Fire alarm linkage system for fire engineering based on Internet of Things

Abstract

The invention discloses a fire alarm linkage system for fire engineering based on the Internet of Things. The operation process of the system specifically comprises the following steps: generating a dynamic risk probability distribution diagram based on a building three-dimensional point cloud model and multi-source data, and deploying a differential sensor array; environment multi-physical field parameters are collected in real time through a differential sensor array, and a multi-mode fire behavior sensing signal set is generated; building a fire situation credibility map: calling a multi-criterion decision module to calculate a linkage equipment weight distribution matrix based on the fire situation credibility map, and generating a multi-stage linkage control instruction set in combination with the building topological structure model and the personnel distribution thermodynamic diagram; and performing multi-cross-system cooperative execution and closed-loop optimization. The method has the following advantages and effects that the inherent limitation of a traditional system can be broken through by constructing an intelligent prevention and control framework combining risk quantification, situation deduction and decision closed loop, the personnel evacuation efficiency is effectively improved, and damage caused by major accidents such as detonation is reduced.

Description

Technical Field

[0001] This invention relates to the field of fire protection technology, and in particular to a fire alarm linkage system for fire protection engineering based on the Internet of Things. Background Technology

[0002] In modern high-rise buildings, large complexes, and industrial facilities, fire prevention and control systems are critical infrastructure for protecting lives and property. Traditional fire alarm and linkage systems generally employ a response mechanism that combines threshold triggering with fixed pre-programmed sequences, relying on independent alarm signals from a single type of sensor to trigger a pre-programmed sequence of device actions. However, this mechanism suffers from a fundamental flaw: an irreconcilable disconnect exists between its static response logic and the multidimensional spatiotemporal coupling characteristics of the dynamic evolution of a fire. Specifically, traditional systems, lacking risk modeling dimensions, cannot quantify the heterogeneous risk distribution in building spaces. Especially in newly constructed scenarios lacking historical fire data, the hidden risks of high-temperature gas accumulation areas (such as ventilation duct inflection points) and high-risk deflagration nodes (such as equipment mezzanines) cannot be captured by statically and uniformly deployed sensor arrays. At the same time, insufficient situational awareness leads to single-modal data (such as only temperature or smoke concentration) severing the coupling relationship of fire evolution. Typical examples include the accelerated chain reaction state where the temperature does not reach the threshold but the free radical concentration rises sharply, or the hidden deflagration risk where the combustible gas exceeds the standard but the airflow diffusion does not trigger the smoke alarm. The system misjudges the true threat level due to the lack of multi-physics field feature fusion capabilities. More importantly, the broken decision-making adaptive closed loop makes it difficult for fixed linkage plans to adapt to the spatial differences in the dynamic evolution of fire and personnel evacuation needs. For example, when the fire spreads along the rapid spread channel, the failure to prioritize the suppression of key nodes leads to loss of control, or the failure to optimize emergency lighting and broadcast path guidance strategies in high-density personnel areas in real time. The root cause of these systemic deficiencies lies in the fact that traditional solutions neither integrate building topology, historical disaster statistics, or cross-domain physical rules (such as the correlation characteristics of electrical faults migrating to fires) to construct dynamic risk models, nor establish reliable quantifiable maps based on thermodynamic threats, chemical reaction intensity, and deflagration tendency coefficients. Furthermore, they lack a real-time feedback optimization mechanism between coordinated commands and fire suppression effects, ultimately leading to delayed risk response or even erroneous actions. Therefore, there is an urgent need for an intelligent prevention and control architecture that combines quantifiable risks, predictable situations, and closed-loop decision-making to overcome the inherent limitations of traditional systems. Summary of the Invention

[0003] The purpose of this invention is to provide a fire alarm linkage system for fire protection engineering based on the Internet of Things (IoT) to solve the problems mentioned in the background art.

[0004] The above-mentioned technical objective of the present invention is achieved through the following technical solution:

[0005] A fire alarm linkage system for fire protection engineering based on the Internet of Things (IoT) includes the following steps in its operation:

[0006] S100. A dynamic risk probability distribution map is generated based on the three-dimensional point cloud model of the building and multi-source data. Differentiated sensor arrays are deployed in the marked high-temperature gas accumulation area, rapid spread channel and high-risk deflagration node. The generation of the dynamic risk probability distribution map is combined with the statistical association rules of the effective historical fire data in the fire association database. If there is a lack of effective historical fire data, it is achieved by combining cross-domain risk rule migration operation.

[0007] S200. The deployed differentiated sensor array collects environmental multi-physics parameters in real time and generates a multimodal fire perception signal set.

[0008] S300. Constructing a Fire Situation Credibility Map: Perform multimodal feature fusion and spatiotemporal alignment processing on the multimodal fire perception signal set, and combine the three-dimensional point cloud model of the building with the fire dynamics prior rule library to generate a spatially gridded fire situation credibility map; wherein, the fire situation credibility map includes a fusion quantitative index of thermodynamic threat level value, chemical chain reaction intensity and deflagration tendency coefficient;

[0009] S400. Based on the fire situation credibility map, the multi-criteria decision module is called to calculate the weight allocation matrix of the linkage equipment, and combined with the building topology model and personnel distribution heat map, a multi-level linkage control instruction set is generated; wherein, the multi-level linkage control instruction set includes equipment start-stop sequence, power regulation parameters and emergency channel activation strategy.

[0010] S500, Multi-system Collaborative Execution and Closed-Loop Optimization: The multi-level linkage control instruction set is sent to the edge computing gateway cluster through the industrial Internet of Things protocol to drive the fire sprinkler system, smoke control system, emergency lighting system and evacuation broadcast system to perform linkage operations; at the same time, the equipment response status data and fire suppression effect indicators are collected in real time, and the spatial sensitive area division rules and decision engine weight parameters are iteratively updated until the fire situation credibility converges to the safe threshold range.

[0011] By adopting the above technical solutions, and using dynamic risk probability distribution maps to guide the deployment of differentiated sensor arrays, targeted monitoring capabilities are achieved for high-temperature gas accumulation areas, rapid spread channels, and high-risk deflagration nodes, significantly improving the accuracy of early fire hazard identification. The multimodal fire perception signal set integrates environmental multi-physics parameters, overcoming the limitations of single sensors and enabling multi-dimensional dynamic perception of fire conditions. The constructed fire situation credibility map integrates three-dimensional quantitative indicators such as thermodynamic threat level, chemical chain reaction intensity, and deflagration tendency coefficient, transforming abstract fire conditions into visualized scientific decision-making basis through spatial grid modeling. The multi-criteria decision module combines building topology and population distribution heat... The system strives to generate a weight allocation matrix for interconnected devices to ensure precise scheduling of fire-fighting resources according to threat priority. The multi-level linkage control command set maximizes emergency response efficiency through the collaborative design of equipment start-stop sequences, power control parameters, and emergency channel activation strategies. The multi-system collaborative execution mechanism driven by industrial IoT protocols enables millisecond-level linkage response of sprinkler, smoke control, lighting, and broadcasting systems. The closed-loop optimization process dynamically updates the spatial sensitive area division rules and decision weight parameters by collecting real-time equipment status and fire suppression effect data, enabling the system to have continuous adaptive evolution capabilities. Ultimately, it promotes the convergence of fire situation credibility to a safe threshold, forming a closed-loop intelligent prevention and control system.

[0012] A further setting is that S100 specifically includes the following steps:

[0013] Based on the spatial topological features of the building's 3D point cloud model and the statistical association rules of valid historical fire data in the fire association database, a dynamic risk probability distribution map is generated; wherein, the dynamic risk probability distribution map contains the mapping relationship between spatial coordinate positions and fire occurrence probability values;

[0014] High-temperature gas accumulation areas, rapid spread channels, and high-risk deflagration nodes are marked on the dynamic risk probability distribution map.

[0015] A multimodal sensor fusion cluster is deployed in the high-temperature gas accumulation zone; wherein the multimodal sensor fusion cluster includes a combination of distributed temperature sensors, infrared thermal imaging sensors, and flame spectral sensors, and satisfies that the spatial deployment density is positively correlated with the local risk probability value; a redundant gas sensor array is deployed at the key nodes of the fluid channel in the rapid propagation channel; wherein the redundant gas sensor array includes at least two sets of heterogeneous combustible gas concentration sensors, and the deployment orientation covers the normal vector direction of the key nodes of the fluid channel; an explosion-proof pressure sensor array is deployed at the high-risk deflagration node; wherein the explosion-proof pressure sensor array includes pressure surge sensors and shock wave sensors, and the deployment location satisfies that the Euclidean distance from the potential deflagration center is less than a preset safety threshold.

[0016] By adopting the above technical solutions, the differentiated deployment strategy enables the use of multi-modal sensor fusion clusters for monitoring high-temperature gas accumulation areas. Cross-validation of temperature, infrared, and spectral data significantly improves the reliability of thermal disaster identification, and the positive correlation between deployment density and local risk probability optimizes the allocation of limited sensing resources. The redundant gas sensor array deployed at key nodes in the rapid spread channel effectively overcomes the monitoring blind spots caused by the uncertainty of gas diffusion direction through heterogeneous sensor networking and normal vector direction coverage. The dual-channel redundancy design significantly enhances data robustness. The close-range deployment of explosion-proof pressure sensor arrays at high-risk deflagration nodes (Euclidean distance less than the safety threshold) combined with dual-modal monitoring of pressure mutation and shock wave enables rapid capture before deflagration, securing a critical response time window.

[0017] A further setting involves generating a dynamic risk probability distribution map, which includes the following steps:

[0018] If there is no valid historical fire data in the fire association database, a theoretical risk probability distribution map is generated by multiphysics coupling numerical simulation based on the geometric constraint features and structural thermal features of the building's three-dimensional point cloud model. The geometric constraint features include volume curvature distribution features and ventilation cross-sectional area ratio features, and the structural thermal features include material thermal conductivity distribution features and combustible material distribution density features.

[0019] Perform cross-domain risk rule migration operations: extract spatial probability density functions from electrical equipment fault location distribution data in non-fire-related databases and map them to fire location probability distribution features; extract turbulence intensity parameters from fluid dynamic diffusion models in non-fire-related databases and convert them into fire spread rate distribution features; generate migration risk probability distribution maps based on migration mapping rules;

[0020] The theoretical risk probability distribution map and the migration risk probability distribution map are integrated using a Bayesian-physical fusion modeling framework.

[0021] By adopting the above technical solutions, the dynamic risk map generation mechanism, in the absence of historical fire data, extracts the geometric constraints and thermal characteristics of buildings through multi-physics field coupled numerical simulation to generate theoretical risk maps, ensuring the physical authenticity of the basic risk assessment; the cross-domain risk rule transfer operation maps the electrical equipment fault location density function to the fire probability distribution and transforms the fluid turbulence intensity parameter into fire spread characteristics, breaking through domain data barriers to achieve knowledge transfer; the Bayesian-physics fusion modeling framework organically integrates the physical rigor of theoretical simulation with the empirical value of transferred data, generating a dynamic risk probability distribution map that combines theoretical support and practical adaptability.

[0022] A further setting is that the Bayesian-physical fusion modeling framework performs the following operations:

[0023] The theoretical risk probability distribution diagram is used as the prior probability distribution;

[0024] Use the migration risk probability distribution map as the input to the likelihood function;

[0025] The dynamic risk probability distribution map is generated by calculating the posterior probability distribution through Markov chain Monte Carlo sampling.

[0026] By adopting the above technical solution, the Yesse-physical fusion modeling uses the theoretical risk map as the prior probability distribution and the migration risk map as the input of the likelihood function. The posterior probability distribution is calculated through Markov chain Monte Carlo sampling, so that the dynamic risk probability distribution map inherits the inherent physical characteristics of the building and incorporates cross-domain statistical laws, which significantly improves the statistical confidence and spatial resolution of risk assessment.

[0027] A further setting is that S200 specifically includes the following steps:

[0028] The deployed differentiated sensor array collects environmental multi-physics parameters in real time; wherein, the environmental multi-physics parameters include thermodynamic parameters, gas composition parameters, and free radical concentration parameters.

[0029] Multi-level signal conversion and fusion processing is performed on the environmental multi-physics parameters to generate standardized physical field feature vectors; wherein, the multi-level signal conversion and fusion processing includes:

[0030] The thermodynamic parameters are converted into thermodynamic risk feature vectors using a nonlinear temperature field mapping model; the gas composition parameters are converted into chemical risk feature vectors using a gas diffusion coupling algorithm; and the free radical concentration parameters are converted into deflagration risk feature vectors using a chain reaction kinetic model.

[0031] The thermodynamic risk feature vector, chemical risk feature vector and deflagration risk feature vector are time-stamped and matched with spatial coordinates by the spatiotemporal alignment module to generate a fused feature tensor.

[0032] A multimodal fire perception signal set is generated based on the fused feature tensor; wherein the multimodal fire perception signal set includes a triplet of time series labels, spatial location codes, and physical field feature values, and the physical field feature values ​​include normalized thermodynamic risk feature values, chemical risk feature values, and deflagration risk feature values.

[0033] By adopting the above technical solutions, the signal processing method converts thermodynamic parameters into risk feature vectors through nonlinear temperature field mapping, avoiding error accumulation caused by linear simplification; gas composition parameters restore the chemical risk propagation law in the real environment through diffusion coupling algorithm; free radical concentration parameters are converted through chain reaction kinetic model to accurately quantify the evolution trend of deflagration risk; the spatiotemporal alignment module eliminates the acquisition delay and spatial misalignment of multi-source signals, and the generated fusion feature tensor ensures the spatiotemporal consistency of subsequent analysis data; standardized physical field feature value triples (time series label, spatial location code, normalized feature value) provide a structured data foundation for fire situation modeling.

[0034] A further setting is that S300 specifically includes the following steps:

[0035] The physical field feature values ​​are encoded and mapped to the corresponding grid cells of the building's 3D point cloud model according to the spatial location. For grid cells covered by multiple sensors, the weighted average of the feature values ​​is calculated based on the Deloni triangulation algorithm. The weight allocation follows the positive correlation rule between sensor spatial deployment density and local risk probability.

[0036] Perform dynamic fusion of multimodal features: calculate the thermodynamic threat level, chemical chain reaction intensity coefficient, and deflagration tendency coefficient for each grid cell;

[0037] Thermodynamic threat level values, chemical chain reaction intensity coefficients, and deflagration tendency coefficients are integrated into a three-dimensional tensor matrix based on spatial grid units. The missing data in areas where no sensors are deployed are filled in by spatiotemporal interpolation algorithms to generate a spatial gridded fire situation reliability map covering the entire building's three-dimensional point cloud model.

[0038] By adopting the above technical solutions, the fire situation map construction method calculates the weighted average of the multi-sensor coverage grid using the Deloni triangulation algorithm. The design principle that the weights are positively correlated with the deployment density and risk probability strengthens the authority of data in high-risk areas. The dynamic fusion of multi-modal features synchronously outputs the thermodynamic threat level, chemical reaction intensity, and deflagration tendency coefficient at the grid cell level, realizing the collaborative analysis of multiple fire elements. The spatiotemporal interpolation algorithm fills the data gaps in unmonitored areas, generating a continuous reliability map covering the entire area and eliminating the blind spots of traditional point monitoring.

[0039] A further setting is that the execution of multimodal feature dynamic fusion specifically includes the following steps:

[0040] The thermodynamic risk feature value is input into a pre-trained convolutional neural network model, which outputs a thermodynamic threat level value in the interval [0,1]. The convolutional neural network model is trained by the mapping relationship between temperature gradient distribution values ​​and actual fire spread speed in historical fire cases.

[0041] The chemical risk characteristic value and free radical concentration parameter are input into the chain reaction kinetic function to solve the relationship function between the free radical concentration change rate and the activation energy threshold, and the chemical chain reaction strength coefficient is output.

[0042] The thermodynamic threat level and chemical chain reaction intensity coefficient are input into the deflagration critical surface model to calculate the Mahalanobis distance between the current environmental parameters and the theoretical deflagration critical point, which is mapped to the deflagration tendency coefficient in the [0,1] interval.

[0043] By adopting the above technical solutions, the convolutional neural network in multimodal feature fusion mines the deep mapping relationship between historical fire temperature gradients and spread rates, making the prediction of thermodynamic threat levels reliable with case support; the chain reaction kinetic function outputs the chemical chain reaction intensity coefficient based on the relationship between the free radical concentration change rate and the activation energy threshold, accurately quantifying the risk of runaway combustion chain reactions; the deflagration critical surface model transforms the complex multivariate interactions into an intuitive deflagration tendency coefficient by calculating the Mahalanobis distance between environmental parameters and the theoretical deflagration point, providing a quantitative basis for deflagration suppression decisions.

[0044] A further setting is that S400 specifically includes the following steps:

[0045] The thermodynamic threat level, chemical chain reaction intensity coefficient, and deflagration tendency coefficient of each spatial grid cell in the fire situation confidence map are extracted and input into a pre-constructed multidimensional threat quantification function to output a normalized comprehensive fire threat index; wherein, the multidimensional threat quantification function is trained by the mapping relationship between fire parameters and actual disaster losses in historical fire cases;

[0046] The multi-criteria decision-making module is invoked to match the comprehensive fire threat index with a preset fire equipment efficiency database. This database stores the coverage radius, response delay time, and suppression efficiency parameters of fire sprinkler systems, smoke extraction systems, emergency lighting systems, and evacuation broadcast systems. Based on the matching results and the connectivity constraints of the building topology model, the spatial coupling degree between equipment deployment nodes is calculated. A linear weighted scoring method is used to dynamically generate a linkage equipment weight allocation matrix, which includes the priority score of each fire equipment within the spatial grid unit.

[0047] The system acquires real-time data on population density distribution and movement trajectory prediction from a heat map of population distribution; calculates the theoretical evacuation time for each area based on the exit location and passage width parameters of the building topology model using a path planning algorithm; inputs population density, movement speed deviation, and theoretical evacuation time into an evacuation risk assessment model, and outputs a population evacuation influencing factor; wherein, the population evacuation influencing factor is positively correlated with the urgency of the linkage strategy.

[0048] The weight allocation matrix of the linked equipment and the personnel evacuation influencing factors are input into the dynamic instruction generation module.

[0049] By adopting the above technical solutions, the coordinated decision-making mechanism integrates three major quantitative indicators through a multi-dimensional threat quantification function to generate a comprehensive fire threat index. The mapping relationship of historical case training ensures the correlation between the index and actual disaster losses. The matching of fire equipment efficiency database and spatial coupling degree calculation ensure that the equipment deployment location and the fire hot zone form the optimal spatial mapping. The weight allocation matrix generated by the linear weighted scoring method dynamically reflects the priority of each equipment's fire suppression contribution in a specific grid unit. The personnel distribution heat map combined with the evacuation impact factor output by the path planning algorithm embeds personnel life safety elements into the equipment scheduling logic to achieve a people-centered emergency response.

[0050] A further setting involves inputting the weight allocation matrix of the linked equipment and the personnel evacuation influencing factors into the dynamic instruction generation module, after which the dynamic instruction generation module performs the following operations:

[0051] If the overall fire threat index is greater than the preset high-risk threshold and the personnel evacuation impact factor exceeds the critical value, then the core response objective is set as the highest priority personnel evacuation.

[0052] If the overall fire threat index is in the medium threat range and the deflagration tendency coefficient continues to rise, then the core response objective is set as core chemical suppression.

[0053] Otherwise, the core response target is set as local thermodynamic suppression;

[0054] Based on the core response target and the priority score in the weight allocation matrix of the linked equipment, a multi-level linkage control instruction set is generated, which includes equipment start-stop sequence, power control parameters, and emergency passage activation strategy. The equipment start-stop sequence specifies the spatiotemporal order of activation of the fire sprinkler system, smoke control system, emergency lighting system, and evacuation broadcast system. The power control parameters include the water pressure gradient value of the sprinkler system and the air volume adjustment value of the smoke control system. The emergency passage activation strategy includes the lighting intensity distribution scheme of the emergency lighting system and the voice guidance path of the evacuation broadcast system.

[0055] By adopting the above technical solutions, the instruction generation strategy dynamically switches core response targets (including personnel evacuation, chemical suppression, and thermodynamic suppression) based on threshold judgments of the comprehensive threat index and evacuation impact factors, ensuring that the response strategy strictly matches the real-time risk situation; the equipment start-up and shutdown sequence design follows the priority ranking of the weight allocation matrix to optimize the spatiotemporal deployment order of limited fire-fighting resources; power parameters such as water pressure gradient value and air volume adjustment value achieve dynamic matching between fire extinguishing intensity and fire intensity; and the collaborative design of emergency lighting intensity distribution scheme and voice guidance path constructs a dual-channel evacuation guidance system of physical and information.

[0056] A further setting is that the S500 simultaneously collects equipment response status data and fire suppression effect indicators in real time, iteratively updates the spatial sensitive area division rules and decision engine weight parameters until the fire situation credibility converges to a safe threshold range; including:

[0057] Real-time acquisition of equipment response status data, including valve opening feedback values, actual fan speed values, measured lighting brightness values, and broadcast coverage intensity values; real-time acquisition of fire suppression effect indicators through the differentiated sensor array; wherein, the fire suppression effect indicators include temperature gradient descent rate, combustible gas concentration change slope, free radical concentration attenuation coefficient, and fire spread rate change; based on the equipment response status data and the fire suppression effect indicators, the following closed-loop optimization operations are performed:

[0058] The fire suppression effect index is compared with the expected suppression target value, and the suppression deviation coefficient of the spatial grid is calculated. If the suppression deviation coefficient of a certain spatial grid cell is continuously greater than the preset tolerance threshold, the cell is marked as a high-sensitivity area and its monitoring weight in the fire situation credibility map is increased.

[0059] The weighting coefficients of the thermodynamic threat level, chemical chain reaction intensity coefficient, and deflagration tendency coefficient in the multidimensional threat quantification function are adjusted by the gradient descent algorithm to maximize the correlation coefficient between the predicted value of the comprehensive fire threat index and the actual suppression effect.

[0060] If the global mean variance of the fire situation confidence map decreases less than the preset convergence threshold and the comprehensive fire threat index remains stable within the safe threshold range for three consecutive monitoring time windows, then the loop terminates; otherwise, return to S300.

[0061] By adopting the above technical solutions, the closed-loop optimization mechanism identifies high-deviation grid cells and increases their monitoring weight by comparing the fire suppression effect index with the expected target value in real time, thereby enhancing the system's dynamic tracking capability for complex fires. The gradient descent algorithm dynamically adjusts the weighting coefficients of the multi-dimensional threat quantification function, so that the predicted value of the comprehensive threat index continuously approaches the actual suppression effect. The dual convergence criteria of the global mean variance reduction rate and the safety threshold interval ensure the effectiveness of the response while avoiding excessive resource consumption, thus achieving autonomous performance evolution and improved response accuracy of the system.

[0062] In summary, the present invention has the following beneficial effects: by constructing an intelligent prevention and control architecture that combines quantifiable risks, predictable situations, and closed-loop decision-making, it can break through the inherent limitations of traditional systems, thereby effectively improving the efficiency of personnel evacuation and reducing the damage caused by major accidents such as deflagrations. Attached Figure Description

[0063] Figure 1 This is the main flowchart of an embodiment;

[0064] Figure 2 This is a schematic diagram of the overall process of S100 in the embodiment;

[0065] Figure 3 This is a schematic diagram of the overall process of S200 in the embodiment;

[0066] Figure 4 This is a flowchart illustrating S300 in the embodiment;

[0067] Figure 5 This is a schematic diagram of the overall process of S400 in the embodiment. Detailed Implementation

[0068] The present invention will be further described in detail below with reference to the accompanying drawings.

[0069] As attached Figures 1 to 5 As shown;

[0070] This embodiment discloses a fire alarm linkage system for fire protection engineering based on the Internet of Things. The system operation process specifically includes the following steps:

[0071] S100. A dynamic risk probability distribution map is generated based on the three-dimensional point cloud model of the building and multi-source data. Differentiated sensor arrays are deployed in the marked high-temperature gas accumulation area, rapid spread channel and high-risk deflagration node. The generation of the dynamic risk probability distribution map is combined with the statistical association rules of effective historical fire data in the fire association database. If effective historical fire data is lacking, it is achieved by combining cross-domain risk rule migration operation.

[0072] In this embodiment, the generation of the dynamic risk probability distribution map and the deployment of the differentiated sensor array are achieved through the following process: First, based on the spatial topological features and structural thermal features of the building's 3D point cloud model, a dynamic risk probability distribution map is generated by combining effective historical fire data from the fire association database with statistical association rules. If the fire association database lacks effective historical data (such as newly built buildings or buildings that have not experienced fires), a cross-domain risk rule migration operation is initiated: specifically, this includes extracting the spatial probability density function from the electrical equipment fault location distribution data and converting it into fire location probability distribution features through a mapping function; simultaneously, turbulence intensity parameters are extracted from the fluid dynamics diffusion model, and fire spread rate distribution features are generated through transformation rules to form the migrated risk probability distribution map. Furthermore, the theoretical risk probability distribution map (generated by multi-physics coupled numerical simulation) and the migrated risk probability distribution map are integrated through a Bayesian-physics fusion modeling framework: the theoretical risk probability distribution map is used as the prior probability distribution.

[0073] The migration risk probability distribution map is used as a likelihood function. The Markov chain Monte Carlo sampling algorithm is used to calculate the posterior probability distribution. The final output is a dynamic risk probability distribution map covering the entire building space. This map stores the coordinate-probability value mapping relationship with three-dimensional grid cells as the basic unit.

[0074] The core of this step lies in constructing a dynamic risk probability distribution map and deploying differentiated sensor arrays in key areas based on this map. The specific implementation process consists of three stages:

[0075] The first phase involves generating a dynamic risk probability distribution map. Based on the spatial topological features of the building's 3D point cloud model and statistical association rules with valid historical fire data from a fire association database, a mapping relationship between spatial coordinates and fire occurrence probability values ​​is generated—the dynamic risk probability distribution map. If historical data is missing, a cross-domain risk rule migration operation is initiated: This involves using multiphysics coupled numerical simulation to calculate the building's geometric constraints (including volume curvature distribution and ventilation cross-sectional area ratio) and structural thermal characteristics (including material thermal conductivity distribution and combustible material density distribution) to generate a theoretical risk probability distribution map. During the cross-domain risk rule migration operation, the probability density function f is extracted from the electrical equipment fault location distribution data. elec (x,y,z), mapped to the fire location probability P fire =k1·f elec Where k1 is the migration coefficient. The turbulence intensity ∈ in the fluid dynamics model turb Converted to fire spread rate Where k2 is the migration coefficient, effectively applicable to ventilation-dominated fires (such as building corridors and industrial pipelines). The migration risk probability is calculated for each coordinate point (x, y, z). Among them, Pfire (x,y,z) represents the probability of a fire location at the coordinates, which can be understood as the probability of a fire starting; V spread (x,y,z) represents the fire spread rate at coordinates; V max Indicates the threshold for the rate of fire spread; w p and w v These are preset weights. Output P corresponding to the full-space coordinates (x, y, z). mig The set of values ​​is the probability distribution map of migration risk.

[0076] Phase two is Bayesian-physical fusion modeling;

[0077] During the dynamic risk probability distribution chart annotation phase, three types of high-risk areas are identified:

[0078] High-temperature gas accumulation zone labeling operation: Calculate the thermal convection intensity gradient of each enclosed space in the 3D point cloud model of the building, and label continuous regions with thermal convection intensity gradients greater than a preset threshold and spatial curvature radii less than a critical value as high-temperature gas accumulation zones; wherein, the thermal convection intensity gradient is solved simultaneously using a gas diffusion model and thermodynamic equations; for example, the thermal convection intensity gradient calculation uses a gas diffusion-thermodynamic simultaneous equation: in, α represents the thermal convection intensity gradient; α represents the material's thermal conductivity; β represents the gas diffusion rate coefficient; C represents the concentration of combustible gas. Represents the second derivative in temperature space; Indicates the rate of change of combustible gas concentration over time; when Furthermore, when the radius of curvature Rc < 2m, it is marked as a high-temperature gas accumulation zone. Rapid propagation channel annotation operation: Identify multi-story passage structures (such as stairwells, elevator shafts, ventilation ducts, etc.) in the 3D point cloud model of a building, and calculate the theoretical propagation coefficient based on the aspect ratio of the passage section and the material flammability index. For example, the theoretical propagation coefficient... Where k3 represents the shape factor, a correction coefficient determined by the channel geometry (1.0 for rectangular channels, 0.8 for circular channels); L channel W represents the total height of the vertical passageway connecting the floors; channel The minimum characteristic dimension of the channel cross-section (take the minimum inner diameter or minimum clear width); α comb This indicates the flammability rating of the material used to construct the inner wall of the passage (0-1, with 1.0 for flammable materials and 0.3 for flame-retardant materials); when C... spread >C thres At that time, C thres A value of 1.8 is used to mark this linear region as a rapid spread channel. High-risk deflagration node marking operation: Locate the local maximum point P of the fire occurrence probability value in the dynamic risk probability distribution map. maxThe judgment condition must be met simultaneously: P fire (P max )>P explode ;where P explode This represents the deflagration probability threshold, with a value of 0.7; d <d safe Where d is the Euclidean distance to the nearest densely populated area of ​​combustible materials; d safe This represents the safe radius for material volatility, taken as 1.5 meters. The theoretical risk probability distribution is used as the prior probability P. prior The migration risk distribution map generated from cross-domain migration is transformed into a likelihood function L, and P is solved using Markov chain Monte Carlo sampling. fire ∝P prior ·L; where P fire The posterior probability is then used to output the final dynamic risk distribution map.

[0079] Phase three is differentiated sensor deployment:

[0080] In high-temperature gas accumulation zones, a multimodal sensor fusion cluster is deployed, including distributed temperature sensors (accuracy ±0.5℃), infrared thermal imaging sensors (resolution 640×480), and flame spectral sensors (wavelength range 300-1100nm). The deployment density is positively correlated with the local risk probability value. In the rapid propagation channel, redundant gas sensor arrays are deployed at critical fluid nodes, with at least two sets of heterogeneous sensors (catalytic combustion type + infrared absorption type combustible gas sensors), symmetrically arranged along the channel normal direction with a spacing ≤0.5m. At high-risk deflagration nodes, explosion-proof pressure sensor arrays are deployed, including piezoelectric pressure surge sensors (range 0-10kPa) and shock wave sensors (frequency response 0.1-10kHz). The deployment locations must satisfy the requirement that the Euclidean distance from the potential deflagration center is less than a preset safety threshold, which is 0.3 meters.

[0081] S200: The deployed differentiated sensor array collects environmental multi-physics parameters in real time and generates a multimodal fire perception signal set.

[0082] The core of this step is to achieve real-time acquisition and fusion processing of environmental multi-physics parameters through differentiated sensor arrays, generating a standardized set of multimodal fire sensing signals. The specific implementation process is as follows:

[0083] Thermodynamic parameters are acquired in real time by a multimodal sensor fusion cluster (including distributed temperature sensors, infrared thermal imaging sensors, and flame spectral sensors) deployed in the high-temperature gas accumulation zone. Acquired parameters include: temperature gradient distribution values, obtained through spatial difference calculation using the distributed temperature sensor array; infrared radiation intensity values, measured by infrared thermal imaging sensors of the object's surface thermal radiation; and flame spectral characteristic values, captured by flame spectral sensors capturing specific wavelength peaks of the flame. Gas composition parameters are acquired by a redundant gas sensor array (at least two heterogeneous sensors) deployed in the rapid propagation channel. Acquired parameters include: combustible gas concentration values, measured by catalytic combustion sensors; oxygen concentration values, monitored using electrochemical sensors; and smoke particulate matter concentration values, acquired by particle sensors based on laser scattering principles. Free radical concentration monitoring sensor arrays are deployed at high-risk deflagration nodes. Acquired parameters include hydroxyl radical concentration values ​​and hydroperoxygen radical concentration values. A hydroxyl radical concentration higher than 50 ppb indicates accelerated chain reactions; a hydroperoxygen radical concentration higher than 30 ppb increases the probability of deflagration. All sensor data are sampled at a frequency of 10Hz and aligned to a unified timing reference via a timestamp synchronization module.

[0084] Thermodynamic parameters are converted into thermodynamic risk feature vectors using a nonlinear temperature field mapping model: when the temperature gradient distribution value exceeds 10℃ / m, the risk value is calculated according to the rule of "gradual increase in the low temperature region and steep increase in the high temperature region"; the infrared radiation intensity value is scaled proportionally, for example, to 500W / m. 2 The highest risk value is 1.0; the flame spectral feature values ​​match the spectral features of typical fires, and the higher the similarity, the greater the risk value. Gas composition parameters are converted into a chemical risk feature vector using a gas diffusion coupling algorithm: combustible gas risk value. Where C fuel This indicates the current concentration of combustible gas; 25 is the lower explosion limit threshold set according to the gas; when C fuel When C is ≥25%, the combustible gas risk value is forcibly set to 1; fuel When the concentration of combustible gas is less than 25%, the risk value for combustible gas is calculated proportionally; the risk value for oxygen deficiency is... in This indicates the current oxygen concentration value, with 21 being the normal oxygen concentration baseline value; when When the risk of hypoxia is 0; when At that time, the risk value of hypoxia increases as the concentration decreases; when At that time, the oxygen deficiency risk value is forcibly set to 0. (Smoke risk value) Where C smoke This indicates the current concentration of particulate matter in the smoke; 100 represents the maximum permissible concentration of smoke; when C smoke When C is ≥100, the smoke risk value is 1; when C smokeWhen the value is <100, the smoke risk value is calculated proportionally. The free radical concentration parameter is converted into a deflagration risk feature vector through a chain reaction kinetic model: the value of hydroxyl radical risk is the current hydroxyl radical concentration value divided by 50 ppb; the value of hydrogen peroxide radical risk is the current hydrogen peroxide radical concentration value divided by 30 ppb.

[0085] The spatiotemporal alignment module performs timestamp synchronization and spatial coordinate matching to generate a fused feature tensor. Based on this fused feature tensor, a multimodal fire sentiment perception signal set is generated. This multimodal fire sentiment perception signal set includes a triplet of time series labels, spatial location codes, and physical field feature values. The physical field feature values ​​include normalized thermodynamic risk feature values, chemical risk feature values, and deflagration risk feature values. Outlier detection is performed on the physical field feature values. If the standard deviation of the thermodynamic risk feature value and the chemical risk feature value under a certain time series label exceeds a preset fluctuation threshold, Kalman filtering smoothing is initiated. Z-score standardization is performed on the normalized deflagration risk feature values.

[0086] S300. Construct a fire situation credibility map: Perform multimodal feature fusion and spatiotemporal alignment processing on the multimodal fire perception signal set, and combine the three-dimensional point cloud model of the building with the fire dynamics prior rule library to generate a spatially gridded fire situation credibility map; wherein, the fire situation credibility map includes a fusion quantitative index of thermodynamic threat level value, chemical chain reaction intensity and deflagration tendency coefficient.

[0087] This embodiment details the construction process of the fire situation credibility map. First, the physical field characteristic values ​​(including normalized thermodynamic risk characteristic values, chemical risk characteristic values, and deflagration risk characteristic values) in the multimodal fire perception signal set generated by S200 are spatially encoded and mapped to the corresponding grid cells of the building's 3D point cloud model. Each grid cell represents an independent spatial block (the side length is typically set to 0.5m × 0.5m × 0.5m). If the same grid cell is covered by multiple sensors (e.g., a multimodal sensor fusion cluster in a high-temperature gas accumulation area and a redundant gas sensor array in a rapid spread channel simultaneously cover a certain grid), the Deloni triangulation algorithm is used to calculate the weighted average of the characteristic values ​​within that grid. The weight allocation rule strictly follows the positive correlation between sensor deployment density and local dynamic risk probability values: for every 0.1 increase in local risk probability value, the sensor weight increases by 15%. For example, if the local risk probability value of a certain grid cell is 0.8, the weight coefficient of its internal sensors is 2.2 times the baseline value; while for a grid cell with a risk probability value of 0.3, the sensor weight is only 0.7 times the baseline value. This design ensures that monitoring data from high-risk areas carries greater weight in decision-making.

[0088] After spatial mapping is completed, a multimodal feature dynamic fusion operation is performed on each grid cell, which is implemented in three steps:

[0089] Thermodynamic threat level calculation: The thermodynamic risk feature values ​​of the grid cells are input into a pre-trained convolutional neural network model, which is trained based on a historical fire case database. The input layer receives temperature gradient distribution values, infrared radiation intensity values, and flame spectral feature values. Spatiotemporal correlation features are extracted through three convolutional layers, and the fully connected layer outputs the thermodynamic threat level value T in the interval [0,1]. threa For example, when the temperature gradient exceeds 15℃ / m and the flame spectral similarity is >90%, the output T threa A value ≥0.9 indicates an extremely high thermodynamic threat. Calculation of the chemical chain reaction strength coefficient: Input the chemical risk characteristic value and free radical concentration parameter into the chain reaction kinetic function, solve the relationship function between the free radical concentration change rate and the activation energy threshold, and output the chemical chain reaction strength coefficient.

[0090]

[0091] Among them, S chem The value represents the strength coefficient of a chemical chain reaction; k4 is the reaction rate constant, with a value of 0.25; R OH The normalized concentration of hydroxyl radicals, C OH ; This represents the normalized concentration of hydroperoxy radicals. and R represents the rate of change in free radical concentration. fuel Indicates the risk value of combustible gas; R represents the risk value of hypoxia. fuel and All are normalized values; when S chem When the value is greater than 1.2, the chain reaction is considered to be accelerated, posing a risk of flashover. This represents the geometric mean of the risks of combustible gas and oxygen deficiency, balancing their impact on the chain reaction (high combustible gas levels in high-temperature zones promote the reaction, while oxygen deficiency may inhibit it but accumulates energy leading to flashover). It is combined with the thermodynamic threat level value T. threa And the chemical chain reaction strength coefficient S chem Input the deflagration critical surface model: Among them B p σ represents the deflagration tendency coefficient; σ represents the smoothness coefficient; d M This represents the Mahalanobis distance between the current environmental parameters and the theoretical deflagration critical point. Where T c and S cThe coordinates of the deflagration critical point are defined by a priori rules base of fire dynamics. Finally, the thermodynamic threat level value, chemical chain reaction intensity coefficient, and deflagration tendency coefficient are integrated according to spatial grid units to form a three-dimensional tensor matrix; the missing data in areas where no sensors are deployed are filled by spatiotemporal interpolation algorithm to generate a spatial gridded fire situation credibility map covering the entire building's three-dimensional point cloud model.

[0092] S400. Based on the fire situation credibility map, the multi-criteria decision module is called to calculate the weight allocation matrix of the linkage equipment, and combined with the building topology model and personnel distribution heat map, a multi-level linkage control instruction set is generated; wherein, the multi-level linkage control instruction set includes equipment start-stop sequence, power regulation parameters and emergency channel activation strategy.

[0093] This embodiment generates a multi-level linkage control instruction set based on a fire situation credibility map. Its core process includes four stages: comprehensive fire threat index calculation, generation of a linkage equipment weight allocation matrix, analysis of personnel evacuation influencing factors, and dynamic instruction generation. The specific implementation process is as follows:

[0094] The thermodynamic threat level value, chemical chain reaction intensity coefficient and deflagration tendency coefficient are input into the multidimensional threat quantification function, and the normalized comprehensive fire threat index is output.

[0095] I threat =w T ·T threa +w s ·S chem +w B ·B P

[0096] Among them, W T w s and W B These are preset weights; for example, in a chemical fire scenario, w s The weighting has been increased to 0.6 (emphasizing chemical risks), for ordinary building fires. T The weight is set to 0.7 (emphasizing heat spread); w T +w s +w B =1.

[0097] An example of a fire equipment performance database is shown in Table 1;

[0098] Table 1 Example of Fire Equipment Performance Database

[0099] Equipment type Coverage radius Response delay Suppression efficiency parameter Fire sprinkler system 8m ≤3s Heat threat suppression rate: 85% Smoke control and exhaust system 12m ≤4s Chemical threat suppression rate: 75% Emergency lighting system 12m ≤1s Evacuation guidance efficiency: 90% Evacuation broadcast system 20m ≤2s Path correction rate 80%

[0100] Based on the building topology model, the connectivity between equipment deployment nodes and the threat mesh is analyzed. If a fire door or obstacle exists on the path from a device node to the mesh, the coupling is reduced by 50%. The process of dynamically generating the weight allocation matrix for linked devices using a linear weighted scoring method is as follows:

[0101] Based on expert experience or historical data analysis, fixed weight coefficients are pre-set for key evaluation indicators that influence equipment priority. These key evaluation indicators include at least:

[0102] Threat Matching Weight Coefficient: Measures the relevance of the ability of this type of fire protection equipment to suppress the dominant threat (thermodynamic threat level value, chemical chain reaction intensity coefficient, deflagration tendency coefficient) of the current grid cell.

[0103] Spatial proximity weight coefficient: measures the weight of the spatial distance (or accessibility) between the deployment node of the fire-fighting equipment and the current spatial grid cell; the closer the distance, the higher the weight.

[0104] Remaining Response Capability Weighting Coefficient: This measures the current availability status of the fire-fighting equipment (whether it is online or occupied); the higher the availability, the higher the weight.

[0105] Next, for each spatial grid cell i and each type of fire protection equipment j (a total of N types, such as sprinkler system, smoke control system, emergency lighting system, and evacuation broadcast system), calculate its initial priority score in that grid cell. Among them, T ij This represents the threat matching degree (a value between 0 and 1) of device type j in grid cell i, calculated through table lookup or simple rules. For example, when the dominant threat is a thermodynamic threat level, the sprinkler system is set to 1.0, and the smoke control system is set to 0.3; w T P represents the threat matching weight coefficient; ij This represents the spatial proximity (a value between 0 and 1) between the deployment node of device type j and the mesh cell i, based on the distance d between the device deployment node and the center of the mesh cell. ij calculate, And normalize to [0-1]; w P R represents the spatial proximity weighting coefficient; ij This represents the remaining responsiveness of device type j in grid cell i (a value between 0 and 1), calculated based on device status, obtaining a value between 0 and 1 according to device status (0 = unavailable, 1 = available) or device occupancy (1 - device occupancy); w R This represents the weighting coefficient for residual response capability.

[0106] For each spatial grid cell, the corrected priority score for each type of fire protection equipment is normalized (e.g., divided by the sum of the corrected priority scores for all equipment types in that grid cell) to obtain the final weight value for that equipment type in that grid cell. The final weight values ​​of all spatial grid cells and all fire protection equipment types are integrated to form a coordinated equipment weight allocation matrix. This matrix has dimensions [M×N], where M is the total number of spatial grid cells, N is the total number of fire protection equipment types, and the matrix elements represent the deployment weight (normalized priority score) of a specific fire protection equipment type in a specific spatial grid cell.

[0107] Real-time acquisition of population density distribution data and movement trajectory prediction data from heatmaps; calculation of theoretical evacuation time for each area based on exit locations and passage width parameters from building topology models using path planning algorithms. Where L represents the path length; v avg C represents the average speed of movement of a crowd; congestion This represents the congestion coefficient of the passageway, calculated based on the real-time population density ρ, with units of (people / m²). 2 );For example Seconds. Input personnel density, movement speed deviation, and theoretical evacuation time into the evacuation risk assessment model, and output the personnel evacuation influencing factor; E impact =β1·ρ+β2·|v actual -v theory |·+β3·t theory ; where E impact Indicates the influencing factor of personnel evacuation; |v actual -v theory | represents the deviation between the actual moving speed and the theoretical speed; β1, β2 and β3 are preset weighting coefficients, with values ​​of 0.6, 0.2 and 0.2 respectively.

[0108] If I threat >0.7 and E impact If the threshold is >0.6, the core response objective is set as the highest priority personnel evacuation; if it is 0.4... threat ≤0.7 and B p If the response continues to rise, the core response target is set as core chemical inhibition; otherwise, the core response target is set as local thermodynamic inhibition.

[0109] S500, Multi-system Collaborative Execution and Closed-Loop Optimization: The multi-level linkage control instruction set is sent to the edge computing gateway cluster through the industrial Internet of Things protocol to drive the fire sprinkler system, smoke control system, emergency lighting system and evacuation broadcast system to perform linkage operations; at the same time, the equipment response status data and fire suppression effect indicators are collected in real time, and the spatial sensitive area division rules and decision engine weight parameters are iteratively updated until the fire situation credibility converges to the safe threshold range.​

[0110] This embodiment details the multi-system collaborative execution and closed-loop optimization process. Its core lies in distributing multi-level linkage control command sets to an edge computing gateway cluster via industrial IoT protocols, driving the fire sprinkler system, smoke extraction system, emergency lighting system, and evacuation broadcast system to perform coordinated operations. Simultaneously, it collects real-time equipment response status data and fire suppression effect indicators, iteratively updating the spatially sensitive area division rules and decision engine weight parameters to dynamically optimize the fire situation credibility map until its global mean converges to a safe threshold range. The specific implementation process includes three stages: equipment response and fire suppression data acquisition, closed-loop optimization operation based on suppression deviation, and convergence condition judgment and loop termination mechanism. First, the system transmits the multi-level linkage control command set to the edge computing gateway cluster via industrial IoT protocols (such as MQTT or CoAP). This cluster is deployed at key nodes in the building and is responsible for real-time command parsing and driving fire equipment to perform coordinated operations. For example, the fire sprinkler system adjusts the nozzle opening based on the water pressure gradient value (e.g., 0.5MPa to 2.0MPa range) in the command, and the smoke extraction system adjusts the airflow based on the air volume adjustment value (e.g., 500m³ / h). 3 / h to 2000m 3 The system adjusts the fan speed, activates the LED array of the emergency lighting system according to the brightness distribution scheme (e.g., 1000 lux in the core area and 500 lux in the secondary area), and plays directional alarms based on voice guidance paths. Simultaneously, the system activates the data acquisition module to acquire real-time equipment response status data, including valve opening feedback values ​​(expressed as a percentage, e.g., 0% for closed, 100% for fully open), actual fan speed values ​​(unit: rpm, measured by a Hall sensor), measured lighting brightness values ​​(unit: lux, collected by a photosensitive sensor), and broadcast coverage intensity values ​​(unit: dB, monitored by a sound pressure meter). It also uses a differentiated sensor array to synchronously collect fire suppression effect indicators, specifically covering the temperature gradient decrease rate (calculated by distributed temperature sensors as the slope of temperature change per unit time, e.g., ℃ / min), combustible gas concentration change slope (monitored by a catalytic combustion sensor as the rate of concentration decrease, e.g., % / min), free radical concentration decay coefficient (obtained by a hydroxyl radical sensor as the concentration decay ratio), and fire spread rate change (based on an infrared thermal imaging sensor tracking the movement rate of the fire boundary). These data are sampled at a frequency of 10Hz and aligned to a unified timing reference through a timestamp synchronization module to form a spatiotemporally aligned data stream for subsequent optimization analysis.

[0111] During the closed-loop optimization phase, the system calculates the suppression deviation coefficient based on equipment response status data and fire suppression effect indicators using spatial gridding. Specifically, for each spatial grid cell in the building's 3D point cloud model, the fire suppression effect indicator is compared with the expected suppression target value; the expected suppression target value is predefined by the multi-level linkage control instruction set (e.g., the target temperature gradient decrease rate of the fire sprinkler system ≥ 20℃ / min, and the target combustible gas concentration change slope of the smoke control system ≤ -5% / min). The formula for calculating the suppression deviation coefficient is:

[0112]

[0113] Among them, E actual E represents the actual collected fire suppression effectiveness index value (such as the current temperature gradient descent rate). target The expected suppression target value (e.g., a preset 20℃ / min) is represented by Db, which is the normalized suppression deviation coefficient. If the suppression deviation coefficient of a certain spatial grid cell continues to exceed the preset tolerance threshold (e.g., 15%, based on the tolerance of equipment response errors in historical fire cases), the system marks the cell as a high-sensitivity area and increases its monitoring weight in the fire situation credibility map (e.g., the weight coefficient is increased from the baseline value of 1.0 to 1.5); this mechanism ensures that optimization resources are focused on areas with poor suppression effects. Subsequently, the system dynamically adjusts the weight parameters in the decision engine through the gradient descent algorithm; specifically targeting the weighting coefficients of the thermodynamic threat level value, chemical chain reaction intensity coefficient, and deflagration tendency coefficient in the multidimensional threat quantification function. The optimization objective is to minimize the correlation coefficient deviation between the predicted value of the comprehensive fire threat index and the actual suppression effect. During the convergence judgment phase, the system monitors the global dynamic changes of the fire situation credibility map. Within three consecutive monitoring time windows (each window lasting 5 seconds), the global mean variance reduction rate of the map is calculated. in and V represents the mean and variance of all grid cells in the fire situation confidence map for the t-th and (t-1)-th time windows, respectively. d The stability of the optimization effect was quantified. If V d If the fire threat index is less than the preset convergence threshold (e.g., 0.05, indicating a variance decrease rate of less than 5%) and the overall fire threat index is stable within the safe threshold range (e.g., 0.2 to 0.4), the system determines that the fire situation has converged and terminates the optimization loop; otherwise, it returns to step S300 to regenerate the fire situation credibility map and starts a new round of closed-loop optimization. This process is executed in real time through the edge computing gateway cluster to ensure the coordinated improvement of energy consumption balance and suppression efficiency in the dynamic fire environment until the fire is completely controlled.

[0114] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A fire alarm linkage system for fire protection engineering based on the Internet of Things, characterized in that, The system operation process specifically includes the following steps: S100. A dynamic risk probability distribution map is generated based on the three-dimensional point cloud model of the building and multi-source data. Differentiated sensor arrays are deployed in the marked high-temperature gas accumulation area, rapid spread channel and high-risk deflagration node. The generation of the dynamic risk probability distribution map is combined with the statistical association rules of the effective historical fire data in the fire association database. If there is a lack of effective historical fire data, it is achieved by combining cross-domain risk rule migration operation. S200. The deployed differentiated sensor array collects environmental multi-physics parameters in real time and generates a multimodal fire perception signal set. S300. Constructing a Fire Situation Credibility Map: Perform multimodal feature fusion and spatiotemporal alignment processing on the multimodal fire perception signal set, and combine the three-dimensional point cloud model of the building with the fire dynamics prior rule library to generate a spatially gridded fire situation credibility map; wherein, the fire situation credibility map includes a fusion quantitative index of thermodynamic threat level value, chemical chain reaction intensity and deflagration tendency coefficient; S400. Based on the fire situation credibility map, the multi-criteria decision module is called to calculate the weight allocation matrix of the linkage equipment, and combined with the building topology model and personnel distribution heat map, a multi-level linkage control instruction set is generated; wherein, the multi-level linkage control instruction set includes equipment start-stop sequence, power regulation parameters and emergency channel activation strategy. S500, Multi-system Collaborative Execution and Closed-Loop Optimization: The multi-level linkage control instruction set is sent to the edge computing gateway cluster through the industrial Internet of Things protocol to drive the fire sprinkler system, smoke control system, emergency lighting system and evacuation broadcast system to perform linkage operations; at the same time, the equipment response status data and fire suppression effect indicators are collected in real time, and the spatial sensitive area division rules and decision engine weight parameters are iteratively updated until the fire situation credibility converges to the safe threshold range.

2. The fire alarm linkage system for fire protection engineering based on the Internet of Things as described in claim 1, characterized in that: S100 specifically includes the following steps: Based on the spatial topological features of the building's 3D point cloud model and the statistical association rules of valid historical fire data in the fire association database, a dynamic risk probability distribution map is generated; wherein, the dynamic risk probability distribution map contains the mapping relationship between spatial coordinate positions and fire occurrence probability values; High-temperature gas accumulation areas, rapid spread channels, and high-risk deflagration nodes are marked on the dynamic risk probability distribution map. A multimodal sensor fusion cluster is deployed in the high-temperature gas accumulation zone; wherein the multimodal sensor fusion cluster includes a combination of distributed temperature sensors, infrared thermal imaging sensors, and flame spectral sensors, and satisfies that the spatial deployment density is positively correlated with the local risk probability value; a redundant gas sensor array is deployed at the key nodes of the fluid channel in the rapid propagation channel; wherein the redundant gas sensor array includes at least two sets of heterogeneous combustible gas concentration sensors, and the deployment orientation covers the normal vector direction of the key nodes of the fluid channel; an explosion-proof pressure sensor array is deployed at the high-risk deflagration node; wherein the explosion-proof pressure sensor array includes pressure surge sensors and shock wave sensors, and the deployment location satisfies that the Euclidean distance from the potential deflagration center is less than a preset safety threshold.

3. A fire alarm linkage system for fire protection engineering based on the Internet of Things as described in claim 2, characterized in that: Generating a dynamic risk probability distribution map includes the following steps: If there is no valid historical fire data in the fire association database, a theoretical risk probability distribution map is generated by multiphysics coupling numerical simulation based on the geometric constraint features and structural thermal features of the building's three-dimensional point cloud model. The geometric constraint features include volume curvature distribution features and ventilation cross-sectional area ratio features, and the structural thermal features include material thermal conductivity distribution features and combustible material distribution density features. Perform cross-domain risk rule migration operations: extract spatial probability density functions from electrical equipment fault location distribution data in non-fire-related databases and map them to fire location probability distribution features; extract turbulence intensity parameters from fluid dynamic diffusion models in non-fire-related databases and convert them into fire spread rate distribution features; generate migration risk probability distribution maps based on migration mapping rules; The theoretical risk probability distribution map and the migration risk probability distribution map are integrated using a Bayesian-physical fusion modeling framework.

4. A fire alarm linkage system for fire protection engineering based on the Internet of Things as described in claim 3, characterized in that: The Bayesian-physical fusion modeling framework performs the following operations: Use the theoretical risk probability distribution as the prior probability distribution; Use the migration risk probability distribution map as the input to the likelihood function; The dynamic risk probability distribution map is generated by calculating the posterior probability distribution through Markov chain Monte Carlo sampling.

5. A fire alarm linkage system for fire protection engineering based on the Internet of Things as described in claim 1, characterized in that: S200 specifically includes the following steps: The environmental multiphysics parameters are collected in real time by the deployed differential sensor array; wherein the environmental multiphysics parameters include thermodynamic parameters, gas composition parameters and free radical concentration parameters. Multi-level signal conversion and fusion processing is performed on the environmental multi-physics parameters to generate standardized physical field feature vectors; wherein, the multi-level signal conversion and fusion processing includes: The thermodynamic parameters are converted into thermodynamic risk feature vectors using a nonlinear temperature field mapping model; the gas composition parameters are converted into chemical risk feature vectors using a gas diffusion coupling algorithm; and the free radical concentration parameters are converted into deflagration risk feature vectors using a chain reaction kinetic model. The thermodynamic risk feature vector, chemical risk feature vector and deflagration risk feature vector are time-stamped and matched with spatial coordinates by the spatiotemporal alignment module to generate a fused feature tensor. A multimodal fire perception signal set is generated based on the fused feature tensor; wherein the multimodal fire perception signal set includes a triplet of time series labels, spatial location codes, and physical field feature values, and the physical field feature values ​​include normalized thermodynamic risk feature values, chemical risk feature values, and deflagration risk feature values.

6. A fire alarm linkage system for fire protection engineering based on the Internet of Things as described in claim 1, characterized in that: S300 specifically includes the following steps: The physical field feature values ​​are encoded and mapped to the corresponding grid cells of the building's 3D point cloud model according to the spatial location. For grid cells covered by multiple sensors, the weighted average of the feature values ​​is calculated based on the Deloni triangulation algorithm. The weight allocation follows the positive correlation rule between sensor spatial deployment density and local risk probability. Perform dynamic fusion of multimodal features: calculate the thermodynamic threat level, chemical chain reaction intensity coefficient, and deflagration tendency coefficient for each grid cell; Thermodynamic threat level values, chemical chain reaction intensity coefficients, and deflagration tendency coefficients are integrated into a three-dimensional tensor matrix based on spatial grid units. The missing data in areas where no sensors are deployed are filled in by spatiotemporal interpolation algorithms to generate a spatial gridded fire situation reliability map covering the entire building's three-dimensional point cloud model.

7. A fire alarm linkage system for fire protection engineering based on the Internet of Things as described in claim 6, characterized in that: The process of performing dynamic fusion of multimodal features specifically includes the following steps: The thermodynamic risk feature value is input into a pre-trained convolutional neural network model, which outputs a thermodynamic threat level value in the interval [0,1]. The convolutional neural network model is trained by the mapping relationship between temperature gradient distribution values ​​and actual fire spread speed in historical fire cases. The chemical risk characteristic value and free radical concentration parameter are input into the chain reaction kinetic function to solve the relationship function between the free radical concentration change rate and the activation energy threshold, and the chemical chain reaction strength coefficient is output. The thermodynamic threat level and chemical chain reaction intensity coefficient are input into the deflagration critical surface model to calculate the Mahalanobis distance between the current environmental parameters and the theoretical deflagration critical point, which is mapped to the deflagration tendency coefficient in the [0,1] interval.

8. A fire alarm linkage system for fire protection engineering based on the Internet of Things according to claim 1, characterized in that, The S400 specifically includes the following steps: The thermodynamic threat level, chemical chain reaction intensity coefficient, and deflagration tendency coefficient of each spatial grid cell in the fire situation confidence map are extracted and input into a pre-constructed multidimensional threat quantification function to output a normalized comprehensive fire threat index; wherein, the multidimensional threat quantification function is trained by the mapping relationship between fire parameters and actual disaster losses in historical fire cases; The multi-criteria decision-making module is invoked to match the comprehensive fire threat index with a preset fire equipment efficiency database. This database stores the coverage radius, response delay time, and suppression efficiency parameters of fire sprinkler systems, smoke extraction systems, emergency lighting systems, and evacuation broadcast systems. Based on the matching results and the connectivity constraints of the building topology model, the spatial coupling degree between equipment deployment nodes is calculated. A linear weighted scoring method is used to dynamically generate a linkage equipment weight allocation matrix, which includes the priority score of each fire equipment within the spatial grid unit. The system acquires real-time data on population density distribution and movement trajectory prediction from a heat map of population distribution; calculates the theoretical evacuation time for each area based on the exit location and passage width parameters of the building topology model using a path planning algorithm; inputs population density, movement speed deviation, and theoretical evacuation time into an evacuation risk assessment model, and outputs a population evacuation influencing factor; wherein, the population evacuation influencing factor is positively correlated with the urgency of the linkage strategy. The weight allocation matrix of the linked equipment and the personnel evacuation influencing factors are input into the dynamic instruction generation module.

9. A fire alarm linkage system for fire protection engineering based on the Internet of Things as described in claim 8, characterized in that: After the weight allocation matrix of the linkage equipment and the personnel evacuation influencing factor are input into the dynamic instruction generation module, the dynamic instruction generation module performs the following operations: If the overall fire threat index is greater than the preset high-risk threshold and the personnel evacuation impact factor exceeds the critical value, then the core response objective is set as the highest priority personnel evacuation. If the overall fire threat index is in the medium threat range and the deflagration tendency coefficient continues to rise, then the core response objective is set as core chemical suppression. Otherwise, the core response target is set as local thermodynamic suppression; Based on the core response target and the priority score in the weight allocation matrix of the linked equipment, a multi-level linkage control instruction set is generated, which includes equipment start-stop sequence, power control parameters, and emergency passage activation strategy. The equipment start-stop sequence specifies the spatiotemporal order of activation of the fire sprinkler system, smoke control system, emergency lighting system, and evacuation broadcast system. The power control parameters include the water pressure gradient value of the sprinkler system and the air volume adjustment value of the smoke control system. The emergency passage activation strategy includes the lighting intensity distribution scheme of the emergency lighting system and the voice guidance path of the evacuation broadcast system.

10. A fire alarm linkage system for fire protection engineering based on the Internet of Things as described in claim 1, characterized in that: The S500 system simultaneously collects real-time data on device response status and fire suppression effectiveness indicators, iteratively updates spatially sensitive area delineation rules and decision engine weight parameters until the fire situation credibility converges to a safe threshold range; including: Real-time acquisition of equipment response status data, including valve opening feedback values, actual fan speed values, measured lighting brightness values, and broadcast coverage intensity values; real-time acquisition of fire suppression effect indicators through the differentiated sensor array; wherein, the fire suppression effect indicators include temperature gradient descent rate, combustible gas concentration change slope, free radical concentration attenuation coefficient, and fire spread rate change; based on the equipment response status data and the fire suppression effect indicators, the following closed-loop optimization operations are performed: The fire suppression effect index is compared with the expected suppression target value, and the suppression deviation coefficient of the spatial grid is calculated. If the suppression deviation coefficient of a certain spatial grid cell is continuously greater than the preset tolerance threshold, the cell is marked as a high-sensitivity area and its monitoring weight in the fire situation credibility map is increased. The weighting coefficients of the thermodynamic threat level, chemical chain reaction intensity coefficient, and deflagration tendency coefficient in the multidimensional threat quantification function are adjusted by the gradient descent algorithm to maximize the correlation coefficient between the predicted value of the comprehensive fire threat index and the actual suppression effect. If the global mean variance of the fire situation confidence map decreases less than the preset convergence threshold and the comprehensive fire threat index remains stable within the safe threshold range for three consecutive monitoring time windows, then the loop terminates; otherwise, return to S300.

Images