Fire spreading path evaluation system combining BIM and AI algorithms

Abstract

The invention relates to the technical field of fire prevention and control, in particular to a BIM and AI algorithm-combined fire spreading path evaluation system, which comprises a multi-source data fusion module, a fire source positioning module, a spreading path prediction module, a risk evaluation module and an intelligent linkage module. A three-dimensional dynamic model is generated by fusing sensor data and BIM data, accurate fire source positioning and fire spreading path prediction are achieved, the regional risk level is quantified in combination with building information, and a control instruction is sent to fire fighting equipment to isolate a high-risk region and delay fire spreading. The fire source positioning precision, the fire behavior prediction reliability and the emergency response efficiency can be improved, and scientific decision support is provided for building fire prevention and control.

Description

Technical Field

[0001] The present invention belongs to the field of fire safety and intelligent building technology, and specifically provides a fire spread path assessment system that combines BIM and AI algorithms. Background Art

[0002] A fire spread path assessment system that combines BIM with AI algorithms is a fire emergency management system that can construct a real-time dynamic three-dimensional model by fusing multi-source sensor data with building information model (BIM) data, and uses intelligent algorithms to predict the path and speed of fire spread. This system primarily addresses the need for fire spread path assessment in complex building environments, encompassing technologies such as data collection, model building, risk assessment, and fire protection coordination. Its core operating method involves integrating multi-source sensor data such as temperature, smoke, and camera data with BIM building information to form a real-time, dynamic three-dimensional model. It then uses machine learning algorithms to analyze historical fire data (such as temperature changes, smoke diffusion speed, and the combustion characteristics of building materials) to accurately locate fire sources, predict fire spread paths, and assess regional risk levels.

[0003] Some existing systems rely on processing video data and other sensor data, but their deep integration of building structure information is limited, resulting in insufficient performance in locating fire sources and predicting fire spread paths. For example, patent CN118999689B, a video data processing system based on a large AI model, was published on March 25, 2025. This patent involves fire information monitoring and processing. A fire information monitoring module collects indoor fire-related data and generates fire determination factors to assess fire severity. However, this technical solution primarily focuses on processing video and sensor data and fails to fully integrate building structure information, limiting its applicability in complex building environments. Furthermore, the system's ability to assess dynamic changes during fire development is limited, making it difficult to fully consider the impact of the specific internal building environment (such as ventilation systems and the distribution of flammable materials) on risk areas.

[0004] Another type of technology focuses on macro-level fire development trend analysis and emergency plan deduction. For example, the publication number CN118821058B is an AI-based digital urban fire emergency plan deduction method and system, with a publication date of January 7, 2025. This patent collects multimodal data and uses historical fire data to train a situation prediction model to generate a response strategy. However, this technical solution focuses on macro-trend analysis and does not deeply integrate BIM data to accurately assess the fire spread path. It also lacks sufficient consideration of dynamic changes within the building (such as the migration of the fire source location and the direction of smoke diffusion). At the same time, there is room for improvement in the active isolation and delay of fire spread in high-risk areas.

[0005] The above issues demonstrate that existing fire monitoring and emergency management systems still have limitations when it comes to assessing fire spread paths using BIM data, particularly in areas such as deep integration of building structural information, precise location of fire sources, prediction of fire spread paths, and dynamic assessment of high-risk areas. Therefore, this invention aims to improve the accuracy and efficiency of fire emergency management by fusing multi-source sensor data with BIM building information to construct a real-time, dynamic three-dimensional model. This model, combined with machine learning algorithms, can predict fire spread paths and speeds, and assess the risk levels of different areas in real time. Summary of the Invention

[0006] The purpose of this invention is to solve the problems existing in the existing technology, such as insufficient integration of building structure information, inaccurate fire source positioning, and limited ability to predict fire spread paths. A fire spread path assessment system combining BIM and AI algorithms is proposed.

[0007] To achieve the above objectives, the present invention adopts the following technical solutions: A fire spread path assessment system combining BIM and AI algorithms includes: The multi-source data fusion module collects real-time data from the fire scene through temperature sensors, smoke sensors, and cameras, and matches the real-time data with the building structure data in the Building Information Model (BIM) to generate a real-time updated 3D dynamic model; The fire source location module is based on a real-time updated 3D dynamic model. It uses spatial topology analysis to extract the distribution characteristics of abnormal points in fire data. Combined with the spatial layout information inside the building, it determines the specific location of the fire source and generates fire source location results. The fire spread path prediction module uses an integrated machine learning algorithm to train historical fire data sets, which include parameters such as temperature change curves, smoke diffusion speed, and building material combustion characteristics. Based on real-time data from the current fire scene, it derives the fire spread path and speed within a specified time period in the future and generates fire spread path prediction results. The risk assessment module combines BIM building information with ventilation system layout, flammable material distribution, and real-time fire data to quantitatively analyze the risk levels of different areas within the building and generate regional risk level assessment results. The intelligent linkage module sends instructions to fire-fighting equipment based on the regional risk level assessment results, controls the operating status of equipment such as automatic sprinkler systems and smoke exhaust devices, and generates isolation instructions for high-risk areas and measures to slow the spread of fire.

[0008] As a further solution of the present invention, the steps of generating a real-time updated three-dimensional dynamic model are specifically as follows: Temperature sensors, smoke sensors, and cameras installed inside the building collect temperature field distribution, smoke concentration distribution, and video surveillance data from the fire scene. Temperature sensors are arranged at fixed intervals at key nodes inside the building, smoke sensors are arranged in a grid pattern to cover the main areas inside the building, and cameras are installed on the top of the building or in key passages to obtain a panoramic view. The collected temperature field distribution, smoke concentration distribution and video surveillance data are transmitted to the central processing unit through the data communication interface. The central processing unit synchronizes the multi-source data based on the timestamp to generate a fire scene data set under a unified time base; Building structure data, including floor plans, floor heights, wall materials, door and window locations, etc., is extracted from the Building Information Model (BIM). The fire scene dataset is matched with the building structure data, and the missing data areas are filled through interpolation algorithms to generate a real-time updated 3D dynamic model.

[0009] As a further solution of the present invention, the steps of determining the specific location of the fire source are specifically as follows: Based on a real-time updated 3D dynamic model, outliers in the temperature field distribution and smoke concentration distribution are extracted. Outliers are defined as areas where the temperature or smoke concentration exceeds a preset threshold. The connectivity between outliers is calculated using spatial topology analysis methods to generate an outlier connectivity graph, where each outlier is represented as a node in the graph, and the edge weights between nodes are determined by the physical distance between the outliers and the temperature or smoke concentration difference. Combined with the spatial layout information inside the building, including wall locations, door and window opening directions, and ventilation duct distribution, the abnormal point connectivity graph is optimized to screen out the most likely fire source location and generate the fire source location result.

[0010] As a further solution of the present invention, the steps of generating the fire spread path prediction result are specifically as follows: Construct a historical fire dataset that includes parameters such as temperature change curves, smoke diffusion speed, and combustion characteristics of building materials for multiple fire events. Each fire event is annotated with its corresponding fire spread path and spread speed. The historical fire dataset is fed into a machine learning algorithm for training, and a deep neural network model is used to extract characteristic patterns in fire scenes to generate a fire spread prediction model. Based on the real-time data of the current fire scene, including temperature field distribution, smoke concentration distribution and fire source location results, the fire spread prediction model is input to deduce the fire spread path and speed within the specified time period in the future, and generate the fire spread path prediction results.

[0011] As a further solution of the present invention, the steps of generating the regional risk level assessment result are specifically as follows: Extract ventilation system layout information from the Building Information Model (BIM), including ventilation duct direction, air outlet location, and wind speed parameters, as well as combustible material distribution information, including combustible material type, distribution density, and combustion characteristics; Combined with real-time fire data, including temperature field distribution, smoke concentration distribution, and fire spread path prediction results, the fire risk index of different areas inside the building is calculated. The fire risk index is determined by the temperature rise rate, smoke concentration growth rate, flammable material distribution density, and ventilation conditions. Different areas inside the building are graded according to the fire risk index to generate regional risk level assessment results, where high-risk areas are defined as areas where the fire risk index exceeds a preset threshold.

[0012] As a further solution of the present invention, the steps of generating high-risk area isolation instructions and fire spread delay measures are specifically as follows: Based on the regional risk level assessment results, identify high-risk areas inside the building and generate high-risk area signs; Send a start command to the automatic sprinkler system through the communication interface to control the sprinkler head to start first in the high-risk area. At the same time, send an adjustment command to the smoke exhaust device to adjust the opening of the smoke exhaust port to change the direction of smoke diffusion. Combined with the fire spread path prediction results, measures to delay the spread of fire are generated, including closing fire doors in specific areas, starting local ventilation systems, and adjusting the air supply direction of air conditioners, generating isolation instructions for high-risk areas and measures to delay the spread of fire.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are: This invention fuses multi-source sensor data with Building Information Model (BIM) data to construct a real-time, dynamic three-dimensional model that comprehensively reflects changes in fire scenarios within a building. By extracting the distribution characteristics of outliers based on spatial topology analysis and combining this with information about the building's internal spatial layout, it accurately locates the fire source, resolving the inaccurate fire source location issue in existing technologies. By integrating machine learning algorithms to train historical fire datasets and combining them with real-time data from current fire scenarios to deduce the fire spread path and speed, the accuracy and reliability of fire spread path predictions are significantly improved. By combining BIM building information with ventilation system layout and flammable material distribution, the risk levels of different areas within the building are quantitatively analyzed, providing a scientific basis for firefighting decision-making. By sending control commands to firefighting equipment, high-risk areas are actively isolated and fire spread is slowed, improving the efficiency and effectiveness of fire emergency response. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 Schematic diagram of the overall architecture of the system of the present invention; Figure 2 Generate flow charts for real-time updated 3D dynamic models; Figure 3 This is the workflow diagram of the fire source location module; Figure 4 This is the algorithm flow chart of the fire spread path prediction module; Figure 5 This is the calculation logic diagram of the regional risk level assessment module; Figure 6 Generate flow charts for control instructions of intelligent linkage modules; Figure 7 This is a schematic diagram of the fire spread path prediction results.

[0015] The accompanying drawings are numbered as follows: 1. Multi-source data fusion module; 2. Fire source location module; 3. Spread path prediction module; 4. Risk assessment module; 5. Intelligent linkage module; 6. Temperature sensor; 7. Smoke sensor; 8. Camera. DETAILED DESCRIPTION

[0016] The present invention provides a fire spread path assessment system that combines BIM and AI algorithms. Its overall architecture is as follows: Figure 1 As shown in the figure, the system includes a multi-source data fusion module 1, a fire source location module 2, a fire spread path prediction module 3, a risk assessment module 4, and an intelligent linkage module 5. These modules exchange information via a data communication interface and work together to complete functions such as real-time fire scene monitoring, fire source location, fire spread path prediction, risk assessment, and firefighting equipment linkage control. The following detailed description of the specific implementation of the system is provided in conjunction with the accompanying drawings.

[0017] In actual applications, temperature sensors 6, smoke sensors 7, and cameras 8 are installed inside the building. These sensors are respectively arranged at key nodes and main areas of the building. Temperature sensors 6 are arranged at fixed intervals to collect temperature field distribution data inside the building; smoke sensors 7 are arranged in a grid pattern, covering the main areas inside the building, to collect smoke concentration distribution data; cameras 8 are installed at the top of the building or at key passages to obtain panoramic video surveillance data. These sensors transmit the collected data to the central processing unit through the data communication interface. The central processing unit synchronizes the multi-source data based on the timestamp to generate a fire scene data set under a unified time reference, such as Figure 2As shown in the figure, the fire scene dataset is then extracted from the Building Information Model (BIM), along with structural data such as floor plans, floor heights, wall materials, and door and window locations. Interpolation algorithms are used to fill in missing data areas during the matching process, ultimately generating a real-time, dynamic 3D model. This process ensures a comprehensive representation of the fire scene and provides foundational data support for subsequent modules.

[0018] The fire source location module 2 is based on a real-time updated three-dimensional dynamic model and determines the specific location of the fire source by extracting abnormal points in the temperature field distribution and smoke concentration distribution. An abnormal point is defined as an area where the temperature or smoke concentration exceeds a preset threshold, such as Figure 3 As shown. After the outliers are extracted, the connectivity between the outliers is calculated using the spatial topology analysis method to generate an outlier connectivity graph. Each node in the graph represents an outlier, and the edge weights between the nodes are determined by the physical distance between the outliers and the temperature difference or smoke concentration difference. Combined with the spatial layout information inside the building, including the position of the walls, the direction of the door and window openings, and the distribution of ventilation ducts, the outlier connectivity graph is optimized to screen out the most likely fire source location. For example, in a certain actual application scenario, a fire occurs inside a building, and the temperature sensor 6 detects an abnormal increase in the temperature in a certain area, while the smoke sensor 7 detects a significant increase in the smoke concentration in the area. Through the spatial topology analysis method, combined with the wall position and ventilation duct distribution information inside the building, the area is determined to be the fire source location, and a fire source positioning result is generated.

[0019] The spread path prediction module 3 constructs a historical fire data set and trains a machine learning algorithm to deduce the fire spread path and speed within a specified time period in the future. The historical fire data set contains parameters such as temperature change curves, smoke diffusion speed, and combustion characteristics of building materials for multiple fire events. Each fire event is marked with its corresponding fire spread path and speed. Figure 4 As shown, a historical fire dataset is fed into a deep neural network model for training, which extracts characteristic patterns from fire scenarios and generates a fire spread prediction model. In practical applications, the fire spread prediction model is fed with real-time data from the current fire scenario, including temperature field distribution, smoke concentration distribution, and fire source location results, to derive the fire spread path and speed within a specified future time period. For example, in a building fire scenario, spread path prediction module 3, based on real-time data from the current fire scenario, predicts that the fire will spread along a corridor to adjacent rooms within the next 10 minutes at a speed of 0.5 meters per minute, and generates a fire spread path prediction result.

[0020] Risk Assessment Module 4 combines the ventilation system layout, flammable material distribution and real-time fire data in the BIM building information to quantitatively analyze the risk levels of different areas inside the building. Figure 5As shown, ventilation system layout information, including duct routing, air outlet location, and wind speed parameters, as well as combustible material distribution information, including combustible material type, distribution density, and combustion characteristics, is extracted from BIM. Combined with real-time fire data, including temperature field distribution, smoke density distribution, and fire spread path predictions, the fire risk index for different areas within the building is calculated. The fire risk index is determined by the temperature rise rate, smoke density growth rate, combustible material distribution density, and ventilation conditions. Based on the fire risk index, different areas within the building are graded to generate regional risk level assessment results. For example, in a specific application scenario, an area within a building has a rapid temperature rise rate, a significant increase in smoke density, a large amount of combustible materials, and poor ventilation conditions, resulting in a high-risk area.

[0021] The intelligent linkage module 5 generates high-risk area isolation instructions and fire spread delay measures based on the regional risk level assessment results. Figure 6 As shown, based on the regional risk level assessment results, high-risk areas within the building are identified and high-risk area identifiers are generated. A start command is sent to the automatic sprinkler system via a communication interface, controlling sprinkler heads to activate preferentially in high-risk areas. Simultaneously, an adjustment command is sent to the smoke exhaust system to adjust the smoke exhaust vent opening to redirect smoke spread. Combined with the fire spread path prediction results, fire spread mitigation measures are generated, including closing fire doors in specific areas, activating local ventilation systems, and adjusting the air supply direction of air conditioners. For example, in a building fire scenario, the intelligent linkage module 5 identifies a corridor as a high-risk area and immediately sends a start command to the automatic sprinkler system, controlling sprinkler heads in that corridor to activate preferentially. Simultaneously, an adjustment command is sent to the smoke exhaust system to adjust the smoke exhaust vent opening to direct smoke spread to safe areas. Furthermore, based on the fire spread path prediction results, the intelligent linkage module 5 generates fire spread mitigation measures, including closing fire doors at both ends of the corridor, activating local ventilation systems, and adjusting the air supply direction of air conditioners to slow the spread of fire to adjacent rooms.

[0022] In actual applications, when a fire occurs in a building, the system workflow is as follows: First, the temperature sensor 6, smoke sensor 7, and camera 8 collect real-time data from the fire scene and transmit the data to the central processing unit. The central processing unit synchronizes the multi-source data based on the timestamp, generates a fire scene dataset under a unified time base, and matches it with the BIM building structure data to generate a real-time updated 3D dynamic model, such as Figure 7As shown. Subsequently, the fire source location module 2 extracts abnormal points in the temperature field distribution and smoke concentration distribution based on the real-time updated three-dimensional dynamic model, and determines the specific location of the fire source in combination with the spatial layout information inside the building. The spread path prediction module 3 derives the fire spread path and speed within the specified time period in the future based on the real-time data of the current fire scene, and generates the fire spread path prediction result, as shown in Figure 7 As shown in the figure, Risk Assessment Module 4 combines BIM building information, ventilation system layout, flammable material distribution, and real-time fire data to quantitatively analyze the risk levels of different areas within the building and generate regional risk level assessment results. Finally, Intelligent Linkage Module 5 generates high-risk area isolation instructions and fire spread mitigation measures based on the regional risk level assessment results. It controls the operating status of equipment such as automatic sprinkler systems and smoke exhaust devices, and generates high-risk area isolation instructions and fire spread mitigation measures.

[0023] In a typical building fire scenario, temperature sensor 6 detects an abnormally high temperature in a room, smoke sensor 7 detects a significant increase in smoke concentration in the same room, and camera 8 captures the presence of an open flame in the room. The central processing unit synchronizes multi-source data based on timestamps, generating a fire scene dataset based on a unified time base. This dataset is then matched with the BIM building structure data to create a real-time, dynamic 3D model. Fire source location module 2 extracts anomalies in the temperature and smoke concentration distributions and, combined with the building's internal spatial layout information, identifies the room as the fire source. Based on real-time data from the current fire scenario, spread path prediction module 3 predicts that the fire will spread along a corridor to adjacent rooms within the next 10 minutes at a rate of 0.5 meters per minute. Risk assessment module 4 combines BIM building information, ventilation system layout, flammable material distribution, and real-time fire status data to quantitatively analyze the risk levels of different areas within the building, assessing the corridor as a high-risk area. Intelligent linkage module 5 identifies the corridor as a high-risk area and immediately sends a start command to the automatic sprinkler system, prioritizing the sprinkler heads in that corridor. It also sends an adjustment command to the smoke exhaust system, adjusting the smoke vent opening to direct smoke to safe areas. Furthermore, based on the fire spread path prediction, intelligent linkage module 5 generates fire mitigation measures, including closing the fire doors at both ends of the corridor, activating the local ventilation system, and adjusting the air supply direction of the air conditioner to slow the spread of the fire to adjacent rooms.

[0024] In order to better enable relevant personnel in this technical field to fully understand and implement the present invention, the implementation principle of the present invention is supplemented below with reference to a specific application scenario.

[0025] In a certain building fire scene, after the system is started, it first collects real-time data of the fire scene through temperature sensors 6, smoke sensors 7 and cameras 8. Temperature sensors 6 are arranged at key nodes inside the building at fixed intervals to detect the temperature field distribution inside the building; smoke sensors 7 are arranged in a grid to cover the main area to monitor the smoke concentration distribution; cameras 8 are installed on the top of the building or at key passages to obtain panoramic video surveillance images. These sensors transmit the collected data to the central processing unit through the data communication interface. The central processing unit synchronously processes multi-source data based on timestamps to generate a fire scene data set under a unified time base, and extracts building structure data such as building floor plans, floor heights, wall materials, door and window positions from the building information model (BIM). Subsequently, the interpolation algorithm is used to fill in possible data missing areas, match the fire scene data set with the building structure data, and generate a real-time updated three-dimensional dynamic model, such as Figure 2 This process ensures the comprehensiveness and dynamism of the fire scenario and provides basic support for the work of subsequent modules.

[0026] The fire source positioning module 2 extracts abnormal points in the temperature field distribution and smoke concentration distribution based on the three-dimensional dynamic model updated in real time. An abnormal point is defined as an area where the temperature or smoke concentration exceeds a preset threshold. For example, in a certain actual scenario, the temperature sensor 6 detects an abnormal increase in the temperature in a room, and the smoke sensor 7 detects a significant increase in the smoke concentration in the room. The fire source positioning module 2 uses a spatial topology analysis method to calculate the connectivity between abnormal points and generate an abnormal point connectivity graph. Each node in the graph represents an abnormal point, and the edge weights between nodes are determined by the physical distance and temperature difference or smoke concentration difference between the abnormal points. Combined with the spatial layout information inside the building, including the location of walls, the direction of door and window openings, and the distribution of ventilation ducts, the abnormal point connectivity graph is optimized to screen out the most likely fire source location. This determines that the room is the specific location of the fire source, and generates a fire source positioning result, such as Figure 3 shown.

[0027] The spread path prediction module 3 derives the fire spread path and speed within a specified time period in the future based on the real-time data of the current fire scene. First, the temperature change curves, smoke diffusion speed, building material combustion characteristics and other parameters of multiple fire events are extracted from the historical fire data set, and each fire event is marked with its corresponding fire spread path and spread speed. The historical fire data set is input into the deep neural network model for training, and the characteristic patterns in the fire scene are extracted to generate a fire spread prediction model. Subsequently, based on the real-time data of the current fire scene, including the temperature field distribution, smoke concentration distribution and fire source positioning results, the fire spread prediction model is input to derive the fire spread path and speed within a specified time period in the future. For example, in a certain building fire scene, the spread path prediction module 3 predicts that the fire will spread along a corridor to the adjacent room within the next 10 minutes based on the real-time data of the current fire scene, and the spread speed is 0.5 meters per minute, and generates a fire spread path prediction result, such as Figure 4 shown.

[0028] The risk assessment module 4 combines the ventilation system layout, flammable material distribution and real-time fire data in the BIM building information to quantitatively analyze the risk levels of different areas inside the building. The ventilation system layout information, including the direction of the ventilation duct, the location of the air outlet and the wind speed parameters, as well as the flammable material distribution information, including the type of flammable material, distribution density and combustion characteristics, are extracted from the BIM. Combined with the real-time fire data, including the temperature field distribution, smoke concentration distribution and the fire spread path prediction results, the fire risk index of different areas inside the building is calculated. The fire risk index is determined by the temperature rise rate, the smoke concentration growth rate, the flammable material distribution density and the ventilation conditions. Different areas inside the building are graded according to the fire risk index to generate regional risk level assessment results. For example, in a certain actual scenario, the temperature rise rate in a certain area inside a building is relatively fast, the smoke concentration increases significantly, and there are a large number of flammable materials in the area, and the ventilation conditions are poor. Therefore, it is rated as a high-risk area. Figure 5 shown.

[0029] The intelligent linkage module 5 generates high-risk area isolation instructions and fire spread delaying measures based on the regional risk level assessment results. Based on the regional risk level assessment results, the high-risk areas inside the building are identified and high-risk area identifiers are generated. A start-up instruction is sent to the automatic sprinkler system through the communication interface to control the sprinkler heads to start first in the high-risk areas. At the same time, an adjustment instruction is sent to the smoke exhaust device to adjust the smoke exhaust port opening to change the direction of smoke diffusion. Combined with the fire spread path prediction results, measures to delay the spread of fire are generated, including closing the fire doors in specific areas, starting the local ventilation system, and adjusting the air supply direction of the air conditioner. For example, in a certain building fire scene, the intelligent linkage module 5 identifies a corridor as a high-risk area, and then sends a start-up instruction to the automatic sprinkler system to control the sprinkler heads in the corridor to start first. At the same time, an adjustment instruction is sent to the smoke exhaust device to adjust the smoke exhaust port opening to guide the smoke to spread to a safe area. In addition, based on the fire spread path prediction results, the intelligent linkage module 5 generates fire spread delaying measures, including closing the fire doors at both ends of the corridor, starting the local ventilation system, and adjusting the air supply direction of the air conditioner to delay the spread of fire to adjacent rooms, such as Figure 6 shown.

[0030] The operating principle of the system in the above process can be further explained as follows: Data collected by temperature sensor 6 and smoke sensor 7 are synchronized and processed using the central processing unit's timestamps to generate a fire scene dataset based on a unified time base. When matching this dataset with the BIM building structure data, an interpolation algorithm is used to fill in missing data areas to ensure the integrity of the three-dimensional dynamic model. The fire source location module 2 optimizes the connectivity graph of outliers using spatial topology analysis methods. Combined with the building's internal spatial layout information, this effectively eliminates interference points and accurately locates the fire source. The spread path prediction module 3 extracts characteristic patterns of the fire scene using a deep neural network model. Combined with real-time data from the current fire scene, it accurately derives the fire spread path and speed. The risk assessment module 4 comprehensively considers the temperature rise rate, smoke concentration growth rate, flammable material distribution density, and ventilation conditions to quantitatively analyze the risk levels of different areas within the building, providing a scientific basis for firefighting decision-making. The intelligent linkage module 5 sends control commands to firefighting equipment via a communication interface, actively isolating high-risk areas and slowing fire spread, thereby improving the efficiency and effectiveness of fire emergency response.

[0031] In summary, the present invention integrates multi-source sensor data with BIM building information to construct a real-time updated three-dimensional dynamic model, combines machine learning algorithms to predict the path and speed of fire spread, and assesses the risk level of different areas in real time, thereby improving the accuracy and efficiency of fire emergency management.

Claims

1. A fire spread path assessment system combining BIM and AI algorithm, characterized by: The system comprises: The multi-source data fusion module (1) collects real-time data of the fire scene through temperature sensors (6), smoke sensors (7) and cameras (8), and matches the real-time data with the building structure data in the building information model (BIM) to generate a three-dimensional dynamic model that is updated in real time; The fire source location module (2) is based on a real-time updated three-dimensional dynamic model and uses spatial topology analysis methods to extract the distribution characteristics of abnormal points in the fire data. Combined with the spatial layout information inside the building, it determines the specific location of the fire source and generates a fire source location result. The spread path prediction module (3) integrates machine learning algorithms to train historical fire data sets, which include temperature change curves, smoke diffusion speed, and building material combustion characteristic parameters. Based on the real-time data of the current fire scene, it deduces the fire spread path and speed within a specified time period in the future and generates fire spread path prediction results; The risk assessment module (4) combines the ventilation system layout, flammable material distribution and real-time fire data in the BIM building information to quantitatively analyze the risk levels of different areas within the building and generate regional risk level assessment results; The intelligent linkage module (5) sends instructions to the fire-fighting equipment according to the regional risk level assessment results, controls the operation status of the automatic sprinkler system and smoke exhaust device equipment, and generates high-risk area isolation instructions and measures to slow down the spread of fire.

2. The fire spread path assessment system combining BIM and AI algorithm according to claim 1 is characterized in that: The steps for generating a real-time updated 3D dynamic model are as follows: The temperature field distribution, smoke concentration distribution and video monitoring data of the fire scene are collected by temperature sensors (6), smoke sensors (7) and cameras (8) installed inside the building, wherein the temperature sensors (6) are arranged at fixed intervals at key nodes inside the building, the smoke sensors (7) are arranged in a grid pattern to cover the main areas inside the building, and the cameras (8) are installed at the top of the building or at key passages to obtain a panoramic view; The collected temperature field distribution, smoke concentration distribution and video surveillance data are transmitted to the central processing unit through the data communication interface. The central processing unit synchronizes the multi-source data based on the timestamp to generate a fire scene data set under a unified time base; The building floor plan, floor height, wall material, and door and window location information are extracted from the Building Information Model (BIM), the fire scene dataset is matched with the building structure data, and the missing data areas are filled through the interpolation algorithm to generate a real-time updated 3D dynamic model.

3. The fire spread path assessment system combining BIM and AI algorithm according to claim 1 is characterized in that: The steps to determine the specific location of the fire source are as follows: Based on a real-time updated 3D dynamic model, outliers in the temperature field distribution and smoke concentration distribution are extracted, where outliers are defined as areas where the temperature or smoke concentration exceeds a preset threshold. The connectivity between outliers is calculated using spatial topology analysis methods to generate an outlier connectivity graph, where each outlier is represented as a node in the graph, and the edge weights between nodes are determined by the physical distance between the outliers and the temperature or smoke concentration difference. Combined with the spatial layout information inside the building, including wall locations, door and window opening directions, and ventilation duct distribution, the abnormal point connectivity graph is optimized to screen out the most likely fire source location and generate the fire source location result.

4. The fire spread path assessment system combining BIM and AI algorithm according to claim 1 is characterized in that: The steps to generate fire spread path prediction results are as follows: Construct a historical fire dataset that includes temperature change curves, smoke diffusion rates, and building material combustion characteristic parameters for multiple fire events. Each fire event is annotated with its corresponding fire spread path and spread rate. The historical fire dataset is fed into a machine learning algorithm for training, and a deep neural network model is used to extract characteristic patterns in fire scenes to generate a fire spread prediction model. Based on the real-time data of the current fire scene, including temperature field distribution, smoke concentration distribution and fire source location results, the fire spread prediction model is input to deduce the fire spread path and speed within the specified time period in the future, and generate the fire spread path prediction results.

5. The fire spread path assessment system combining BIM and AI algorithm according to claim 1 is characterized in that: The specific steps for generating regional risk level assessment results are: Extract ventilation system layout information from the Building Information Model (BIM), including ventilation duct direction, air outlet location, and wind speed parameters, as well as combustible material distribution information, including combustible material type, distribution density, and combustion characteristics; Combined with real-time fire data, including temperature field distribution, smoke concentration distribution, and fire spread path prediction results, the fire risk index of different areas inside the building is calculated. The fire risk index is determined by the temperature rise rate, smoke concentration growth rate, flammable material distribution density, and ventilation conditions. Different areas inside the building are graded according to the fire risk index to generate regional risk level assessment results.

6. The fire spread path assessment system combining BIM and AI algorithm according to claim 1 is characterized in that: The specific steps for generating high-risk area isolation instructions and fire spread mitigation measures are: Based on the regional risk level assessment results, identify high-risk areas inside the building and generate high-risk area signs; Send a start command to the automatic sprinkler system through the communication interface to control the sprinkler head to start first in the high-risk area. At the same time, send an adjustment command to the smoke exhaust device to adjust the opening of the smoke exhaust port to change the direction of smoke diffusion. Combined with the fire spread path prediction results, measures to delay the spread of fire are generated, including closing fire doors in specific areas, activating local ventilation systems, and adjusting the air supply direction of air conditioners.

7. The fire spread path assessment system combining BIM and AI algorithm according to claim 2 is characterized in that: The interpolation algorithm uses linear interpolation or bilinear interpolation methods to fill in the data missing areas to ensure the matching accuracy between the fire scene dataset and the building structure data.

8. The fire spread path assessment system combining BIM and AI algorithm according to claim 4 is characterized in that: The training process of a deep neural network model consists of the following steps: The historical fire dataset is divided into a training set and a test set. The deep neural network model is trained using the training set, and the prediction accuracy of the model is verified using the test set. The model parameters are adjusted through the back propagation algorithm to optimize the model performance and generate a fire spread prediction model.

Images