Fire alarm inspection system and method based on big data

Abstract

The invention discloses a fire alarm inspection system and method based on big data, and relates to the technical field of intelligent fire protection and public safety. The problems that a traditional fire fighting system is serious in data island phenomenon, single in risk perception dimension, static and rigid in early warning model, high in false alarm rate and low in inspection efficiency are solved. According to the method, a fire-fighting theme database is constructed, and based on a multi-dimensional dynamic analysis model, comprehensive analysis of space aggregation risks, time sequence evolution trends and equipment linkage logic is realized, and dynamic fire danger hidden danger indexes are calculated; a key monitoring object is positioned through a visual thermodynamic diagram, grading alarm is triggered, differentiated inspection tasks and an optimal route are generated based on the intelligence, and emergency disposal linkage is supported; closed-loop management from global risk perception, intelligent analysis and early warning to precise inspection treatment is realized, and early recognition and early warning accuracy and resource scheduling efficiency of fire hazards are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of smart fire protection and public safety technology, specifically a fire alarm inspection system and method based on big data. Background Technology

[0002] With the acceleration of urbanization and the continuous emergence of high-rise and super-large complexes, fire safety has become a top priority in safeguarding public safety and protecting people's lives and property. Traditional fire alarm and equipment inspection systems mainly rely on independent sensors with preset thresholds and periodic manual on-site inspections. However, this model has exposed many limitations in the increasingly complex modern building environment, such as information silos and response delays, and can no longer meet the development needs of smart fire protection.

[0003] The existing technology has the following problems: alarm judgment is singular, false alarm rate is high and early warning capability is lacking; risk perception is fragmented and lacks global quantitative assessment; inspection plan is rigid and lacks efficiency and pertinence; system response is passive and lacks intelligent decision support. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a fire alarm inspection system and method based on big data to solve the above-mentioned technical problem.

[0005] The first aspect of the present invention provides a fire alarm inspection system based on big data, comprising the following modules: Data acquisition module: Collects real-time data from fire sensors and static building information; Big data processing module: processes the collected data to form a fire safety-themed database; Risk Analysis Module: Based on a fire safety database, a risk assessment model is built to conduct multi-dimensional comprehensive analysis, calculate the dynamic fire hazard index, and generate a structured risk report; Visualized early warning module: Analyzes structured risk reports, generates heat maps of fire hazards, locates key monitoring targets, and triggers tiered alarm information; Intelligent inspection planning module: Based on the fire hazard index and structured risk briefing, and integrating hierarchical alarm information, it generates an inspection task list and inspection route plan, and pushes them to the mobile inspection terminal.

[0006] Preferably, collecting real-time data from fire sensors and static building information includes the following steps: Multiple fire sensors are installed on the construction site to continuously collect real-time data and simultaneously collect the static attributes of the sensors. The static data and sensor data are then transmitted to the back end through a data communication network. By calling the building information database API, static building information is obtained, including three-dimensional dimensional data and component attribute data of building structural components, and fire protection equipment information. Among them, building structural components include walls, columns, floor slabs, and stairs. The three-dimensional dimensional data includes the spatial coordinates, geometry, and dimensions of the components, and the component attribute data includes the material properties and fire resistance rating of the components. It also accesses the pre-set fire emergency response plan database to obtain standardized emergency response plan data associated with building functional zoning and equipment types. Historical sensor data sequences and historical fire event records are collected from fire protection system logs and the fire management platform.

[0007] Preferably, the collected data is processed to form a fire safety-themed database, including the following steps: The collected data is cleaned and its quality is monitored. Using predefined association identifiers, real-time and historical sensor data, static sensor data, three-dimensional dimension data and component attribute data of building structural components, fire equipment information and logical topology relationships are associated. By logically integrating these multi-source heterogeneous data, a unified data view that supports association queries is formed. Among them, the logical topology relationship of fire equipment is constructed based on the physical location information of fire equipment and the preset functional linkage rules. Based on historical sensor data and other equipment status information in the data view, historical baseline data is generated through statistical learning to characterize the normal operation status of various fire-fighting equipment containing sensors. Data views, historical baseline data, historical fire incident records, and standardized emergency response plan data are stored in a hierarchical storage system according to different access needs, and a fire-themed database is constructed.

[0008] Preferably, based on a fire safety database, a risk assessment model is constructed to conduct multi-dimensional comprehensive analysis, calculate a dynamic fire hazard index, and generate a structured risk report, including the following steps: From the data view of the fire protection database, obtain real-time sensor data, historical sensor data, three-dimensional dimension data and component attribute data of building structural components, sensor static attributes, and fire protection equipment information; Based on the data relationships established in the data view, spatial features, temporal features, and linkage state features are constructed to form feature vector data for model input. Construct a risk assessment model based on graph neural network with multi-task output capability and pre-train it; input real-time feature vector data into the trained risk assessment model to perform multi-dimensional comprehensive analysis including spatial dimension, time series and equipment linkage, and output risk heat map data, list of key prevention and control areas, key risk factor sequence and risk trend prediction, equipment status diagnosis conclusion, linkage verification results and fire confirmation degree value. Based on the sequence of key risk factors, the list of key prevention and control areas, and the diagnostic conclusions of equipment status, specific fire-fighting suggestions are matched and generated from the data of standardized emergency response plans. Based on the results of multi-dimensional comprehensive analysis of the risk assessment model, a quantitative dynamic fire hazard index is calculated by assigning weights to each indicator and performing linear weighting through a combined weight fusion algorithm based on the analytic hierarchy process. The results of multi-dimensional comprehensive analysis, fire-fighting recommendations, and dynamic fire hazard index are packaged together into a machine-readable document with a fixed format, namely a structured risk briefing.

[0009] Preferably, the risk assessment model performs a multi-dimensional comprehensive analysis, including the following steps: Input real-time feature vector data into the risk assessment model for multi-dimensional comprehensive analysis: Spatial dimension analysis: Abnormal readings are identified based on the comparison results between real-time sensor data and historical baseline data; at the same time, the alarm status of the sensor is determined by comparing whether the real-time reading of the sensor exceeds the preset threshold in its static data; abnormal readings and alarm status are used as risk factors; combined with the physical location of the sensor, the spatial clustering effect of risk factors is evaluated to generate risk heat map data; based on the three-dimensional dimension data and component attribute data of building structural components, the potential spread path and range of risk factors in the building structural network are simulated to generate a list of key prevention and control areas; Time series analysis: By comparing real-time sensor data and historical baseline data, short-term abnormal data points and long-term trend-deteriorating sensor parameters are identified and serialized into key risk factor sequences; based on the changing patterns revealed by historical baseline data, risk trend predictions of key risk factor sequences are generated. Equipment linkage analysis: Based on the logical topology of fire protection equipment in the data view, equipment groups are constructed. By verifying the consistency of the equipment status within the group with the preset linkage logic in terms of causality and timing, potential causal relationships are inferred. By combining the consistency verification results and potential causal relationships, the fire confirmation degree value is quantified, and equipment failures, system false alarms and real fires are identified, generating equipment status diagnosis conclusions and linkage verification results.

[0010] Preferably, the calculation of the dynamic fire hazard index includes the following steps: For the risk heatmap data generated by spatial dimension analysis, calculate the average risk value of all data points as the comprehensive spatial risk value (SRV). Based on the key risk factor sequence obtained from time series analysis, the deviation of each factor reading from its historical baseline data is calculated to obtain a comprehensive severity index (ADR). Based on the change trajectory obtained by regression analysis in time series analysis, the average rate of change of this trajectory function in the future prediction time interval is calculated as the slope index M characterizing the trend of risk evolution. The fire confirmation degree (FCD) value is directly taken from the output of the equipment linkage analysis. Based on historical fire incident records in the fire protection subject database, the static historical risk value H is determined by the number of historical fires and the area burned by fires. The spatial risk value, severity of key risk factor sequences, slope of risk trend prediction, and fire confirmation value of the risk heat map data output by static historical risk values ​​and multi-dimensional comprehensive analysis are normalized to the numerical range of [0,1], and weight coefficients are assigned to them respectively, with the sum of each weight coefficient being 1. The dynamic fire hazard index DFRI is obtained by linearly weighting and summing each index value with its corresponding weight coefficient.

[0011] Preferably, the process of parsing structured risk reports, generating a fire hazard heat map, locating key monitoring targets, and triggering tiered alarm information includes the following steps: The structured risk briefing is analyzed to extract risk heat map data, a list of key prevention and control areas, a sequence of key risk factors, equipment status diagnosis conclusions, fire handling suggestions, and risk trend predictions. A three-dimensional spatial model of the building is constructed based on the three-dimensional dimensional data of building structural components in the fire protection theme database. The extracted risk heat map data is overlaid with the three-dimensional spatial model of the building to generate an initial heat map, and the risk evolution trend is marked on the initial heat map based on the risk trend predictions. The list of key prevention and control areas is located in the three-dimensional spatial model of the building. At the same time, the associated risk factors and equipment information are obtained based on the key risk factor sequence and equipment status diagnosis conclusion. A key area and its associated risk factors and equipment information are defined as a key monitoring object and assigned a unique tracking ID. The dynamic fire hazard index is used as the comprehensive alarm score, and a preset scoring threshold range and corresponding alarm level are set; the comprehensive alarm score is compared with the preset scoring threshold to determine the final alarm level. The alarm level, the set of key monitoring object IDs, and the associated fire response suggestions are packaged into a standard alarm information package; depending on the alarm level, it is pushed to the corresponding mobile terminal through the communication channel.

[0012] Preferably, based on the fire hazard index and structured risk briefing, and integrating graded alarm information, a differentiated inspection task list and inspection route plan are generated and pushed to the mobile inspection terminal, including the following steps: Analyze structured risk briefings and graded alarm information to extract key monitoring objects and their tracking IDs, fire response suggestions, equipment status diagnosis conclusions, alarm levels, and three-dimensional dimensional data of building structural components; Based on alarm levels, equipment status diagnostic conclusions, and fire response recommendations, inspection priorities are assigned to each key monitoring object, and standardized inspection instructions are generated, thus forming an inspection task list sorted by priority. A spatial navigation network is constructed based on the three-dimensional dimensional data of building structural components, and the spatial distance of inspection points, inspection priority, and evacuation route access status are taken as cost factors for path planning. All points in the inspection task list are taken as necessary nodes. The optimal inspection route is solved by the path planning algorithm to ensure that high-priority points are placed at the beginning of the route sequence and generate an electronic route map with directional navigation guidance. The inspection task list and electronic route map are packaged into a mobile inspection task package, pushed to the designated mobile inspection terminal, and a task status synchronization mechanism is established to receive inspection feedback and status updates.

[0013] A second aspect of the present invention provides a fire alarm inspection method based on big data, comprising the following steps: S1: Collects real-time data from fire sensors and static building information; S2: Process the collected data to form a unified fire protection-themed database; S3: Based on the fire protection database, a risk assessment model is constructed to conduct multi-dimensional comprehensive analysis, calculate the dynamic fire hazard index, and generate a structured risk briefing; S4: Analyze the structured risk briefing, generate a fire hazard heat map, locate key monitoring targets, and trigger tiered alarm information; S5: Based on the fire hazard index, structured risk briefing, and graded alarm information, generate differentiated inspection task lists and optimal inspection routes, and push them to mobile inspection terminals.

[0014] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention effectively identifies false alarms and achieves early warning through multi-dimensional comprehensive analysis and cross-validation; This invention generates dynamic risk indices and visualized heat maps based on a unified data view, realizing the transformation from fragmented information to global quantitative assessment; This invention dynamically generates differentiated inspection tasks and optimal routes based on real-time data and alarm levels, realizing the upgrade from rigid planning to precise and efficient scheduling; This invention improves emergency response capabilities by providing structured briefings that integrate risk diagnosis and handling suggestions, transforming passive response into proactive decision support. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the module flow of the present invention.

[0016] Figure 2 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Please see Figure 1 This invention is a fire alarm inspection system based on big data, comprising the following modules: Data acquisition module: Collects real-time data from fire sensors and static building information; Big data processing module: processes the collected data to form a fire safety-themed database; Risk Analysis Module: Based on a fire safety database, a risk assessment model is built to conduct multi-dimensional comprehensive analysis, calculate the dynamic fire hazard index, and generate a structured risk report; Visualized early warning module: Analyzes structured risk reports, generates heat maps of fire hazards, locates key monitoring targets, and triggers tiered alarm information; Intelligent inspection planning module: Based on the fire hazard index and structured risk briefing, and integrating hierarchical alarm information, it generates an inspection task list and inspection route plan, and pushes them to the mobile inspection terminal.

[0019] Specifically, multiple fire sensors are installed on the building site to continuously collect real-time data and simultaneously collect static sensor attributes. The data is transmitted to the backend via a data communication network. The building information database API is called to obtain static building information. The pre-set fire response plan database is accessed to obtain standardized response plan data. Historical sensor data sequences and historical fire event records are collected from the fire system logs and fire management platform.

[0020] Data is cleaned and its quality is monitored. Using predefined association identifiers, the cleaned real-time and historical sensor data, sensor static data, and building static data are logically associated to form a data view. A historical baseline for normal equipment operation is generated based on historical data. Data is stored in layers in time-series databases, relational databases, and data lakes according to access characteristics to build a fire protection-themed database.

[0021] Based on the feature vectors constructed from data associations, representing spatial, temporal, and linkage states, these vectors are input into a pre-trained risk assessment model built on a graph neural network for multi-dimensional comprehensive analysis. The model outputs data including risk heatmaps, key prevention and control areas, key risk factors and risk trend predictions, equipment status diagnoses, linkage verification results, and fire confirmation rates. By combining the above results and using a weighted fusion algorithm, a dynamic fire hazard index is calculated and packaged together with matching fire response suggestions and multi-dimensional comprehensive analysis results into a structured risk briefing.

[0022] The structured risk briefing is analyzed, and the heat map data is overlaid and rendered with the 3D building model to generate a fire hazard heat map. Areas with rising risks are dynamically marked. Key prevention and control areas are located, and their associated risk factors and equipment information are labeled to form key monitoring objects. The alarm level is determined based on the dynamic fire hazard index and preset thresholds, and the alarm information and handling suggestions are pushed to the relevant personnel terminals.

[0023] Based on risk briefings and alarm information, inspection priorities are assigned to key targets, standardized inspection instructions are generated, and an inspection task list is formed. Based on the building space network, the optimal inspection route is planned by comprehensively considering spatial distance, inspection priority, and evacuation route access status to ensure that high-priority points are handled first. The task list and electronic route map are packaged and pushed to mobile terminals to achieve closed-loop management from risk perception to inspection execution.

[0024] In one embodiment of the present invention, the acquisition of real-time data from fire sensors and static building information includes the following steps: Multiple fire sensors are installed on the construction site to continuously collect real-time data and simultaneously collect the static attributes of the sensors. The static data and sensor data are then transmitted to the back end through a data communication network. By calling the building information database API, static building information is obtained, including three-dimensional dimensional data and component attribute data of building structural components, and fire protection equipment information. Among them, building structural components include walls, columns, floor slabs, and stairs. The three-dimensional dimensional data includes the spatial coordinates, geometry, and dimensions of the components, and the component attribute data includes the material properties and fire resistance rating of the components. It also accesses the pre-set fire emergency response plan database to obtain standardized emergency response plan data associated with building functional zoning and equipment types. Historical sensor data sequences and historical fire event records are collected from fire protection system logs and the fire management platform.

[0025] Specifically, various types of fire sensors are deployed inside the building to form a three-dimensional monitoring network. These include: smoke detectors, installed in most indoor areas such as rooms and corridors, for early detection of smoke particles; heat detectors, installed in kitchens, electrical rooms, and other places where high temperatures may exist under normal circumstances, serving as a supplement to smoke detectors or as the main detector in specific areas; water pressure sensors, installed in the main pipes, branch pipes, and ends of the fire sprinkler system to monitor the water pressure status of the pipeline network in real time; linear heat detectors, laid in linear areas such as above transformers and shelves in large warehouses to monitor the temperature of continuous spaces; and combustible gas detectors, installed in gas valve rooms, laboratories, and other locations where combustible gas leaks may occur. These sensors continuously collect real-time environmental data, such as smoke concentration, temperature, and pressure, as well as their own static attributes, such as device ID, model, and installation location. Both real-time and static sensor data are transmitted to the back-end data processing center via a wireless communication network.

[0026] The building information database invoked by this system is a real and accessible database. It is not part of this invention but rather an existing infrastructure utilized by this invention. By calling the standard API of this database, static building information is obtained, including three-dimensional dimensional data and component attribute data of building structural components, and fire equipment information. Among them, building structural components include walls, columns, floor slabs, and stairs; three-dimensional dimensional data includes the spatial coordinates, geometry, and dimensions of the components; component attribute data includes the material properties and fire resistance rating of the components; fire equipment information includes the location and model of fire hydrants, sprinkler heads, alarms, and other fire equipment; at the same time, based on the three-dimensional dimensional data of building structural components, the spatial topological relationships of the building structure are obtained, that is, the connectivity, adjacency, and containment logic of the internal spaces of the building, for example, clearly identifying which door connects a certain room to the evacuation corridor.

[0027] The fire emergency response plan database is pre-built and accessible, typically developed by fire management departments based on national standards, building risk assessments, and specific fire protection system designs. The system accesses standardized emergency response plan data associated with specific functional areas and equipment types within the current building through a database query interface. These plans are stored in a structured format, clearly defining the response procedures, equipment linkage sequences, and safety precautions for different fire scenarios.

[0028] Collect historical sensor data sequences and historical fire incident records from the building's original fire protection system logs and fire management platform. The historical fire incident records include the time, precise location, and final handling conclusions of past actual fires, equipment failures, false alarms, and other events.

[0029] In one embodiment of the present invention, the collected data is processed to form a fire protection-themed database, including the following steps: The collected data is cleaned and its quality is monitored. Using predefined association identifiers, real-time and historical sensor data, static sensor data, three-dimensional dimension data and component attribute data of building structural components, fire equipment information and logical topology relationships are associated. By logically integrating these multi-source heterogeneous data, a unified data view that supports association queries is formed. Among them, the logical topology relationship of fire equipment is constructed based on the physical location information of fire equipment and the preset functional linkage rules. Based on historical sensor data and other equipment status information in the data view, historical baseline data is generated through statistical learning to characterize the normal operation status of various fire-fighting equipment containing sensors. Data views, historical baseline data, historical fire incident records, and standardized emergency response plan data are stored in a hierarchical storage system according to different access needs, and a fire-themed database is constructed.

[0030] Specifically, the collected raw data undergoes preprocessing, including noise filtering, outlier removal, and data normalization for real-time sensor data; and integrity verification and format standardization are performed on sensor static data and building static information. Throughout this process, key quality indicators such as data reception delay, packet loss rate, and outlier ratio are monitored simultaneously; when any indicator falls below a preset threshold, a data quality alarm will be automatically triggered.

[0031] A predefined global association identifier system is established, comprising spatial location IDs and unique device codes. These identifiers logically link real-time and historical sensor data, static sensor data, 3D dimensional data and component attribute data of building structural members, fire equipment information and their logical topological relationships, and standardized emergency response plan data across sources, forming a unified data view that supports association queries. Querying this view allows obtaining, for example, the recent hourly readings of all smoke detectors located in the fire compartment of the 5th floor east zone, along with their associated sprinkler head models and the fire resistance rating of the components in the room. The spatial location ID is directly adopted from the predefined spatial code in the building static information, serving as the primary association key to link sensors, fire equipment, and building components in the same physical space. The unique device code is obtained by fusing the device's factory serial number and logical address code, used to associate real-time fire sensor data, static sensor data, and the corresponding 3D dimensional data of the building structural members in the building static information.

[0032] The logical topology relationship construction process for fire protection equipment is as follows: First, basic data is obtained, including the physical location information of the fire protection equipment and pre-set functional linkage rules. Equipment types include smoke detectors, heat detectors, manual alarm buttons, audible and visual alarms, smoke exhaust valves, etc. Second, pre-set functional linkage rules are obtained. These rules are derived from national fire protection design codes with legal force and engineering guidance significance (such as the "Code for Fire Protection Design of Buildings" GB50016), building fire protection design documents, and emergency plans approved by the fire protection authorities. After the obtained rules are structured, they are stored in the system's rule base. The rules use the explicit format "When [triggering equipment type / number] alarms in [area], [target equipment type / number] should be activated in [target area]" to define the functional linkage logic between fire protection equipment.

[0033] After acquiring the basic data, spatial clustering analysis is performed. Based on the physical location information of fire protection equipment, the DBSCAN density clustering algorithm is used to automatically group devices that are physically adjacent or located in the same enclosed space into physical device groups. For example, all smoke detectors and heat detectors in the same room are grouped into one group. Subsequently, rule matching and logical group construction are performed. By parsing the trigger conditions and linkage targets in the preset function linkage rules, the rules are matched with the formed physical device groups to establish the logical topological relationship between the devices. For example, according to the rule "when a smoke detector alarms in the east corridor of the 5th floor, the audible and visual alarms in that area and adjacent areas should be activated and the electric fire dampers on this floor should be closed," the system first uses spatial clustering to find all smoke detectors located in the specified area to form physical group A, and the audible and visual alarms and electric fire dampers on this floor in the corresponding area to form physical groups B and C, respectively. Then, logical groups are created according to the rules, defining any detector in group A as the trigger source and the devices in groups B and C as linkage targets.

[0034] Based on historical sensor data in the data view, a statistical process control method is used to establish a dynamic historical baseline for each sensor. This baseline is a dynamic data profile that adapts to different time periods. The construction process of the historical baseline includes: first, grouping and aggregating the historical data of each sensor according to the time dimension to construct a dataset under multiple time slices; then, using the statistical process control method, calculating the central trend and dispersion of the data for each time slice; finally, based on these statistics, defining a dynamic normal fluctuation range and more warning threshold upper and lower limits for each sensor at different time periods.

[0035] Data views, historical baseline data, historical fire incident records, and standardized emergency response plan data are stored in a hierarchical storage system according to different access requirements, while a fire-themed database is also constructed. Real-time sensor data is stored in a time-series database to meet the performance requirements of real-time monitoring and rapid backtracking; data requiring complex relational queries and analyses, including historical baseline data, three-dimensional dimension data and component attribute data of building structural components, fire equipment information and its logical topological relationships, and standardized emergency response plan data, are stored in a relational database; raw historical data is imported into a data lake for low-cost storage and used for long-term trend analysis, model retraining, and offline computing tasks such as accident backtracking.

[0036] In one embodiment of the present invention, a risk assessment model is constructed based on a fire safety database to conduct multi-dimensional comprehensive analysis, calculate a dynamic fire hazard index, and generate a structured risk report, including the following steps: From the data view of the fire protection database, obtain real-time sensor data, historical sensor data, three-dimensional dimension data and component attribute data of building structural components, sensor static attributes, and fire protection equipment information; Based on the data relationships established in the data view, spatial features, temporal features, and linkage state features are constructed to form feature vector data for model input. Construct a risk assessment model based on graph neural network with multi-task output capability and pre-train it; input real-time feature vector data into the trained risk assessment model to perform multi-dimensional comprehensive analysis including spatial dimension, time series and equipment linkage, and output risk heat map data, list of key prevention and control areas, key risk factor sequence and risk trend prediction, equipment status diagnosis conclusion, linkage verification results and fire confirmation degree value. Based on the sequence of key risk factors, the list of key prevention and control areas, and the diagnostic conclusions of equipment status, specific fire-fighting suggestions are matched and generated from the data of standardized emergency response plans. Based on the results of multi-dimensional comprehensive analysis of the risk assessment model, a quantitative dynamic fire hazard index is calculated by assigning weights to each indicator and performing linear weighting through a combined weight fusion algorithm based on the analytic hierarchy process. The results of multi-dimensional comprehensive analysis, fire-fighting recommendations, and dynamic fire hazard index are packaged together into a machine-readable document with a fixed format, namely a structured risk briefing.

[0037] Specifically, real-time sensor data, historical sensor data, 3D dimensional data of building structural components, component attribute data, sensor static attributes, fire equipment information, and their logical topological relationships are extracted from the data view of the fire protection database. Based on this associated data, three types of features are constructed in parallel: spatial feature construction integrates the physical coordinates of sensors with the 3D dimensional data and component attribute data of building structural components, employing spatial analysis and gridding techniques to quantify the spatial distribution of risk points and their impact on the building environment; temporal feature construction dynamically compares real-time sensor data with historical baseline data of the same equipment, using statistical process control and trend fitting methods to identify short-term abnormal fluctuations and long-term deterioration trends of data points; and linkage status feature construction utilizes the logical topological relationships of fire equipment to verify the consistency of readings within functionally linked equipment groups, identifying equipment malfunctions or system false alarms. These constructed features are combined into feature vectors for model input.

[0038] The specific consistency verification process includes defining a preset causal-temporal constraint rule for each group of interconnected devices. This rule contains two core elements: causal constraints, used to check whether the states of associated devices conform to logical causal relationships. For example, the rule can be defined as follows: when a smoke detector alarms, its linked audible and visual alarms should become active within a specified time. Temporal constraints, used to check whether the device state changes occur within a preset time window. By monitoring real-time data streams, it is determined whether the states of all devices in the group meet the preset rules within the time window: if all are met, it is considered consistent; if any device's state does not meet the rules or its response times out, it is considered inconsistent. Finally, the system quantifies this inconsistency into a specific feature value, which serves as an important basis for identifying device failures, communication interruptions, or system false alarms.

[0039] A risk assessment model with multi-task output capability based on graph neural network is constructed. During the training phase, historical feature vector data and their corresponding historical fire event conclusions are used as supervision signals to learn the high-dimensional mapping relationship between complex features and multiple risk outcomes.

[0040] By utilizing a pre-trained risk assessment model, a multi-dimensional comprehensive analysis of real-time feature vectors is performed, outputting structured results in three dimensions: In the spatial dimension, risk heat map data is output to quantitatively display the risk density distribution within the building, and a list of key prevention and control areas is generated accordingly; In the temporal dimension, key risk factor sequences are identified and output, and risk trend predictions are generated; In the equipment linkage dimension, equipment status diagnosis conclusions and fire confirmation values ​​are output, effectively distinguishing between real fires and equipment malfunctions.

[0041] Based on the list of key risk factors and key prevention and control areas output by the risk assessment model, specific fire response recommendations are automatically matched and generated from the standard emergency response plan database. Simultaneously, using the analytic hierarchy process (AHP), reasonable weights are assigned to multiple indicators such as spatial risk value, severity of risk factors, trend slope, and fire confirmation rate. Through linear weighted fusion, a comprehensive and quantitative dynamic fire hazard index is calculated.

[0042] The results of multi-dimensional comprehensive analysis, fire-fighting recommendations, and dynamic fire hazard index are packaged together into a machine-readable document with a fixed schema, namely a structured risk briefing.

[0043] In one embodiment of the present invention, the risk assessment model performs a multi-dimensional comprehensive analysis, including the following steps: Input real-time feature vector data into the risk assessment model for multi-dimensional comprehensive analysis: Spatial dimension analysis: Abnormal readings are identified based on the comparison results between real-time sensor data and historical baseline data; at the same time, the alarm status of the sensor is determined by comparing whether the real-time reading of the sensor exceeds the preset threshold in its static data; abnormal readings and alarm status are used as risk factors; combined with the physical location of the sensor, the spatial clustering effect of risk factors is evaluated to generate risk heat map data; based on the three-dimensional dimension data and component attribute data of building structural components, the potential spread path and range of risk factors in the building structural network are simulated to generate a list of key prevention and control areas; Time series analysis: By comparing real-time sensor data and historical baseline data, short-term abnormal data points and long-term trend-deteriorating sensor parameters are identified and serialized into key risk factor sequences; based on the changing patterns revealed by historical baseline data, risk trend predictions of key risk factor sequences are generated. Equipment linkage analysis: Based on the logical topology of fire protection equipment in the data view, equipment groups are constructed. By verifying the consistency of the equipment status within the group with the preset linkage logic in terms of causality and timing, potential causal relationships are inferred. By combining the consistency verification results and potential causal relationships, the fire confirmation degree value is quantified, and equipment failures, system false alarms and real fires are identified, generating equipment status diagnosis conclusions and linkage verification results.

[0044] Specifically, inputting real-time feature vector data into the risk assessment model for multi-dimensional comprehensive analysis includes the following: Spatial dimension analysis includes the following: First, the real-time readings of each sensor are compared with the historical baseline data of the same type of equipment in the same location in the fire protection database to identify abnormal readings that significantly deviate from the normal pattern; the preset thresholds in the static attributes of the sensor, determined by national standards or on-site commissioning (e.g., the alarm threshold for a smoke detector is 15%obs / m), are read, and it is determined whether the real-time sensor data exceeds the preset threshold to determine its alarm status; both abnormal readings and alarm status are defined as risk factors and bound to the physical coordinates of the sensor.

[0045] The kernel density estimation method is used to assess the degree of clustering of risk factors in geospatial space. Specifically, a decay function is defined with each risk factor as the center, and its density contribution in the surrounding space is calculated. Finally, by superimposing the contributions of all risk factors, a continuous risk density distribution surface is generated. According to the preset gradient, the risk density distribution surface is quantized into visualized risk heat map data, which is rendered on the interface using a color spectrum from green (low risk) to red (high risk).

[0046] Based on the three-dimensional dimensional data and component attribute data of building structural components, the building space is abstracted into a building network model, with spatial units such as rooms as nodes and connecting paths such as doors and corridors as edges. A graph traversal algorithm is used to simulate the potential spread path and range of smoke or heat in the building network model, starting from the nodes where the identified risk factors are located, while fully considering the blocking effect of components such as fire compartments.

[0047] The analysis involves overlaying high-density areas in the risk heatmap with areas predicted to be rapidly affected in the spread simulation. For example, an area with moderate risk density but located in a critical passageway, and whose spread simulation shows would affect the entire floor, would be designated as a priority area. Based on this, a list of key prevention and control areas is generated, with each area in the list associated with the core risk factors that led to its designation as a priority area.

[0048] Time series analysis includes the following: multi-timescale analysis of real-time sensor data, comparing real-time sensor data with historical baseline data from devices of the same location and type; at the short-term scale, using statistical process control methods to monitor whether real-time data points exceed upper and lower control limits calculated based on historical data to identify sudden abnormal data points; at the long-term scale, using regression analysis techniques to fit data from the past few weeks or months to obtain a trajectory of change, and analyzing whether the slope direction of the fitted data remains consistent within a continuous time window to determine whether the device exhibits a stable deterioration trend. For example, if a water pressure sensor reading shows a downward trend over several consecutive analysis periods, and its rate of decline (slope) remains around 0.1 Bar per month, it can be determined that the decline is stable.

[0049] The identified short-term abnormal data points and sensors exhibiting significant deterioration trends, along with their physical parameters, are structured and serialized into a key risk factor sequence. Simultaneously, based on the long-term change trajectory revealed by regression analysis, numerical extrapolation is performed to extend this trajectory function forward, thereby generating a quantitative risk trend prediction for this risk factor over a future period, such as predicting that the water pressure will fall below the minimum allowable value in 12 hours.

[0050] Equipment linkage analysis includes the following: Based on the associated fire protection equipment information and their logical topology relationships in the data view, sensors and other fire protection equipment that are physically adjacent or functionally linked are automatically grouped into equipment groups. The readings or states of the equipment within the group are then verified to ensure they conform to predefined causal and temporal constraints. For example, a constraint rule can be defined as: when smoke detector A alarms, its linked smoke exhaust valve B should open within 30 seconds, and the audible and visual alarm C should activate within 10 seconds. These state changes are monitored within a preset time window. If all associated equipment operates as expected, it is considered consistent; otherwise, it is marked as inconsistent. Finally, a quantitative consistency score is output, such as the proportion of satisfied constraint rules to the total number of rules.

[0051] The temporal causal discovery algorithm is used to analyze the temporal order and statistical dependence of historical data changes of each device in the group. By calculating the conditional probability of subsequent devices responding to changes in the state of the leading device within a preset time window in historical events, the causal confidence is quantified. For example, in historical data, if smoke detector A alarms 100 times, and heat detector B follows suit and alarms within 30 seconds 92 times, then the causal confidence of A causing B is 0.92.

[0052] A weighted fusion algorithm is used to calculate a fire confirmation score between 0 and 1, combining the consistency score and causal confidence score. The higher the score, the more likely the signal is a real fire. Finally, based on the fire confirmation score, the system combines the specific results of the consistency check with the relationships inferred from causal reasoning, and follows predefined decision rules (e.g., if the fire confirmation score is >0.8 and there are no non-critical equipment failures, it is determined to be a real fire). The system then identifies isolated equipment failures, false alarms, and real fires, and outputs structured equipment status diagnostic conclusions and linkage verification results.

[0053] In one embodiment of the present invention, calculating the dynamic fire hazard index includes the following steps: For the risk heatmap data generated by spatial dimension analysis, calculate the average risk value of all data points as the comprehensive spatial risk value (SRV). Based on the key risk factor sequence obtained from time series analysis, the deviation of each factor reading from its historical baseline data is calculated to obtain a comprehensive severity index (ADR). Based on the change trajectory obtained by regression analysis in time series analysis, the average rate of change of this trajectory function in the future prediction time interval is calculated as the slope index M characterizing the trend of risk evolution. The fire confirmation degree (FCD) value is directly taken from the output of the equipment linkage analysis. Based on historical fire incident records in the fire protection subject database, the static historical risk value H is determined by the number of historical fires and the area burned by fires. The spatial risk value, severity of key risk factor sequences, slope of risk trend prediction, and fire confirmation value of the risk heat map data output by static historical risk values ​​and multi-dimensional comprehensive analysis are normalized to the numerical range of [0,1], and weight coefficients are assigned to them respectively, with the sum of each weight coefficient being 1. The dynamic fire hazard index DFRI is obtained by linearly weighting and summing each index value with its corresponding weight coefficient.

[0054] Specifically, for the risk heatmap data generated by spatial dimension analysis, the average risk value of all data points is calculated as the comprehensive spatial risk value (SRV). Based on the key risk factor sequence obtained from time series analysis, the deviation of each factor reading from its historical baseline data is calculated, and the deviation of all factors is weighted and summed to obtain the comprehensive severity index ADR. Based on the change trajectory obtained by regression analysis in time series analysis, the average rate of change of this trajectory function in the future prediction time interval is calculated as the slope index M characterizing the risk trend prediction. The fire confirmation degree (FCD) value is directly taken from the output of the equipment linkage analysis. Based on historical fire incident records in a fire safety database, the static historical risk value H is determined by the number of historical fires and the historical average fire area. The specific process is as follows: First, the number of fires occurring in the assessment area within a preset statistical period (e.g., the past 5 years) is statistically analyzed to calculate the annual average fire frequency F (unit: times / year); and the burned area of ​​each fire is statistically analyzed to calculate the annual average burned area S (unit: square meters / year). Then, the two indicators F and S, which have different dimensions, are mapped to the [0,1] interval using a min-max normalization method, transforming them into dimensionless values. The calculation formula is: The normalized value of fire frequency was obtained. ,in and The minimum and maximum values ​​of F are defined in the region being evaluated; calculated using the formula: The normalized value of the burned area was obtained. ,in and The minimum and maximum values ​​of S are defined for the evaluated region. Finally, a static historical risk value H is obtained by linear weighted fusion using preset weights. α and β are weighting coefficients, and α+β=1. The weighting coefficients α and β are preset values ​​determined based on the correlation and contribution of historical fire frequency, burned area and overall risk level through historical data analysis, and in accordance with the principle of frequency taking precedence over severity in public safety risk assessment. For example, in one embodiment, they are set to α=0.6 and β=0.4.

[0055] The spatial risk value, severity of key risk factor sequences, slope of risk trend prediction, and fire confirmation value of the risk heat map data output by static historical risk values ​​and multi-dimensional comprehensive analysis are normalized to the numerical range of [0,1]. Based on historical data analysis, a weight configuration reflecting the degree of influence of each indicator on fire hazard is obtained, resulting in a set of weight coefficients that meet the normalization condition. For example, in one embodiment, the weight configuration scheme can be: W1=0.25, W2=0.20, W3=0.15, W4=0.30, W5=0.10, and the sum of each weight coefficient is 1.

[0056] The dynamic fire hazard index is obtained by linearly weighting and summing the values ​​of each indicator with their corresponding weight coefficients. The calculation formula is: DFRI=(SRV·W1)+(ADR·W2)+(M·W3)+(FCD·W4)+W5·(H·e^(-λt)); The Dynamic Fire Risk Index (DFRI) is obtained, where SRV represents the spatial risk value; ADR represents the severity; M represents the slope; FCD represents the fire confirmation value; H represents the static historical risk value; t represents the time interval in years from the most recent historical fire; λ represents the predefined attenuation coefficient; and W1, W2, W3, W4, and W5 are the weighting coefficients of the corresponding indicators.

[0057] In one embodiment of the present invention, parsing the structured risk briefing, generating a fire hazard heat map, locating key monitoring targets, and triggering tiered alarm information includes the following steps: The structured risk briefing is analyzed to extract risk heat map data, a list of key prevention and control areas, a sequence of key risk factors, equipment status diagnosis conclusions, fire handling suggestions, and risk trend predictions. A three-dimensional spatial model of the building is constructed based on the three-dimensional dimensional data of building structural components in the fire protection theme database. The extracted risk heat map data is overlaid with the three-dimensional spatial model of the building to generate an initial heat map, and the risk evolution trend is marked on the initial heat map based on the risk trend predictions. The list of key prevention and control areas is located in the three-dimensional spatial model of the building. At the same time, the associated risk factors and equipment information are obtained based on the key risk factor sequence and equipment status diagnosis conclusion. A key area and its associated risk factors and equipment information are defined as a key monitoring object and assigned a unique tracking ID. The dynamic fire hazard index is used as the comprehensive alarm score, and a preset scoring threshold range and corresponding alarm level are set; the comprehensive alarm score is compared with the preset scoring threshold to determine the final alarm level. The alarm level, the set of key monitoring object IDs, and the associated fire response suggestions are packaged into a standard alarm information package; depending on the alarm level, it is pushed to the corresponding mobile terminal through the communication channel.

[0058] Specifically, the process involves analyzing structured risk reports with fixed patterns to extract risk heatmap data, a list of key prevention and control areas, sequences of key risk factors, equipment status diagnostic conclusions, fire response recommendations, and risk trend prediction data. Subsequently, a 3D spatial model of the building is constructed based on the 3D dimensional data of building structural components from a fire safety database. The extracted risk heatmap data is then spatially overlaid on the 3D model, mapping the geographic coordinates (X, Y) of each risk data point to the corresponding ground coordinates (X, Y, Z) of the 3D model. Based on the risk intensity value (Z risk), a continuous color gradient is rendered on the surface of the 3D model using vertex shading technology to generate an initial heatmap. The colors follow a continuous spectrum from green to red, representing low to high risk. Simultaneously, based on the risk trend prediction data, dynamic visual markers with flashing boundaries are added to areas on the heatmap where the risk is increasing, to indicate the evolving risk trend.

[0059] The key prevention and control areas are precisely located in the three-dimensional spatial model of the building, and the key risk factor sequence that led to the area being listed as a key area is associated with the equipment status diagnosis conclusions. Through spatial clustering algorithm, high-risk areas that are spatially continuous and have similar risk factors are aggregated into a logical unit. This logical unit (i.e., a key area), the key risk factor sequence that led to its listing as a key area, and the relevant equipment status diagnosis conclusions in the area are packaged together and defined as a key monitoring object, and a unique tracking ID is assigned to each object.

[0060] The dynamic fire hazard index, which has been normalized and ranges from [0,1], is used as the comprehensive alarm score. A three-level alarm mechanism is implemented based on the preset scoring threshold range: Level 1 alarm (DFRI≥0.8), Level 2 alarm (0.6≤DFRI<0.8), and Level 3 alarm (0.4≤DFRI<0.6). The final alarm level is automatically determined by comparing the comprehensive score with the threshold range in real time.

[0061] The alarm level, the set of key monitoring object IDs, and the associated fire response suggestions are packaged into a standard alarm information package. Depending on the alarm level, the information is distributed through differentiated communication channels: Level 1 alarms trigger SMS, in-app push notifications, and voice calls simultaneously; Level 2 alarms enable SMS and in-app push notifications; Level 3 alarms only send in-app push notifications. Alarm information will be automatically routed to the terminals of emergency personnel at the corresponding level and with the appropriate responsibilities, and a confirmation receipt mechanism will be established to ensure that the information is delivered.

[0062] In one embodiment of the present invention, based on the fire hazard index and structured risk briefing, and integrating graded alarm information, an inspection task list and inspection route plan are generated and pushed to the mobile inspection terminal, including the following steps: Analyze structured risk briefings and graded alarm information to extract key monitoring objects and their tracking IDs, fire response suggestions, equipment status diagnosis conclusions, alarm levels, and three-dimensional dimensional data of building structural components; Based on alarm levels, equipment status diagnostic conclusions, and fire response recommendations, inspection priorities are assigned to each key monitoring object, and standardized inspection instructions are generated, thus forming an inspection task list sorted by priority. A spatial navigation network is constructed based on the three-dimensional dimensional data of building structural components, and the spatial distance of inspection points, inspection priority, and evacuation route access status are taken as cost factors for path planning. All points in the inspection task list are taken as necessary nodes. The optimal inspection route is solved by the path planning algorithm to ensure that high-priority points are placed at the beginning of the route sequence and generate an electronic route map with directional navigation guidance. The inspection task list and electronic route map are packaged into a mobile inspection task package, pushed to the designated mobile inspection terminal, and a task status synchronization mechanism is established to receive inspection feedback and status updates.

[0063] Specifically, it receives and parses structured risk briefings and graded alarm information, extracting key monitoring objects and their unique tracking IDs, fire response suggestions, equipment status diagnosis conclusions, alarm levels, and three-dimensional dimensional data of building structural components.

[0064] Based on alarm levels as the core decision-making criterion and incorporating the urgency of faults reflected in equipment status diagnostic conclusions, inspection priorities are assigned to each key monitoring object. Combining the associated fire response recommendations for each key monitoring object with the specific fault descriptions in the equipment status diagnostic conclusions, executable standardized inspection instructions are automatically generated. For example, for diagnosing a fluctuating reading of a combustible gas detector, the instruction might be on-site calibration and recording of the reading. All key monitoring objects are sorted according to their assigned priorities, and their corresponding standardized inspection instructions are integrated to form a clearly structured, task-specific, and descending-priority inspection task list.

[0065] A spatial navigation network within the building is constructed based on the 3D dimensional data of the building's structural components. This network abstracts key locations such as rooms and equipment installation points as nodes, and the passageways and porches connecting them as edges. All points in the inspection task list are set as the sequence of nodes that must be visited in the path planning. Subsequently, a genetic algorithm is used to solve for the optimal path. The specific process is as follows: First, possible inspection paths are encoded as chromosomes, where the gene sequence represents the visiting order of each point. The fitness function is used to evaluate the merits of each path. Spatial distance is calculated to evaluate path efficiency. Priority compliance is evaluated, that is, the ranking position of high-priority points in the inspection sequence is evaluated to optimize the visiting order of key points. Finally, passage feasibility is verified, and the fitness value of paths using prohibited passages is significantly reduced. Based on the comprehensive evaluation of spatial distance, inspection priority, and evacuation passage access status, the algorithm continuously evolves the population through selection, crossover, and mutation operations. After multiple generations of iteration, it converges to the optimal solution, obtaining an optimal inspection route that, under the premise of satisfying all access constraints, shortens the total journey as much as possible and prioritizes visiting high-risk points. The passage status of evacuation routes is determined by integrating real-time fire alarm information with the route type from component attribute data.

[0066] The calculated optimal inspection route is overlaid and rendered onto the 3D building space model through spatial coordinate mapping, generating an electronic route map that includes turn-by-turn guidance, point markers, and key prompts.

[0067] The inspection task list and electronic route map are packaged together into a lightweight mobile inspection task package, which is pushed to the designated mobile inspection terminal via the network. A two-way communication task status synchronization mechanism is established between the system and the mobile terminal; when the inspection personnel perform tasks on the terminal, the status information is transmitted back to the system backend in real time, ensuring that the management end has full control and traceability over the inspection progress and on-site conditions.

[0068] Please see Figure 2 As shown, this invention is a fire alarm inspection method based on big data, comprising the following steps: S1: Collects real-time data from fire sensors and static building information; S2: Process the collected data to form a fire protection-themed database; S3: Based on the fire protection database, a risk assessment model is constructed to conduct multi-dimensional comprehensive analysis, calculate the dynamic fire hazard index, and generate a structured risk briefing; S4: Analyze the structured risk briefing, generate a fire hazard heat map, locate key monitoring targets, and trigger tiered alarm information; S5: Based on the fire hazard index, structured risk briefing, and graded alarm information, generate an inspection task list and optimal inspection route, and push them to the mobile inspection terminal.

[0069] The above embodiments are only used to illustrate the technical methods 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 methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A big data based fire alarm inspection system, characterized in that, The system comprises the following modules: Data acquisition module: collect real-time data of fire sensors and static information of buildings; Big data processing module: process the collected data to form a fire theme database; Risk analysis module: based on the fire theme database, build a risk assessment model for multi-dimensional comprehensive analysis, calculate the dynamic fire risk index, and generate a structured risk report; Visual early warning module: analyze the structured risk report to generate a fire risk hidden danger heat map, locate key monitoring objects, and trigger hierarchical alarm information; Intelligent inspection planning module: based on the fire risk index and the structured risk report, and combined with the hierarchical alarm information, generate an inspection task list and an inspection route plan, and push them to a mobile inspection terminal.

2. The fire alarm inspection system based on big data according to claim 1, wherein, Collect real-time data of fire sensors and static information of buildings, including the following steps: Install various fire sensors at the building site to continuously collect real-time data, and simultaneously collect the static attributes of the sensors. Then, transmit the static data and sensor data of the sensors to the backend through a data communication network; Call the building information database API to obtain the static information of the building, including the three-dimensional size data and component attribute data of the building structure components, and the fire equipment information. The building structure components include walls, columns, floors, and stairs. The three-dimensional size data includes the spatial coordinates, geometric shape, and size of the components. The component attribute data includes the material properties and fire resistance rating of the components. Access the pre-set fire disposal plan database to obtain standardized disposal plan data associated with the building functional zoning and equipment types. Collect the historical period sensor data sequence and historical fire event records from the fire system log and the fire management platform.

3. The fire alarm inspection system based on big data according to claim 2, characterized in that, Process the collected data to form a fire theme database, including the following steps: Clean and quality monitor the collected data; use pre-defined association identifiers to associate real-time and historical sensor data, sensor static data, three-dimensional size data and component attribute data of building structure components, fire equipment information, and logical topology relationship. Form a unified data view that supports associated queries by logically integrating these multi-source heterogeneous data. The logical topology relationship of the fire equipment is derived based on the physical location information of the fire equipment and the pre-set functional linkage rules; Based on the historical sensor data and other equipment state information in the data view, generate historical baseline data representing the normal operating state of various fire equipment containing sensors through statistical learning; Store the data view, historical baseline data, historical fire event records, and standardized disposal plan data in a hierarchical storage system according to different access requirements, and build a fire theme database.

4. The fire alarm inspection system based on big data according to claim 3, characterized in that, Based on the fire theme database, build a risk assessment model for multi-dimensional comprehensive analysis, calculate the dynamic fire risk index, and generate a structured risk report, including the following steps: From the data view of the fire theme database, obtain real-time sensor data, historical sensor data, three-dimensional size data and component attribute data of building structure components, sensor static attributes, and fire equipment information; Based on the data correlation established in the data view, spatial feature construction, time series feature construction, and linkage state feature construction are performed; and a feature vector data for model input is formed; A risk assessment model with multi-task output capability based on a graph neural network is constructed and pre-trained; real-time feature vector data is input into the trained risk assessment model for multi-dimensional comprehensive analysis including spatial dimension, time series, and equipment linkage, and risk heat map data, key prevention and control area list, key risk factor sequence, risk trend prediction, equipment state diagnosis conclusion, linkage verification result, and fire condition confirmation value are output; Based on the key risk factor sequence, key prevention and control area list, and equipment state diagnosis conclusion, specific fire fighting disposal suggestions are matched and generated from standardized disposal plan data; Based on the results of multi-dimensional comprehensive analysis of the risk assessment model, a combined weight fusion algorithm based on the analytic hierarchy process is used to assign weights to each index and perform linear weighting, and a quantitative dynamic fire risk hidden danger index is calculated; The multi-dimensional comprehensive analysis results, fire fighting disposal suggestions, and dynamic fire risk hidden danger index are collectively packaged into a machine-readable document with a fixed mode, i.e., a structured risk brief.

5. The fire alarm inspection system based on big data according to claim 4, characterized in that, The risk assessment model is constructed for multi-dimensional comprehensive analysis, including the following steps: Real-time feature vector data is input into the risk assessment model for multi-dimensional comprehensive analysis: Spatial dimension analysis: abnormal readings are identified based on the comparison results of real-time sensing data and historical baseline data; at the same time, the alarm state of the sensor is determined by comparing whether the real-time reading of the sensor exceeds the pre-set threshold in the static data; the abnormal reading and the alarm state are taken as risk factors; the spatial aggregation effect of the risk factors is evaluated in combination with the physical position of the sensor to generate risk heat map data; based on the three-dimensional size data and component attribute data of the building structure components, the potential spread path and range of the risk factors in the building structure network are simulated to generate a key prevention and control area list; Time series analysis: short-term abnormal data points and long-term trend deterioration of sensor parameters are identified by comparing real-time sensing data and historical baseline data, and are sequenced into a key risk factor sequence; based on the change law revealed by the historical baseline data, a risk trend prediction of the key risk factor sequence is generated; Equipment linkage analysis: equipment groups are constructed based on the logical topological relationship of fire fighting equipment in the data view; the consistency of the equipment state and the pre-set linkage logic in causality and time sequence is verified, and the potential causal relationship is inferred; based on the consistency verification result and the potential causal relationship, a fire condition confirmation value is quantified, and equipment failure, system false alarm, and real fire are identified to generate equipment state diagnosis conclusion and linkage verification result.

6. The fire alarm inspection system based on big data according to claim 5, characterized in that, The dynamic fire risk hidden danger index is calculated, including the following steps: The risk value average of all data points of the risk heat map data generated by spatial dimension analysis is calculated as the comprehensive spatial risk value SRV; Based on the key risk factor sequence obtained by time series analysis, the deviation degree of each factor reading relative to its historical baseline data is calculated to obtain a comprehensive severity index ADR; Based on the change trajectory fitted by the regression analysis technique in time series analysis, the average change rate of the trajectory function in the future prediction time interval is calculated as the slope index M representing the risk evolution trend; The fire confirmation degree value FCD is directly taken from the output of the equipment linkage analysis; Based on the historical fire event records in the fire theme database, the static historical risk value H is determined by the number of historical fires and the average loss; The static historical risk value and the spatial risk value of the risk heat map data, the severity of the key risk factor sequence, the slope of the risk trend prediction, and the fire confirmation degree value are normalized to the value interval of [0, 1] respectively; and each weight coefficient is assigned, and the sum of each weight coefficient is 1; and the linear weighted sum of each index value and the corresponding weight coefficient is taken to obtain the dynamic fire risk hidden danger index DFRI.

7. The fire alarm inspection system based on big data according to claim 6, characterized in that, Analyzing the structured risk brief, generating a fire risk hidden danger heat map, positioning key monitoring objects, triggering hierarchical alarm information, including the following steps: Analyzing the structured risk brief, extracting risk heat map data, key prevention and control area list, key risk factor sequence, equipment state diagnosis conclusion, fire fighting disposal suggestion, risk trend prediction; constructing a three-dimensional space model of the building based on the three-dimensional size data of the building structure components in the fire theme database; superimposing the extracted risk heat map data on the three-dimensional space model of the building to generate an initial heat map, and based on the risk trend prediction, marking the risk evolution trend on the initial heat map; Positioning the key prevention and control area list in the three-dimensional space model of the building, and based on the key risk factor sequence and the equipment state diagnosis conclusion, obtaining the associated risk factors and equipment information; defining a key area and the associated risk factors and equipment information as a key monitoring object, and assigning it a unique tracking ID; Taking the dynamic fire risk hidden danger index as the comprehensive alarm score, and presetting the score threshold interval and the corresponding alarm level; comparing the comprehensive alarm score with the preset score threshold to determine the final alarm level; Packaging the alarm level, key monitoring object ID set, and associated fire fighting disposal suggestion into a standard alarm information package; according to the different alarm levels, pushing to the corresponding mobile terminal through the communication channel.

8. The fire alarm inspection system based on big data according to claim 7, characterized in that, Based on the fire risk hidden danger index and the structured risk brief, and integrating the hierarchical alarm information, generating a patrol task list and a patrol route plan, and pushing to the mobile patrol terminal, including the following steps: Analyzing the structured risk brief and the hierarchical alarm information, extracting the key monitoring objects and their tracking IDs, fire fighting disposal suggestions, equipment state diagnosis conclusions, alarm levels, and three-dimensional size data of building structure components; Based on the alarm level, equipment state diagnosis conclusion, and fire fighting disposal suggestion, assigning a patrol priority to each key monitoring object and generating a standardized patrol instruction, thereby forming a patrol task list sorted by priority; Based on the three-dimensional size data of the building structure components, constructing a spatial navigation network, taking the spatial distance of the patrol point, the patrol priority, and the passage state of the evacuation passage as the path planning cost factors; All points in the inspection task list are taken as necessary nodes, and the optimal inspection route is solved by a path planning algorithm to ensure that high-priority points are in the front of the route sequence, and an electronic route map containing turn-by-turn navigation instructions is generated; The inspection task list and the electronic route map are packaged as a mobile inspection task package, which is pushed to the designated mobile inspection terminal, and a task state synchronization mechanism is established to receive inspection feedback and state updates.

9. A fire alarm inspection method based on big data, characterized in that, The method comprises the following steps: S1: collecting real-time data of fire sensors and static information of buildings; S2: processing the collected data to form a unified fire theme database; S3: based on the fire theme database, constructing a risk assessment model for multi-dimensional comprehensive analysis, calculating a dynamic fire hazard index, and generating a structured risk report; S4: analyzing the structured risk report to generate a fire hazard heat map, locating key monitoring objects, and triggering hierarchical alarm information; S5: based on the fire hazard index, the structured risk report, and the hierarchical alarm information, generating a differentiated inspection task list and an optimal inspection route, and pushing them to the mobile inspection terminal.

Images