Fire risk remote dynamic assessment method and system based on multi-source risk perception

Abstract

This invention relates to the field of fire early warning technology and discloses a method and system for remote dynamic assessment of fire risk based on multi-source risk perception. The method includes: performing real-time synchronous acquisition and standardized preprocessing of multi-source heterogeneous fire data; constructing a dynamic directed graph topology network of the building space and calculating the environmental propagation coefficient; adaptively allocating weight coefficients for various data modes and calculating the primary risk intensity of local risk sources; measuring the comprehensive risk potential of a target assessment point within the space after being affected by the combined influence of all risk sources; comparing the comprehensive risk potential and primary risk intensity with the comprehensive safety threshold and primary safety threshold respectively, classifying the risk level of the assessment point, and executing a multi-level linkage response based on the risk level determination results. This invention can quantitatively reflect the dynamic impact of the real-time physical state inside a building on the cross-regional spread of fire risk, and improves emergency evacuation efficiency by reconstructing dynamic evacuation paths to prevent evacuees from accidentally entering high-risk propagation channels.

Description

Technical Field

[0001] This invention relates to the field of fire early warning technology, specifically to a method and system for remote dynamic assessment of fire risks based on multi-source risk perception. Background Technology

[0002] As urban building scale continues to expand, the number of large commercial complexes and complex industrial parks is increasing day by day. These places have intricate internal spatial structures and are often densely populated, so fire early warning inside these buildings has become an important aspect.

[0003] Fire risk assessment methods aim to monitor and warn of potential fire hazards by collecting environmental data within buildings, thereby providing a basic guarantee for personnel evacuation and emergency rescue. In existing fire safety management, conventional fire risk assessment methods typically employ static zoning and threshold triggering mechanisms. These systems rely primarily on sensors deployed in various areas for monitoring; once environmental parameters reach fixed alarm thresholds, the system is triggered and capable of synchronous alarms across the entire area. Simultaneously, fixed evacuation signs are usually installed inside buildings to provide escape route guidance to people on site when the alarm sounds.

[0004] However, the existing fire risk assessment methods have some shortcomings: although they have the ability to provide synchronous alarms across the entire area, they only solve the problem of hazard notification, ignoring the dynamic impact of internal building conditions such as the opening and closing status of fire doors and the status of HVAC ventilation on the cross-area spread of fire risk. This static mechanism cannot predict the direction of fire spread or dynamically assess the degree of danger in various physical spaces within the building, which can easily lead to people colliding head-on with spreading high-temperature toxic gases while escaping according to fixed signs, making it impossible for them to evacuate in time or even causing danger. Therefore, there is an urgent need for a remote dynamic fire risk assessment method and system based on multi-source risk perception to solve the above problems. Summary of the Invention

[0005] To address the problems in related technologies, this invention provides a remote dynamic assessment method for fire risk based on multi-source risk perception, thereby overcoming the aforementioned technical problems in existing related technologies.

[0006] To solve the aforementioned technical problem, the present invention is achieved through the following technical solution: In a first aspect, embodiments of the present invention provide a remote dynamic assessment method for fire risk based on multi-source risk perception, specifically including: collecting static topological data and continuous physical quantities of the target building, simultaneously smoothing and denoising the collected continuous physical quantities and determining dynamic baselines to obtain a relative deviation value reflecting the degree of danger; constructing a dynamic directed graph topological network of the building space based on the static topological data, and superimposing airflow velocity and physical barrier attenuation factors in the connected paths to dynamically calculate the environmental propagation coefficient reflecting the equivalent spatial penetration scale of risk spread; adaptively allocating weight coefficients of various data modes according to the relative deviation value, individually assessing the fire occurrence probability of each node, and calculating the primary risk intensity of local risk sources; calculating the comprehensive risk potential of the target assessment point in the space after being affected by the combined influence of all risk sources according to the environmental propagation coefficient and the primary risk intensity; comparing the comprehensive risk potential and the primary risk intensity with the comprehensive safety threshold and the primary safety threshold respectively, remotely and dynamically classifying the risk level of the assessment point, and executing a multi-level linkage response according to the risk level determination result.

[0007] As a preferred embodiment of the fire risk remote dynamic assessment method based on multi-source risk perception described in this invention, the formula for calculating the relative deviation value is as follows: ; In the formula, For nodes First Physical quantities in The relative deviation value at time, For the feature normalized span value, For nodes Data mode At the present moment The effective physical quantity value, For nodes Data mode exist The dynamic safety baseline threshold at any given time.

[0008] As a preferred embodiment of the remote dynamic assessment method for fire risk based on multi-source risk perception described in this invention, the formula for calculating the environmental propagation coefficient is as follows: ; In the formula, For nodes To the node exist Environmental propagation coefficient at any given time For nodes arrive The fundamental propagation scale constant, For those caused by HVAC systems or natural wind point to airflow speed in the direction, The characteristic time constant; The physical barrier attenuation factor is taken as the value when the fire door is fully open. When fully closed and well-sealed, take a value close to minimum value .

[0009] As a preferred embodiment of the fire risk remote dynamic assessment method based on multi-source risk perception described in this invention, the formula for calculating the basic propagation scale constant is as follows: ; In the formula, For nodes arrive The fundamental propagation scale constant, The natural diffusion reference distance under standard reference conditions. For connecting nodes and The actual physical net cross-sectional area of ​​the passage between them The area of ​​the doorway is a preset reference standard. This is the scaling index for spatial morphology.

[0010] As a preferred embodiment of the remote dynamic assessment method for fire risk based on multi-source risk perception described in this invention, the risk caused by the HVAC system or natural wind... point to airflow speed in the direction The calculation formula is: In the formula, The current three-dimensional wind field velocity vector. , These are the three-dimensional absolute coordinate vectors of the target node and the risk source node, respectively.

[0011] As a preferred embodiment of the fire risk remote dynamic assessment method based on multi-source risk perception described in this invention, the calculation formula for the primary risk intensity is as follows: In the formula, For nodes exist The initial risk intensity at any given moment, For sensor data mode sets, These are the prior weight coefficients for various types of data.

[0012] As a preferred embodiment of the fire risk remote dynamic assessment method based on multi-source risk perception described in this invention, the formula for calculating the comprehensive risk potential is as follows: In the formula, As assessment point exist The comprehensive risk potential at any moment The total number of risk sources with a non-zero primary risk intensity. For specific risk sources The hazard level coefficient, Spatial path distance, This is the topological attenuation ratio constant.

[0013] As a preferred embodiment of the fire risk remote dynamic assessment method based on multi-source risk perception described in this invention, the remote dynamic classification of assessment points into risk levels includes: setting a comprehensive safety threshold. and primary security threshold If comprehensive risk potential If so, it is considered a safe zone; if And the initial risk intensity If it is, then it is determined to be a high-risk area; if And the initial risk intensity If so, it is determined to be an area about to spread.

[0014] As a preferred embodiment of the fire risk remote dynamic assessment method based on multi-source risk perception described in this invention, the multi-level linkage response includes a linkage prevention mechanism and evacuation route reconstruction; the linkage prevention mechanism includes closing fire doors and reversing the HVAC air supply direction; the evacuation route reconstruction includes marking comprehensive risk potential. Exceeding the comprehensive safety threshold The topological edges and nodes are dynamic restricted areas, which in turn change the direction of the intelligent evacuation signs.

[0015] Secondly, embodiments of the present invention provide a remote dynamic assessment system for fire risk based on multi-source risk perception, comprising: a data acquisition and preprocessing module for acquiring static topological data and continuous physical quantities of the target building to obtain a relative deviation value reflecting the degree of danger; a topology and environmental dynamics module for constructing a dynamic directed graph topological network of the building space and dynamically calculating the environmental propagation coefficient reflecting the equivalent spatial penetration scale of risk spread; a primary risk assessment module for adaptively allocating weight coefficients of various data modes based on the relative deviation value to assess the primary risk intensity of local risk sources; a global risk field strength calculation module for introducing the environmental propagation coefficient and the primary risk intensity to calculate the comprehensive risk potential of the target assessment point in the space after being affected by the synergistic influence of all risk sources; and a risk classification and linkage response module for comparing the comprehensive risk potential and the primary risk intensity with the comprehensive safety threshold and the primary safety threshold, respectively, classifying the risk level of the assessment point, and executing a multi-level linkage response based on the risk level determination result.

[0016] The present invention has the following beneficial effects: 1. This invention, through dynamic calculation of the environmental propagation coefficient, can realistically reflect the dynamic impact of real-time physical states inside a building, such as increased spread due to ventilation airflow or blocked spread due to closed fire doors, on the cross-regional spread of fire risk during fire early warning. It provides environmental driving parameters for predicting the direction of fire spread. From the perspective of the global physical field, it shows the nonlinear attenuation law of fire and high-temperature toxic gases in the complex three-dimensional topology of a building under the combined effects of fire door blocking state, ventilation convection, spatial channel path, and intrinsic fire load. It enables the judgment of the probability of fire occurrence in the area and dynamically quantifies the spatiotemporal distribution of risk cross-regional spread.

[0017] 2. This invention introduces a dynamic adaptive dual-threshold judgment based on historical environmental noise and topology, combined with multi-level linkage response. This not only curbs the spread of fire in the early stages of a fire from a physical perspective, but also prevents escaping people from accidentally entering high-risk transmission channels such as high-temperature toxic gas accumulation areas through dynamic evacuation path reconstruction, thereby improving the emergency evacuation efficiency and the level of personnel safety protection in complex building environments.

[0018] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, the drawings can be obtained from these drawings without creative effort.

[0020] Figure 1The present invention provides a flowchart of a method for remote dynamic assessment of fire risk based on multi-source risk perception.

[0021] Figure 2 This is a schematic diagram of the S5 process provided by the present invention.

[0022] Figure 3 This invention provides a schematic diagram of a remote dynamic assessment system for fire risks based on multi-source risk perception. Detailed Implementation

[0023] 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.

[0024] Example 1 In fire safety management of large commercial complexes or complex industrial parks, existing fire assessment methods typically employ static zoning and fixed threshold triggering mechanisms. While these methods can provide synchronous alarms across the entire area, they only address the issue of hazard notification and neglect the dynamic impact of internal building conditions, such as the opening and closing status of fire doors and the ventilation status of HVAC systems, on the cross-regional spread of fire risk. They cannot predict the direction of fire spread or dynamically assess the degree of danger in each physical space within the building. This can easily lead to people colliding head-on with the spreading high-temperature toxic gases when escaping according to fixed signs, making it impossible for them to evacuate in time or even causing them danger.

[0025] To solve the above technical problems, such as Figure 1 As shown, Embodiment 1 of the present invention provides a remote dynamic assessment method for fire risk based on multi-source risk perception. Specifically, Embodiment 1 takes the fire protection Internet of Things transformation scenario of a large multi-story commercial complex as an example. On-site, multi-source sensors covering temperature, smoke, and electrical parameters, as well as door magnetic switches and HVAC wind speed and direction sensors are deployed.

[0026] In the specific implementation of Example 1: First, real-time synchronous acquisition of multi-source heterogeneous fire protection data is performed to obtain static topological basic data and continuous physical quantities. Simultaneously, the acquired continuous physical quantities are smoothed, denoised, and dynamically baselined to obtain a relative deviation value reflecting the degree of danger. This method eliminates the differences in sampling frequency, transmission delay, and dimensions among various sensor heterogeneous data, providing a highly time-consistent data source for the subsequent construction of a dynamic directed graph. Second, a dynamic directed graph topology network of the building space is constructed based on the static topological basic data. Airflow velocity and physical barrier factors are superimposed on the connected paths to dynamically calculate the environmental propagation coefficient reflecting the equivalent spatial penetration scale of risk spread. This method can realistically and quantitatively reflect the dynamic impact of real-time physical states such as intensified airflow spread within the building or closed fire doors on the cross-regional evolution of fire risk during fire early warning. Then, the weight coefficients of various data modes are adaptively allocated based on the relative deviation value, and the fire occurrence probability of each node is individually evaluated to calculate the primary risk intensity of the local risk source. This method accurately reflects the fire occurrence probability exhibited by a single node without considering spatial propagation. Next, by introducing the environmental propagation coefficient and primary risk intensity, the global dynamic risk field strength based on topological tension and physical field coupling is calculated, obtaining the comprehensive risk potential of the target assessment point within space after being affected by the combined influence of all risk sources. This method dynamically quantifies the spatiotemporal distribution and nonlinear attenuation law of risk spread across regions. Finally, the comprehensive risk potential and primary risk intensity are compared with dynamically generated comprehensive safety thresholds and primary safety thresholds, respectively, to remotely and dynamically classify the risk level of the assessment point, and to execute multi-level linkage response based on the risk level determination results. This method not only physically curbs the spread of fire in the early stages of a fire, but also prevents escaping people from accidentally entering high-risk propagation channels such as high-temperature toxic gas accumulation areas through dynamic evacuation path reconstruction, thereby improving the emergency evacuation efficiency and the level of personnel safety protection in complex building environments.

[0027] Furthermore, to better illustrate the technical solution of Embodiment 1 of the present invention, a detailed description of the remote dynamic assessment method for fire risk based on multi-source risk perception is provided, specifically including the following: S1. Perform real-time synchronous acquisition and standardized preprocessing of multi-source heterogeneous fire protection data, specifically including the following sub-steps: S11. Collect the spatial three-dimensional coordinate node set of each smoke control zone and room. And the physical connectivity between nodes (edge ​​set) This serves as the static topology foundation data. The specific implementation steps are as follows: S111. Obtain the Building Information Model (BIM) file (e.g., IFC format) or 3D CAD drawings of the target building, and extract the geometric data and semantic information of the building's interior through the model parsing engine on the cloud server. The semantic information includes the spatial layout and attribute labels of walls, doors, windows, stairwells, elevator shafts, and HVAC ventilation ducts.

[0028] S112. Based on the resolved geometric data, the building's interior space is divided into multiple independent three-dimensional spatial polyhedra according to physical partitions and smoke control zones. For each independent three-dimensional spatial polyhedron (such as an independent room, a section of corridor, or a designated smoke control zone within a large open space), its three-dimensional geometric center point is extracted as a topological node using a volume integral calculation algorithm. Each node is assigned a globally unique identification code (ID), and its corresponding three-dimensional absolute coordinates are recorded. All nodes combine to form a static set of building topology nodes. .

[0029] S113. Determine the node set through spatial Boolean operations and semantic association analysis. Any two adjacent nodes and Does a physical connection exist between them? If the spaces represented by two nodes are directly adjacent through a physical door, an operable window, a shared corridor opening, or a connecting HVAC duct and have a channel for air / heat exchange, then a connection is established between them in the graph model. and undirected edge Traverse all nodes to form a basic edge set that records the original physical connectivity of the building. .

[0030] S114. For edge sets Each connected edge in The static physical properties of the connected path are extracted based on the model's semantic information and then assigned values. Specifically, this includes: Connection type label: Indicates whether this edge belongs to a normally closed fire door, normally open fire door, ordinary wooden door, ventilation duct, or an open boundary without physical barriers.

[0031] Basic flow cross-sectional area: Extract the maximum physical cross-sectional area of ​​the connecting channel (e.g., the area of ​​a door or the cross-sectional area of ​​a ventilation duct) as the basic propagation scale constant for subsequent calculations. Reference basis.

[0032] S115. The node set constructed above Edge set The three-dimensional coordinates and static attributes bound to the target building are encapsulated in a graph database format to generate a basic static topology map of the target building. This map is stored on a remote server as the system foundation, providing a reference coordinate system and topological constraints for subsequent overlay of dynamic sensing data (such as door sensor status and wind speed) to evolve into a dynamic directed graph.

[0033] S12. Real-time collection of continuous physical quantities, including temperature at each node, through distributed IoT sensor nodes. Smoke concentration Cable temperature in critical electrical circuits Simultaneously, dynamic discrete state variables are collected, including the opening and closing status of normally open / normally closed fire doors between nodes, and the wind speed and direction of HVAC supply and exhaust vents. The specific implementation steps are as follows: S121. Deploy multi-source heterogeneous IoT sensor devices (including temperature sensors, photoelectric smoke detectors, residual current transformers, door magnetic switches, and anemometers) at the construction site. Uniquely bind the physical MAC address or IP address of each sensor device to the three-dimensional absolute coordinates of its installation location, and map it to the static topology node set constructed in S11. or edge set In particular, environmental measurement sensors are mapped to nodes. Door magnetic and ventilation sensors are mapped to the connected edge .

[0034] S122, Regarding temperature Smoke concentration Temperature of cables in critical electrical circuits Dynamic, continuous physical quantities that are susceptible to interference from the on-site electromagnetic environment and airflow fluctuations are controlled by the edge computing gateway to operate at a fixed frequency (e.g., High-frequency oversampling is performed to obtain the original discrete sequence data. To eliminate transient thermal noise from the sensor itself and environmental spikes, the edge gateway incorporates a time-series smoothing and denoising algorithm to perform weighted calculations on the original sequence. The discrete-time exponential weighted smoothing formula is as follows: ; In the formula, For nodes Data mode (like At the current discrete time The effective physical quantity value after smoothing and denoising will be used as the basis for subsequent calculations of the initial risk intensity. The basis for input; For sensors In history Each sampling time The original physical quantity measurements obtained; This represents the total number of sampling points within the sliding data cleaning window. These represent the current time and the historical sampling time, respectively. This is the characteristic smoothing time constant for this type of sensor.

[0035] S123. For the opening and closing status of normally open / normally closed fire doors between nodes, an event-driven mechanism is adopted. Only when the physical opening / closing state of the door magnetic switch changes is the sensor actively pushing a status change message and hardware interrupt signal to the upper-level gateway, reducing communication channel occupancy. For the wind speed and direction of HVAC supply and exhaust vents, a lower frequency (e.g., ...) is used. The periodic inspection mechanism of the wind turbine system obtains its current discrete state quantity because the inertia of the wind turbine system is relatively large and the state changes relatively slowly.

[0036] S124. Due to differences in sampling frequency, transmission delay, and communication protocols among different types of sensors, the edge computing gateway utilizes the Network Time Protocol (NTP) to process all received data. And discrete state variables are uniformly timestamped to evaluate the system's computation cycle. Based on this, asynchronously arriving heterogeneous data are aligned to the same time segment using a linear interpolation method to form the data at that moment. A panoramic multi-source state matrix.

[0037] The S125 edge computing gateway encapsulates the aligned panoramic multi-source state matrix into a standard JSON format and sends it to the remote cloud server in real time via a wired fiber optic private network for subsequent dynamic graph evolution and physical field strength calculation. When transient congestion or disconnection is detected in the uplink network, the edge gateway temporarily stores the state matrix in local NVRAM (non-volatile random access memory). After the network is restored, it appends the original timestamp and performs high-priority breakpoint resumption, ensuring that the cloud-based dynamic risk assessment model does not experience time sequence gaps or misjudgments due to communication interruptions.

[0038] S13. Normalize the collected continuous physical quantities according to their maximum safe baseline based on their physical characteristics to eliminate dimensional differences and provide basic input for subsequent primary intensity assessment. The specific implementation steps are as follows: S131. Because different functional areas within a building (such as electrical rooms, underground parking garages, and catering kitchens) have drastically different safety tolerances for the same physical quantity, the system first uses the architectural spatial semantic tags parsed in S11 to define the safety tolerance for each node. Configure an independent static physical feature baseline matrix. For any node The first For sensor-like modes, extract their corresponding upper limit reference values ​​(i.e., static safety baseline thresholds) representing the physical quantity in a normal and safe state. ) and the typical numerical range (i.e., the characteristic normalized range value) representing the development of this physical quantity from the initial warning to full danger. ).

[0039] S132. Considering that seasonal changes, diurnal temperature variations, or normal business operations in the building's environment can cause slow drift in background physical quantities (e.g., the base temperature of top-floor rooms is generally higher than that of basements in winter), using a static baseline is highly likely to cause false alarms. The system compensates for low-frequency environmental drift on the static baseline by designing a dynamic safety baseline function: ; In the formula, For nodes Data mode exist Dynamic safety baseline threshold at any given time; This is the low-frequency drift compensation coefficient, with a value range of... This is used to control the weighting of the background environment's influence on the alarm baseline. For absolute hazardous quantities such as smoke concentration, which should not be compromised by the environment, this applies. For physical quantities such as temperature that are easily affected by the environment, ; The background sampling window width for ultra-long time series is the average of the data over the past 24 hours. The smoothed effective physical quantity output in S12; This is the standard reference background value for this region / node.

[0040] S133, High-frequency smoothing effective physical quantity based on S12 output The dynamic safety baseline calculated by S132 Dimensionless calculations are performed to map all heterogeneous physical quantities into pure numerical values ​​that reflect the degree of danger, which can be represented by the following formula: ; In the formula, For nodes First Physical quantities in The relative deviation at time; For the feature normalized span value, For nodes Data mode At the present moment The effective physical quantity value.

[0041] S134, in the output Previously, if the underlying sensor experienced a disconnection, short circuit, or malicious damage, it would often output an extremely high full-scale fault value (such as a temperature display). The system sets the upper limit of hardware constraints. (For example, set as) ).when And if the duration exceeds the hardware failure determination period, the system will terminate the sensor. Forced to It also reports equipment damage maintenance work orders separately to the cloud, thereby preventing the entire area assessment model from crashing or experiencing a chain of false alarms due to input extreme outliers caused by the physical failure of a single sensor.

[0042] In this embodiment 1, a cloud-based parsing engine is used to extract three-dimensional spatial geometric data and semantic information from the building information model, and to construct a set of nodes. With edge set The static topology base is composed of an edge computing gateway that aggregates continuous physical quantities and discrete state quantities collected by multi-source heterogeneous IoT sensors, synchronizes timestamps using a network time protocol and performs interpolation alignment to form a panoramic multi-source state matrix; combined with a low-frequency drift compensation coefficient Calculate the dynamic safety baseline threshold The effective physical quantity after smoothing and denoising is converted into a dimensionless relative deviation value. This method integrates the static spatial structural features of a building with real-time evolving dynamic fire sensing data, eliminating differences in sampling frequency, transmission delay, and dimensions among heterogeneous data from various sensors. This provides a highly time-consistent data source for subsequent construction of dynamic directed graphs and calculation of primary risk intensity.

[0043] S2. Construct a dynamic directed graph topology network for the building space and dynamically calculate the environmental propagation coefficient, specifically including the following sub-steps: S21. Model the building's interior space as a dynamic directed graph. For any connected path in the graph (i.e., from the risk source node) To the target evaluation node (path), dynamically calculate its environmental propagation coefficient .

[0044] S22, Environmental Propagation Coefficient The calculation formula is: ; In the formula, For nodes To the node exist The environmental propagation coefficient at any given time represents the equivalent spatial penetration scale of risk spread; It is the basic propagation scale constant, representing the reference scale for natural thermal radiation and flue gas diffusion; For those caused by HVAC systems or natural wind point to The speed of airflow in the direction; The physical barrier attenuation factor is taken as the value when the fire door is fully open. When fully closed and well-sealed, take a value close to minimum value ; The characteristic time constant represents the physical time difference for the system to make a complete assessment of changes in environmental state. It is determined by the sum of the physical response time constants of the heterogeneous front-end sensors and the total clock cycles of a single system calculation and network transmission. Specifically: S221, Strongly correlated with the physical form of the connected path, the basic flow cross-sectional area extracted from S114 is used to calculate the basic propagation scale constant. : ; In the formula, For nodes arrive The fundamental propagation scale constant; The natural diffusion reference distance under standard reference conditions was obtained through standard combustion chamber experiments. For the connection nodes extracted by S114 and The actual net physical cross-sectional area of ​​passageways (such as doorways and corridors); This is the preset reference standard doorway area; It is a spatial scaling index, which characterizes the degree of nonlinear influence of area changes on diffusion distance; Specifically, the natural diffusion baseline distance Source nodes are extracted from BIM data using the equivalent volume method. Three-dimensional physical volume Calculation yielded: .

[0045] Specifically, spatial form scaling index Based on the target node The physical boundary model is obtained through spatial orthogonal feature scaling dimensionality reduction, and its expression is as follows: ; In the formula, , , To be based on the target node The feature lengths of the minimum circumscribed cuboid extracted from the corresponding 3D polyhedron along the three orthogonal physical axes, satisfying the sorting constraints. .

[0046] For example: In this embodiment, the spatial form scaling index The above formula is used for dynamic adaptive calculation, as detailed below: One-dimensional strongly directional spaces (such as narrow corridors and HVAC ventilation ducts): are characterized by When the fluid is strongly constrained by the sidewalls and propels in a piston-like manner, the formula calculation result will be... Cross-sectional area The increase of the value directly and linearly increases the propagation distance, which conforms to the physical laws of one-dimensional confined jets; Two-dimensional, confined, flat spaces (such as conventional office areas and standard rooms): Their characteristics are as follows When the fluid mainly diffuses in a horizontal plane, the formula calculation result will be... This is equivalent to reducing the dimensionality by taking the square root of the area in the original formula, mapping it to a linear diffusion scale; Three-dimensional freely expanding spaces (such as grand atriums and spacious lobbies): their characteristics are as follows When the fluid exhibits free diffusion as a three-dimensional axisymmetric plume, the formula calculation results will be... This is equivalent to the dimensionality reduction of the volume of space by the cube root.

[0047] The above calculation method makes it possible to The value of is in Achieving a continuous and smooth transition within the range, driven by BIM geometric data, ensures the universal applicability and robustness of the system model in heterogeneous and complex architectural spaces.

[0048] S222. Combining the wind direction sensor data obtained from the inspection in S123 with the three-dimensional coordinates of the nodes in S112, extract the effective wind speed component and calculate the airflow velocity. : ; In the formula, This is the current three-dimensional wind field velocity vector collected by the anemometer in S123. , This represents the three-dimensional absolute coordinate vectors of the target node and the risk source node extracted from S112. Subtracting the two yields the result from... point to The unit direction vector of the path is obtained by dividing the direction vector by its L2 norm (distance).

[0049] In this embodiment 1, the static three-dimensional physical space connectivity path inside the building is deeply coupled with dynamic environmental perception variables, and the foundation propagation scale constant is calculated by extracting the foundation flow cross-sectional area. And superimposed by airflow speed Characteristic time constant and physical barrier attenuation factor The dynamic convection term, which is jointly determined, is used to solve the environmental propagation coefficient from the risk source node to the target assessment node in real time. Specifically, for example: setting a basic propagation scale. Characteristic time constant If two adjacent nodes and The fire door in the room is in the open position, that is... And HVAC systems are becoming The wind speed from Blowing towards ,Right now Then the calculation yields This indicates that under favorable wind and unobstructed conditions, the risk penetration capability is significantly enhanced. If the fire door is closed at this time ( ),but The equivalent penetration scale drops sharply. This method overcomes the shortcomings of existing fire assessment methods that rely solely on static zoning and static distance. It can realistically and quantitatively reflect the dynamic impact of real-time physical conditions inside the building (such as ventilation airflow intensifying the spread or fire doors closing to block the spread) on the cross-regional evolution of fire risk, and provides environmental driving parameters for subsequent prediction of fire spread direction and calculation of global risk field strength.

[0050] S3. Dynamically calculate the primary risk intensity of local risk sources. Specifically, this involves assessing a single node without considering spatial propagation. In itself The probability of fire occurrence at any given moment reflects the initial risk intensity and is represented by the following formula: In the formula, For nodes exist The initial risk intensity at any given moment, and its range. ; A set of sensor data modes; Let be the prior weight coefficients for each type of data and satisfy the following conditions: .

[0051] For example, in this embodiment 1, the system automatically performs dynamic weight redistribution based on the anomaly level of real-time data from each heterogeneous sensor at the current moment. The relative deviation value output in step S133 is then invoked. Calculate the prior weight coefficients for each type of data: ; In the formula, For nodes First Various sensor modes (such as temperature) ,smoke ,electric )exist Dynamic adaptive weights at any given time; , This refers to the dimensionless relative deviation value of the individual physical quantity obtained in step S133. Specifically, for example: the temperature at a certain node... An error occurred, settings Dynamic security baseline Feature normalized span value When the measured temperature At that time, calculate ;calculate .

[0052] S4. Calculate the global dynamic risk field strength based on the coupling of topological tension and physical field. Specifically, this involves a cloud server combined with the environmental propagation coefficient. Compared with primary risk intensity Any node in the computation space The formula for calculating the comprehensive risk potential after the combined influence of all risk sources within the building is as follows: In the formula, As assessment point exist The comprehensive risk potential at any given moment; The total number of risk sources with a non-zero primary risk intensity; For specific risk sources The hazard level coefficient; Spatial path distance; This is the topological attenuation ratio constant.

[0053] For example, the above parameters , and The method for determining it is as follows: S41, Spatial Path Distance The directed graph of digital topology of building space is based on the S11 step. The actual spread physical distance is calculated using a graph search algorithm. Specifically, Dijkstra's algorithm can be used to calculate the spread distance in the edge set. Searching for risk source nodes To the target node The shortest connected path can be expressed by the following formula: In the formula, This is the set of all possible connected paths that connect two nodes in a topological directed graph. For a single physical edge in a path obtained based on BIM 3D coordinates The Euclidean length.

[0054] S42, Topological attenuation proportionality constant The spatial tortuosity of the spread path can be represented by the three-dimensional absolute coordinates of the node and the actual spatial path distance. The geometric dimensionality reduction ratio can be obtained through the following formula: ; In the formula, To start from the risk source node To the target node The topological attenuation ratio constant; Let be the straight-line Euclidean distance between the three-dimensional coordinates of the two nodes.

[0055] S43, Specific Risk Sources Hazard level coefficient The severity of a fire breaking out at a specific node depends on the total energy contained in the combustible materials within that space, i.e., the fire load density in fire protection engineering. This parameter is determined by the room functional semantics in the BIM model, and can be specifically extracted by reading the target building's BIM model database and then... The functional labels of the corresponding spaces (e.g., paper archives, open lobby, underground parking garage) are mapped to the intrinsic fire load density corresponding to the national mandatory fire protection standards, and then globally normalized. This can be represented by the following formula: ; In the formula, For nodes Equivalent fire load density after multi-source correction; The global standard reference fire load density is preset by the system. In this embodiment, it is anchored to the benchmark value of fire load density in a standard low-risk office scenario. For example, equivalent fire load density This can be reflected by the following formula: ; In the formula, For nodes Equivalent fire load density after multi-source correction; , These are the lower and upper limits of the national standard fire load density, respectively, matched through BIM semantic tags. If local semantics are missing in the BIM, the system automatically assigns the average baseline safety value for this type of building. This is a space utilization index, with a value range of [value range missing]. This parameter can be dynamically assigned by the visual AI of security cameras (such as recognizing the height of stacked goods) or by UWB personnel positioning density, when the space is completely empty. Take the lower limit value when fully loaded. Take the upper limit value; For the first External multi-source correction coefficients, for example, when a node is found When there is illegal storage of flammable materials, A penalty coefficient greater than 1.0 is applied if a node is configured with an excessive automatic sprinkler system. Choose an attenuation coefficient less than 1.0.

[0056] In this embodiment 1, the cloud server uses the constructed digital topology directed graph Based on this, a graph search algorithm is used to extract the actual physical distance of risk source nodes from target assessment points. Furthermore, the topological attenuation ratio constant, which characterizes spatial tortuosity, was calculated using the geometric dimensionality reduction ratio of the three-dimensional absolute coordinates to the physical distance. Simultaneously, the hazard level coefficients are derived from the semantic mapping of room functions based on the BIM model. Primary risk intensity at each node and environmental transmission coefficient The comprehensive risk potential after the synergistic impact of all non-zero risk sources within the entire building. Perform a global overlay calculation. Specifically, for example: assume the risk source node... place (High intensity), hazard factor Target assessment points distance path distance Let constants be used. In the following scenarios: Scenario A: Fire door closed: Calculations show... .but Comprehensive risk potential received by the point This indicates that the risk has been effectively isolated.

[0057] Scenario B, the door is open and there is a tailwind: Calculations show that... .but Comprehensive risk potential received by the point The calculation results indicate that even the target area The sensors themselves are normal, but due to the change in physical topology (door open, downwind), the risk of spread has increased dramatically.

[0058] The above scenario demonstrates that the method breaks through the bottleneck of traditional fire protection systems that can only rely on single-point isolated perception. From the perspective of the global physical field, it can show the non-linear attenuation law of fire spread, high-temperature toxic gas under the combined effects of fireproof door blockage, ventilation convection, spatial passage paths, and intrinsic fire load in the three-dimensional topology of complex buildings, and can judge the probability of fire occurrence in this area. At the same time, it dynamically quantifies the spatio-temporal distribution trend of risk cross-regional spread, providing a theoretical basis for subsequent dynamic escape path planning.

[0059] S5. Remote dynamic risk level division and multi-level linkage response, referring to Figure 2 shown below, specifically including the following steps: S51. The remote server divides the calculated global distribution matrix into multiple risk level intervals (such as safe, highly dangerous, about to spread), and generates a panoramic risk dynamic heat map. Specifically, the risk level division method is as follows: The system sets a comprehensive safety threshold and a primary safety threshold . If , it is determined as a safe area; if and the primary risk intensity , it is determined as a highly dangerous area; if and the primary risk intensity , it is determined as an about-to-spread area, and a multi-level linkage response of the linkage prevention mechanism and the reconstruction of the evacuation path for risk avoidance is executed to contain the spread of the fire.

[0060] S52. Linkage prevention mechanism: If the status of a certain node is determined as an about-to-spread area, the system sends a control instruction to the Internet of Things to remotely close the fireproof door between and (that is, force ), and at the same time reverse the HVAC air supply direction to reduce , and contain the spread of the fire.

[0061] S53. Dynamic evacuation path reconstruction for risk avoidance: While sounding the global audible and visual alarm, the system, based on the dynamic distribution of the global , marks the topological edges and nodes with a comprehensive risk potential exceeding the safety threshold as dynamic no-go areas, cuts off dangerous escape paths before the smoke arrives, and realizes dynamic avoidance. At the same time, it联动controls the on-site intelligent evacuation indication signs to dynamically change the direction of the escape arrows, guiding the crowd to avoid the spread channels with a high transmission coefficient , and avoids colliding head-on with the incoming high-temperature toxic gas on the evacuation passage, realizing dynamic evacuation.

[0062] Exemplarily: In this Example 1, the primary risk intensity This reflects the probability of a fire occurring at a single node. Since sensors in different areas exhibit varying background noise levels, using a uniform, fixed value does not reflect objective realities. Therefore, we employ Laida's rule from statistics (…). The criteria (based on the node's own long-term historical security baseline data) dynamically and adaptively generate thresholds, and the calculation formula is as follows: ; In the formula, For nodes exist The dynamic primary safety threshold at any given moment; For nodes In the past, the background sampling window was very long. (e.g., the historical arithmetic average of primary risk intensity over the past 24 hours); For nodes In the same time window The standard deviation of the primary risk intensity represents the degree of fluctuation in environmental noise in the area; To calculate the confidence coefficient, according to Laida's rule, we directly use an absolute objective constant here. When the deviation from the mean exceeds 3 standard deviations, under the assumption of normal distribution, the probability that it belongs to normal environmental fluctuations is less than [a certain value]. This can be objectively determined as an abnormal critical point.

[0063] In this embodiment 1, a node Whether a node is in a state of imminent spread or high danger should be considered the limit of its spatial physical topology defense. Specifically, this should be considered a comprehensive security threshold. Defined as: when the most dangerous adjacent node with which it has a direct physical connection just reaches the primary fire outbreak critical point (i.e., reaches the critical point of fire outbreak). When the physical barrier between the two (such as a fire door) is in the most unfavorable fully open state (windless environment), the target node The theoretical limit of the risk field that can be withstood can be expressed by the following formula: ; In the formula, For nodes exist The overall safety threshold at any given moment; To analyze the semantic associations based on the parsed geometric data, and the nodes The set of adjacent nodes that have a direct physical connection interface (such as a door, window, or corridor opening); Adjacent nodes Intrinsic fire load density normalization coefficient based on the semantic mapping of room functions in Building Information Modeling (BIM); Adjacent nodes The current primary security threshold; The shortest physical distance between the target node and its neighboring nodes; This is the topology attenuation ratio constant obtained using the absolute coordinates and physical distance of the nodes; The fundamental propagation scale constant is calculated based on the extracted fundamental flow cross-sectional area. In this embodiment 1, the remote server calculates the overall risk potential of the entire domain. The distribution matrix is ​​divided into multiple risk level intervals, generating a panoramic dynamic risk heat map; in this process, the system abandons the traditional unified fixed alarm threshold, and instead uses... The criteria adaptively generate primary safety thresholds by combining long-term historical data. And based on the spatial physical topology defense limit, the comprehensive security threshold of the target node is calculated. By comparing the current parameters of nodes with the aforementioned dynamic dual thresholds, the system determines in real time whether each area is in a safe, highly dangerous, or imminent spread state. Once identified as an imminent spread zone, the system immediately issues an IoT command to remotely close fire doors and reverse the HVAC airflow direction to contain the fire. Simultaneously, the system marks high-risk topology edges as restricted areas, and in conjunction with this, changes the direction of arrows on the on-site intelligent evacuation signs, reconstructing escape routes. Specifically, for example: Suppose a corridor node in a multi-story commercial complex. Adjacent catering kitchen nodes A fire has occurred. The system is based on corridor nodes. The dynamic primary safety threshold is calculated based on environmental fluctuations over the past 24 hours. Simultaneously, based on the fire load density of the catering kitchen and the physical connection characteristics between the two, the comprehensive safety threshold that the corridor can withstand is calculated. At the current moment, although the relative deviation value measured by the smoke and temperature sensors in the corridor is relatively low, the actual real-time primary risk intensity is... (less than) However, due to the fierce fire in the kitchen and the fact that the fire door was open, the system calculated the overall risk potential of the corridor. (greater than) Based on this, the system determines the corridor nodes. Located in an area where the fire is about to spread; the system immediately triggers the linkage prevention mechanism, remotely forcibly closing the fire door between the corridor and the kitchen, and controlling the corridor HVAC system to switch to positive pressure ventilation mode to form an airflow barrier; at the same time, the system marks the corridor as a dynamic restricted area, sends instructions to the smart evacuation signs in the floor, dynamically reverses the escape arrows that originally pointed to the corridor, and guides people to evacuate through the backup safety passage away from the fire source.

[0064] The above examples demonstrate that by introducing a dynamic adaptive dual-threshold determination based on historical environmental noise and topology, and combining it with multi-level linkage response, this method not only physically curbs the spread of fire in the early stages of a fire, but also prevents evacuees from accidentally entering high-risk transmission channels such as high-temperature toxic gas accumulation areas through dynamic evacuation path reconstruction, thereby improving the efficiency of emergency evacuation and the level of personnel safety protection in complex building environments.

[0065] Example 2 As a second embodiment of the present invention, such as Figure 3 As shown in Example 1, this example also discloses a remote dynamic assessment system for fire risk based on multi-source risk perception, which specifically includes: a data acquisition and preprocessing module, a topology network and propagation calculation module, a primary risk assessment module, a risk field strength calculation module, and a risk classification and linkage response module.

[0066] The data acquisition and preprocessing module is used to perform real-time synchronous acquisition of multi-source heterogeneous fire protection data, obtain static topology data and continuous physical quantities, and simultaneously perform smoothing and noise reduction and dynamic baseline determination on the acquired continuous physical quantities to obtain a relative deviation value reflecting the degree of danger. ; The topology network and propagation calculation module is used to construct a dynamic directed graph topology network of building space based on static topology data, and to superimpose airflow velocity and physical barrier factors into the connected paths to dynamically calculate the environmental propagation coefficient, which reflects the equivalent spatial penetration scale of risk spread. ; The primary risk assessment module is used based on the relative deviation value. The system adaptively assigns weight coefficients to various data modalities, individually assesses the fire occurrence probability of each node, and calculates the primary risk intensity of local risk sources. ; Among them, the risk field strength calculation module is used to introduce the environmental propagation coefficient. Compared with primary risk intensity The global dynamic risk field strength based on the coupling of topological tension and physical field is calculated to obtain the comprehensive risk potential of the target assessment point in space after being affected by the synergistic influence of all risk sources. ; Among them, the risk classification and linkage response module is used to integrate comprehensive risk potential. and primary risk intensity Each with dynamically generated comprehensive security threshold and primary safety threshold The risk levels of assessment points are dynamically classified remotely through comparison, and multi-level linkage responses are executed based on the risk level determination results.

[0067] In the specific implementation of Embodiment 2 above, firstly, the system integrates the static spatial structural characteristics of the building with the real-time evolving dynamic fire perception data through a data acquisition and preprocessing module. This method eliminates the differences in sampling frequency, transmission delay, and dimensions of heterogeneous data from various sensors, providing a highly time-consistent data source for the subsequent construction of a dynamic directed graph. Secondly, by deeply coupling static connectivity paths and dynamic environmental perception variables through a topology network and a propagation calculation module, the environmental propagation coefficient reflecting the equivalent spatial penetration scale of risk spread is dynamically calculated. This method can realistically and quantitatively reflect the dynamic impact of real-time physical conditions, such as intensified airflow within buildings or blocked fire doors, on the cross-regional evolution of fire risk during fire early warning. Next, the primary risk intensity of local risk sources is calculated by adaptively assigning prior weights to various sensors through the primary risk assessment module. This method objectively and accurately reflects the fire occurrence probability of a single node without considering spatial propagation. Then, the comprehensive risk potential of the target assessment point under the combined influence of all risk sources is calculated through a global overlay calculation module using the risk field strength calculation module. This method dynamically quantifies the spatiotemporal distribution and nonlinear decay law of risk spread across regions. Finally, the risk classification and linkage response module executes a multi-level linkage response based on dynamic adaptive dual threshold judgment, including closing fire doors, reversing air supply, and reconstructing escape routes. This method not only physically curbs the fire in the initial stage but also prevents escaping people from accidentally entering areas where high-temperature toxic gases accumulate, improving the efficiency of emergency evacuation and the level of personnel safety protection in complex building environments.

[0068] Example 3 This is the third embodiment of the present invention, addressing the situation where some older or existing buildings lack Building Information Models (BIM) or 3D CAD drawings. Based on Embodiment 1, this embodiment supplements Embodiment 1 with a method for automatic identification and simplified deployment of building structures using laser scanning or visual mapping. This embodiment replaces the static topology data acquisition method in Embodiment 1, while the remaining dynamic risk assessment logic remains consistent with Embodiment 1.

[0069] In the specific implementation of Example 3: By employing 3D laser scanning and simultaneous localization mapping (SLAM) technology, the 3D geometry of the target building can be reconstructed without BIM data, and the topological nodes, connected edges, and physical parameters required in Example 1 can be extracted. The specific steps for obtaining static topological basic data are as follows: S11' Reconstruction of Static Topological Base Data Based on Scanning and Mapping: The S111 uses a portable scanning device equipped with a LiDAR and a visual depth camera, and employs SLAM (Simultaneous Localization and Mapping) algorithm to perform a full-coverage traversal scan inside the target building. The system records the device's trajectory in real time and outputs high-precision 3D point cloud data with a globally unified coordinate system.

[0070] S112. Upload the acquired 3D point cloud data to the cloud or a local computing workstation. Using classic 3D point cloud segmentation algorithms such as region growing or voxelization, remove point clouds of clutter such as interior furniture, and segment the building's internal physical space into multiple independent bounding volumes (such as rooms, corridors, and lobbies). For each segmented independent bounding volume, extract its core representative point as a topological node through geometric centroid calculation. And record its corresponding three-dimensional absolute coordinates. All extracted nodes are combined to form a static set of building topology nodes. .

[0071] S113' Employ a point cloud semantic segmentation network (such as PointNet) to extract features from the common interfaces between adjacent independent bounding volumes, automatically identifying physical connectivity features such as wall openings, doorways, or connecting corridors. If two adjacent nodes are detected... and If a physical opening exists between the two, it is determined that there is a channel for air and heat exchange, and a connection is established in the graphical model. and undirected edge This leads to the formation of basic edge sets. .

[0072] S114`, For edge sets Each connected edge in The minimum bounding box algorithm is used to fit the 3D boundary of the point cloud doorway opening, and the connection nodes are automatically calculated. and The actual physical net cross-sectional area of ​​the passage between them .

[0073] Regarding the normalized coefficient of intrinsic fire load density that needs to be obtained from BIM semantic tags in Example 1 and static safety baseline threshold In this simplified deployment strategy, the system generates an electronic sketch containing the spatial planar outline. On-site construction or maintenance personnel then use mobile devices to map each node according to its actual function (such as office area, power distribution room, catering kitchen). Standard function tags are distributed in batches, and the system backend automatically matches and assigns corresponding tags. and .

[0074] S115', The above-mentioned node set based on point cloud reconstruction. Edge set 3D coordinates and the calculated actual physical net cross-sectional area These parameters are encapsulated into a standard graph database format to generate a static topology map of the target building, and then integrated into S12 and subsequent steps in Example 1. (Subsequent environmental propagation coefficient) and comprehensive risk potential The dynamic calculation process is consistent with that in Example 1.

[0075] In this embodiment, mature LiDAR and SLAM technologies are used to fill the gaps in the lack of digital building information models for some buildings. This solution directly extracts the actual geometric dimensions of the objective environment to calculate the actual physical net cross-sectional area. By using three-dimensional absolute coordinates and combining them with a minimalist label mapping system assisted by human intervention, the system enables the rapid construction of an environmental base, reducing the deployment threshold and implementation cost of IoT-based fire protection upgrades for existing buildings.

[0076] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0077] The preferred embodiments of the invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A remote dynamic assessment method for fire risk based on multi-source risk perception, characterized in that, include: The static topological data and continuous physical quantities of the target building are collected. At the same time, the collected continuous physical quantities are smoothed and denoised and dynamic baseline is determined to obtain the relative deviation value that reflects the degree of danger. Based on the static topological data, a dynamic directed graph topological network of building space is constructed, and air flow velocity and physical barrier attenuation factor are superimposed in the connected path to dynamically calculate the environmental propagation coefficient that reflects the equivalent spatial penetration scale of risk spread. Based on the relative deviation value, the weight coefficients of various data modes are adaptively assigned, the fire occurrence probability of each node is evaluated individually, and the primary risk intensity of the local risk source is calculated. Based on the environmental propagation coefficient and the primary risk intensity, the comprehensive risk potential of the target assessment point in space after being affected by the synergistic influence of all risk sources is calculated. The comprehensive risk potential and the primary risk intensity are compared with the comprehensive safety threshold and the primary safety threshold, respectively, and the risk level of the assessment point is remotely and dynamically classified. Multi-level linkage response is executed according to the risk level determination result.

2. The method for remote dynamic assessment of fire risk based on multi-source risk perception according to claim 1, characterized in that, The formula for calculating the relative deviation value is: ； In the formula, For nodes First Physical quantities in The relative deviation value at time, For the feature normalized span value, For nodes Data mode At the present moment The effective physical quantity value, For nodes Data mode exist The dynamic safety baseline threshold at any given time.

3. The method for remote dynamic assessment of fire risk based on multi-source risk perception according to claim 2, characterized in that, The formula for calculating the environmental propagation coefficient is as follows: ； In the formula, For nodes To the node exist Environmental propagation coefficient at any given time For nodes arrive The fundamental propagation scale constant, For those caused by HVAC systems or natural wind point to airflow speed in the direction, The characteristic time constant; The physical barrier attenuation factor is taken as the value when the fire door is fully open. When fully closed and well-sealed, take a value close to minimum value .

4. The method for remote dynamic assessment of fire risk based on multi-source risk perception according to claim 3, characterized in that, The formula for calculating the fundamental propagation scale constant is as follows: ； In the formula, For nodes arrive The fundamental propagation scale constant, The natural diffusion reference distance under standard reference conditions. For connecting nodes and The actual physical net cross-sectional area of ​​the passage between them The area of ​​the doorway is a preset reference standard. This is the scaling index for spatial morphology.

5. The method for remote dynamic assessment of fire risk based on multi-source risk perception according to claim 3, characterized in that, The above caused by HVAC systems or natural wind point to airflow speed in the direction The calculation formula is: In the formula, The current three-dimensional wind field velocity vector. , These are the three-dimensional absolute coordinate vectors of the target node and the risk source node, respectively.

6. The method for remote dynamic assessment of fire risk based on multi-source risk perception according to claim 3, characterized in that, The formula for calculating the primary risk intensity is as follows: In the formula, For nodes exist The initial risk intensity at any given moment, For sensor data mode sets, These are the prior weight coefficients for various types of data.

7. The method for remote dynamic assessment of fire risk based on multi-source risk perception according to claim 6, characterized in that, The formula for calculating the comprehensive risk potential is as follows: In the formula, As assessment point exist The comprehensive risk potential at any moment The total number of risk sources with a non-zero primary risk intensity. For specific risk sources The hazard level coefficient, Spatial path distance, This is the topological attenuation ratio constant.

8. The method for remote dynamic assessment of fire risk based on multi-source risk perception according to claim 7, characterized in that, The risk levels of the remotely dynamically assigned assessment points include: setting a comprehensive safety threshold. and primary security threshold If comprehensive risk potential If so, it is considered a safe zone; if And the initial risk intensity If it is, then it is determined to be a high-risk area; if And the initial risk intensity If so, it is determined to be an area about to spread.

9. The method for remote dynamic assessment of fire risk based on multi-source risk perception according to claim 8, characterized in that, The multi-level coordinated response includes a coordinated prevention mechanism and the reconstruction of evacuation routes; the coordinated prevention mechanism includes closing fire doors and reversing the direction of HVAC air supply; the reconstruction of evacuation routes includes marking comprehensive risk potential. Exceeding the comprehensive safety threshold The topological edges and nodes are dynamic restricted areas, which in turn change the direction of the intelligent evacuation signs.

10. A remote dynamic assessment system for fire risk based on multi-source risk perception, employing the remote dynamic assessment method for fire risk based on multi-source risk perception as described in any one of claims 1 to 9, characterized in that, include: The data acquisition and preprocessing module is used to collect static topological data and continuous physical quantities of the target building to obtain relative deviation values ​​that reflect the degree of danger. The Topology and Environmental Dynamics module is used to construct a dynamic directed graph topology network of building space and dynamically calculate the environmental propagation coefficient that reflects the equivalent spatial penetration scale of risk spread. The primary risk assessment module is used to adaptively allocate weight coefficients for various data modes based on relative deviation values ​​to assess the primary risk intensity of local risk sources. The global risk field strength calculation module is used to incorporate the environmental propagation coefficient and the primary risk intensity to calculate the comprehensive risk potential of the target assessment point in space after being affected by the combined influence of all risk sources. The risk classification and linkage response module is used to compare the comprehensive risk potential and primary risk intensity with the comprehensive safety threshold and primary safety threshold, respectively, to classify the risk level of the assessment point, and to execute multi-level linkage response based on the risk level determination results.

Images