A park fire early warning management system and method for property management

Abstract

The application relates to the technical field of intelligent fire fighting, and discloses a park fire fighting early warning management system and method for property management, which comprises the following steps: constructing a conduction accessibility matrix to quantify the physical correlation strength of sensors in a building heat conduction network; when a sensor alarms, a conduction path correlation device is activated to cross-verify data and output a three-level credibility label; a fixed plan is deconstructed into basic units of a detection element, a control element and an inhibition element, a plan execution chain containing an execution sequence is dynamically assembled based on real-time fire simulation, and new park response rules are intelligently migrated by matching a building topology vector with a historical case library. The application breaks through the limitation of a traditional system which relies on a single sensor threshold value for judgment, gives the sensor network self-explanation ability through a physical conduction model, eliminates the hijacking effect of false alarms on system decision-making, reconstructs a linear process of "alarm-confirmation-response" into a multi-path parallel decision mechanism, and avoids the secondary disaster risk caused by the false start of a smoke exhaust system in a traditional scheme.

Description

Technical Field

[0001] This invention relates to a fire early warning management system and method for property management, belonging to the field of fire early warning management technology for property management parks. Background Technology

[0002] With the continuous advancement of smart city construction, the demand for fire early warning management systems in property management and building safety is increasing. Traditional fire early warning systems typically rely on independent alarm information from sensors, using set thresholds to determine the existence of fire risk. However, this method, based on a single sensor threshold, is susceptible to environmental interference and sensor malfunctions, leading to a high false alarm rate and affecting the system's decision-making accuracy. For example:

[0003] 1. Existing fire alarm systems typically employ centralized monitoring, using multiple sensors to detect fire-related signals, including temperature and smoke concentration. However, alarm signals from a single sensor cannot accurately reflect the true extent of a fire. Due to the complex heat conduction and airflow changes within buildings, sensors in certain areas may generate false alarms due to heat radiation or other factors, leading to misjudgments of a fire and triggering unnecessary responses. This not only wastes resources but may also increase the risk of secondary disasters.

[0004] 2. To reduce false alarms, some systems have introduced multi-sensor data fusion and intelligent algorithms to optimize alarm judgment. However, existing multi-sensor fusion methods still face some problems. For example, most systems rely on simple weighted average algorithms to fuse data from multiple sensors, but this method fails to fully consider the physical correlation and spatiotemporal factors between sensors, and may still result in high false alarm and false negative rates. In addition, traditional fire emergency response procedures are usually linear, processing alarm information sequentially through an "alarm-confirmation-response" process, lacking sufficient flexibility and intelligent decision-making capabilities, and unable to cope with complex and ever-changing fire scenarios.

[0005] To overcome these problems, some solutions attempt to incorporate models of heat conduction and airflow within buildings, combined with the physical correlations of sensor networks, to improve the accuracy of early warnings. However, these systems often rely too heavily on complex physical models, limiting their response speed and real-time performance. Furthermore, the execution chain of existing systems is often fixed, lacking flexible adjustment mechanisms and unable to quickly generate optimal response plans based on real-time changes in the fire situation. Summary of the Invention

[0006] This invention provides a fire early warning management system and method for property management parks, the main purpose of which is to solve the problems of resource waste and secondary disaster risks caused by false alarms and response delays in existing systems.

[0007] To achieve the above objectives, the present invention provides a property management park fire early warning management system and method, comprising:

[0008] The conduction topology modeling module is used to generate a three-dimensional heat conduction network based on the building floor plan, and encodes the wall thermal conductivity and ventilation path parameters into a conduction reachability matrix. The matrix stores the physical association strength of each sensor node and the expected conduction time threshold.

[0009] The dynamic verification engine module connects to the park's fire sensor network. When any sensor triggers an alarm, it activates the associated sensors along its transmission path to perform data cross-verification. Based on the transmission reachability matrix, it calculates the physical consistency between the alarm signal and the associated sensor data and outputs three levels of credibility labels, including physical credibility, logical suspicion, and physical conflict.

[0010] The response element dynamic assembly module includes a pre-stored library of basic units for detection elements, control elements, and suppression elements. The detection elements define fire feature identification rules, the control elements define equipment linkage logic, and the suppression elements define fire extinguishing resource scheduling strategies. When the credibility tag is physically credible, the response element combination is matched based on real-time fire boundary prediction to generate a contingency plan execution chain that includes execution order and triggering conditions.

[0011] The topology migration adaptation module encodes building floor height and fire compartment area parameters into topology vectors, matches them with the historical park case library through similarity calculation, and outputs the initial combination recommendation scheme of response elements to the dynamic assembly module of response elements.

[0012] The incremental learning module is connected to the output of the dynamic verification engine module. It records the deviation between the transmission model prediction value and the actual sensor data in each alarm event and optimizes the time threshold parameter in the transmission reachability matrix through offline training.

[0013] In a preferred embodiment, in the conductive topology modeling module, the thermal conductivity of the wall is dynamically corrected by the infrared thermal imaging detection value of the building material, and the ventilation path parameters are updated by the airflow velocity data in the operation log of the park's fresh air system.

[0014] As a preferred implementation, the triggering conditions of the dynamic verification engine module include: when the sensor alarm signal is opposite to the data fluctuation direction of at least two associated nodes in the transmission reachability matrix, it is automatically downgraded to a logically suspicious label and a manual review process is initiated.

[0015] As a preferred embodiment, the detector element includes a smoldering feature recognition rule, specifically: when the temperature rise rate is detected to be ≤2℃ / min in a confined space and the direction of smoke concentration gradient change is opposite to the ventilation path, an inert gas injection command is triggered and automatic spraying is prohibited.

[0016] In a preferred embodiment, in the topology migration adaptation module, the similarity calculation adopts Euclidean distance metric, and assigns a weight coefficient of 0.6 to the floor height parameter and a weight coefficient of 0.4 to the fire compartment area parameter.

[0017] As a preferred implementation, the generation logic of the contingency plan execution chain includes: when the fire boundary prediction shows that it will spread to the adjacent zone within 30 seconds, an automatic addition of a delayed start command for the smoke exhaust system is added, and the smoke exhaust operation is executed after the open flame is confirmed.

[0018] As a preferred implementation, the optimization process of the incremental learning module includes: adopting a sliding window mechanism, retaining only the alarm event data of the most recent 90 days for matrix parameter training, and converting historical data outside the window into a case library for storage.

[0019] As a preferred implementation, it also includes a visualization decision sandbox module, which displays the heat map of the transmission accessibility matrix and the status of the contingency plan execution chain in real time, and provides a manual intervention interface to cover the automatically generated response element combination.

[0020] As a preferred implementation, in the visual decision sandbox module, manual intervention operations include: manually adjusting the time threshold parameter of the transmission reachability matrix, or forcibly specifying the priority order of response element combinations.

[0021] A fire early warning management method for property management parks includes the following steps:

[0022] Step 1: Conductive topology modeling. Generate a three-dimensional heat conduction network based on the building floor plan. Encode the wall thermal conductivity and ventilation path parameters into a conduction reachability matrix. The matrix stores the physical association strength and expected conduction time threshold of each sensor node.

[0023] Step 2: Dynamic verification. Monitor the park's fire sensor network in real time. When any sensor triggers an alarm, activate the associated sensors on its transmission path for data cross-verification. Calculate the physical consistency between the alarm signal and the associated sensor data based on the transmission reachability matrix, and output three levels of credibility labels including physically credible, logically suspicious, and physically conflicting.

[0024] Step 3: Dynamic assembly of response elements. When the credibility tag is physically credible, based on real-time fire boundary prediction, the response element combination is matched from the pre-stored basic unit library of detection elements, control elements, and suppression elements to generate a contingency plan execution chain containing the execution order and triggering conditions. Among them, the detection element defines the fire feature identification rules, the control element defines the equipment linkage logic, and the suppression element defines the fire extinguishing resource scheduling strategy.

[0025] Step 4: Topology migration adaptation. The building floor height and fire compartment area parameters are encoded into topology vectors. The similarity is used to match the historical park case library and output the initial combination recommendation scheme of the response element to guide the initial configuration of the dynamic assembly module of the response element.

[0026] Step 5: Incremental learning. Record the deviation between the propagation model's predicted value and the actual sensor data in each alarm event. Optimize the time threshold parameter in the propagation reachability matrix through offline training to improve the model's prediction accuracy.

[0027] Compared to the problems described in the background art, the beneficial effects of this invention are: constructing a logical approach that is both "physically reliable" and "logically reliable," eliminating the hijacking effect of false alarms from a single sensor on system decision-making through spatiotemporal constraint analysis of the reachability matrix, realizing the self-explanatory capability of the sensor network, and reconstructing the linear process of alarm-confirmation-response in the traditional system into a multi-path parallel decision-making mechanism by deconstructing fixed plans into composable basic units of detection, control, and suppression elements, and generating the optimal execution chain by combining real-time fire spread simulation, thereby avoiding the secondary disaster risks caused by the accidental activation of the smoke exhaust system in traditional solutions. Attached Figure Description

[0028] Figure 1 This is a system architecture diagram of the present invention.

[0029] Figure 2 This is a flowchart of the dynamic verification process for this invention.

[0030] Figure 3 This is a flowchart illustrating the fire response plan generation process of the present invention, which involves multi-module interaction.

[0031] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0032] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0033] This application provides a fire early warning management system for property management parks. It includes: a conduction topology modeling module, used to generate a three-dimensional heat conduction network based on building floor plans, encoding wall thermal conductivity and ventilation path parameters into a conduction reachability matrix, wherein the matrix stores the physical association strength and expected conduction time threshold of each sensor node;

[0034] The dynamic verification engine module connects to the park's fire sensor network. When any sensor triggers an alarm, it activates the associated sensors along its transmission path to perform data cross-verification. Based on the transmission reachability matrix, it calculates the physical consistency between the alarm signal and the associated sensor data and outputs three levels of credibility labels, including physical credibility, logical suspicion, and physical conflict.

[0035] The response element dynamic assembly module includes a pre-stored library of basic units for detection elements, control elements, and suppression elements. The detection elements define fire feature identification rules, the control elements define equipment linkage logic, and the suppression elements define fire extinguishing resource scheduling strategies. When the credibility tag is physically credible, the response element combination is matched based on real-time fire boundary prediction to generate a contingency plan execution chain that includes execution order and triggering conditions.

[0036] The topology migration adaptation module encodes building floor height and fire compartment area parameters into topology vectors, matches them with the historical park case library through similarity calculation, and outputs the initial combination recommendation scheme of response elements to the dynamic assembly module of response elements.

[0037] The incremental learning module is connected to the output of the dynamic verification engine module. It records the deviation between the transmission model prediction value and the actual sensor data in each alarm event and optimizes the time threshold parameter in the transmission reachability matrix through offline training.

[0038] In a preferred embodiment, in the conductive topology modeling module, the thermal conductivity of the wall is dynamically corrected by the infrared thermal imaging detection value of the building material, and the ventilation path parameters are updated by the airflow velocity data in the operation log of the park's fresh air system.

[0039] As a preferred implementation, the triggering conditions of the dynamic verification engine module include: when the sensor alarm signal is opposite to the data fluctuation direction of at least two associated nodes in the transmission reachability matrix, it is automatically downgraded to a logically suspicious label and a manual review process is initiated.

[0040] As a preferred embodiment, the detector element includes a smoldering feature recognition rule, specifically: when the temperature rise rate is detected to be ≤2℃ / min in a confined space and the direction of smoke concentration gradient change is opposite to the ventilation path, an inert gas injection command is triggered and automatic spraying is prohibited.

[0041] In a preferred embodiment, in the topology migration adaptation module, the similarity calculation adopts Euclidean distance metric, and assigns a weight coefficient of 0.6 to the floor height parameter and a weight coefficient of 0.4 to the fire compartment area parameter.

[0042] As a preferred implementation, the generation logic of the contingency plan execution chain includes: when the fire boundary prediction shows that it will spread to the adjacent zone within 30 seconds, an automatic addition of a delayed start command for the smoke exhaust system is added, and the smoke exhaust operation is executed after the open flame is confirmed.

[0043] As a preferred implementation, the optimization process of the incremental learning module includes: adopting a sliding window mechanism, retaining only the alarm event data of the most recent 90 days for matrix parameter training, and converting historical data outside the window into a case library for storage.

[0044] As a preferred implementation, it also includes a visualization decision sandbox module, which displays the heat map of the transmission accessibility matrix and the status of the contingency plan execution chain in real time, and provides a manual intervention interface to cover the automatically generated response element combination.

[0045] As a preferred implementation, in the visual decision sandbox module, manual intervention operations include: manually adjusting the time threshold parameter of the transmission reachability matrix, or forcibly specifying the priority order of response element combinations.

[0046] A fire early warning management method for property management parks includes the following steps:

[0047] Step 1: Conductive topology modeling. Generate a three-dimensional heat conduction network based on the building floor plan. Encode the wall thermal conductivity and ventilation path parameters into a conduction reachability matrix. The matrix stores the physical association strength and expected conduction time threshold of each sensor node.

[0048] Step 2: Dynamic verification. Monitor the park's fire sensor network in real time. When any sensor triggers an alarm, activate the associated sensors on its transmission path for data cross-verification. Calculate the physical consistency between the alarm signal and the associated sensor data based on the transmission reachability matrix, and output three levels of credibility labels including physically credible, logically suspicious, and physically conflicting.

[0049] Step 3: Dynamic assembly of response elements. When the credibility tag is physically credible, based on real-time fire boundary prediction, the response element combination is matched from the pre-stored basic unit library of detection elements, control elements, and suppression elements to generate a contingency plan execution chain containing the execution order and triggering conditions. Among them, the detection element defines the fire feature identification rules, the control element defines the equipment linkage logic, and the suppression element defines the fire extinguishing resource scheduling strategy.

[0050] Step 4: Topology migration adaptation. The building floor height and fire compartment area parameters are encoded into topology vectors. The similarity is used to match the historical park case library and output the initial combination recommendation scheme of the response element to guide the initial configuration of the dynamic assembly module of the response element.

[0051] Step 5: Incremental learning. Record the deviation between the propagation model's predicted value and the actual sensor data in each alarm event. Optimize the time threshold parameter in the propagation reachability matrix through offline training to improve the model's prediction accuracy.

[0052] See Figure 1 , Figure 2 and Figure 3 ,like Figure 1As shown, the system receives a sensor alarm signal and activates relevant nodes along the transmission path for verification. Next, the system performs data consistency calculations to determine if the alarm signal conforms to the requirements of the transmission model. If the data conforms to the transmission model, the system marks it as physically reliable and proceeds to the subsequent response process. If the data does not conform to the transmission model, it is marked as logically suspicious, and the system initiates a manual review process to ensure the accuracy of the verification results. This process effectively reduces the impact of false alarms and ensures that the system can make a more accurate response based on the verification results. Figure 2 The system receives a sensor alarm signal and activates relevant nodes along the transmission path for verification. Next, the system performs data consistency calculations to determine if the alarm signal conforms to the requirements of the transmission model. If the data conforms to the transmission model, the system marks it as physically reliable and proceeds to the subsequent response process. If the data does not conform to the transmission model, it is marked as logically suspicious, and the system initiates a manual review process to ensure the accuracy of the verification results. This process effectively reduces the impact of false alarms and ensures that the system can make a more accurate response based on the verification results; for example... Figure 3 The diagram shows the fire response plan generation process of the present invention, which involves multi-module interaction. It illustrates the dynamic plan generation process of the fire response system and includes the following core steps: The fire simulation module sequentially requests feature recognition rules from the detection element library and obtains temperature / smoke rules; requests equipment linkage logic from the control element library and receives smoke exhaust / sprinkler commands; requests resource scheduling strategies from the suppression element library and obtains extinguishing agent delivery paths; and dynamically assembles the above response elements (including temperature / smoke rules, smoke exhaust / sprinkler commands, and extinguishing agent delivery paths) through the execution chain generator. Finally, a plan execution chain containing a complete operation sequence is generated. This chain accurately guides the coordinated work of various fire-fighting equipment through a sequence of timed instructions. All of these are extended implementation methods known to those skilled in the art.

[0053] Example 1: In this example, based on building floor plan data, the system creates a 3D heat conduction network through a conductive topology modeling module. The core task of this module is to form a conductive accessibility matrix based on building structural parameters, such as the thermal conductivity of wall materials and airflow velocity in ventilation paths. Each element in the matrix represents the physical correlation strength and conduction time threshold between sensor nodes. For example, the conductive accessibility matrix is ​​a matrix-like data structure used to describe the heat conduction relationship between different sensor nodes within the building; wall thermal conductivity is obtained through infrared thermal imaging detection technology, as each building material has different thermal conductivity characteristics; ventilation path parameters can be based on real-time operating data of the park's ventilation system, such as airflow velocity data, obtained through monitoring device logs. These parameters will affect the processing and response of sensor data. When any sensor detects a fire alarm signal, the dynamic verification engine module starts and activates the connected sensor nodes for data cross-verification. This process verifies the consistency between the alarm signal and the data from adjacent sensors based on the physical consistency calculation in the conductive accessibility matrix.

[0054] In terms of operational path, physical consistency verification can be used to determine the rationality of alarm signals and avoid false alarms. For example, when multiple sensor data show significant inconsistencies, the verification engine will downgrade the alarm information to a "logically suspicious" state and trigger manual review. This process can effectively reduce the interference of false alarms on system decision-making and improve the accuracy of fire response through verification. When the alarm signal is determined to be "physically reliable," the system will enter the response phase. The response element dynamic assembly module selects appropriate response elements from the pre-stored library of detection elements, control elements, and suppression elements based on real-time fire simulation results to generate a contingency plan execution chain. Detection elements include fire feature identification rules such as temperature rise rate and smoke concentration; control elements include logic defining equipment linkage, such as the linkage between smoke exhaust systems and sprinkler systems; and suppression elements include scheduling strategies for fire extinguishing resources, such as the deployment path and timing of fire extinguishers. The selection of each response element is based on the prediction results of the current fire model to ensure that the actions taken conform to the actual situation of fire spread. For new campuses, the topology migration adaptation module plays a role. This module encodes parameters such as building height and fire compartment area within the park to form a topological vector, and then performs similarity matching with park cases in a historical case library. By calculating similarity, the system can recommend the most suitable initial response element combination. The system uses Euclidean distance to calculate similarity and compares the structural characteristics of different parks to quickly adapt the fire response plan for the new park.

[0055] In practical implementation, data from each alarm event is continuously monitored and recorded. With the help of the incremental learning module, the system continuously optimizes the time threshold parameters in the transmission reachability matrix by analyzing the deviation of historical alarm events. The incremental learning process involves comparing the predicted value of the transmission model with the actual sensor data for each alarm event, calculating the deviation, and using a sliding window mechanism to retain only the alarm data from the most recent 90 days for training. Historical data outside the window is used to update the case library. Offline training optimizes the time threshold, enabling the transmission model to continuously improve its predictive accuracy.

[0056] To ensure the transparency and operability of the system's decision-making process, a visual decision-making sandbox module is provided. This module displays a heatmap of the reachability matrix in real time and shows the status of the contingency plan execution chain. Simultaneously, the system provides an interface for manual intervention, allowing operators to adjust time threshold parameters in the matrix or change the priority of response element combinations. The manual intervention process allows operators to adjust the model output according to the actual situation, overriding automatically generated response element combinations to ensure the flexibility and accuracy of response measures.

[0057] Example 2: In the conductive topology modeling module, a three-dimensional heat conduction network can be generated using building floor plans. Specifically, parameters such as the thermal conductivity of the walls and the airflow velocity of the ventilation path are used as inputs to calculate the physical correlation strength between sensor nodes within the building. To ensure the accuracy of these parameters, the following methods can be used: Wall thermal conductivity: The thermal conductivity of building materials is measured using infrared thermal imaging technology. The thermal conductivity of each building material is obtained after actual measurement; common building materials (such as concrete, steel, and glass) have different thermal conductivity properties. When modeling the conductive topology, the actual value for each building area should be used for the specific material. Ventilation path parameters: These parameters are obtained through airflow velocity data from the park's fresh air system and dynamically updated based on actual wind speed sensor data. Real-time monitoring of airflow velocity can effectively correct the airflow model within the building, thereby affecting the prediction of fire propagation paths. Through these parameters, a conductive reachability matrix is ​​established, which stores the physical correlation strength and expected conduction time threshold of each sensor node within the building. The conduction time between each sensor and other sensors is obtained based on actual measurements. It's important to note that this data comes from real-time building sensor data and historical building characteristics, and the matrix is ​​continuously adjusted and optimized based on real-time monitoring data. When any sensor triggers an alarm, the dynamic verification engine module cross-verifies the data with other associated sensors using the accessibility matrix. During this process, if the alarm signal's fluctuation direction is opposite to that of at least two associated sensors, the system will automatically mark the alarm as logically suspicious and trigger a manual review process.

[0058] To further clarify the calculation method of this process, the data consistency calculation is first completed through the following steps: the alarm signal of each sensor is compared with the signals of its associated sensors, and the consistency between these signals is calculated based on the physical conduction model. Assuming sensor A detects a fire alarm signal, its associated sensors B and C will have their data consistency evaluated using the conduction time threshold and association strength in the matrix. If the signal of sensor A differs significantly from the signals of B and C within the expected time range, and these differences cannot be explained by normal environmental fluctuations, the system automatically marks it as logically suspicious. This method effectively eliminates interference caused by false alarms from a single sensor, improving the accuracy and reliability of the alarm system.

[0059] In the dynamic assembly module of response elements, we dynamically select response elements (including detection elements, control elements, and suppression elements) and generate the optimal execution chain based on the real-time fire simulation results of the transmission topology model. The selection of response elements is based on the signal characteristics of each sensor and its correlation with other sensors. During this process, the system dynamically assembles the optimal response plan based on physically reliable labels and fire simulation predictions. For example, when the system judges the speed of fire spread based on the fire simulation, it will adjust the activation sequence of the smoke extraction system and the fire extinguishing system in real time. The fire boundary prediction calculation model will adjust the activation timing of the smoke extraction system based on information such as the current fire temperature rise rate and smoke concentration. For example, if the fire is expected to spread to the adjacent fire compartment within 30 seconds, the system will postpone the smoke extraction operation and activate it only after the open flame is confirmed. During this process, the generation of the contingency plan execution chain requires comprehensive judgment based on the feedback information of the sensor network, the fire spread simulation results, and the emergency response strategies in the historical case library. An important goal of this module is to avoid the secondary disaster risks caused by the accidental activation of the smoke extraction system in traditional solutions. The incremental learning module, through offline training, records the deviation between the predicted values ​​of the transmission model and the actual sensor data for each alarm event. To avoid over-reliance on historical data while ensuring model accuracy, a sliding window mechanism is employed, which only retains alarm data from the most recent 90 days for training. Data outside the window is transferred to a case library for adaptation to new areas and subsequent optimization. During each training iteration, the system updates the time threshold parameters in the transmission accessibility matrix based on new alarm data. This incremental learning process ensures the system can continuously optimize with new fire event data, improving the accuracy and responsiveness of the prediction model. To ensure greater flexibility and accuracy when handling complex situations, the system also provides a visual decision-making sandbox module. This module not only displays a heatmap of the transmission accessibility matrix in real time but also shows the execution status of the emergency response chain. Through this module, operators can manually intervene, adjusting parameters or priorities to address complex fire situations. For example, in certain special circumstances (such as probe failure or abnormal environmental factors), operators can manually adjust the time threshold parameters of the conduction reachability matrix or specify the priority of certain response elements to ensure that the system can make appropriate responses. These are all extended implementation methods known to those skilled in the art.

[0060] Example 3: In this example, the thermal conductivity of the walls is measured using infrared thermal imaging technology to ensure that the thermal conductivity characteristics of each building material are accurately quantified. Different types of wall materials (such as concrete, steel, and glass) are configured differently based on their thermal conductivity. Ventilation path parameters: These parameters are dynamically updated based on real-time monitoring of the airflow velocity of the park's fresh air system and actual operating data. Airflow velocity data measured by wind speed sensors allows for precise modeling of the airflow characteristics of the ventilation path, and combined with sensor data, predicts fire spread paths. When generating the conduction reachability matrix, all parameter values ​​are based on actual monitoring and measurement, and are periodically calibrated and updated. Each element in the matrix represents the physical correlation strength between sensor nodes and the expected conduction time threshold; these values ​​are calculated using the aforementioned parameters. To ensure accuracy, data from each sensor is verified during matrix calculation to avoid false alarms due to errors or inaccurate measurements.

[0061] In the dynamic verification engine, when any sensor triggers an alarm signal, the system compares the data with that of other associated sensors based on the transmission reachability matrix. Specifically, if the alarm signal has the opposite direction to or significantly different from the signals of at least two associated sensors, the system marks the alarm as "logically suspicious" and initiates a manual review process based on a set physical consistency standard. The core calculation method for this process is data consistency calculation: by comparing the direction and amplitude of signal data fluctuations between the alarm sensor and associated sensors, a consistency index is calculated. If this index is lower than a predetermined threshold (set based on sensor type and building characteristics), the system considers the data inconsistent and marks it as "logically suspicious." This optimization not only significantly reduces the probability of false alarms but also ensures high system reliability in complex environments. The implementation path for this step is clear, and the specific calculation process can be automatically completed by the system's internal dynamic verification module.

[0062] In the dynamic assembly module of response elements, fire boundary prediction: Based on real-time fire simulation, the system predicts the direction and speed of fire spread. If the fire is expected to spread to adjacent areas within 30 seconds, the system will prioritize activating the smoke extraction system. This decision is based on simulations of fire growth rate and smoke concentration changes monitored by sensors. Response element combination: Based on the results of the fire simulation, the system selects appropriate response elements from a pre-stored library of detection elements, control elements, and suppression elements. Detection elements determine the fire type based on smoke concentration and temperature rise rate; control elements define equipment linkage rules (such as the linkage between smoke extraction and sprinkler systems); and suppression elements formulate action plans through fire extinguishing resource scheduling strategies. When generating the execution chain, the system dynamically adjusts the priority and execution order of each response element in conjunction with real-time fire simulation data. For example, if the fire spread rate is determined to be fast, the activation command for the smoke extraction system will be delayed to avoid secondary disasters caused by accidental activation.

[0063] The incremental learning module optimizes the model by analyzing the deviation between the transmission model prediction and the actual sensor data in each alarm event. This embodiment specifically illustrates the optimization steps and methods, namely data acquisition and training: Sensor data for each alarm event is recorded in the system and compared with the system's predicted values. Through deviation analysis, the system gradually adjusts the time threshold parameters in the transmission reachability matrix to ensure that the prediction model gradually improves its accuracy; Sliding window mechanism: To avoid over-reliance on historical data, this embodiment adopts a sliding window mechanism, retaining only the data from the most recent 90 days for training. Data outside the window is used to update the case library so that the system can quickly adapt to new park and building environments. Through this mechanism, the system can continuously adjust and optimize the model based on feedback from actual events, improving the system's real-time response capability.

[0064] This embodiment adds a visualization decision-making sandbox module to display the execution status of the transmission reachability matrix and response element combinations in real time. Operators can manually intervene through this module to adjust the system's response strategy. The manual intervention interface allows operators to manually adjust the time threshold parameters in the transmission reachability matrix or reorder the execution order of the response element combinations when the automatically generated response element combinations do not conform to the actual situation. This function ensures that the system can flexibly respond to various emergencies in complex environments, and all of these are extended implementation methods known to those skilled in the art.

[0065] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A fire early warning management system for property management parks, characterized in that, include: The conduction topology modeling module is used to generate a three-dimensional heat conduction network based on the building floor plan, and encodes the wall thermal conductivity and ventilation path parameters into a conduction reachability matrix. The matrix stores the physical association strength of each sensor node and the expected conduction time threshold. The dynamic verification engine module connects to the park's fire sensor network. When any sensor triggers an alarm, it activates the associated sensors along its transmission path to perform data cross-verification. Based on the transmission reachability matrix, it calculates the physical consistency between the alarm signal and the associated sensor data and outputs three levels of credibility labels, including physical credibility, logical suspicion, and physical conflict. Data consistency calculation: By comparing the direction and amplitude of signal data fluctuations between alarm sensors and associated sensors, a consistency index is calculated. When the direction of data fluctuations of a sensor alarm signal is opposite to that of at least two associated nodes in the transmission reachability matrix, it is automatically downgraded to a logically suspicious label and a manual review process is initiated. The response element dynamic assembly module includes a pre-stored library of basic units for detection elements, control elements, and suppression elements. The detection elements define fire feature identification rules, the control elements define equipment linkage logic, and the suppression elements define fire extinguishing resource scheduling strategies. When the credibility tag is physically credible, the response element combination is matched based on real-time fire boundary prediction to generate a contingency plan execution chain that includes execution order and triggering conditions. The topology migration adaptation module encodes building floor height and fire compartment area parameters into topology vectors, matches them with the historical park case library through similarity calculation, and outputs the initial combination recommendation scheme of response elements to the dynamic assembly module of response elements. The incremental learning module is connected to the output of the dynamic verification engine module. It records the deviation between the transmission model prediction value and the actual sensor data in each alarm event and optimizes the time threshold parameter in the transmission reachability matrix through offline training.

2. The property management park fire early warning management system as described in claim 1, characterized in that, In the conductive topology modeling module, the thermal conductivity of the wall is dynamically corrected by the infrared thermal imaging detection value of the building material, and the ventilation path parameters are updated by the airflow velocity data in the operation log of the park's fresh air system.

3. The property management park fire early warning management system as described in claim 1, characterized in that, The detection element includes a smoldering characteristic recognition rule, specifically: when the temperature rise rate is ≤2℃ / min and the direction of smoke concentration gradient change is opposite to the ventilation path in a confined space, an inert gas injection command is triggered and automatic spraying is prohibited.

4. The property management park fire early warning management system as described in claim 1, characterized in that, In the topology migration adaptation module, the similarity calculation adopts Euclidean distance metric, and assigns a weight coefficient of 0.6 to the floor height parameter and a weight coefficient of 0.4 to the fire compartment area parameter.

5. The property management park fire early warning management system as described in claim 1, characterized in that, The logic for generating the contingency plan execution chain includes: when the fire boundary prediction shows that it will spread to the adjacent zone within 30 seconds, an automatic delay start command for the smoke extraction system is added, and the smoke extraction operation is executed after the open flame is confirmed.

6. The property management park fire early warning management system as described in claim 1, characterized in that, The optimization process of the incremental learning module includes: adopting a sliding window mechanism, retaining only the alarm event data of the most recent 90 days for matrix parameter training, and converting historical data outside the window into a case library for storage.

7. The property management park fire early warning management system as described in claim 1, characterized in that, It also includes a visualization decision sandbox module, which displays a heatmap of the transmission accessibility matrix and the status of the contingency plan execution chain in real time, and provides a manual intervention interface to cover the automatically generated response element combination.

8. The property management park fire early warning management system as described in claim 7, characterized in that, In the visualization decision-making sandbox module, manual intervention operations include: manually adjusting the time threshold parameter of the transmission reachability matrix, or forcibly specifying the priority order of response element combinations.

9. A method for fire early warning management in property management parks, characterized in that, Includes the following steps: Step 1: Conductive topology modeling. Generate a three-dimensional heat conduction network based on the building floor plan. Encode the wall thermal conductivity and ventilation path parameters into a conduction reachability matrix. The matrix stores the physical association strength and expected conduction time threshold of each sensor node. Step 2: Dynamic verification. Monitor the park's fire sensor network in real time. When any sensor triggers an alarm, activate the associated sensors on its transmission path for data cross-verification. Calculate the physical consistency between the alarm signal and the associated sensor data based on the transmission reachability matrix, and output three levels of credibility labels including physically credible, logically suspicious, and physically conflicting. Data consistency calculation: By comparing the direction and amplitude of signal data fluctuations between alarm sensors and associated sensors, a consistency index is calculated. When the direction of data fluctuations of a sensor alarm signal is opposite to that of at least two associated nodes in the transmission reachability matrix, it is automatically downgraded to a logically suspicious label and a manual review process is initiated. Step 3: Dynamic assembly of response elements. When the credibility tag is physically credible, based on real-time fire boundary prediction, the response element combination is matched from the pre-stored basic unit library of detection elements, control elements, and suppression elements to generate a contingency plan execution chain containing the execution order and triggering conditions. Among them, the detection element defines the fire feature identification rules, the control element defines the equipment linkage logic, and the suppression element defines the fire extinguishing resource scheduling strategy. Step 4: Topology migration adaptation. The building floor height and fire compartment area parameters are encoded into topology vectors. The similarity is used to match the historical park case library and output the initial combination recommendation scheme of the response element to guide the initial configuration of the dynamic assembly module of the response element. Step 5: Incremental learning. Record the deviation between the propagation model's predicted value and the actual sensor data in each alarm event. Optimize the time threshold parameter in the propagation reachability matrix through offline training to improve the model's prediction accuracy.

Images