Fire safety management supervision platform system

Abstract

The invention relates to the technical field of fire safety management, and discloses a fire safety management supervision platform system. The system comprises a building information acquisition module for acquiring building structure data and environment monitoring data of a target building in real time; the risk area division module is used for dividing a plurality of independent fire risk areas according to the building structure data and distributing area identifiers; the historical data indexing module is used for indexing a historical fire accident data set and a fire-fighting facility maintenance record set of each region according to the region identifier; the dynamic risk analysis module fuses the environment monitoring data and the historical fire accident data set to generate a real-time fire risk coefficient sequence; the fire-fighting resource scheduling module is used for generating a fire-fighting resource allocation scheme based on the real-time fire risk coefficient sequence and the facility maintenance record set; and the emergency response decision module generates a multi-stage fire emergency response strategy according to the resource allocation scheme. The system improves the timeliness and pertinence of fire safety management.

Description

Technical Field

[0001] This invention relates to the field of fire safety management technology, specifically a fire safety management and supervision platform system. Background Technology

[0002] Fire safety is a crucial component of public safety, directly impacting the lives and property of the people and social stability. With accelerating urbanization, buildings are expanding in scale, becoming increasingly complex in structure, and diversifying in function. Traditional fire safety management models are no longer sufficient to meet the fire safety needs of modern buildings, leading to a continuous increase in fire prevention and control pressure. In existing technologies, building fire safety management largely relies on manual inspections and periodic checks, which suffers from problems such as untimely data collection, limited coverage, and delayed risk assessment. For example, in complex building scenarios such as large commercial complexes, high-rise buildings, and industrial parks, manual inspections struggle to monitor real-time changes in the building's internal environment, such as abnormally high temperatures, excessive smoke concentrations, and aging electrical wiring—potential fire hazards that often lead to delayed fire detection and missed opportunities for optimal response. Regarding risk zone delineation, existing technologies typically use simple methods based on building function or area, failing to fully integrate building structural data for refined delineation. This results in vague risk zone definitions and an inability to accurately identify high-risk areas. This crude approach leads to unreasonable allocation of fire safety management resources, insufficient control over key areas, and an increased probability of fire accidents. Regarding the utilization of historical data, existing fire safety management systems lack an effective mechanism for integrating and indexing historical fire accident data and fire protection facility maintenance records. Key information such as fire accident characteristics and facility failure patterns in different areas is stored in a scattered manner, making it difficult to achieve precise correlation queries through area identification. This results in the inability to uncover risk evolution patterns from historical data and to formulate targeted prevention strategies. For example, an area may have experienced multiple electrical fires, but because historical data was not effectively indexed and analyzed, the frequency of electrical facility inspections and maintenance standards in that area were not adjusted in a timely manner, leading to the recurrence of similar accidents. The dynamic risk analysis process has significant shortcomings. Existing systems mostly use static assessment models, relying solely on inherent building attributes for risk rating without integrating real-time environmental monitoring data for dynamic updates. Fire risk evolves dynamically with environmental changes, and static assessment results cannot reflect real-time risk status, easily leading to inaccurate risk warnings, false alarms, and missed alarms, thus affecting the efficiency of fire safety management. In terms of fire resource dispatch, existing technologies largely rely on experience-based decision-making, lacking a scientific dispatch mechanism based on real-time risk data and facility status. The lack of linkage between fire facility maintenance records and risk levels leads to situations where fire facilities are poorly maintained and emergency supplies are insufficient in areas with high fire hazards, while low-risk areas are over-allocated resources, resulting in resource waste. During fire incidents, this irrational resource allocation often leads to slow emergency response and inefficient deployment of rescue equipment, delaying firefighting efforts. The emergency response decision-making process also has shortcomings. The existing system's emergency response strategies are relatively simplistic and fail to develop tiered response strategies based on factors such as fire risk level, building structural characteristics, and resource allocation. When a fire occurs, it is difficult to quickly match the optimal emergency response plan, resulting in weakly targeted emergency measures, ineffective control of the fire's spread, and increased losses caused by the fire. Summary of the Invention

[0003] The purpose of this invention is to provide a fire safety management and supervision platform system to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides a fire safety management and supervision platform system, the system comprising: The building information acquisition module is used to acquire the building structure data and environmental monitoring data of the target building in real time. The risk zone division module is used to divide the building structure data into several independent fire risk zones and assign a zone identifier to each fire risk zone. The historical data indexing module is used to index the fire accident data set and fire facility maintenance record set of each fire risk area within a historical time period based on the area identifier; The dynamic risk analysis module is used to integrate the environmental monitoring data and the fire accident data set to generate a real-time fire risk coefficient sequence. The fire resource scheduling module is used to generate a fire resource allocation scheme based on the real-time fire risk coefficient sequence and the fire facility maintenance record set; The emergency response decision module is used to generate multi-level fire emergency response strategies based on the fire resource allocation scheme.

[0005] Preferably, the system further includes: The fire evolution prediction module is used to construct a fire spread time series model based on fire spread path data in the historical fire accident dataset. The strategy correction module is used to input the current environmental monitoring data into the fire spread time series model, output the fire spread direction prediction result, and correct the multi-level fire emergency response strategy based on the result. The verification and execution module is used to execute the modified multi-level fire emergency response strategy under simulated fire scenarios and generate a strategy execution performance report. The closed-loop update module is used to update the fire spread time series model based on the strategy execution performance report.

[0006] 3. The fire safety management and supervision platform system according to claim 1, characterized in that, the step of integrating the environmental monitoring data and the fire accident data set to generate a real-time fire risk coefficient sequence includes: Extract temperature distribution data, smoke concentration data, and airflow velocity data from the environmental monitoring data; Extract a set of environmental feature vectors from the historical fire incident data set. Calculate the deviation distances between the temperature distribution data, smoke concentration data, and airflow velocity data and the environmental feature vector set, respectively; A real-time fire risk coefficient is generated based on the weighted sum of each deviation distance; Multiple real-time fire risk coefficients are arranged in order of timestamp to form a real-time fire risk coefficient sequence.

[0007] Preferably, calculating the deviation distances between the temperature distribution data, smoke concentration data, and airflow velocity data and the environmental feature vector set includes: The set of environmental feature vectors is clustered to generate multiple environmental feature cluster centers; Calculate the temperature dimension Euclidean distance between the temperature distribution data and the temperature dimension of each environmental feature cluster center; Calculate the Manhattan distance (concentration dimension) between the smoke concentration data and the cluster centers of each environmental feature; Calculate the inverse cosine similarity of the airflow velocity data with the velocity dimension of each environmental feature cluster center; The deviation distance is generated by weighting and fusing the Euclidean distance in the temperature dimension, the Manhattan distance in the concentration dimension, and the inverse cosine similarity in the velocity dimension corresponding to the cluster centers of each environmental feature.

[0008] Preferably, the step of generating a fire resource allocation scheme based on the real-time fire risk coefficient sequence and the fire protection facility maintenance record set includes: Extract fire equipment availability status data from the fire protection facility maintenance record set; The real-time fire risk coefficient sequence is trend-fitted to generate the slope of fire risk change; Risk level intervals are defined based on the slope of the fire risk change. Match the risk level range of each fire risk area with the corresponding fire equipment availability status data; The number of fire extinguishing equipment, the deployment location of firefighters, and the priority of emergency exit activation are allocated according to the matching results.

[0009] Preferably, the construction of the fire spread time series model includes: Analyze the fire source location data, building material ignition point data, and building structure topology data in the historical fire accident data set; A three-dimensional spatial mesh is constructed based on the building structure topology data; Map the fire source location data to the initial grid nodes of the three-dimensional spatial grid; The time interval for the spread of fire in adjacent grid nodes is calculated based on the ignition point data of the building materials. Connect all the spreading grid nodes in chronological order to form a time-series chain of fire spread paths.

[0010] Preferably, the step of inputting current environmental monitoring data into the fire spread time series model and outputting the fire spread direction prediction result includes: Acquire real-time temperature gradient data and smoke diffusion direction data at the current fire location; Match the historical fire node in the fire spread path time sequence chain that is closest to the real-time temperature gradient data; Extract the subsequent spread path branches of the historical fire nodes; Calculate the angle between the smoke diffusion direction data and the direction of each subsequent diffusion path branch; The branch of the subsequent diffusion path with the smallest directional angle is selected as the predicted result of the fire diffusion direction.

[0011] Preferably, the step of modifying the multi-level fire emergency response strategy based on the result includes: Extract the high-risk spread area identifiers from the fire spread direction prediction results; Obtain population distribution density data within the high-risk spread area; Adjust the evacuation route planning scheme in the aforementioned multi-level fire emergency response strategy; The weight of fire-fighting resource allocation in the high-risk spread areas will be dynamically increased.

[0012] Preferably, the dynamic increase of the fire resource allocation weight in the high-risk spread area includes: Obtain the building load-bearing structure data of the high-risk sprawl area; Calculate the ratio of the predicted fire spread rate to the fire resistance coefficient of the building's load-bearing structure; The weight increment for fire resource allocation is set according to the ratio; The fire resource allocation weight increment is superimposed on the fire resource allocation scheme.

[0013] Preferably, the step of executing the modified multi-level fire emergency response strategy under the simulated fire scenario and generating a strategy execution performance report includes: Construct a digital twin building model that includes predictions of the direction of fire spread; Virtual fire parameters are injected into the digital twin building model; Execute the revised multi-level fire emergency response strategy and record the virtual fire resource dispatch path; Compare the overlap between the virtual fire resource scheduling path and the preset optimal path; A strategy execution performance report is generated based on the overlap deviation value.

[0014] Preferably, updating the fire spread time series model based on the strategy execution performance report includes: Extract the fire control delay time data from the strategy execution performance report; Obtain the time difference between the actual arrival time of fire-fighting resources and the virtual dispatch time; Adjust the spread time interval parameter in the fire spread timing model according to the time difference; Recalculate the time interval for the fire to spread to adjacent grid nodes.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This fire safety management and supervision platform system achieves a comprehensive upgrade in fire safety management through the coordinated operation of its various modules. The building information acquisition module acquires real-time building structure data and environmental monitoring data of the target building, breaking through the limitations of traditional data acquisition that is lagging and static. This allows managers to promptly grasp the dynamic changes in the building structure and abnormal factors in the environment, providing fresh and accurate basic data for subsequent risk analysis. The risk zone delineation module divides the building into several independent fire risk zones based on building structure data and assigns zone identifiers, changing the previous method of roughly dividing risk zones. This precise delineation based on building structure characteristics makes the management of each risk zone more targeted, enabling the development of differentiated regulatory strategies according to the structural characteristics of different zones, and avoiding the average distribution and waste of management resources. The historical data indexing module indexes the historical fire accident data sets and fire protection facility maintenance record sets of each risk area based on the regional identifier, linking the scattered historical data with specific risk areas. By sorting through the historical data, the patterns of fire occurrence, common hazards, and operational status of fire protection facilities in different areas can be clearly presented, providing a longitudinal time dimension reference for understanding the regional risk characteristics and helping to discover long-term potential risks. The dynamic risk analysis module integrates environmental monitoring data with historical fire accident data to generate a real-time fire risk coefficient sequence, enabling dynamic risk assessment. It no longer relies on fixed-period assessments but continuously updates risk levels based on real-time changes in environmental factors and historical risk patterns. This allows managers to stay informed about the risk dynamics in each area, promptly identify areas with rising risks, and take intervention measures. The fire resource dispatch module generates resource allocation plans based on real-time fire risk coefficient sequences and fire facility maintenance records, moving away from experience-based dispatching. By combining real-time risk levels and facility maintenance status, it can prioritize the allocation of fire resources to areas with higher risks and better facility conditions, ensuring that resources are used where they are most needed and improving the rationality of resource utilization. The emergency response decision-making module generates multi-level fire emergency response strategies based on resource allocation plans, making emergency response more flexible and targeted. Different levels and content of response strategies are formulated for different risk levels and resource distributions, enabling the rapid activation of appropriate emergency measures when a fire occurs, improving the efficiency and effectiveness of emergency response, and better addressing various fire hazards. Attached Figure Description

[0016] Figure 1 This is a sequence diagram of the fire safety management and supervision platform system described in this invention; Figure 2 A flowchart for fire safety management that includes fire prediction and strategy optimization; Figure 3 A flowchart for calculating the deviation distance between environmental data and historical characteristics; Figure 4 A flowchart for constructing a time-series model of fire spread. Detailed Implementation

[0017] 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.

[0018] Please see Figure 1 This invention provides a fire safety management and supervision platform system, the system comprising: The building information acquisition module collects environmental monitoring data in real time through temperature sensors, smoke detectors, and anemometers deployed inside the building, while simultaneously importing building structural data from the BIM system. The risk zone delineation module uses graph theory algorithms to divide the building floor plan into several connected subgraphs as fire risk zones, each assigned a unique QR code identifier. The historical data indexing module uses blockchain technology to establish an index linking zone identifiers to the fire safety database, ensuring the traceability of accident records and maintenance logs. The dynamic risk analysis module uses machine learning algorithms to match and analyze real-time environmental data with historical accident characteristics. The fire resource scheduling module uses a multi-objective optimization algorithm to calculate the optimal resource allocation scheme. The emergency response decision-making module automatically generates hierarchical strategies including evacuation routes, equipment scheduling, and communication command based on a contingency plan knowledge base.

[0019] Example 1: See Figure 2 The fire evolution prediction module uses a long short-term memory neural network to process multi-dimensional time-series information from historical fire accident datasets. The module's data processing workflow begins with the standardization and organization of historical fire thermal imaging data. Historical thermal imaging data is sliced ​​according to time series, with each slice spaced 5 seconds apart, covering the entire process from the initial fire to the end of extinguishing. Data preprocessing includes filtering to remove sensor noise and normalizing temperature values; all temperature readings are converted to a relative value range of 0 to 1. A dynamic grid method is used for building space division, automatically adjusting the grid cell size based on the complexity of the building structure. 75cm×75cm cells are used in open areas, while 25cm×25cm cells are used in densely populated equipment areas. Each grid cell is bound to attribute labels including the thermal conductivity of the building material and the density of combustible materials. The neural network architecture uses a three-layer gated recurrent unit. The number of hidden layer nodes is adaptively set according to the total number of grid cells. The input layer receives the grid temperature distribution status for 20 consecutive time segments, and the output layer generates the ignition probability of each cell in the subsequent 10 time segments. The network training process employs an adaptive moment estimation algorithm optimizer, and the loss function combines cross-entropy error and time continuity constraints.

[0020] The strategy correction module deploys a prediction process based on real-time environmental monitoring data. This module continuously receives the latest temperature distribution map from the building's infrared sensor network, with the data sampling frequency synchronized with the historical data slice frequency. Upon acquiring real-time data, it immediately performs coordinate registration, mapping the sensor coordinates to a preset grid coordinate system. The mapping process considers building structural deformation parameters and uses affine transformation to correct for viewing deviations caused by equipment installation locations. The transformed latest temperature state vector, along with the historical states from the previous 19 time segments, forms a complete input sequence, which is then input into a pre-trained fire evolution prediction model. The model outputs a temperature change gradient field and a fire probability distribution map for the prediction period. The prediction results are encoded using a three-channel matrix: the red channel represents the fire probability, the green channel represents the temperature change rate, and the blue channel represents the thermal radiation influence range. Regional clustering analysis is performed on this output matrix to identify high-risk propagation zones where the fire probability exceeds a threshold. The strategy correction operation is implemented based on the spatial distribution characteristics of these high-risk zones. On floors where the coverage density of high-risk propagation zones exceeds a preset threshold, the control level of evacuation routes is automatically increased, and the optimal deployment location of isolation barriers is marked in the building's 3D model.

[0021] The verification and execution module establishes a fully digital testing environment to evaluate the effectiveness of strategy adjustments. This module integrates Building Information Modeling (BIM) and Fire Dynamics Modeling (FDM) to construct a physically realistic digital twin. In the virtual environment, initial fire source parameters, including heat release rate growth curves and combustion product generation patterns, are configured based on real fire case files. When executing the revised multi-level emergency strategy, the module simulates the coordinated response of multiple emergency units: the fire equipment dispatch system generates equipment movement paths based on resource allocation plans; the personnel evacuation system calculates crowd flow trajectories based on the adjusted evacuation plan; and the environmental control system executes preset smoke control and exhaust commands. All simulation processes employ a discrete event progression mechanism, updating the status data of each system every 0.5 seconds. The performance evaluation phase sets three types of key recording points: key fire development nodes record the timestamp when the heat release rate reaches the MW level; emergency response nodes record the time taken for the first batch of fire equipment to arrive; and key control points record the time when the integrity of major fire compartments is compromised. Quantifiable performance indicators are generated by comparing the temporal relationships of these three types of nodes.

[0022] The closed-loop update module implements an iterative optimization mechanism for model parameters. This module collects the actual response data sequences recorded during the validation execution phase and aligns them temporally with the output sequences of the prediction model. The alignment process uses a dynamic time warping algorithm to eliminate time axis scaling differences and establishes a correspondence between the actual and predicted values ​​at each prediction time point. The parameter update algorithm employs gradient descent to calculate the error gradient of each neuron in the network output layer and updates the hidden layer weight matrix through backpropagation. To address the temporal dependency characteristics in the fire propagation model, a special forget gate tuning factor modification rule is set for the gating unit. When the fire spread delay error in a specific direction exceeds the tolerance threshold in three consecutive validation tests, a dedicated tuning procedure for the propagation path parameters in that area is triggered. Regression testing is performed immediately after all parameter update operations are completed to ensure that the model's predictive stability in the baseline fire scenario meets operational requirements. The updated model parameters are version-managed, retaining the five most recent valid versions for comparative analysis.

[0023] Example 2: See Figure 3 The dynamic risk analysis module establishes a multi-source heterogeneous data processing flow when performing environmental data fusion. This flow includes three stages: sensor data preprocessing, historical feature matching, and risk coefficient calculation. Raw data collected by the temperature sensor network first undergoes outlier detection and removal, using a sliding window outlier detection algorithm to identify and filter abrupt changes caused by equipment failure. Effective temperature readings are used to generate a three-dimensional temperature field distribution inside the building through spatial interpolation algorithms. The interpolation process considers the influence of the building structure on heat conduction, setting thermal resistance boundary conditions at wall locations. The smoke concentration data acquisition system includes both photoelectric and ionization sensors. Sensor type calibration is performed before data fusion, and the response curves of the two sensors are fitted using the least squares method. Airflow velocity monitoring uses an ultrasonic anemometer array; the measured data is converted into standard ventilation parameters after turbulence correction.

[0024] The construction of the historical environmental feature vector set employs time-series segmentation technology, dividing environmental data from the 30 minutes preceding fire accidents recorded over the past five years into feature segments at 5-minute intervals. Each feature segment contains three components: a temperature distribution matrix, a smoke concentration vector, and an airflow velocity tensor. Principal component analysis (PCA) is used to process these three components separately. After expanding the temperature matrix into a one-dimensional vector, the eigenvalues ​​of the covariance matrix are calculated, retaining the top k principal components with a cumulative contribution rate exceeding 85%. Due to the nonlinear characteristics of the smoke concentration data, kernel PCA is used for feature extraction. The airflow velocity tensor is first decomposed into magnitude and direction components, which are then subjected to PCA separately and merged into a comprehensive feature vector.

[0025] The deviation distance calculation process establishes a multi-dimensional measurement system. For the temperature dimension, an improved Euclidean distance calculation method is used, incorporating a heat conduction attenuation factor when calculating temperature differences between grid cells. A dedicated nonlinear function is designed for smoke concentration distance measurement, considering the varying contributions of different concentration ranges to fire early warning. Airflow velocity similarity analysis introduces vortex feature descriptors, assessing the similarity of airflow patterns through flow field morphology matching. An adaptive weight adjustment mechanism is designed in the weighted fusion stage, dynamically updating weight coefficients based on the real-time change rate of environmental parameters. When the temperature change rate exceeds a threshold, the weight of the temperature dimension is automatically increased.

[0026] The real-time fire risk factor is generated using the following calculation formula:

[0027] in: This indicates the fire risk coefficient at the current moment; Represents the temperature deviation distance; Indicates the distance of smoke concentration deviation; For airflow pattern differences; These are dynamic weighting coefficients for temperature, smoke, and airflow parameters. These coefficients are automatically adjusted according to changes in environmental conditions to meet [the following requirements]. The constraints.

[0028] Temperature deviation distance Calculate the Mahalanobis distance between the current temperature field distribution and the historical pre-fire temperature characteristics, considering the differences in thermal inertia between different building areas. Smoke concentration deviation distance. The logarithmic ratio method is used to measure the degree of difference between current concentrations and historical characteristics. (Airflow pattern difference) It is obtained by converting the current flow field vorticity distribution with the historical model.

[0029] The subsequent processing of the risk coefficient sequence includes two steps: temporal smoothing and spatial clustering. Temporal smoothing employs an exponentially weighted moving average method, with the smoothing window width adaptively adjusted according to the gradient of risk coefficient changes. Spatial clustering analysis divides the building plan into several risk zones, using a density clustering algorithm to identify high-risk clusters, and assigning a comprehensive risk level label to each zone. The final generated risk coefficient sequence is stored in a time-series database with dual indexes of timestamp and spatial location, supporting queries based on zone identifiers and time ranges.

[0030] The integrated processing of fire protection facility maintenance records includes equipment status analysis and maintenance timeliness assessment. Equipment status data extracts availability indicators from inspection records, including parameters such as fire extinguisher pressure, fire water tank level, and alarm sensitivity. Maintenance timeliness analysis calculates the number of days between the last maintenance time and the current time, and determines the effective maintenance period based on the equipment type. The data association phase establishes a mapping relationship between risk coefficients and equipment status, and specially marks faulty equipment in high-risk areas.

[0031] The risk level range is defined using a dynamic threshold method, with threshold boundaries dynamically calculated based on building usage and personnel density parameters. Different risk level classification standards are applied to office and warehouse areas, with a lower risk assessment threshold appropriately applied to densely populated areas. The classification results are visualized using color coding and updated in real-time on the building's electronic floor plan, supporting multi-level zooming for detailed viewing.

[0032] The generation process of the fire resource allocation plan employs a multi-objective optimization algorithm, simultaneously considering risk control requirements and resource scheduling costs. The optimization objective function comprises three components: maximizing risk coverage, minimizing response time, and minimizing resource consumption. The solution process utilizes an improved particle swarm optimization algorithm, adding constraint handling mechanisms to the standard algorithm to ensure the plan meets the basic requirements of fire protection codes. The final generated allocation plan details the types and quantities of fire extinguishing equipment required for each area, the locations and shift schedules of firefighters, and the control levels for emergency exits.

[0033] Example 3: See Figure 4 The construction of the fire spread time series model begins with the graph theory transformation of the building structure topology data. This module abstracts the building floor plan into a weighted directed graph structure, where independent room units of the building space are transformed into graph nodes, and porches, passageways, and ventilation ducts are transformed into connecting edges. Each node is associated with a set of attributes including geometric dimensions and spatial volume, and each edge is labeled with connection type and accessibility index. The ignition point data of building materials is obtained by querying the material database and transformed into the fire resistance time attribute value of the nodes, which is determined by the intersection of the material thermal decomposition temperature curve and the standard temperature rise curve. The fire source location data is mapped to the nearest neighbor grid nodes through a spatial coordinate transformation algorithm. The mapping process takes into account the actual measurement error of the building structure and uses the nearest neighbor interpolation method to optimize the positioning accuracy.

[0034] The propagation time interval calculation establishes a mathematical model for the heat conduction process between adjacent nodes. This process considers three core thermodynamic parameters: the thermal conductivity coefficient λ of the building material reflects the heat transfer capacity in a solid medium; the air convection coefficient η characterizes the influence of fluid motion on heat exchange; and the material's heat capacity ρc determines the heat required per unit temperature rise. The discretized solution of the unsteady-state heat conduction equation is implemented using the finite volume method, treating nodes as control volumes and connecting edges as heat exchange surfaces. The calculation formula is as follows:

[0035] The parameters are defined as follows: This represents the minimum time interval between the spread of fire from the current node to the adjacent node. The density of building materials; Indicates the specific heat capacity of building materials; It is the spatial volume of the current node; Set as the minimum temperature rise required for ignition; It is the heat transfer efficiency correction factor between node i and its neighboring node k; The thermal conductivity coefficient of the building component; This represents the heat conduction contact area from node i to node k; It is the initial temperature difference between nodes i and k; This represents the amount of heat exchanged due to air convection.

[0036] The construction of the fire spread path time sequence chain employs an improved path search strategy. Based on the spread time interval matrix, the system starts from the initial fire source node and uses a priority queue breadth-first search algorithm to traverse the entire building map structure. During the search process, two key data structures are maintained in real time: a time accumulation stack records the estimated time when each node reaches the ignition temperature; and a path backtracking pointer records the optimal propagation path direction. In multi-path selection scenarios with the same time cost, the algorithm adds a material flammability weight factor, prioritizing propagation paths containing flammable materials. The final generated path time sequence chain adopts a three-level storage structure: the first level records the fire development stages in seconds; the second level stores the set of ignition nodes included in each stage; and the third level labels the spread mode parameters between adjacent stages.

[0037] The fire spread direction prediction process receives real-time monitoring data streams. Temperature gradient data at the current fire point is collected via a thermocouple array deployed around the fire area; the raw data is converted into a three-dimensional temperature gradient vector field through spatial difference operations. Smoke spread direction monitoring employs computer vision technology, with a monitoring network composed of explosion-proof cameras analyzing the smoke trajectory in real time. The smoke direction vector extraction algorithm combines optical flow and feature point tracking techniques, outputting a predicted main spread direction value every 0.5 seconds. All the above dynamic data undergoes data consistency verification; when the angle between the temperature gradient and the smoke direction vector exceeds a reasonable range, a sensor calibration procedure is triggered.

[0038] Historical fire node matching is performed in a time-series database. The system retrieves historical path segments that match the current fire development stage in chronological order. The matching degree is calculated using two features: temperature gradient similarity is measured using the cosine similarity of the vector angle; spatial distribution similarity is evaluated using Hausdorff distance. For historical nodes with matching degrees exceeding a threshold, a set of their subsequent spread path branches is extracted. This set contains prediction data in three dimensions: spatial dimension describing the set of possible physical locations for spread, temporal dimension indicating the estimated ignition time at each location, and intensity dimension predicting the heat release rate change curve.

[0039] The calculation of the directional angle employs a multi-scale analysis method. The smoke diffusion direction vector is decomposed into a macroscopic architectural-level directional angle θ and a microscopic spatial-level directional angle φ. At the architectural level, the global spread direction vector for each subsequent path branch is calculated and compared with the measured macroscopic directional angle θ in planar angle. At the spatial level, the direction of the local path segment vectors of each branch is analyzed and a dot product operation is performed with the measured microscopic directional angle φ. The final comprehensive matching score formula is as follows:

[0040] In the formula: This represents the matching score of the k-th branch. and For the measured direction components, and For the branch prediction direction component, the weighting coefficients , The direction prediction is dynamically adjusted based on the building's spatial structural characteristics. The output is the branch path with the highest comprehensive score, along with associated spatiotemporal intensity prediction parameters.

[0041] The prediction results are presented using a multi-layer overlay technique. The base layer displays the building's floor plan; the path layer uses gradient-colored arrows to indicate the predicted spread direction; the thermal layer uses semi-transparent color spots to represent the probability distribution of fire ignition at different times; and the marker layer highlights the locations of high-risk load-bearing components. All layers support drag-and-drop browsing along a timeline, dynamically displaying the fire's development and evolution over the next 5-30 minutes. The system updates its predictions every 10 seconds, and automatically prompts for manual review when the difference in direction between three consecutive predictions exceeds a threshold. The prediction data is also written to a distributed cache system for real-time analysis by other emergency decision-making modules.

[0042] Example 4: The analysis and processing of fire spread direction prediction results begins with the identification of critical path nodes. The system loads structural topology data from the building information model and, combined with the fire spread trajectory marked in the prediction results, automatically marks structurally vulnerable points along the propagation path. Taking the prediction of fire spreading from the east office area to the west shopping mall via the connecting corridor as an example, the system first marks the steel structure nodes at the connection between the connecting corridor and the main building, and then extracts the building material parameters and load-bearing design loads for that area. The fire compartment status detection module acquires the closure status data of fireproof roller shutters and firewalls in real time. When a mechanical failure is detected in fireproof roller shutter No. 3, a flashing red border is displayed as a warning on the building floor plan.

[0043] The collection of personnel distribution density data employed multi-source fusion technology. The indoor positioning system acquired personnel's mobile phone signal strength via a Bluetooth beacon network, updating location data every 10 seconds, and cross-validated using contour recognition technology from the video surveillance system. The data processing center performed spatial grid mapping, dividing the building plan into 2-meter precision grid cells and calculating the number of people within each cell. In the shopping mall scenario, the system detected 48 people in real-time in the second-floor children's play area, reaching a density per unit area that met the warning threshold, and marked this area as an orange-level key protection zone.

[0044] The dynamic revision of the evacuation route planning scheme was implemented using a phased execution strategy. The initial scheme used the principle of the shortest straight path to connect exits and areas where people gathered. The revision process introduced a fire propagation barrier mechanism. When the prediction showed that the atrium area would be blocked by fire within 8 minutes, the system recalculated the set of feasible paths for each floor. The path evaluation function included three dynamic variables: real-time temperature readings of the path segment, smoke concentration monitoring values, and structural stability coefficients. Taking the third floor of the North Building as an example, the original evacuation route via the east side passage was replaced by a detour route via the fire escape due to the predicted fire situation. After the new planned route was verified by the strength of the load-bearing walls, it was sent to the emergency broadcast system for navigation guidance.

[0045] The dynamic adjustment of fire resource allocation weights employs a hierarchical decision-making model. The basic layer assigns fixed weight values ​​based on the fire intensity prediction coefficient, while the enhanced layer integrates multi-dimensional real-time factors. Model inputs include: the fire resistance limit of building components in the direction of fire spread, the current operating mode of the ventilation system, and the material reserves in the emergency resource depot. Using a case study of fire spread to the basement, the system identifies the underground electrical substation in the fire path and immediately upgrades the protection level of that area. The adjustment algorithm executes the following logical sequence: extracting the thickness data of the concrete roof slab of the electrical substation, comparing the current fire temperature with the temperature difference at the concrete bursting critical point, and calculating the heat flux intensity by combining the opening values ​​of ventilation duct valves.

[0046] The incremental calculation of resource allocation weights incorporates a time decay factor and a cascading risk coefficient. This mechanism sets the initial weight adjustment magnitude to be positively correlated with the fire prediction certainty, decaying exponentially over time. When hazardous materials storage is identified in high-risk areas, a cascading risk factor is added. Table 1 illustrates the weight adjustment process for the underground warehouse area.

[0047] Table 1: Weight adjustment process for the underground warehouse area.

[0048] Area Identification Basic risk level Structural fragility index speed of fire spread Dangerous goods categories Initial adjustment coefficient Corrected weights B2F-08 orange color 0.82 0.4m / s Class C combustibles 1.2x 132% B2F-12 red 0.91 0.7m / s Class D flammable materials 1.5x 187% B2F-15 red 0.96 1.2m / s Class B explosives 2.0x 240% The scheduling scheme is restructured to implement a resource reallocation algorithm. The system reorders regional priorities based on corrected weights and generates equipment deployment instructions: for high-risk areas with a weight increase exceeding 50%, mobile fire-fighting robot deployment instructions are automatically added; when the increase exceeds 100%, smoke extraction vehicles are simultaneously dispatched to enhance on-site smoke extraction capabilities. When selecting deployment locations, the system automatically adjusts personnel configuration based on the physical isolation status of fire compartments, and automatically switches to unmanned equipment deployment mode for areas that have lost contact. The activation priority of emergency exits is dynamically sorted according to the traffic capacity index, temporarily converting freight elevator passages with high carrying capacity into evacuation routes.

[0049] A strategy coordination mechanism ensures the consistency of revised strategies. The system maintains a strategy version tree structure to record each adjustment trajectory. When a contradiction is detected between key parameters of previous and subsequent versions, a manual review process is initiated. In high-rise residential fire cases, the system automatically verifies the coordination between evacuation routes and smoke exhaust directions before implementing revised plans, avoiding design conflicts where evacuation paths coincide with smoke diffusion directions. All revision commands are transmitted to on-site mobile terminals via encrypted data links, and the equipment control center simultaneously receives the operation authorization key, forming a complete closed-loop control system. Consistency checks are performed every 2 minutes during plan execution. When the deviation between on-site environmental monitoring data and the prediction model exceeds the preset tolerance, a strategy rollback mechanism is automatically triggered.

[0050] Example 5: The construction process of a digital twin building model begins with the acquisition and fusion of high-precision 3D point cloud data. A 3D laser scanner establishes a multi-station observation point network inside the building, forming a spatial coordinate point set covering the entire area, with the accuracy of each point controlled within a 2mm error range. The scanned data is spatially registered with the building information model, mapping component attribute information to the corresponding spatial areas. The physical property parameters of structural components, including concrete compressive strength and steel yield strength, are imported through material testing reports to form a virtual building entity with material mechanical parameters. Non-structural components such as ceilings and partitions are supplemented with integrity data based on construction drawings, and destructible characteristic thresholds are marked in the model. The location and status parameters of various fire protection facilities are mapped in real time through an IoT data interface to ensure the synchronous correspondence between the digital model and the physical building.

[0051] The virtual fire parameter injection system sets parameter groups based on fire dynamics principles. The heat release rate parameter selects a standard growth curve according to the type of combustible material, considering the differences in development patterns among different types of fires. The smoke generation rate parameter determines the concentration ratio of key toxic substances through chemical composition analysis and is linked to a material thermal decomposition characteristic database. The flame propagation characteristics set a heat radiation diffusion pattern based on the geometric features of the building space, while also importing the influence model of ventilation system operating parameters. The parameter setting interface supports hierarchical configuration; the basic layer configures the main characteristic parameters, while advanced layers allow adjustment of detailed variables such as turbulence intensity and local oxygen concentration.

[0052] The strategy execution simulation environment establishes a discrete-time progression mechanism. The system decomposes the emergency response strategy into a sequence of executable instructions, including three main categories: equipment activation instructions, personnel movement instructions, and facility operation instructions. In the fire-fighting equipment dispatch simulation, each fire truck is treated as an independent intelligent agent, and its movement path is calculated based on the actual width of the building passageway for passability. The equipment deployment time is determined according to the standard procedures in the operation manual. The personnel behavior simulation adopts a group dynamics model, considering the gradient of the impact of panic on movement speed. The system records the processing time node, execution status, and resource consumption data of each instruction instance, and all operations proceed according to actual physical laws on the virtual timeline.

[0053] The tracking of fire resource dispatch routes employs a multi-dimensional recording system. Movement trajectories are recorded using a three-dimensional coordinate sequence, forming a complete path curve from the equipment's assembly point to its operational position. The operation timeline records key nodes, including equipment departure time, vehicle arrival time, hose connection completion time, and water cannon activation time. Resource consumption measurement includes the cumulative value of extinguishing agent usage, water supply pipeline pressure fluctuation curves, and changes in equipment wear and tear. Recorded data is stored categorized by resource unit, supporting playback of the complete operational process of a specific unit along a timeline.

[0054] The generation of the preset optimal route relies on a digital navigation model of the building. This model quantitatively assesses the internal accessibility of the building, comprehensively considering physical constraints such as passageway width, corner radius, ground slope, and doorway height. The route planning algorithm combines the accessibility index and historical scheduling efficiency data to establish a dynamic cost function that includes a direction change penalty and a deceleration coefficient for narrow passages. The optimal route storage format contains both spatial coordinate sequences and time prediction sequences as benchmarks for evaluation.

[0055] The path overlap analysis implements a quantitative evaluation process for differences. In the time dimension, a segmented alignment algorithm is used to handle speed differences, calculating the cumulative time offset of corresponding nodes. In the spatial dimension, the Fraser distance metric is used to analyze the spatial approximation between the actual and theoretical paths. Directional consistency detection calculates the angular deviation at key turning points. After dimensionless processing of the three indicators, a comprehensive deviation index is synthesized according to preset weight ratios. The deviation grading display system maps the deviation index to a seven-level spectral color scale, visually presenting problem areas in the 3D building model.

[0056] The extraction of fire control delay time employs the critical event comparison method. The system identifies three baseline event points: the time the fire front arrives at the fire compartment, the time fire-fighting equipment deploys water, and the turning point where the fire significantly weakens. The time difference between the actual events and the simulated predictions is calculated to obtain two sets of core data: thermal development delay parameters and response execution delay parameters. A distribution map is used to display the delay values ​​in different areas, identifying systemic delay points.

[0057] The parameter updates for the fire spread time-series model employ an incremental learning approach. The system establishes a parameter sensitivity analysis model, identifying four core parameters most sensitive to delay time: the building material thermal conductivity adjustment factor, the air convection intensity correction value, the radiative heat absorption ratio coefficient, and the structural heat capacity compensation parameter. The update process employs a two-stage adjustment: primary adjustment proportionally to the average delay; secondary adjustment targeting specific delay-exceeding areas for directional parameter optimization. After each adjustment, a regression test set is run to verify the model's stability in typical fire scenarios and prevent overfitting. Historical parameter versions are saved through metadata management, supporting longitudinal comparative analysis of model performance. The complete update process forms a closed-loop control architecture, outputting a record of prediction accuracy improvement at each decision cycle.

[0058] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0059] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A fire safety management monitoring platform system, characterized in that, The method comprises the following steps: a building information acquisition module is used to acquire building structure data and environmental monitoring data of a target building in real time; a risk area division module is used to divide a plurality of independent fire-fighting risk areas according to the building structure data, and to assign a region identifier to each fire-fighting risk area; a historical data indexing module is used to index a fire accident data set and a fire-fighting facility maintenance record set of each fire-fighting risk area in a historical time according to the region identifier; a dynamic risk analysis module is used to fuse the environmental monitoring data and the fire accident data set to generate a real-time fire risk coefficient sequence; a fire-fighting resource scheduling module is used to generate a fire-fighting resource allocation scheme based on the real-time fire risk coefficient sequence and the fire-fighting facility maintenance record set; an emergency response decision module is used to generate a multi-level fire emergency response strategy according to the fire-fighting resource allocation scheme.

2. The fire safety management monitoring platform system according to claim 1, wherein, Further comprising: a fire evolution prediction module is used to construct a fire spread time sequence model according to fire spread path data in the historical fire accident data set; a strategy correction module is used to input current environmental monitoring data into the fire spread time sequence model, output a fire spread direction prediction result, and correct the multi-level fire emergency response strategy according to the result; a verification execution module is used to execute the corrected multi-level fire emergency response strategy in a simulated fire scene to generate a strategy execution efficiency report; a closed-loop update module is used to update the fire spread time sequence model according to the strategy execution efficiency report.

3. The fire safety management monitoring platform system of claim 1, wherein, The fusion of the environmental monitoring data and the fire accident data set to generate a real-time fire risk coefficient sequence comprises: extracting temperature distribution data, smoke concentration data and air flow velocity data from the environmental monitoring data; extracting an environmental feature vector set at the time of a historical fire from the fire accident data set; calculating the deviation distance of the temperature distribution data, the smoke concentration data and the air flow velocity data from the environmental feature vector set, respectively; generating a real-time fire risk coefficient according to the weighted sum of the deviation distances; arranging a plurality of real-time fire risk coefficients in timestamp order to form a real-time fire risk coefficient sequence.

4. The fire safety management monitoring platform system according to claim 2, wherein, The calculation of the deviation distance of the temperature distribution data, the smoke concentration data and the air flow velocity data from the environmental feature vector set, respectively, comprises: performing clustering processing on the environmental feature vector set to generate a plurality of environmental feature clustering centers; calculating the temperature dimension Euclidean distance of the temperature distribution data from each environmental feature clustering center, respectively; calculating the concentration dimension Manhattan distance of the smoke concentration data from each environmental feature clustering center, respectively; calculating the velocity dimension cosine similarity inverse of the air flow velocity data from each environmental feature clustering center, respectively; performing weighted fusion on the temperature dimension Euclidean distance, the concentration dimension Manhattan distance and the velocity dimension cosine similarity inverse corresponding to each environmental feature clustering center to generate the deviation distance.

5. The fire safety management monitoring platform system of claim 1, wherein, The generation of a fire-fighting resource allocation scheme based on the real-time fire risk coefficient sequence and the fire-fighting facility maintenance record set comprises: extracting fire-fighting equipment availability state data from the fire-fighting facility maintenance record set; trend fitting is performed on the real-time fire risk coefficient sequence to generate a fire risk change slope; a risk level interval is divided according to the fire risk change slope; the risk level interval of each fire risk area is matched with the corresponding fire equipment availability state data; the number of fire extinguishing equipment, the deployment position of fire personnel and the priority of emergency passage are allocated according to the matching result.

6. The fire safety management monitoring platform system of claim 2, wherein, The construction of the fire spread timing model includes: analyzing the fire source location data, building material ignition point data and building structure topology data in the historical fire accident data set; constructing a three-dimensional space grid according to the building structure topology data; mapping the fire source location data to the initial grid node of the three-dimensional space grid; calculating the spread time interval of fire in adjacent grid nodes based on the building material ignition point data; connecting all the spread grid nodes in time sequence to form a fire spread path timing chain.

7. The fire safety management monitoring platform system of claim 5, wherein, The current environment monitoring data is input into the fire spread timing model to output the fire spread direction prediction result, which includes: obtaining real-time temperature gradient data and smoke diffusion direction data of the current fire point; matching the historical fire node closest to the real-time temperature gradient data in the fire spread path timing chain; extracting the subsequent spread path branch of the historical fire node; calculating the direction angle between the smoke diffusion direction data and each subsequent spread path branch; selecting the subsequent spread path branch with the smallest direction angle as the fire spread direction prediction result.

8. The fire safety management monitoring platform system of claim 2, wherein, The multi-level fire emergency response strategy is corrected according to the result, which includes: extracting the high-risk spread area identifier in the fire spread direction prediction result; obtaining the personnel distribution density data in the high-risk spread area; adjusting the evacuation route planning scheme in the multi-level fire emergency response strategy; dynamically increasing the fire resource allocation weight of the high-risk spread area.

9. The fire safety management monitoring platform system of claim 7, wherein, The fire resource allocation weight of the high-risk spread area is dynamically increased, which includes: obtaining the building bearing structure data of the high-risk spread area; calculating the ratio of the predicted fire spread speed to the fire resistance coefficient of the building bearing structure; setting the fire resource allocation weight increment according to the ratio; superimposing the fire resource allocation weight increment in the fire resource allocation scheme.

10. The fire safety management monitoring platform system of claim 2, wherein, The corrected multi-level fire emergency response strategy is executed in the simulated fire scene to generate a strategy execution efficiency report, which includes: constructing a digital twin building model containing the fire spread direction prediction result; injecting virtual fire parameters into the digital twin building model; executing the corrected multi-level fire emergency response strategy and recording the virtual fire resource scheduling path; comparing the overlap degree of the virtual fire resource scheduling path and the preset optimal path; generating a strategy execution efficiency report according to the overlap deviation value.

11. The fire safety management monitoring platform system of claim 9, wherein, The fire spread timing model is updated according to the strategy execution efficiency report, which includes: extracting the fire control delay time data in the strategy execution efficiency report; obtaining the time difference between the actual fire resource arrival time and the virtual scheduling time; adjusting the spread time interval parameter in the fire spread timing model according to the time difference; recalculating the spread time interval of fire in adjacent grid nodes.

Images