computing nodes. The cloud server calls the global building digital twin model, and combines real-time wind speed, HVAC status and building material combustion characteristics to use computational fluid dynamics algorithms to perform high-precision three-dimensional simulation of fire spread and smoke flow, generating a dynamic fire spread prediction model. The evacuation planning unit is used to divide the area into different types based on the functional use of each area of ​​the enterprise. The area types include densely populated areas, high-value areas, ordinary areas, and uninhabited areas. Priority evacuation weights are assigned to the area types, and the safety weights of each evacuation route are further optimized based on the distance between the area distribution location and the precise three-dimensional coordinates of the fire point. The path generation unit is used to calculate the safety weight of each evacuation route in real time according to the dynamic fire spread prediction model, generate a dynamic optimal evacuation route that avoids the spread of fire and toxic smoke, and send control commands to the edge computing node, which drives the intelligent evacuation signs in the building to dynamically change their direction.

Images