Subway fire safety monitoring and early warning method and system under multi-data fusion

Abstract

The invention relates to the related technical field of safety monitoring and early warning, in particular to a subway fire safety monitoring and early warning method and system under multi-data fusion, and the method comprises the steps: constructing a risk evolution graph through a first fire characteristic index and a second fire characteristic index; according to the method, initial fire source heat release rates under different combustible load densities are taken as core correlation characteristics, a fire influence range is subjected to multi-scale evaluation, and a dynamic early warning threshold adjustment strategy is used for graded warning, so that the problem of lack of systematic risk modeling taking a fire development stage as a core is solved; according to the method, the technical problems of poor potential risk area pre-judgment and chain failure influence analysis capability and difficulty in meeting subway fire prevention and control requirements in the prior art are solved, multi-dimensional monitoring data are integrated, a risk evolution graph taking a fire development stage as a node and heat release rate and smoke concentration time sequence coupling as a constraint is constructed, a fire spreading track is dynamically tracked, and the fire prevention and control requirements are met. And the reliability and systematicness of subway fire prevention and control are comprehensively enhanced.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring and early warning technology, specifically to a method and system for monitoring and warning of subway fire safety under multi-data fusion. Background Technology

[0002] Subways are characterized by large passenger flow, enclosed spaces, complex evacuation routes, and dense equipment. Once a fire occurs, the fire can spread rapidly and the smoke can be difficult to disperse, which can easily cause casualties and significant property damage, posing a serious threat to urban public safety. Fire safety monitoring and early warning is the core link of the subway operation safety assurance system. Its early warning accuracy, response speed, and scientific risk assessment directly determine the efficiency of fire response and the effectiveness of personnel evacuation.

[0003] Current methods for monitoring and early warning of fire safety in subways often rely on isolated monitoring data such as smoke and temperature. This results in one-sided setting of fire characteristic indicators, which can easily lead to false alarms and missed alarms. Furthermore, they lack risk modeling centered on the stages of fire development, making it impossible to dynamically track the fire spread trajectory and risk escalation process. They also lack sufficient prediction of potential risk areas and analysis of the impact of chain failures, making it difficult to meet the needs of subway fire prevention and control.

[0004] In summary, existing technologies suffer from a lack of systematic risk modeling centered on the fire development stages, weak capabilities in predicting potential risk areas and analyzing the impact of cascading failures, and are therefore unable to meet the technical requirements for subway fire prevention and control. Summary of the Invention

[0005] This application provides a method and system for monitoring and early warning of subway fire safety through multi-data fusion, aiming to solve the technical problems in the existing technology, such as the lack of systematic risk modeling centered on the fire development stage, weak ability to predict potential risk areas and analyze the impact of chain failures, which makes it difficult to meet the needs of subway fire prevention and control.

[0006] In view of the above problems, the technical solution to achieve the present application is as follows:

[0007] In its first aspect, this application provides a method for monitoring and early warning of subway fire safety under multi-data fusion. The method includes: receiving multi-dimensional monitoring data based on subway lines in operation, comparing this data with the subway's safe operating conditions, and setting a first fire characteristic index and a second fire characteristic index; constructing a risk evolution map using the first and second fire characteristic indices, with fire development stages as nodes and the temporal coupling relationship between heat release rate and smoke concentration as edge constraints; and, based on the smoke spread inhibition factor and using the initial fire source heat release rate under different combustible load densities as the core correlation feature, conducting a multi-scale assessment of the fire impact range, and using a dynamic early warning threshold adjustment strategy for graded alarms.

[0008] In one possible implementation, passenger flow density logs are associated with each area node of the subway's safe operation condition, and a first fire characteristic indicator is set; smoke sensor and carbon monoxide detector records are associated with each equipment node of the subway's safe operation condition, and a second fire characteristic indicator is set.

[0009] In possible implementations, a video flame recognition unit, a gas monitoring unit, and a ventilation status detection unit are integrated to construct a dynamic environmental perception network and acquire subway operating environment data. Based on the equipment-dense sub-sections associated with the risk impact coefficient, and combined with the subway operating environment data, an environment-risk coupling matrix is ​​constructed to dynamically correct the smoke spread inhibition factor in the risk evolution map.

[0010] In a possible implementation, the rows of the environment and risk coupling matrix correspond to environmental risk factor categories, including ventilation efficiency, combustible gas concentration, and ambient temperature rise rate; the columns of the environment and risk coupling matrix correspond to spatial risk areas, including combustible material sub-areas in station halls, densely packed cable sub-areas in tunnels, and high-temperature sub-areas in equipment rooms; the environment and risk coupling matrix is ​​quantified and dynamically updated, specifically including: determining the smoke suppression coefficient of ventilation efficiency for each area based on the smoke diffusion rate of the video flame recognition unit; determining the fire ignition factor of combustible gas for each area based on the carbon monoxide concentration gradient of the gas monitoring unit; determining the fire spread acceleration coefficient of ambient temperature rise rate for each area based on the ambient temperature time-series data of the ventilation status detection unit; for the first spatial risk area, the risk potential energy is superimposed using the smoke suppression coefficient, fire ignition factor, and fire spread acceleration coefficient to obtain the first comprehensive risk impact factor; based on the first spatial risk area and the first comprehensive risk impact factor, each spatial risk area is traversed to obtain the comprehensive risk impact factor of each spatial risk area; the comprehensive risk impact factor of each spatial risk area is used as the corresponding element value of the environment and risk coupling matrix for dynamic updating.

[0011] In a possible implementation, a sliding time window is used to determine the directed edges between adjacent fire development stage nodes in the risk evolution map, connecting each fire evolution stage; and the preceding causal relationships, time-dependent parameters, and multi-source signal correlation information between each fire evolution stage are labeled.

[0012] In a possible implementation, the directed edges of the risk evolution graph are dynamically weighted, and the pre-cause relationships, time-dependent parameters and multi-source signal correlation information between each fire development stage are fused based on LSTM time-series perception. The smoke spread inhibition factor is then output to the ventilation control system, and the smoke spread inhibition factor is used to dynamically adjust the opening of the air valve.

[0013] In a possible implementation, a time-series feature analysis is performed using a start threshold and a stop threshold, which are used to mark the smoldering stage, the initial open flame stage, and the flashover risk stage; the dynamic early warning threshold adjustment strategy is configured based on the time-series relationship corresponding to the start threshold and the stop threshold.

[0014] In possible implementations, a subway fire early warning knowledge graph is established by associating fire types, historical fire cases, and the starting and ending thresholds; based on the subway fire early warning knowledge graph, potential risk areas are identified, and a topological expansion analysis of the fire impact range is performed.

[0015] In one possible implementation, an attention-based graph convolutional network is set up, with input node features including historical fire alarm frequency, equipment aging index, and combustible material accumulation frequency, to obtain fire propagation weights between each fire development stage; based on the fire propagation weights between each fire development stage, key channels for smoke diffusion and fire spread paths are determined; based on the subway fire early warning knowledge graph and key channels for smoke diffusion, the fire spread paths are used to simulate the process of obstructed personnel evacuation and equipment cascading failures, to determine the potential risk areas.

[0016] In a second aspect, this application provides a subway fire safety monitoring and early warning system based on multi-data fusion. The system includes: a multi-dimensional monitoring data receiving module: receiving multi-dimensional monitoring data based on subway lines in operation, comparing it with subway safety operation conditions, and setting a first fire characteristic indicator and a second fire characteristic indicator; a risk evolution map construction module: constructing a risk evolution map with fire development stages as nodes and the temporal coupling relationship between heat release rate and smoke concentration as edge constraints using the first and second fire characteristic indicators; and a graded alarm module: in the risk evolution map, based on the smoke spread inhibition factor and with the initial fire source heat release rate under different combustible load densities as the core correlation feature, performing multi-scale assessment of the fire impact range, and using a dynamic early warning threshold adjustment strategy for graded alarms.

[0017] In summary, one or more technical solutions provided in this application integrate multi-dimensional monitoring data to construct a risk evolution map with fire development stages as nodes and heat release rate and smoke concentration time-series coupling as constraints, dynamically track the fire spread trajectory, adapt to different fire development stages, and comprehensively enhance the reliability and systematic technical effect of subway fire prevention and control. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0019] Figure 1 This application provides a flowchart illustrating a method for monitoring and early warning of subway fire safety under multi-data fusion.

[0020] Figure 2 This application provides a structural schematic diagram of a subway fire safety monitoring and early warning system based on multi-data fusion.

[0021] Figure labeling: Multi-dimensional monitoring data receiving module M100, risk evolution map construction module M200, hierarchical alarm module M300. Detailed Implementation

[0022] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0023] Example 1: The present application will be described in detail below with reference to the accompanying drawings, as follows... Figure 1 As shown, this application provides a method for monitoring and early warning of subway fire safety under multi-data fusion, wherein the method includes:

[0024] S1: Based on the subway lines in operation, receive multi-dimensional monitoring data and compare it with the subway's safe operation conditions to set a first fire characteristic indicator and a second fire characteristic indicator; S2: Based on the first fire characteristic indicator and the second fire characteristic indicator, construct a risk evolution map with the fire development stage as the node and the temporal coupling relationship between heat release rate and smoke concentration as the edge constraint.

[0025] Specifically, multi-dimensional monitoring data refers to data from different types of sensors and monitoring equipment, including temperature sensors, smoke sensors, carbon monoxide detectors, and passenger flow density logs, reflecting the real-time status of the subway environment from multiple perspectives; subway safe operation conditions refer to the baseline data of various environmental parameters and equipment operating status during normal subway operation, such as temperature change curves and smoke concentration ranges during normal operation; the first fire characteristic index and the second fire characteristic index are key parameters used to characterize fire characteristics, extracted based on the comparison of multi-dimensional monitoring data and subway safe operation conditions.

[0026] Fire development stages refer to the various stages from the onset to the extinguishing of a fire, including the smoldering stage, the initial open flame stage, and the flashover risk stage; heat release rate refers to the amount of heat released by a fire per unit time, which is an important indicator for measuring fire intensity; the temporal coupling relationship of smoke concentration refers to the relationship between smoke concentration and time, and the correlation between this change and the heat release rate; the risk evolution map is a model used to describe the evolution of fire risk over time, with fire development stages as nodes and the temporal coupling relationship between heat release rate and smoke concentration as edge constraints, which can dynamically reflect the changing trend of fire risk.

[0027] Execution steps: Based on the operational status of the subway line, collect multi-dimensional monitoring data from various sensors such as distributed temperature sensing components, smoke sensors, and carbon monoxide detectors. Compare this monitoring data with safe operating conditions such as temperature rise benchmarks and smoke concentration benchmarks under normal subway operation, and extract deviation information from standard temperature rise curves and standard smoke concentration curves. Based on this deviation information, combined with passenger flow density logs and equipment node monitoring data, set a first fire characteristic index and a second fire characteristic index. For example, the temperature change curve during normal operation is the benchmark curve. When the distributed temperature sensing component detects that the temperature rise rate in a certain area deviates significantly from the benchmark curve, the first fire characteristic index is set based on the passenger flow density log of that area. At the same time, if the smoke concentration and carbon monoxide concentration recorded by the smoke sensor and carbon monoxide detector also show abnormalities, the second fire characteristic index is set based on the equipment node monitoring data.

[0028] Using two fire characteristic indicators, the first and second, a risk evolution map is constructed. Fire development stages are used as nodes, such as dividing fires into smoldering, initial open flame, and flashover risk stages. The temporal coupling relationship between heat release rate and smoke concentration is used as edge constraints. Specifically, the connection relationships and weights between nodes are determined based on the changes in heat release rate and smoke concentration over time at different fire stages. Furthermore, in the smoldering stage, the heat release rate is low, and the smoke concentration gradually increases, resulting in lower connection weights between nodes. As the fire progresses to the initial open flame stage, the heat release rate increases sharply, and the smoke concentration also rises rapidly, leading to a corresponding increase in connection weights between nodes. The constructed risk evolution map can dynamically track the fire spread trajectory and adapt to different fire development stages.

[0029] S3: In the risk evolution map, based on the smoke spread inhibition factor, the initial fire source heat release rate under different combustible load densities is used as the core correlation feature to conduct multi-scale assessment of the fire impact range, and a dynamic early warning threshold adjustment strategy is used for graded alarms.

[0030] Specifically, the smoke spread inhibition factor refers to the parameter used to control the speed and range of smoke diffusion during a fire through ventilation systems or other means. Its value is usually related to factors such as the opening degree of the ventilation system's dampers and the wind speed. Furthermore, by adjusting the ventilation system, the spread of smoke can be effectively controlled, thereby mitigating the impact of the fire on other areas. Combustible material load density refers to the mass of combustible material per unit area or unit volume, which directly affects the intensity and spread speed of a fire. Furthermore, the combustible material load density varies in different areas; for example, the distribution and density of combustible materials in areas such as station halls, tunnels, and equipment rooms have their own characteristics.

[0031] The initial heat release rate (IPR) refers to the rate at which heat is released from the fire source in the early stages of a fire. It is an important indicator for measuring fire intensity and is closely related to the speed of fire development and the scope of its impact. The higher the IPR, the faster the fire spreads. Multi-scale assessment refers to evaluating the scope of fire impact from different spatial scales to comprehensively understand the potential hazards caused by the fire and to more accurately determine the potential impact area. Furthermore, different spatial scales include local areas, the entire station, and sections of the line. The dynamic warning threshold adjustment strategy refers to adjusting the warning threshold based on the dynamic development of the fire and real-time monitoring data. Compared with conventional fixed threshold warnings, dynamic warning thresholds can better adapt to changes in fire scenarios and improve the accuracy and timeliness of warnings.

[0032] Implementation steps: Introduce a smoke spread inhibition factor into the risk evolution map, and dynamically adjust the opening of the air valves through the ventilation control system to control the spread rate and range of smoke. Specifically, when a fire occurs, the ventilation system adjusts the opening of the air valves according to the smoke spread inhibition factor to reduce the spread rate of smoke in a specific area, thereby mitigating the impact of the fire on other areas. Using the initial ignition heat release rate under different combustible load densities as the core of the assessment, a multi-scale assessment of the fire's impact range is conducted. Specifically, based on the combustible load density and initial ignition heat release rate of the fire-occurring area, combined with the fire development stage information in the risk evolution map, the impact range of the fire at different spatial scales, such as local areas, the entire station, and the track section, is assessed.

[0033] For example, if a fire occurs in an equipment room with a high density of combustible materials, the initial heat release rate of the ignition source is high, and the fire may spread rapidly to adjacent areas. In this case, the focus of the assessment is the fire impact range of the equipment room and its surrounding areas. If a fire occurs in the station hall, the density of combustible materials is low, the initial heat release rate of the ignition source is relatively low, and the fire spreads slowly. In this case, the focus of the assessment is the fire impact range inside the station hall and evacuation routes.

[0034] Based on the multi-scale assessment results, a dynamic early warning threshold adjustment strategy is used for tiered alarms. The dynamic early warning threshold is adjusted according to real-time data on the fire development stage, smoke spread inhibition factor, and fire impact range. Specifically, during the smoldering stage, the early warning threshold is set lower, and a low-level alarm is issued when the monitoring data approaches the threshold, reminding staff to pay attention to the fire risk. As the fire enters the initial open flame stage, the early warning threshold is raised accordingly, and a high-level alarm is issued when the monitoring data exceeds the threshold, prompting emergency evacuation and firefighting measures. By dynamically adjusting the early warning threshold, changes in fire risk can be reflected more accurately, enabling precise assessment and dynamic early warning of the fire impact range, avoiding false alarms and missed alarms, and improving the initiative and reliability of subway fire prevention and control.

[0035] Furthermore, by setting a first fire characteristic indicator and a second fire characteristic indicator, the method of this application includes:

[0036] The passenger flow density logs are associated with each area node of the subway's safe operation condition to set a first fire characteristic indicator; the smoke sensor and carbon monoxide detector records are associated with each equipment node of the subway's safe operation condition to set a second fire characteristic indicator.

[0037] Specifically, passenger flow density logs refer to data logs that record the number and distribution of passengers in different areas of the subway at different time periods. They are usually obtained through video surveillance, turnstile records or other sensors and can reflect the density of people in subway stations and carriages. The various area nodes of the subway safe operation conditions refer to areas in the subway system with specific functions and safety requirements, such as station halls, platforms, tunnel sections and equipment rooms. Each area node has its own range of safe parameters during normal operation, such as temperature and smoke concentration.

[0038] The first fire characteristic indicator is a key parameter used to characterize fire risk, set after correlating passenger flow density logs with the subway's safe operation conditions. It mainly reflects the fire risk characteristics of densely populated areas. The smoke sensor and carbon monoxide detector records refer to the real-time smoke and carbon monoxide concentration data detected by these two sensors in the subway environment. These are important characteristic parameters when a fire occurs, reflecting the initial signs and severity of the fire. The second fire characteristic indicator is a key parameter used to characterize fire risk, set after correlating smoke sensor and carbon monoxide detector records with the subway's safe operation conditions. It mainly reflects the fire risk characteristics of equipment areas.

[0039] Execution steps: Associate passenger flow density logs with various regional nodes of the subway's safe operation conditions. Specifically, by analyzing passenger flow density logs in areas such as station halls and platforms, and combining them with the safety parameter ranges of these areas during normal operation, set the first fire characteristic indicator. In particular, if the passenger flow density in a certain area is much higher than the normal level, and the temperature or smoke concentration is slightly abnormal, it indicates that there is a high fire risk in that area. At this time, these abnormal data are used as the first fire characteristic indicator.

[0040] The smoke sensors and carbon monoxide detectors are linked to various equipment nodes in the subway's safe operation. Specifically, in densely populated equipment areas such as equipment rooms and tunnel sections, smoke and carbon monoxide concentration data are analyzed and combined with the normal operating parameter ranges of these areas to set a second fire characteristic indicator. If the smoke or carbon monoxide concentration in a certain equipment area exceeds the normal range, it may indicate that there is a fire hazard in that area. In this case, these abnormal data are used as the second fire characteristic indicator.

[0041] By analyzing the correlation of multi-dimensional data, fire characteristic indicators are set to improve the accuracy and reliability of fire early warning. The first fire characteristic indicator, based on passenger flow density logs, can effectively reflect the fire risk in densely populated areas and avoid false alarms or missed alarms caused by crowds. Furthermore, in the station hall area during peak hours, where passenger flow density is high, the combination of passenger flow density data can more accurately determine the fire risk and issue early warnings in a timely manner if abnormal temperature or slight smoke occurs.

[0042] Meanwhile, the second fire characteristic indicator, based on records from smoke sensors and carbon monoxide detectors, can accurately capture fire signs in equipment areas, ensuring timely response to fire risks in densely populated equipment areas. Specifically, in equipment rooms, abnormal changes in smoke and carbon monoxide concentrations may indicate fires caused by electrical equipment malfunctions. By setting the second fire characteristic indicator, fire hazards can be quickly identified and measures taken to prevent the fire from spreading. Setting the first and second fire characteristic indicators can comprehensively cover the fire risk characteristics of different areas in the subway environment, effectively improving the initiative and reliability of subway fire prevention and control.

[0043] Furthermore, this application's method also includes constructing a risk evolution map with fire development stages as nodes and the temporal coupling relationship between heat release rate and smoke concentration as edge constraints:

[0044] The system integrates a video flame recognition unit, a gas monitoring unit, and a ventilation status detection unit to construct a dynamic environmental perception network and acquire subway operating environment data. Based on the equipment-dense sub-sections associated with the risk impact coefficient, and combined with the subway operating environment data, an environment-risk coupling matrix is ​​constructed to dynamically correct the smoke spread inhibition factor in the risk evolution map.

[0045] Specifically, the video flame recognition unit refers to a device that uses a video surveillance system combined with a flame recognition algorithm to detect the presence of flames in the subway environment in real time. It can quickly identify flame characteristics, thereby promptly detecting early signs of fire. Furthermore, flame characteristics include color, shape, and flashing frequency. The gas monitoring unit refers to a sensor device used to detect the concentration of various gas components in the subway environment. It can monitor changes in gas concentration in real time. Furthermore, various gas components in the subway environment include smoke, carbon monoxide, and carbon dioxide.

[0046] A ventilation status detection unit refers to equipment used to monitor the operating status of the subway ventilation system, including parameters such as valve opening, wind speed, and wind direction. The ventilation status directly affects the diffusion and spread of smoke. A dynamic environmental sensing network is a comprehensive monitoring network composed of video flame recognition units, gas monitoring units, and ventilation status detection units, which can acquire multi-dimensional data on the subway operating environment in real time. A risk impact coefficient is a quantitative indicator that measures the degree of fire risk impact in different areas. It is usually related to factors such as the area's combustible load density, personnel density, and equipment importance, and is used to assess the potential hazards of a fire to a specific area.

[0047] Equipment-dense sub-sections refer to areas in the subway environment where equipment is concentrated, such as equipment rooms and cable wells. These areas typically have a higher fire risk because the high density of equipment can lead to rapid fire spread. The environment-risk coupling matrix is ​​used to describe the dynamic correlation between subway operating environment data and fire risk. By quantifying the impact of environmental changes on fire risk through the environment-risk coupling matrix, dynamic assessment and correction of fire risk can be achieved. Dynamic correction refers to updating and adjusting the smoke spread inhibition factor in the risk evolution map in real time based on real-time monitoring data and environmental changes to ensure the accuracy and timeliness of risk assessment.

[0048] Execution steps: Integrate video flame recognition units, gas monitoring units, and ventilation status detection units to construct a dynamic environmental perception network. These units are distributed in various key areas of the subway, such as station halls, platforms, tunnel sections, and equipment rooms. Specifically, the video flame recognition unit can quickly identify the appearance of flames, the gas monitoring unit can detect changes in smoke and harmful gas concentrations, and the ventilation status detection unit can monitor the operating status of the ventilation system, including the opening degree of air valves and wind speed. These data together constitute a real-time profile of the subway operating environment.

[0049] Based on the risk impact coefficient, a coupling matrix between environment and risk is constructed for equipment-dense sub-areas and combined with subway operation environment data. Specifically, risk impact coefficients are calculated according to the characteristics of equipment-dense sub-areas, and these coefficients are combined with real-time monitored environmental data. The characteristics of equipment-dense sub-areas include combustible load density and equipment importance. For example, in the equipment room area, if the gas monitoring unit detects an increase in smoke concentration and the ventilation status detection unit shows that the air valve opening is insufficient, the fire risk level of the area can be calculated by combining the risk impact coefficient of the area.

[0050] The smoke spread inhibition factor in the risk evolution map is dynamically corrected by using the environment and risk coupling matrix. Furthermore, if the environment and risk coupling matrix shows that the risk level of a certain equipment-dense sub-area increases, it may be due to poor ventilation leading to smoke accumulation. At this time, the dynamic environmental perception network will feed this information back to the risk evolution map and dynamically adjust the smoke spread inhibition factor, thereby more accurately assessing the potential impact range and risk level of the fire.

[0051] By utilizing a dynamic environmental perception network and an environment-risk coupling matrix, real-time dynamic assessment and correction of fire risks are achieved. This dynamic correction mechanism can adjust the smoke spread inhibition factor in the fire risk assessment model in a timely manner based on real-time monitoring data and environmental changes, thereby improving the accuracy and timeliness of fire early warning. Furthermore, when ventilation system malfunctions or anomalies occur in densely equipped areas, the dynamic correction of the smoke spread inhibition factor can more accurately predict the smoke diffusion path and the potential impact range of the fire, providing a more scientific basis for fire emergency response and further enhancing the reliability and systematic nature of subway fire prevention and control.

[0052] Furthermore, the method of this application also includes constructing an environment-risk coupling matrix and dynamically correcting the flue gas spread inhibition factor in the risk evolution map.

[0053] The rows of the environment and risk coupling matrix correspond to environmental risk factor categories, including ventilation efficiency, combustible gas concentration, and ambient temperature rise rate; the columns of the environment and risk coupling matrix correspond to spatial risk areas, including combustible material sub-areas in station halls, densely packed cable sub-areas in tunnels, and high-temperature sub-areas in equipment rooms. The environment and risk coupling matrix is ​​quantified and dynamically updated, specifically including: determining the smoke suppression coefficient of ventilation efficiency for each area based on the smoke diffusion rate of the video flame recognition unit; determining the fire ignition coefficient of combustible gas for each area based on the carbon monoxide concentration gradient of the gas monitoring unit; determining the fire spread acceleration coefficient of ambient temperature rise rate for each area based on the ambient temperature time-series data of the ventilation status detection unit; for the first spatial risk area, the risk potential energy is superimposed using the smoke suppression coefficient, fire ignition coefficient, and fire spread acceleration coefficient to obtain the first comprehensive risk impact factor; based on the first spatial risk area and the first comprehensive risk impact factor, each spatial risk area is traversed to obtain the comprehensive risk impact factor for each spatial risk area; the comprehensive risk impact factor of each spatial risk area is used as the corresponding element value of the environment and risk coupling matrix for dynamic updating.

[0054] Specifically, environmental risk factor categories refer to the categories of environmental factors that affect the fire risk of subways, including ventilation efficiency, combustible gas concentration, and ambient temperature rise rate, reflecting the key driving factors for fire development in the subway environment; spatial risk areas refer to areas in the subway system with different fire risk characteristics, including combustible material sub-sections in station halls, densely packed cable sub-sections in tunnels, and high-temperature sub-sections in equipment rooms, which have different fire risk levels due to their functions and environmental characteristics; matrix element quantification refers to quantifying the degree of influence of environmental risk factors on each spatial risk area and converting it into specific numerical values.

[0055] Dynamic updates refer to the continuous updating of element values ​​in the environment and risk coupling matrix based on real-time monitoring data to reflect the current fire risk status; the smoke suppression coefficient is a quantitative indicator of the effect of ventilation efficiency on the suppression of smoke diffusion in each area. The higher the ventilation efficiency, the larger the smoke suppression coefficient, indicating that the ventilation system has a better effect on suppressing smoke diffusion; the fire growth coefficient is a quantitative indicator of the effect of combustible gas concentration on the growth of fire. The higher the combustible gas concentration, the larger the fire growth coefficient, indicating that the fire develops faster.

[0056] The fire spread acceleration coefficient is a quantitative indicator of the accelerating effect of the ambient temperature rise rate on fire spread. The higher the temperature rise rate, the larger the fire spread acceleration coefficient, indicating that the fire spreads faster. Risk potential energy superposition refers to the comprehensive calculation of the impact of different environmental risk factors on a certain area to obtain the comprehensive risk impact factor of the area. Risk potential energy superposition can comprehensively reflect the fire risk status of a certain area.

[0057] Execution steps: Construct an environment and risk coupling matrix. The rows of the environment and risk coupling matrix correspond to environmental risk factor categories, specifically including ventilation efficiency, combustible gas concentration, and ambient temperature rise rate. The columns correspond to spatial risk areas, specifically including combustible material sub-areas in station halls, densely packed cable sub-areas in tunnels, and high-temperature sub-areas in equipment rooms. Quantify and dynamically update the matrix elements of the environment and risk coupling matrix. Specifically, based on the smoke diffusion speed of the video flame recognition unit, determine the smoke suppression coefficient of ventilation efficiency for each area. Furthermore, by monitoring the smoke diffusion speed of a certain area through the video flame recognition unit and combining it with the data from the ventilation status detection unit, calculate the smoke suppression coefficient corresponding to the ventilation efficiency of that area. The smoke suppression coefficient is used to characterize the suppression effect of the ventilation system on smoke diffusion in each area.

[0058] Based on the carbon monoxide concentration gradient of the gas monitoring unit, the fire-increase coefficient of combustible gas in each area is determined. Furthermore, the gas monitoring unit detects the carbon monoxide concentration gradient in a certain area and calculates the fire-increase coefficient for that area. The fire-increase coefficient is used to characterize the fire development under the correlation of combustible gas concentration in each area. Based on the environmental temperature time series data of the ventilation status detection unit, the fire spread acceleration coefficient of the environmental temperature rise rate in each area is determined. Furthermore, the ventilation status detection unit monitors the environmental temperature rise rate in a certain area and calculates the fire spread acceleration coefficient for that area. The fire spread acceleration coefficient is used to characterize the fire spread speed in each area.

[0059] For the first spatial risk area, the risk potential energy is superimposed using the smoke suppression coefficient, fire growth coefficient, and fire spread acceleration coefficient calculated above to obtain the first comprehensive risk impact factor. For example, assuming the smoke suppression coefficient of the combustible material sub-section in the station hall is 0.8, the fire growth coefficient is 1.2, and the fire spread acceleration coefficient is 1.5, then the comprehensive risk impact factor of this area is 0.8 + 1.2 + 1.5 = 3.5. Based on the comprehensive risk impact factor of the first spatial risk area, the comprehensive risk impact factors of other spatial risk areas are calculated separately.

[0060] The comprehensive risk impact factors of each spatial risk area are dynamically updated as corresponding element values ​​in the environment-risk coupling matrix. These updated element values ​​can reflect the fire risk status of each area in real time. Preferably, by quantifying and dynamically updating the environment-risk coupling matrix, a refined assessment and real-time monitoring of subway fire risk can be achieved. By combining key environmental risk factors such as ventilation efficiency, combustible gas concentration, and ambient temperature rise rate with specific spatial risk areas, the fire risk status of each area in the subway environment can be comprehensively and dynamically reflected, ensuring the accuracy and timeliness of fire risk assessment and further improving the reliability of subway fire prevention and control.

[0061] Furthermore, the method of this application also includes:

[0062] By using a sliding time window, directed edges are determined between adjacent fire development stage nodes in the risk evolution map, connecting each fire evolution stage; the preceding causal relationships, time-dependent parameters, and multi-source signal correlation information between each fire evolution stage are labeled.

[0063] Specifically, the sliding time window is used to analyze time series data. By sliding a fixed-length time interval on the time axis, the data features within that interval are extracted to analyze the changing trends and correlations in the dynamic process. Directed edges in the risk evolution map represent the connection relationship from one stage of fire development to another, reflecting the directionality and dynamism of the fire's evolution from one stage to the next.

[0064] Precursor relationships refer to the triggering factors or conditions that cause a fire to evolve from one stage to the next, including rising temperature and increased smoke concentration. These factors are the direct causes of fire evolution. Time-dependent parameters refer to time-related parameters in the fire evolution process, including the time delay from one stage to another and the evolution rate, reflecting the dynamic characteristics of fire evolution. Multi-source signal correlation information refers to the correlation between signals from different sensors or monitoring units, such as the correlation between temperature signals and smoke concentration signals, which is used to comprehensively analyze the fire evolution process.

[0065] Execution steps: Analyze the time series of fire monitoring data through a sliding time window. For example, set the sliding time window length to 3 minutes, sliding once every 1 minute, and extract data features such as temperature, smoke concentration, and carbon monoxide concentration within each time window to observe the changing trends of fire characteristic parameters in different time windows. Based on these time series data, determine the directed edges between adjacent fire development stage nodes in the risk evolution map. Furthermore, if the temperature and smoke concentration rise significantly at the same time within a certain time window and exceed the preset fire characteristic threshold, it can be determined that the fire has evolved from the smoldering stage to the open flame initiation stage, thereby adding a directed edge from the smoldering stage to the open flame initiation stage in the risk evolution map.

[0066] The study identifies the pre-ignition factors, time-dependent parameters, and multi-source signal correlation information for each stage of fire evolution. Furthermore, regarding the pre-ignition factors, the study reveals that the simultaneous increase in temperature and smoke concentration may be a pre-ignition factor in the evolution from the smoldering stage to the initial open flame stage, indicating that these factors trigger further fire development. For time-dependent parameters, the study analyzes time series data to determine the time delay, evolution rate, and percentage increase in smoke concentration from the smoldering stage to the initial open flame stage. Regarding multi-source signal correlation information, the study analyzes the correlation between temperature and smoke concentration signals, finding a positive correlation between the two during fire evolution. The correlation coefficient indicates a close relationship between temperature increases and smoke concentration increases, jointly driving fire evolution.

[0067] By using sliding time windows and multi-source signal analysis, a fire risk evolution map can be dynamically constructed. By labeling the relationships of preceding causes, time-dependent parameters, and multi-source signal correlation information, the evolution path and dynamic characteristics of a fire from one stage to another can be clearly described. Specifically, in the early stages of a fire, by analyzing the changing trends of temperature and smoke concentration, it is possible to predict in advance that the fire may enter the next more dangerous stage, thereby allowing for timely adjustments to early warning strategies.

[0068] Furthermore, based on the smoke propagation inhibition factor and with the initial ignition source heat release rate under different combustible load densities as the core correlation feature, the method of this application also includes:

[0069] Dynamic weights are assigned to the directed edges of the risk evolution graph. Based on LSTM time-series perception, the relationships between the precursory causes, time-dependent parameters, and multi-source signal correlation information between each fire development stage are fused. The smoke spread inhibition factor is then output to the ventilation control system. The smoke spread inhibition factor is used to dynamically adjust the opening of the air valve.

[0070] Specifically, dynamic weight assignment refers to dynamically adjusting the weights of each directed edge in the risk evolution map based on real-time data and characteristics of fire evolution. The weights of the directed edges reflect the probability and risk level of fire evolving from one stage to another. LSTM (Long Short-Term Memory) networks can process and predict long-term dependencies in time series data, and are used to analyze the changing trends of fire characteristic parameters over time, capturing the dynamic characteristics in the fire evolution process.

[0071] Temporal awareness fusion refers to the use of LSTM networks to comprehensively analyze and fuse the pre-existing causes, time-dependent parameters, and multi-source signal correlation information between fire development stages in order to extract key features and dynamic patterns of fire evolution. Smoke spread inhibition factor is a parameter generated through dynamic analysis of the fire evolution process, used to guide the ventilation control system to adjust the opening of air valves to suppress the diffusion and spread of smoke. Ventilation control system refers to an automated control system used in the subway environment to adjust the operating status of the ventilation system, controlling the diffusion path and speed of smoke by adjusting parameters such as air valve opening and wind speed.

[0072] Execution steps: Dynamically assign weights to the directed edges in the risk evolution graph; Based on time series analysis of fire monitoring data, combined with the time-series perception capability of the LSTM network, calculate the weights of the directed edges between each fire development stage; further, through LSTM network analysis of time series data from the smoldering stage to the initial open flame stage, it is found that the rate of increase of temperature and smoke concentration are positively correlated, and the weight of the directed edge is calculated based on these characteristics.

[0073] By using an LSTM network to perform time-series sensing fusion of the pre-existing causes, time-dependent parameters, and multi-source signal correlation information between different stages of fire development, the LSTM network further considers multi-source signals such as the rate of temperature rise, changes in smoke concentration, and carbon monoxide concentration gradient, as well as the time delay and evolution rate from one stage to another. Through this fusion, the LSTM network can extract key features of fire evolution and generate a smoke spread inhibition factor. Specifically, when the LSTM network analyzes that the fire is in the initial stage of open flame and the smoke concentration is rising rapidly, it outputs a high smoke spread inhibition factor, indicating that ventilation control needs to be strengthened to suppress smoke diffusion.

[0074] The generated smoke spread inhibition factor is output to the ventilation control system for dynamically adjusting the damper opening. Based on the received smoke spread inhibition factor, the ventilation control system adjusts the damper opening to optimize ventilation and suppress smoke diffusion. In these steps, through dynamic weighting and LSTM time-series awareness fusion, accurate prediction of fire evolution and effective control of smoke diffusion are achieved. Dynamic weighting reflects the risk level of a fire from one stage to another in real time, while the time-series awareness capability of the LSTM network captures the dynamic characteristics of fire evolution, generating a smoke spread inhibition factor. By outputting this factor to the ventilation control system, the damper opening can be dynamically adjusted to optimize ventilation strategies, thereby effectively suppressing smoke diffusion and reducing the harm of fire to personnel and equipment.

[0075] Furthermore, this application's method includes multi-scale assessment of the fire's impact range and the use of a dynamic early warning threshold adjustment strategy for tiered alarms:

[0076] A time-series feature analysis is performed using a start threshold and a stop threshold, which are used to mark the smoldering stage, the initial open flame stage, and the flashover risk stage; the dynamic early warning threshold adjustment strategy is configured based on the time-series relationship corresponding to the start threshold and the stop threshold.

[0077] Specifically, the initiation threshold and termination threshold refer to the characteristic parameter thresholds used to mark different stages of fire development. The initiation threshold is used to determine the critical point when a fire enters a certain stage, and the termination threshold is used to determine the critical point when a fire leaves that stage and enters the next stage. The initiation threshold and termination threshold are usually set based on fire characteristic parameters such as temperature, smoke concentration, and carbon monoxide concentration. Time-series characteristic analysis refers to determining the development stage of a fire by analyzing the changing trends of fire characteristic parameters over time, and identifying the transition from one stage to another. The dynamic early warning threshold adjustment strategy refers to dynamically adjusting the early warning threshold according to the fire development stage and real-time monitoring data. Compared with fixed threshold early warning, dynamic early warning thresholds can better adapt to different stages of fire development, improving the accuracy and timeliness of early warning.

[0078] Execution steps: Set start and stop thresholds to mark different stages of fire development. For example, for the smoldering stage, the start threshold is set when the temperature rise rate exceeds 0.1℃ / min and the smoke concentration exceeds 10%; for the initial open flame stage, the start threshold is set when the temperature exceeds 50℃ and the smoke concentration exceeds 20%, and the stop threshold is set when the temperature exceeds 70℃ and the smoke concentration exceeds 30%; for the flashover risk stage, the start threshold is set when the temperature exceeds 80℃ and the smoke concentration exceeds 40%. Through time series analysis of monitoring data, determine whether the fire characteristic parameters have reached these thresholds. When the temperature sensor detects a temperature rise and a simultaneous increase in smoke concentration, it is determined that the fire has moved from the smoldering stage to the initial open flame stage.

[0079] Based on the temporal relationship between the starting and ending thresholds, a dynamic early warning threshold adjustment strategy is configured. For example, in the smoldering stage, the early warning threshold is set relatively low. When the rate of temperature rise or smoke concentration approaches the starting threshold, a low-level warning is issued to alert attention to potential fire risks. In the initial open flame stage, the early warning threshold is appropriately increased. When the temperature or smoke concentration approaches the ending threshold, a medium-level warning is issued to prompt the implementation of preliminary emergency measures. In the flashover risk stage, the early warning threshold is further increased. When the temperature or smoke concentration exceeds the starting threshold, a high-level warning is issued to prompt the immediate initiation of emergency evacuation and fire extinguishing measures.

[0080] By employing time-series feature analysis and dynamic early warning threshold adjustment strategies, accurate identification and early warning of different fire development stages can be achieved. The early warning threshold can be dynamically adjusted according to the actual development of the fire, avoiding false alarms and missed alarms caused by fixed threshold early warnings. Furthermore, in the smoldering stage, a lower early warning threshold can promptly capture early signs of the fire and take measures in advance; in the flashover risk stage, a higher early warning threshold can ensure the accuracy of the early warning and avoid false alarms caused by excessively low thresholds, thus significantly improving the reliability and effectiveness of the fire early warning system.

[0081] Furthermore, the method of this application includes:

[0082] A knowledge graph for subway fire early warning is established by associating fire types, historical fire cases, and the aforementioned starting and ending thresholds. Based on the knowledge graph, potential risk areas are identified, and a topological expansion analysis of the fire impact range is performed.

[0083] Specifically, the subway fire early warning knowledge graph is a structured knowledge representation used to integrate various types of information related to subway fires, including fire types, historical fire cases, and starting and ending thresholds. This information is organically connected through the relationships between nodes and edges, forming a knowledge network that can be queried and analyzed. Fire types refer to the classification of the causes and characteristics of fires, including electrical fires, human-caused fires, and equipment failure fires. Each fire type has different characteristics and evolutionary patterns. Historical fire cases refer to records of past subway fire events, including information such as the time, location, cause, development process, handling measures, and final result of the fire. Topology expansion analysis refers to analyzing the possible paths and affected areas of fire spread based on the topological structure of the fire's impact range. The topological structure reflects the connection relationships between various areas in the subway system, predicting the potential impact of the fire on surrounding areas.

[0084] Execution steps: Correlate fire types, historical fire cases, and start and end thresholds to construct a subway fire early warning knowledge graph. Fire type is used as a node category in the knowledge graph, with each fire type connected to the relevant start and end thresholds. At the same time, historical fire cases are used as another node category. Each case records the specific circumstances of the fire and is associated with the corresponding fire type and threshold. Common fire types include electrical fires.

[0085] Based on a knowledge graph for subway fire early warning, potential risk areas are identified. By analyzing fire type characteristics and historical case data in the knowledge graph, combined with current monitoring data, areas that may face fire risks are identified. For example, if current monitoring data shows that the temperature and smoke concentration in a certain equipment room are close to the initiation threshold of an electrical fire, and similar fires have occurred in this area in the past, then that equipment room will be identified as a potential risk area. Topological expansion analysis of the fire impact range is conducted. Based on the topology of the subway system, the possible paths and impact ranges of fire spreading from potential risk areas are analyzed. For example, if a fire occurs in an equipment room, topological expansion analysis can predict that the fire may spread along ventilation ducts and cable shafts to adjacent tunnel sections or platform areas. During the analysis, fire type and historical case data are combined to assess the fire propagation risk on each path, thereby determining the possible scope of the fire's impact.

[0086] Furthermore, based on the aforementioned subway fire early warning knowledge graph, the method of this application identifies potential risk areas, including:

[0087] An attention-based graph convolutional network is set up, with input node features including historical fire alarm frequency, equipment aging index, and combustible material accumulation frequency, to obtain fire propagation weights between each fire development stage; based on the fire propagation weights between each fire development stage, key channels for smoke diffusion and fire spread paths are determined; based on the subway fire early warning knowledge graph and key channels for smoke diffusion, the fire spread paths are used to simulate the process of obstructed personnel evacuation and equipment cascading failures, to determine the potential risk areas.

[0088] Specifically, attention-based graph convolutional networks are graph neural networks that incorporate attention mechanisms to process graph-structured data. By assigning different weights to each node, they can more effectively capture important relationships between nodes, thereby improving the model's ability to represent graph data. Node features refer to the input features of each node in the graph convolutional network, which are used to describe the attributes of the node. Node features include historical fire alarm frequency, equipment aging index, and frequency of combustible material accumulation, reflecting the fire risk characteristics of different areas of the subway.

[0089] Fire spread weight refers to the weight by which fire spreads from one area to another during the fire development stage, reflecting the probability and intensity of fire spread; critical channels for smoke diffusion refer to the paths or areas where smoke is most likely to spread during a fire, usually related to the subway's ventilation system, building structure, etc.; fire spread paths refer to the paths along which a fire spreads from the ignition point to other areas, and these paths can be determined by analyzing fire spread weights; the process of obstructed evacuation and equipment cascading failure refers to the situation that may lead to obstructed evacuation and equipment cascading failure during a fire due to smoke diffusion and fire spread.

[0090] Execution steps: Set up a graph convolutional network based on the attention mechanism. The input node features include historical fire alarm frequency, equipment aging index, and combustible material accumulation frequency, reflecting the fire risk characteristics of each area of ​​the subway. Through the GAT model, calculate the fire propagation weight between each fire development stage. Furthermore, in the process from the initial open flame stage to the flashover risk stage, calculate the weight of fire propagation from the equipment room to adjacent areas based on the input node features. Common input node features include the equipment aging index.

[0091] Based on the obtained fire propagation weights, the critical channels for smoke diffusion and the fire spread paths were identified. Analysis of the fire propagation weights revealed that the fire was most likely to spread from the equipment room to the platform area via ventilation ducts. Simultaneously, ventilation ducts and cable shafts were identified as the critical channels for smoke diffusion, which will become the main paths for smoke diffusion during a fire. Based on the subway fire early warning knowledge graph and the critical channels for smoke diffusion, the fire spread paths were used to simulate the process of obstructed evacuation and equipment cascading failures, identifying potential risk areas. The simulation showed that after the fire spreads to the platform area via ventilation ducts, it may cause the evacuation routes in the platform area to be blocked by smoke. At the same time, a fire in the equipment room may trigger a cascading failure of equipment in the platform area, identifying the platform area as a potential risk area requiring advance measures, such as activating emergency evacuation plans and shutting down relevant equipment.

[0092] In summary, the beneficial effects of the embodiments of this application are:

[0093] This application utilizes a multi-dimensional monitoring data collection system based on operational subway lines, comparing it with subway safety operation conditions to establish a first and second fire characteristic index. Through these indicators, a risk evolution map is constructed, using fire development stages as nodes and the temporal coupling relationship between heat release rate and smoke concentration as boundary constraints. Within this risk evolution map, based on smoke spread inhibition factors and with the initial ignition source heat release rate under different combustible load densities as the core correlation feature, the fire impact range is assessed at multiple scales, and a dynamic early warning threshold adjustment strategy is used for tiered alarms. This application provides a subway fire safety monitoring and early warning method and system based on multi-data fusion. It integrates multi-dimensional monitoring data, constructs a risk evolution map with fire development stages as nodes and the temporal coupling relationship between heat release rate and smoke concentration as constraints, dynamically tracks the fire spread trajectory, adapts to different fire development stages, and comprehensively enhances the reliability and systematic effectiveness of subway fire prevention and control.

[0094] Example 2, based on the same inventive concept as the subway fire safety monitoring and early warning method under multi-data fusion in the foregoing examples, such as... Figure 2 As shown in the figure, this application embodiment provides a subway fire safety monitoring and early warning system based on multi-data fusion, wherein the system includes:

[0095] Multi-dimensional monitoring data receiving module M100: Based on the subway lines in operation, it receives multi-dimensional monitoring data, compares it with the subway's safe operation conditions, and sets the first fire characteristic indicator and the second fire characteristic indicator.

[0096] Risk Evolution Map Construction Module M200: Constructs a risk evolution map with fire development stages as nodes and the temporal coupling relationship between heat release rate and flue gas concentration as edge constraints using the first fire characteristic index and the second fire characteristic index.

[0097] The graded alarm module M300: In the risk evolution map, based on the smoke spread inhibition factor, the initial fire source heat release rate under different combustible load densities is used as the core correlation feature to conduct multi-scale assessment of the fire impact range, and a dynamic early warning threshold adjustment strategy is used to conduct graded alarms.

[0098] Furthermore, the multi-dimensional monitoring data receiving module M100 is used to perform the following method:

[0099] The passenger flow density logs are associated with each area node of the subway's safe operation condition to set a first fire characteristic indicator; the smoke sensor and carbon monoxide detector records are associated with each equipment node of the subway's safe operation condition to set a second fire characteristic indicator.

[0100] Furthermore, the risk evolution map construction module M200 is also used to perform the following methods:

[0101] The system integrates a video flame recognition unit, a gas monitoring unit, and a ventilation status detection unit to construct a dynamic environmental perception network and acquire subway operating environment data. Based on the equipment-dense sub-sections associated with the risk impact coefficient, and combined with the subway operating environment data, an environment-risk coupling matrix is ​​constructed to dynamically correct the smoke spread inhibition factor in the risk evolution map.

[0102] Furthermore, the risk evolution map construction module M200 is also used to perform the following methods:

[0103] The rows of the environment and risk coupling matrix correspond to environmental risk factor categories, including ventilation efficiency, combustible gas concentration, and ambient temperature rise rate; the columns of the environment and risk coupling matrix correspond to spatial risk areas, including combustible material sub-areas in station halls, densely packed cable sub-areas in tunnels, and high-temperature sub-areas in equipment rooms. The environment and risk coupling matrix is ​​quantified and dynamically updated, specifically including: determining the smoke suppression coefficient of ventilation efficiency for each area based on the smoke diffusion rate of the video flame recognition unit; determining the fire ignition coefficient of combustible gas for each area based on the carbon monoxide concentration gradient of the gas monitoring unit; determining the fire spread acceleration coefficient of ambient temperature rise rate for each area based on the ambient temperature time-series data of the ventilation status detection unit; for the first spatial risk area, the risk potential energy is superimposed using the smoke suppression coefficient, fire ignition coefficient, and fire spread acceleration coefficient to obtain the first comprehensive risk impact factor; based on the first spatial risk area and the first comprehensive risk impact factor, each spatial risk area is traversed to obtain the comprehensive risk impact factor for each spatial risk area; the comprehensive risk impact factor of each spatial risk area is used as the corresponding element value of the environment and risk coupling matrix for dynamic updating.

[0104] Furthermore, the multi-dimensional monitoring data receiving module M100 is also used to perform the following method:

[0105] By using a sliding time window, directed edges are determined between adjacent fire development stage nodes in the risk evolution map, connecting each fire evolution stage; the preceding causal relationships, time-dependent parameters, and multi-source signal correlation information between each fire evolution stage are labeled.

[0106] Furthermore, the hierarchical alarm module M300 is also used to perform the following method:

[0107] Dynamic weights are assigned to the directed edges of the risk evolution graph. Based on LSTM time-series perception, the relationships between the precursory causes, time-dependent parameters, and multi-source signal correlation information between each fire development stage are fused. The smoke spread inhibition factor is then output to the ventilation control system. The smoke spread inhibition factor is used to dynamically adjust the opening of the air valve.

[0108] Furthermore, the hierarchical alarm module M300 is used to perform the following method:

[0109] A time-series feature analysis is performed using a start threshold and a stop threshold, which are used to mark the smoldering stage, the initial open flame stage, and the flashover risk stage; the dynamic early warning threshold adjustment strategy is configured based on the time-series relationship corresponding to the start threshold and the stop threshold.

[0110] Furthermore, the hierarchical alarm module M300 is also used to perform the following method:

[0111] A knowledge graph for subway fire early warning is established by associating fire types, historical fire cases, and the aforementioned starting and ending thresholds. Based on the knowledge graph, potential risk areas are identified, and a topological expansion analysis of the fire impact range is performed.

[0112] Furthermore, the hierarchical alarm module M300 is also used to perform the following method:

[0113] An attention-based graph convolutional network is set up, with input node features including historical fire alarm frequency, equipment aging index, and combustible material accumulation frequency, to obtain fire propagation weights between each fire development stage; based on the fire propagation weights between each fire development stage, key channels for smoke diffusion and fire spread paths are determined; based on the subway fire early warning knowledge graph and key channels for smoke diffusion, the fire spread paths are used to simulate the process of obstructed personnel evacuation and equipment cascading failures, to determine the potential risk areas.

[0114] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Figure 1 The subway fire safety monitoring and early warning method and specific examples under multi-data fusion in Embodiment 1 are also applicable to the subway fire safety monitoring and early warning system under multi-data fusion in this embodiment. Through the foregoing detailed description of the subway fire safety monitoring and early warning method under multi-data fusion, those skilled in the art can clearly understand the subway fire safety monitoring and early warning system under multi-data fusion in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0115] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0116] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for monitoring and early warning of subway fire safety based on multi-data fusion, characterized in that, The method includes: Based on the subway lines in operation, receive multi-dimensional monitoring data, compare it with the subway's safe operation conditions, and set the first fire characteristic indicator and the second fire characteristic indicator. Based on the first fire characteristic index and the second fire characteristic index, a risk evolution map is constructed with the fire development stage as the node and the temporal coupling relationship between heat release rate and smoke concentration as the edge constraint. In the risk evolution map, based on the smoke spread inhibition factor, the initial heat release rate of the fire source under different combustible load densities is used as the core correlation feature to conduct multi-scale assessment of the fire impact range, and a dynamic early warning threshold adjustment strategy is used for graded alarms.

2. The subway fire safety monitoring and early warning method based on multi-data fusion as described in claim 1, characterized in that, The method for setting a first fire characteristic indicator and a second fire characteristic indicator includes: The passenger flow density logs are associated with the various regional nodes of the subway's safe operation conditions, and a first fire characteristic indicator is set. The smoke sensor and carbon monoxide detector records are associated with each equipment node of the subway's safe operation conditions, and a second fire characteristic index is set.

3. The subway fire safety monitoring and early warning method based on multi-data fusion as described in claim 1, characterized in that, The method further includes constructing a risk evolution map with fire development stages as nodes and the temporal coupling relationship between heat release rate and smoke concentration as edge constraints. Integrating video flame recognition unit, gas monitoring unit and ventilation status detection unit, a dynamic environmental perception network is constructed to acquire subway operating environment data; Based on the equipment-dense sub-sections associated with the risk impact coefficient, and combined with the subway operating environment data, an environment-risk coupling matrix is ​​constructed to dynamically correct the smoke spread inhibition factor in the risk evolution map.

4. The subway fire safety monitoring and early warning method based on multi-data fusion as described in claim 3, characterized in that, The method further includes constructing an environment-risk coupling matrix and dynamically correcting the flue gas spread inhibition factor in the risk evolution map. The rows of the environment and risk coupling matrix correspond to environmental risk factor categories, including ventilation efficiency, combustible gas concentration, and environmental temperature rise rate. The columns of the environment and risk coupling matrix correspond to spatial risk areas, including the combustible material sub-area in the station hall, the dense cable sub-area in the tunnel, and the high temperature sub-area in the equipment room. The environment and risk coupling matrix is ​​quantified and dynamically updated, specifically including: determining the smoke suppression coefficient of ventilation efficiency for each area based on the smoke diffusion rate of the video flame recognition unit; determining the fire ignition coefficient of combustible gas for each area based on the carbon monoxide concentration gradient of the gas monitoring unit; determining the fire spread acceleration coefficient of the ambient temperature rise rate for each area based on the ambient temperature time series data of the ventilation status detection unit; for the first spatial risk area, the risk potential energy is superimposed using the smoke suppression coefficient, fire ignition coefficient, and fire spread acceleration coefficient to obtain the first comprehensive risk impact factor; based on the first spatial risk area and the first comprehensive risk impact factor, each spatial risk area is traversed to obtain the comprehensive risk impact factor of each spatial risk area; the comprehensive risk impact factor of each spatial risk area is used as the corresponding element value of the environment and risk coupling matrix for dynamic updating.

5. The subway fire safety monitoring and early warning method based on multi-data fusion as described in claim 2, characterized in that, By using a sliding time window, directed edges are determined between adjacent fire development stage nodes in the risk evolution map, connecting each fire evolution stage. The relationships between the precursory causes, time-dependent parameters, and multi-source signal correlation information of each stage of fire evolution are marked.

6. The subway fire safety monitoring and early warning method based on multi-data fusion as described in claim 5, characterized in that, Based on the flue gas propagation inhibition factor, and using the initial ignition source heat release rate under different combustible load densities as the core correlation feature, the method further includes: Dynamic weights are assigned to the directed edges of the risk evolution graph. Based on LSTM time-series perception, the relationships between the precursory causes, time-dependent parameters, and multi-source signal correlation information between each fire development stage are fused. The smoke spread inhibition factor is then output to the ventilation control system. The smoke spread inhibition factor is used to dynamically adjust the opening of the air valve.

7. The subway fire safety monitoring and early warning method based on multi-data fusion as described in claim 1, characterized in that, The method includes conducting multi-scale assessments of the fire's impact range and implementing tiered alerts using a dynamic early warning threshold adjustment strategy. Temporal feature analysis is performed using start-up and end-of-flight thresholds, which are used to mark the smoldering stage, the initial open flame stage, and the flashover risk stage. The dynamic early warning threshold adjustment strategy is configured based on the temporal relationship between the start threshold and the end threshold.

8. The subway fire safety monitoring and early warning method based on multi-data fusion as described in claim 7, characterized in that, The method includes: A knowledge graph for subway fire early warning is established by associating fire types, historical fire cases, and the aforementioned starting and ending thresholds. Based on the aforementioned subway fire early warning knowledge graph, potential risk areas are identified, and a topological expansion analysis of the fire impact range is performed.

9. The subway fire safety monitoring and early warning method based on multi-data fusion as described in claim 8, characterized in that, Based on the aforementioned subway fire early warning knowledge graph, potential risk areas are identified, and the method includes: Set up an attention-based graph convolutional network, with input node features including historical fire alarm frequency, equipment aging index, and combustible material accumulation frequency, to obtain the fire propagation weights between each fire development stage; Based on the fire propagation weights between the various fire development stages, the key channels for smoke diffusion and the fire spread paths are determined. Based on the aforementioned knowledge graph of subway fire early warning and key channels for smoke diffusion, the fire spread path is used to simulate the process of obstructed personnel evacuation and equipment cascading failures, thereby identifying the potential risk areas.

10. A subway fire safety monitoring and early warning system based on multi-data fusion, characterized in that: The system is used to implement the subway fire safety monitoring and early warning method based on multi-data fusion as described in any one of claims 1-9, wherein the system comprises: Multi-dimensional monitoring data receiving module: Based on the subway lines in operation, it receives multi-dimensional monitoring data, compares it with the subway's safe operation conditions, and sets the first fire characteristic indicator and the second fire characteristic indicator. Risk evolution map construction module: Constructs a risk evolution map with fire development stage as nodes and the temporal coupling relationship between heat release rate and smoke concentration as edge constraints through the first fire characteristic index and the second fire characteristic index; Graded alarm module: In the risk evolution map, based on the smoke spread inhibition factor, the initial fire source heat release rate under different combustible load densities is used as the core correlation feature to conduct multi-scale assessment of the fire impact range, and a dynamic early warning threshold adjustment strategy is used to conduct graded alarms.

Images