Construction project fire emergency linkage method based on digital twinning

Abstract

The application discloses a construction project fire emergency linkage method based on digital twinning, relates to the technical field of fire emergency linkage, and comprises the following steps: collecting real-time sensing data such as temperature, smoke, video and personnel position, and dynamically updating a digital twin body synchronous with a physical environment through data assimilation technology. Fire parameters are injected into the digital twin body, and computational fluid dynamics simulation is run to predict fire spread. Based on the simulation results and real-time data, a dynamic threat assessment graph is generated using the analytic hierarchy process, and accordingly, an emergency rule base is matched through an event-driven decision tree algorithm to generate preliminary evacuation and equipment linkage instructions, and finally, an executable instruction set is output after spatial verification with a building information model. The method realizes dynamic high-fidelity mapping of the physical environment, accurate prediction of the fire situation, and intelligent generation and verification of emergency decisions.

Description

Technical Field

[0001] This invention belongs to the field of fire emergency response technology, specifically a construction project fire emergency response method based on digital twins. Background Technology

[0002] Traditional fire emergency systems in construction projects rely on pre-installed independent sensors and fixed plans for response. Existing technologies generally trigger preset alarms and equipment linkage programs when point detectors such as smoke and heat sensors reach a threshold, while evacuation guidance depends on static emergency lighting and broadcast instructions. This is a passive response mode based on discrete event triggers, and its operating logic is based on the assumption of a static building environment.

[0003] Such existing technologies have shortcomings. The system lacks the ability to map and predict the dynamic evolution of a fire in real time. The static physical information model cannot be continuously updated as the fire environment changes, causing the information basis for decision-making to quickly become invalid. At the same time, emergency decisions are based only on local alarm signals and simple rules, which cannot quantitatively assess the comprehensive risks of different areas within the building over time, and are difficult to deal with the complex situation of multiple factors such as fire intensity, smoke spread, and personnel distribution.

[0004] Currently, the field of fire emergency response needs to address two key issues: achieving high-frequency synchronization between virtual models and the real physical environment to ensure the real-time nature and fidelity of the analytical basis; and, based on this, realizing dynamic simulation of fire development and multi-factor fusion-based threat quantification assessment to support accurate emergency decision-making. This requires building an intelligent emergency platform capable of assimilating sensing data in real time and performing highly reliable physical simulations. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes a construction project fire emergency response method based on digital twins, including: Collect real-time sensing data within the construction project, including temperature distribution, smoke concentration, video images, and personnel location information; By using real-time sensing data collected, the digital twin synchronized with the physical space is dynamically updated through data assimilation technology to ensure that the digital twin reflects the latest physical environment status. In the updated digital twin, simulated fire initiation parameters are injected, and a fire development process simulation based on computational fluid dynamics is run to generate dynamic prediction results of fire spread. Based on the dynamic prediction results of fire spread and the environmental and personnel data in the real-time sensing data, the analytic hierarchy process is used to assess the comprehensive threat level of different zones in the building, and a dynamic threat assessment map is obtained. Based on the dynamic threat assessment map and combined with the pre-set emergency rule base, an event-driven decision tree algorithm is used to generate preliminary evacuation guidance and fire equipment linkage instructions. The initial evacuation guidance and fire equipment linkage instructions are spatially logically verified with the building information model in the digital twin. After eliminating spatial conflicts, an executable instruction set is generated.

[0006] Furthermore, the method of dynamically updating the digital twin synchronized with the physical space using real-time sensing data and data assimilation technology includes: Establish a static basic digital twin model that includes building structure, equipment layout, and material properties; Spatial interpolation is performed on the temperature and smoke concentration data in the real-time sensing data to generate corresponding three-dimensional field data; Compare the three-dimensional field data with the preset state values ​​of the corresponding spatial locations in the static basic digital twin model; The Kalman filter algorithm is used to make optimal estimates of the dynamic state variables in the static basic digital twin model, which include the temperature value and smoke concentration value of each region. The optimally estimated dynamic state variables are used to replace the original corresponding values ​​in the static basic digital twin model, completing one update iteration of the digital twin.

[0007] Furthermore, in the updated digital twin, simulated fire initiation parameters are injected, and a fire development process simulation based on computational fluid dynamics is run to generate dynamic prediction results of fire spread, including: One or more initial locations where the temperature rises abnormally or the smoke concentration increases abnormally are identified from real-time sensing data, and these locations are used as the starting points of the simulated fire. Assign initial fire source power and combustible material type parameters to each fire initiation point; Within the three-dimensional geometric mesh of the digital twin, the computational fluid dynamics solver is activated to solve the mass, momentum, energy, and chemical component transport equations. The simulation process is advanced with a preset time step, and the spatial distribution and temporal evolution data of the temperature field, smoke concentration field, toxic gas concentration field and visibility field are calculated at each time step. Record and output dynamic predictions of fire spread over a period of time starting from the current moment.

[0008] Furthermore, based on the dynamic prediction results of fire spread and the environmental and personnel data in the real-time sensing data, the analytic hierarchy process (AHP) is used to assess the comprehensive threat level of different zones within the building, resulting in a dynamic threat assessment map, including: The building plan is divided into multiple assessment zones; A hierarchical assessment index structure is constructed for each assessment zone, including fire thermal threat, smoke toxicity threat, and evacuation obstruction threat. From the dynamic prediction results of fire spread, extract the predicted temperature and predicted smoke concentration data of each assessment zone at key future time points; Extract real-time personnel density data for each assessment zone from the real-time sensing data; The weights of each indicator—fire thermal threat, smoke toxicity threat, and evacuation obstruction threat—were determined using the analytic hierarchy process (AHP). After normalizing the predicted temperature, predicted smoke concentration, and real-time personnel density data for each assessment zone, the comprehensive threat score for each zone is calculated by combining the weights of the corresponding indicators. Based on the overall threat score, each zone is assigned a level label, and the level labels of all zones are overlaid on the building floor plan to form a dynamic threat assessment map.

[0009] Furthermore, based on the dynamic threat assessment map and combined with a pre-set emergency rule base, an event-driven decision tree algorithm is used to generate preliminary evacuation guidance and fire equipment linkage instructions, including: A pre-built emergency rule base contains standard emergency response logic corresponding to different threat levels, different fire source locations, and different building functional areas; The partition with the highest threat level in the dynamic threat assessment graph is used as the key event to trigger the root node of the decision tree; Based on the location of the fire's origin, its current direction of spread, and the distribution of building exits, select the appropriate branch logic at the decision node of the decision tree; Traverse the decision tree until a leaf node is reached. Each leaf node corresponds to a set of predefined preliminary instructions, which include suggested evacuation routes, start and stop of smoke exhaust fans, raising and lowering of fireproof roller shutters, emergency broadcast content, and start and stop of fire pumps.

[0010] Furthermore, the initial evacuation guidance and fire equipment linkage instructions are spatially logically verified with the building information model in the digital twin. After eliminating spatial conflicts, an executable instruction set is generated, including: The instructions involving spatial actions in the initial instructions are mapped to specific components in the building information model; Check whether the recommended evacuation routes pass through areas that have been blocked off by simulated fires or where fireproof roller shutters have been lowered; Check whether the operation of the smoke exhaust fan conflicts with the design logic of adjacent smoke control zones; Check whether the descent sequence of the fireproof roller shutters will disrupt the currently identified main personnel evacuation routes; If a spatial logic conflict is found, the initial instruction is adjusted according to the preset conflict resolution rules to generate a conflict-free alternative instruction. All instructions that have passed verification or adjustment are encoded according to their execution sequence and device address, and packaged into an executable instruction set.

[0011] Furthermore, the method also includes: The executable instruction set is sent to the fire emergency equipment and personnel guidance terminals in the physical space to drive them to perform corresponding actions; During the execution of the action, new real-time sensing data is continuously collected and compared with the simulation expected data at the corresponding moment in the digital twin to calculate the data deviation; When the data deviation exceeds the preset threshold, the model parameter calibration process of the digital twin is triggered, and the key parameters of the simulation model are corrected using new real-time sensing data. The revised key parameters of the simulation model are fed back to the data assimilation technology stage as model parameters in the digital twin update process; The step of sending the executable instruction set to the fire emergency equipment and personnel guidance terminals in the physical space to drive them to perform corresponding actions includes: The executable instruction set is distributed via a dedicated fire protection network or an IoT gateway; For smoke exhaust fans, fireproof roller shutters, and fire pumps, the commands are converted into standard control protocol signals that the equipment controller can recognize; For personnel guidance terminals, instructions are converted into directional control commands for dynamic evacuation indicator lights, voice and text content for emergency broadcasts, and graphic and text information for displays. Monitor the instruction reception confirmation signals of each device and terminal to ensure that the instructions have been delivered.

[0012] Furthermore, during the execution of the action, new real-time sensing data is continuously collected and compared with the simulated expected data at the corresponding moment in the digital twin to calculate the data deviation, including: Record the simulation time elapsed from the issuance of the instruction to the current moment; From the digital twin, the simulation expectation data of the simulation time scale is extracted, and the simulation expectation data includes the expected temperature and expected smoke concentration at key locations; Extract the measured temperature and measured smoke concentration at the same physical location from the newly collected real-time sensing data; Calculate the absolute difference between the measured value and the expected value at the same location; The average and variance of the absolute differences at all monitoring locations are calculated as a measure of overall data deviation.

[0013] Furthermore, when the data deviation exceeds a preset threshold, the model parameter calibration process of the digital twin is triggered, using new real-time sensing data to correct key parameters of the simulation model, including: Set the average and variance thresholds for the data deviation; When the calculated data deviation exceeds any threshold, the calibration process is initiated. Identify the spatial locations with the largest data deviations and use them as the main calibration areas; Use new real-time sensing data within the main calibration area as the calibration target value; Inversely adjust the key parameters in the computational fluid dynamics simulation model that affect the simulation results of the main calibration region. These key parameters include the rate of increase of the heat release rate of the fire source and the pyrolysis characteristics of the wall material. The key parameters are iteratively adjusted so that the simulation output of the adjusted model in the main calibration area is close to the new real-time sensing data. The corrected key parameters are then used as the output of the model parameter calibration process.

[0014] Furthermore, the step of feeding back the corrected simulation model's key parameters to the data assimilation technology stage as model parameters in the digital twin update process includes: When the data assimilation technology dynamically updates the digital twin, the corrected key parameters of the simulation model are used as the process model parameters of the Kalman filter algorithm. Using the updated process model parameters, predict the state of the digital twin at the next time step; By fusing new real-time sensing data with the state predicted using new parameters, the digital twin is updated.

[0015] Compared with the prior art, the beneficial effects of the present invention are: By fusing multi-source sensing data such as temperature, smoke, video, and personnel location in real time, and applying data assimilation algorithms to continuously estimate and correct the state variables of the digital twin, the environmental parameter field in the virtual model can be dynamically updated to reflect real changes in the physical world. This process solves the problem of rapid distortion in traditional static models during a fire, constructing a high-fidelity dynamic mirror that remains synchronized with the physical environment, providing an accurate and up-to-date data foundation for all subsequent analysis and decision-making.

[0016] In a synchronously updated digital twin, high-precision numerical simulations of the fire development process are performed based on computational fluid dynamics principles. The simulation model takes real-time or predicted fire parameters as input, solves fluid motion and heat / mass transfer equations, and thus deduces the three-dimensional spread process, velocity, and concentration distribution of flames and smoke. Combining the physical field data output from the simulation with real-time personnel distribution information, a quantitative assessment model is established using the analytic hierarchy process (AHP). This model comprehensively weights and overlays multiple criteria, including fire intensity, smoke hazard, spread velocity, and personnel density, to generate a dynamic threat assessment map reflecting the real-time comprehensive risk level of each zone within the building. This method achieves scientific prediction of fire situations and quantitative risk assessment.

[0017] Based on the spatiotemporal distribution of risks revealed by the dynamic threat assessment map, an event-driven decision tree algorithm is used to match a pre-set emergency rule base, automatically generating preliminary evacuation route planning and fire equipment linkage strategies. Subsequently, the preliminary instructions are validated against the building information model in the digital twin for spatial logic and topological relationships, identifying and eliminating potential path conflicts, equipment interference, and other contradictions. Finally, a set of spatially feasible and logically consistent executable instructions is output. This represents a shift in emergency response strategies from simple triggering based on fixed rules to intelligent generation based on in-depth dynamic situational analysis and spatial validation. Attached Figure Description Figure 1 This is a flowchart illustrating the steps of the construction project fire emergency response method based on digital twins as described in this invention. Figure 2 A flowchart for dynamically updating a digital twin; Figure 3 A flowchart simulating the fire development process; Figure 4 A composite analysis diagram of the impact of fireproof roller shutters on evacuation flow; Figure 5 This is a multi-dimensional dynamic simulation analysis diagram of the fire development process. Detailed Implementation

[0018] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] See Figure 1This paper describes a construction project fire emergency response method based on digital twins. It collects real-time sensing data within the construction project, including temperature distribution, smoke concentration, video images, and personnel location information. Using data assimilation technology, the collected real-time sensing data is dynamically updated to reflect the latest physical environment. Simulated fire initiation parameters are injected into the updated digital twin, and a fire development process simulation based on computational fluid dynamics is run to generate dynamic predictions of fire spread. Based on the dynamic predictions of fire spread and environmental and personnel data from the real-time sensing data, the analytic hierarchy process (AHP) is used to assess the comprehensive threat level of different zones within the building to obtain a dynamic threat assessment map. Based on the dynamic threat assessment map and a pre-set emergency rule base, an event-driven decision tree algorithm is used to generate preliminary evacuation guidance and fire equipment linkage instructions. These instructions are then spatially logically verified against the building information model in the digital twin, and spatial conflicts are eliminated before generating an executable instruction set.

[0020] See Figure 2 In one embodiment of the present invention, the construction project fire emergency linkage method based on digital twin involves establishing a digital twin synchronized with the physical space, the dynamic updates of which rely on the application of data assimilation technology. Taking the construction project of a multi-story office building as an example, a static basic digital twin model of the building has been pre-established. The static basic digital twin model includes the three-dimensional geometry of all floors of the building, the layout of internal rooms and corridors, the location of ventilation ducts and doors and windows, the material properties of concrete, glass and decorative panels used in building components and their combustion characteristic parameters. At the same time, the static basic digital twin model also integrates the equipment layout information and network addresses of all fire detectors, temperature sensors, cameras and personnel positioning beacons. In the static basic digital twin model, each room and corridor area is preset with temperature and smoke concentration values ​​corresponding to a safe state as initial dynamic state variables. For example, the temperature value is preset to 22 degrees Celsius and the smoke concentration value is preset to 0.01 milligrams per cubic meter.

[0021] In some embodiments, the system acquires real-time sensing data when the sensor network deployed within the building begins operation. Assume that at a certain moment, the system receives readings of 28°C, 31°C, and 26°C from three temperature sensors located in the east corridor of the third floor, and readings of 0.05 mg / m³ and 0.03 mg / m³ from two smoke sensors. The spatial coordinates of these sensors are known in the static underlying digital twin model. Using the real-time sensing data from these discrete points, the system generates three-dimensional temperature field data and three-dimensional smoke concentration field data covering the entire east corridor area of ​​the third floor through a spatial interpolation algorithm. The interpolated three-dimensional field data shows that the interpolated temperature in the central area of ​​the corridor is approximately 30°C, and the interpolated smoke concentration is approximately 0.045 mg / m³.

[0022] In practical implementation, the system compares the generated 3D temperature field data with the preset temperature value of 22 degrees Celsius for all corresponding spatial grid points in the third-level east corridor area of ​​the static basic digital twin model, calculating the temperature difference for each grid point. Similarly, the 3D smoke concentration field data is compared with the preset smoke concentration value of 0.01 milligrams per cubic meter. These difference data are input into the observation update stage of a Kalman filter algorithm. The Kalman filter algorithm treats the static basic digital twin model as a dynamic system, whose system state is composed of dynamic state variables of all regions. The Kalman filter algorithm predicts the dynamic state variables at the next moment based on the preset process model, and then uses the 3D field data obtained by spatial interpolation from real-time sensing data as the observation value for correction, performing optimal estimation of the dynamic state variables. The optimal estimation process is achieved through the state update equation: ; in: This represents the dynamic state variable vector after optimal estimation by the Kalman filter algorithm at time k. This represents the vector of dynamic state variables predicted based on the process model at time k. This represents the Kalman gain matrix at time k. Let represent the observation vector derived from the three-dimensional field data at time k, and H represent the observation model matrix, which describes the mapping relationship from the system state to the observed values. The Kalman filter algorithm, through iterative calculation, finally outputs a set of optimally estimated dynamic state variables, including the optimal estimated temperature value and the optimal estimated smoke concentration value of the third-floor east corridor area, which may be updated to 29.5 degrees Celsius and 0.042 milligrams per cubic meter, respectively.

[0023] Understandably, after completing the optimal estimation, the system directly replaces the original temperature and smoke concentration values ​​of each area in the static basic digital twin model with the dynamically estimated state variables. In specific implementation, the temperature value of all grid points in the third-floor east corridor area of ​​the static basic digital twin model is updated from the preset 22 degrees Celsius to 29.5 degrees Celsius, and the smoke concentration value is updated from the preset 0.01 milligrams per cubic meter to 0.042 milligrams per cubic meter. For other areas that have not received real-time abnormal sensing data, their dynamic state variables, after being optimally estimated by the Kalman filter algorithm, may still be close to or equal to the preset values. This update iteration transforms the environmental state in the digital twin from a static preset state to a state reflecting the latest measured data of the physical space, completing a synchronization between the digital twin and the physical space.

[0024] See Figure 3 In one embodiment of the present invention, the construction project fire emergency response method based on digital twins will initiate a simulation prediction of the fire development process after the digital twin is updated. Continuing with the example of the multi-story office building construction project, assume that the updated digital twin shows anomalies in the optimal estimated temperature and optimal estimated smoke concentration values ​​in the east corridor area of ​​the third floor. The system then continuously analyzes the real-time sensing data stream, identifying that the smoke concentration value from the smoke detector numbered D303 in the east corridor of the third floor rapidly increased from 0.042 mg / m³ to 0.25 mg / m³ within three consecutive sampling periods. Simultaneously, the adjacent temperature sensor T304 identified that the temperature value increased from 29.5 degrees Celsius to 35 degrees Celsius. The system determines that there is an abnormal increase in smoke concentration and an abnormal increase in temperature at this location, and identifies this spatial coordinate point, for example, three-dimensional coordinates (105, 40, 9.5), as a simulated fire initiation point.

[0025] In some embodiments, the system assigns fire initiation parameters to the identified fire initiation point. The combustible material type parameter is set based on the building information model attributes of the space where the coordinate point is located in the digital twin; for example, if the area is labeled "office furniture and paper document storage area," the combustible material type parameter is set to "mixed fuel Class A." The initial ignition source power parameter is estimated based on the initial temperature rise rate and spatial characteristics, for example, set to 1500 kilowatts. These fire initiation parameters, along with the spatial coordinates of the fire initiation point, are injected into the updated digital twin.

[0026] In practical implementation, the digital twin is essentially a virtual environment containing a finely detailed three-dimensional geometric mesh. The system activates a built-in computational fluid dynamics (CFD) solver within this mesh. The CFD solver solves the governing equations describing the flow of fire smoke over a mesh containing the local region of the fire initiation point and potentially affected adjacent regions. These governing equations include mass conservation equations, momentum conservation equations, energy conservation equations, and chemical component transport equations. The CFD solver uses the finite volume method for discretization, where the source term treatment of the energy equation is crucial. Its discretized form involves the heat release rate contribution of the fire source; a simplified expression for the volumetric heat source term is: ; in: This represents the volumetric heat source intensity within the computational grid cell, expressed in watts per cubic meter. It is the rate of volumetric heat release; Indicates the combustion efficiency factor; This represents the mass combustion rate of combustibles within a grid cell, expressed in kilograms per second. The effective calorific value of a combustible material is expressed in joules per kilogram. This represents the volume of the computational grid cell, in cubic meters. This formula converts the ignition source power and combustible material characteristics in the fire initiation parameters into energy inputs for computational fluid dynamics simulations.

[0027] Understandably, the computational fluid dynamics (CFD) solver advances the simulation process at preset time steps, solving a complete set of governing equations within each time step. The system sets the simulation time step to 0.1 seconds, with a total simulation duration of 600 seconds. In the first time step, based on the fire initiation parameters and initial environmental conditions, the CFD solver calculates the values ​​of temperature, smoke concentration, carbon monoxide concentration, and visibility fields within a small area centered on the fire initiation point. In the second time step, using the results from the previous time step as initial conditions and considering factors such as buoyancy, wall friction, and ventilation conditions, the CFD solver calculates the updated spatial distribution of temperature, smoke concentration, toxic gas concentration, and visibility fields as smoke spreads upward and begins to accumulate under the ceiling. This process iterates continuously, generating a set of three-dimensional field data covering the entire simulation area at each time step, recording the spatial distribution and time-varying values ​​of temperature, smoke, toxic gases, and visibility.

[0028] In practice, the computational fluid dynamics solver runs continuously for 6000 time steps, completing a full simulation from the current moment to 600 seconds into the future. The system records and outputs full-field data at every key time point throughout this process, such as the predicted temperature, predicted smoke concentration, predicted carbon monoxide concentration, and predicted visibility at every 3D grid point within the building at 30, 60, 120, 300, and 600 seconds into the future. This time-series organized full-field data set collectively constitutes the dynamic prediction of fire spread over a period of time from the current moment. The dynamic prediction results are stored in data files, explicitly including the direction and speed of fire smoke spread, the expansion range of high-temperature and dense smoke areas, and the timeline of environmental condition deterioration in each area.

[0029] In one embodiment of the present invention, after generating dynamic prediction results of fire spread, the construction project fire emergency linkage method based on digital twins initiates an assessment of the comprehensive threat level of different zones within the building. Taking the construction project of the multi-story office building as an example, the system divides the building plan into multiple assessment zones. For example, the third floor plan is divided into assessment zone A corresponding to the east corridor area, assessment zone B corresponding to the east office area, assessment zone C corresponding to the atrium area, assessment zone D corresponding to the west office area, and assessment zone E corresponding to the west corridor area. A hierarchical assessment index structure is constructed for each assessment zone, including fire thermal threat index, smoke toxicity threat index, and evacuation obstruction threat index. The fire thermal threat index represents the potential for high temperature to harm people and building components, the smoke toxicity threat index represents the risk of harm to people from toxic substances in smoke, and the evacuation obstruction threat index represents the degree of risk of restricted movement or blocked paths for people. The system extracts predicted data for each assessment zone at key future time points from the dynamic prediction results of fire spread. For the time point of 120 seconds in the future, the predicted temperature value for assessment zone A is 85 degrees Celsius and the predicted smoke concentration value is 15 milligrams per cubic meter. The predicted temperature value for assessment zone B is 45 degrees Celsius and the predicted smoke concentration value is 8 milligrams per cubic meter. At the same time, the system extracts real-time personnel density data for each assessment zone from real-time sensing data. For example, the real-time personnel density of assessment zone A is 0.3 people per square meter, the real-time personnel density of assessment zone B is 0.8 people per square meter, and the real-time personnel density of assessment zone C is 1.2 people per square meter, obtained by statistical analysis of personnel positioning terminal signals.

[0030] In some embodiments, the system uses the analytic hierarchy process (AHP) to determine the weights of each assessment indicator. By constructing a pairwise comparison judgment matrix for the fire thermal threat indicator, smoke toxicity threat indicator, and evacuation obstruction threat indicator, and calculating the eigenvector, the weights of the fire thermal threat indicator (0.5), smoke toxicity threat indicator (0.3), and evacuation obstruction threat indicator (0.2) are obtained. The system normalizes the raw data for each assessment zone by dividing the predicted temperature value by the reference temperature value (100 degrees Celsius) to obtain the normalized predicted temperature value, dividing the predicted smoke concentration value by the reference smoke concentration value (20 milligrams per cubic meter) to obtain the normalized predicted smoke concentration value, and dividing the real-time personnel density value by the reference personnel density value (2 people per square meter) to obtain the normalized real-time personnel density value. The comprehensive threat score for each assessment zone is calculated by weighted summation. The formula for calculating the comprehensive threat score is: ; in: This represents the overall threat score for assessment partition i; Indicates the weight of fire thermal threat indicators; This represents the normalized predicted temperature value for evaluation partition i; Indicates the weight of indicators of the threat of flue gas toxicity; This represents the normalized predicted smoke concentration value for evaluation partition i; The weights of indicators representing threats to evacuation are indicated. This represents the normalized real-time population density value for evaluation partition i. For evaluation partition A, the calculation process is as follows: For evaluation partition B, the calculation is as follows: .

[0031] In implementation, the system assigns a level label to each partition based on its overall threat score. A score of 0.6 or higher is classified as "high threat," between 0.3 and 0.6 as "medium threat," and less than 0.3 as "low threat." Partition A has an overall threat score of 0.68 and is therefore labeled "high threat," partition B has a score of 0.425 and is labeled "medium threat," and partition C has a score of 0.55 and is also labeled "medium threat." The system overlays all partition level labels onto the building plan using different colored areas, forming a dynamic threat assessment map. This dynamic threat assessment map is displayed in the digital twin's 3D visualization interface as red for partition A, yellow for partition B, and yellow for partition C. The system has a pre-set emergency rule base, which contains standard emergency response logic corresponding to different threat levels, different fire source locations, and different building functional areas. For example, the rule entry stipulates that "for high-threat-level corridor areas, the standard emergency response logic includes activating the smoke exhaust system in the area, lowering the nearby fireproof roller shutters, and broadcasting directional evacuation instructions."

[0032] In one embodiment of the present invention, after generating preliminary evacuation guidance and fire equipment linkage instructions, the construction project fire emergency linkage method based on digital twin performs spatial logic verification. The system maps the instructions involving spatial actions in the preliminary instructions to specific components in the building information model. For example, the suggested evacuation route "from each room through the west corridor to the west safety staircase" is mapped to a series of ID sequences of corridor and doorway components in the building information model. The instruction "start the third floor east side smoke exhaust fan FAN-E3" is mapped to the corresponding fan equipment model and the geometric range of the smoke control zone it serves in the building information model. The instruction "lower the fireproof roller shutter FIRECURTAIN-EC3" is mapped to the specific roller shutter component located at the junction of the east side and the atrium in the building information model. The system checks the real-time status based on the mapping relationship and the digital twin, verifying whether the suggested evacuation routes pass through areas already blocked by simulated fire or where fireproof roller shutters have been lowered. In the fire development simulation within the digital twin, the prediction results for the next 180 seconds show that part of the western corridor is already covered by high-temperature smoke; therefore, the system determines that a section of the original suggested evacuation route is unusable. The system checks whether the activation of the smoke exhaust fan conflicts with the design logic of adjacent smoke control zones. The building information model shows that smoke exhaust fan FAN-E3 is responsible for smoke control zone Z3-E, while the smoke exhaust fan FAN-C3 in the adjacent smoke control zone Z3-C is in a closed state. Activating FAN-E3 is consistent with the design logic. The system checks whether the descent sequence of the fireproof roller shutters will cut off the currently identified main evacuation flow. Real-time sensing data shows that a large number of people are moving westward in assessment zones B and C. The descent position of the fireproof roller shutter FIRECURTAIN-EC3 is located to its east, and it will not cut off the current main westward evacuation flow.

[0033] In some embodiments, after detecting a path conflict, the system adjusts the initial instructions according to preset conflict resolution rules. These rules include "automatically calculating an alternative safe path when the original path is blocked by fire." The system replans the path, generating conflict-free alternative instructions, adjusting the suggested evacuation path to "from each room through the north passage of the central hall to the north safety staircase." The system encodes all verified or adjusted instructions according to their execution sequence and device address, packaging them into an executable instruction set. The executable instruction set exists as a structured list, containing information such as the target device address, control command, and scheduled execution timestamp for each instruction. See Table 1 for a simplified instruction encoding snippet.

[0034] Table 1: Example Fragments of Executable Instruction Sets CMD-001 Fire controller-03 Start the smoke exhaust fan FAN-E3 0 CMD-002 Fireproof roller shutter controller-12 Fireproof Roller Shutter FIRECURTAIN-EC3 2 CMD-003 Emergency Broadcast Host The audio message reads, "Please evacuate via the north passageway of the central hall to the north exit." 5 CMD-004 Dynamic Indicator Controller - Zone B Set the indicator light direction to point north. 5 In practice, the system distributes executable instruction sets to fire emergency equipment and personnel guidance terminals in the physical space via a dedicated fire protection network or IoT gateway. For smoke exhaust fans, fireproof roller shutters, and fire pumps, the instructions are converted into standard control protocol signals recognizable by the equipment controllers. For example, the instruction "Start Smoke Exhaust Fan FAN-E3" is converted into a Modbus TCP protocol "Write Coil" command sent to the address "Fire Controller-03". For personnel guidance terminals, the instructions are converted into directional control commands for dynamic evacuation indicator lights, voice and text content for emergency broadcasts, and graphic information for displays. The system monitors the instruction reception confirmation signals of each device and terminal. If a success response code is received from "Fire Controller-03" within a predetermined time, it confirms that the instruction "Start Smoke Exhaust Fan FAN-E3" has been delivered and successfully executed.

[0035] It is understandable that during the execution of corresponding actions by the driving devices and terminals, the system continuously collects new real-time sensing data. For example, 10 seconds after the command is issued, a new set of temperature and smoke data is collected. The system compares this new real-time sensing data with the corresponding simulated expected data in the digital twin. The digital twin predicts the expected state of each sensor 10 seconds after the command is issued in the fire development simulation. The system calculates the data deviation, which involves comparing the measured values ​​and expected values ​​at the same location and time point. When the data deviation exceeds a preset threshold, the system triggers the model parameter calibration process of the digital twin, using the new real-time sensing data to correct key parameters of the simulation model, such as correcting the growth rate of the fire source heat release rate. The system feeds back the corrected key parameters of the simulation model to the data assimilation technology stage as model parameters in the next update iteration of the digital twin.

[0036] In practical implementation, after the model parameter calibration process is initiated, the system identifies the spatial location with the largest data deviation, uses the new real-time sensing data within the main calibration area as the calibration target value, and adjusts the key parameters in the computational fluid dynamics simulation model that affect the simulation results of the main calibration area in reverse. This process is achieved through an optimization algorithm. An objective function to guide the direction of parameter adjustment is to minimize the mean square error between the simulation output and the measured data, and its expression is: ; in: Represented by a set of key parameters The objective function value of the independent variable; Indicates the number of observation points used for calibration; Indicates the first New measured values ​​of real-time sensing data at each observation point; Indicates the use of the current set of key parameters. During the simulation, at the first Simulation output values ​​at each observation point. Iterative adjustment of key parameters in the system. This ensures that the adjusted model's simulation output in the main calibration region... Approaching new real-time sensing data This allows us to obtain the corrected key parameters.

[0037] Optionally, redundant links can be used in the network for issuing commands to ensure reliability. Optionally, for critical fire-fighting equipment, a retransmission mechanism can be set up when a confirmation signal is received for monitoring commands; commands will be automatically retransmitted if no confirmation signal is received. After the system completes parameter correction, the corrected key parameters, such as the heat release rate growth rate of the fire source (from 8.0 MW / min to 9.5 MW / min), are transmitted to the data assimilation module. During subsequent dynamic updates of the digital twin using data assimilation technology, the Kalman filter algorithm will use these corrected parameters as its process model parameters, thereby making predictions of future states more accurate and forming a closed loop that includes execution feedback and model optimization.

[0038] See Figure 4 This is a composite analysis diagram showing the impact of fire-resistant roller shutters on evacuation flow. Combining line graphs and area plots, it clearly illustrates the correlation between personnel density, evacuation flow smoothness, and the degree of impact of shutter descent at different shutter positions. This diagram can be directly used for decision support in fire emergency response. The WC2 roller shutter area has high personnel density, low evacuation flow smoothness, and significant roller shutter impact, making it a key control area for emergency evacuation. The descent sequence of the WC2 roller shutter can be adjusted, or evacuation guidance resources can be added to this area to reduce the impact on personnel evacuation. The quantitative data in this diagram can be used for parameter calibration of digital twin models, improving the accuracy of evacuation simulation in fire scenarios. The command execution success rate is generally higher than 98%, verifying the effectiveness of the dedicated fire protection network, redundant links, and retransmission mechanism, meeting the high availability requirements of the fire protection system, and can serve as a quantitative basis for project acceptance and operation and maintenance optimization.

[0039] In one embodiment of the present invention, the construction engineering fire emergency linkage method based on digital twin continuously performs data comparison and model optimization after executing the command. The system records the simulation time from the time the executable command set is issued to the current time. For example, the timer starts from the time the command "start the smoke exhaust fan FAN-E3" is sent. After 30 seconds, the system starts a new round of data comparison process to extract the simulation expected data at the simulation time scale of 30 seconds from the digital twin. The simulation expected data includes the expected temperature and expected smoke concentration of key locations calculated in advance by the fire development process simulation. For example, the expected temperature of point P1 on the east side of the third floor is 72 degrees Celsius and the expected smoke concentration is 8.5 mg / m³. The expected temperature of point P2 in the atrium on the third floor is 48 degrees Celsius and the expected smoke concentration is 3.2 mg / m³. The expected temperature of point P3 on the west side of the third floor is 32 degrees Celsius and the expected smoke concentration is 0.9 mg / m³. The system extracts the measured temperature and measured smoke concentration at the same physical location from the newly collected real-time sensing data. Through the sensor network, it reads the measured temperature of point P1 as 78 degrees Celsius and the measured smoke concentration as 9.8 milligrams per cubic meter, the measured temperature of point P2 as 55 degrees Celsius and the measured smoke concentration as 4.1 milligrams per cubic meter, and the measured temperature of point P3 as 31 degrees Celsius and the measured smoke concentration as 1.0 milligrams per cubic meter.

[0040] In some embodiments, the system calculates the absolute difference between the measured and expected values ​​at the same location. For point P1, the absolute temperature difference is calculated as |78-72| = 6 degrees Celsius, and the absolute difference in smoke concentration is calculated as |9.8-8.5| = 1.3 mg / m³. For point P2, the absolute temperature difference is calculated as |55-48| = 7 degrees Celsius, and the absolute difference in smoke concentration is calculated as |4.1-3.2| = 0.9 mg / m³. For point P3, the absolute temperature difference is calculated as |31-32| = 1 degree Celsius, and the absolute difference in smoke concentration is calculated as |1.0-0.9| = 0.1 mg / m³. The system calculates the average and variance of the absolute differences at all monitoring locations. The arithmetic mean of all six differences (three temperature differences and three concentration differences) is taken as the overall average data deviation. The variance of these six differences relative to this average is calculated as the overall data deviation variance, which serves as a measure of data deviation. The system sets the average threshold and variance threshold for data deviation. For example, the average threshold is set to 0.15 per unit in Celsius and milligrams per cubic meter, and the variance threshold is set to 0.01 per unit.

[0041] It is understandable that the system initiates a calibration process when the calculated data deviation exceeds any threshold. Assuming the average calculated data deviation is 0.18 per unit, exceeding the threshold of 0.15, the system immediately triggers the digital twin model parameter calibration process. The system identifies several spatial locations with the largest data deviations as the primary calibration areas. By sorting the deviation values ​​of each monitoring point, it is found that the temperature and smoke concentration deviations at point P2 are significant; therefore, the atrium area where point P2 is located is designated as the primary calibration area. The system uses new real-time sensing data within the primary calibration area as calibration target values; that is, the measured temperature of 55 degrees Celsius and the measured smoke concentration of 4.1 milligrams per cubic meter at point P2 are set as calibration targets. The system then adjusts the key parameters in the computational fluid dynamics simulation model that affect the simulation results of the primary calibration area. These key parameters include the rate of increase of the heat release rate of the fire source and the pyrolysis characteristics of the wall material. These parameters were initially set as estimates based on standard design values.

[0042] In practice, the model parameter calibration process adjusts key parameters through an iterative optimization algorithm. This algorithm aims to minimize the difference between the simulation output and the measured data. An objective function is defined to quantify this difference; the expression for the objective function is: ; in: Represented by a set of key parameters The objective function value of the independent variable; This indicates the number of observation points within the main calibration area; here, M is 1 (point P2). This represents the measured temperature value at the m-th observation point; Indicates the use of the current set of key parameters. The temperature value output at the m-th observation point during simulation; It is a normalized reference value for temperature; This represents the measured smoke concentration value at the m-th observation point; Indicates the use of the current set of key parameters. The smoke concentration value output at the m-th observation point during simulation; It is a normalized reference value for smoke concentration.

[0043] The system feeds back the corrected key parameters of the simulation model to the data assimilation technology stage, using them as model parameters in the digital twin update process. During the dynamic update of the digital twin by the data assimilation technology, the Kalman filter algorithm uses these corrected key parameters as its process model parameters. The Kalman filter algorithm then uses these updated process model parameters to predict the state of the digital twin at the next moment, more accurately reflecting the dynamic characteristics of fire development. The system fuses the new real-time sensing data with the predicted state using the new parameters, completing one iteration of the digital twin update through the Kalman filter algorithm's update steps, thus achieving model self-correction based on actual execution feedback.

[0044] See Figure 5 This is a multi-dimensional dynamic simulation analysis diagram of the fire development process, integrating three key indicators: fire spread rate, peak temperature, and visibility. It possesses significant engineering and technical value in fire emergency response scenarios. The fire peak at approximately 60 seconds can serve as the dividing point between the "rapid fire development phase" and the "stable spread phase," providing a basis for the timing design of emergency commands. At the fire peak (60 seconds), the linkage effect of equipment such as smoke exhaust fans and fireproof roller shutters can be verified through changes in temperature and visibility, optimizing the equipment start-up and shutdown sequence. The dynamic correlation of multiple indicators verifies the rationality of the computational fluid dynamics (CFD) simulation model and can serve as a quantitative verification basis for the "fire development process simulation" stage of the digital twin. The fluctuation characteristics of fire spread can be used to optimize the threshold settings of fire detection algorithms, reducing false alarms and missed alarms.

[0045] 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 construction project fire emergency response method based on digital twins, characterized in that, include: Collect real-time sensing data within the construction project, including temperature distribution, smoke concentration, video images, and personnel location information; By using real-time sensing data collected, the digital twin synchronized with the physical space is dynamically updated through data assimilation technology to ensure that the digital twin reflects the latest physical environment status. In the updated digital twin, simulated fire initiation parameters are injected, and a fire development process simulation based on computational fluid dynamics is run to generate dynamic prediction results of fire spread. Based on the dynamic prediction results of fire spread and the environmental and personnel data in the real-time sensing data, the analytic hierarchy process is used to assess the comprehensive threat level of different zones in the building, and a dynamic threat assessment map is obtained. Based on the dynamic threat assessment map and combined with the pre-set emergency rule base, an event-driven decision tree algorithm is used to generate preliminary evacuation guidance and fire equipment linkage instructions. The initial evacuation guidance and fire equipment linkage instructions are spatially logically verified with the building information model in the digital twin. After eliminating spatial conflicts, an executable instruction set is generated.

2. The construction project fire emergency response method based on digital twins according to claim 1, characterized in that, The process of dynamically updating a digital twin synchronized with the physical space using real-time sensing data and data assimilation technology includes: Establish a static basic digital twin model that includes building structure, equipment layout, and material properties; Spatial interpolation is performed on the temperature and smoke concentration data in the real-time sensing data to generate corresponding three-dimensional field data; Compare the three-dimensional field data with the preset state values ​​of the corresponding spatial locations in the static basic digital twin model; The Kalman filter algorithm is used to make optimal estimates of the dynamic state variables in the static basic digital twin model, which include the temperature value and smoke concentration value of each region. The optimally estimated dynamic state variables are used to replace the original corresponding values ​​in the static basic digital twin model, completing one update iteration of the digital twin.

3. The construction project fire emergency response method based on digital twins according to claim 1, characterized in that, The process involves injecting simulated fire initiation parameters into the updated digital twin, running a fire development process simulation based on computational fluid dynamics, and generating dynamic prediction results for fire spread, including: One or more initial locations where the temperature rises abnormally or the smoke concentration increases abnormally are identified from real-time sensing data, and these locations are used as the starting points of the simulated fire. Assign initial fire source power and combustible material type parameters to each fire initiation point; Within the three-dimensional geometric mesh of the digital twin, the computational fluid dynamics solver is activated to solve the mass, momentum, energy, and chemical component transport equations. The simulation process is advanced with a preset time step, and the spatial distribution and temporal evolution data of the temperature field, smoke concentration field, toxic gas concentration field and visibility field are calculated at each time step. Record and output dynamic predictions of fire spread over a period of time starting from the current moment.

4. The construction project fire emergency response method based on digital twins according to claim 1, characterized in that, Based on the dynamic prediction results of fire spread and the environmental and personnel data in real-time sensing data, the analytic hierarchy process (AHP) is used to assess the comprehensive threat level of different zones within the building, resulting in a dynamic threat assessment map, including: The building plan is divided into multiple assessment zones; A hierarchical assessment index structure is constructed for each assessment zone, including fire thermal threat, smoke toxicity threat, and evacuation obstruction threat. From the dynamic prediction results of fire spread, extract the predicted temperature and predicted smoke concentration data of each assessment zone at key future time points; Extract real-time personnel density data for each assessment zone from the real-time sensing data; The weights of each indicator—fire thermal threat, smoke toxicity threat, and evacuation obstruction threat—were determined using the analytic hierarchy process (AHP). After normalizing the predicted temperature, predicted smoke concentration, and real-time personnel density data for each assessment zone, the comprehensive threat score for each zone is calculated by combining the weights of the corresponding indicators. Based on the overall threat score, each zone is assigned a level label, and the level labels of all zones are overlaid on the building floor plan to form a dynamic threat assessment map.

5. The construction project fire emergency response method based on digital twins according to claim 1, characterized in that, Based on the dynamic threat assessment map and a pre-set emergency rule base, an event-driven decision tree algorithm is used to generate preliminary evacuation guidance and fire equipment linkage instructions, including: A pre-built emergency rule base contains standard emergency response logic corresponding to different threat levels, different fire source locations, and different building functional areas; The partition with the highest threat level in the dynamic threat assessment graph is used as the key event to trigger the root node of the decision tree; Based on the location of the fire's origin, its current direction of spread, and the distribution of building exits, select the appropriate branch logic at the decision node of the decision tree; Traverse the decision tree until a leaf node is reached. Each leaf node corresponds to a set of predefined preliminary instructions, which include suggested evacuation routes, start and stop of smoke exhaust fans, raising and lowering of fireproof roller shutters, emergency broadcast content, and start and stop of fire pumps.

6. The construction project fire emergency response method based on digital twins according to claim 1, characterized in that, The initial evacuation guidance and fire equipment linkage commands are spatially logically verified with the building information model in the digital twin. After eliminating spatial conflicts, an executable command set is generated, including: The instructions involving spatial actions in the initial instructions are mapped to specific components in the building information model; Check whether the recommended evacuation routes pass through areas that have been blocked off by simulated fires or where fireproof roller shutters have been lowered; Check whether the operation of the smoke exhaust fan conflicts with the design logic of adjacent smoke control zones; Check whether the descent sequence of the fireproof roller shutters will disrupt the currently identified main personnel evacuation routes; If a spatial logic conflict is found, the initial instruction is adjusted according to the preset conflict resolution rules to generate a conflict-free alternative instruction. All instructions that have passed verification or adjustment are encoded according to their execution sequence and device address, and packaged into an executable instruction set.

7. The construction project fire emergency response method based on digital twins according to claim 6, characterized in that, The method further includes: The executable instruction set is sent to the fire emergency equipment and personnel guidance terminals in the physical space to drive them to perform corresponding actions; During the execution of the action, new real-time sensing data is continuously collected and compared with the simulation expected data at the corresponding moment in the digital twin to calculate the data deviation; When the data deviation exceeds the preset threshold, the model parameter calibration process of the digital twin is triggered, and the key parameters of the simulation model are corrected using new real-time sensing data. The revised key parameters of the simulation model are fed back to the data assimilation technology stage as model parameters in the digital twin update process; The step of sending the executable instruction set to the fire emergency equipment and personnel guidance terminals in the physical space to drive them to perform corresponding actions includes: The executable instruction set is distributed via a dedicated fire protection network or an IoT gateway; For smoke exhaust fans, fireproof roller shutters, and fire pumps, the commands are converted into standard control protocol signals that the equipment controller can recognize; For personnel guidance terminals, instructions are converted into directional control commands for dynamic evacuation indicator lights, voice and text content for emergency broadcasts, and graphic and text information for displays. Monitor the instruction reception confirmation signals of each device and terminal to ensure that the instructions have been delivered.

8. The construction project fire emergency response method based on digital twins according to claim 7, characterized in that, During the execution of the action, new real-time sensing data is continuously collected and compared with the corresponding simulated expected data in the digital twin to calculate the data deviation, including: Record the simulation time elapsed from the issuance of the instruction to the current moment; From the digital twin, the simulation expectation data of the simulation time scale is extracted, and the simulation expectation data includes the expected temperature and expected smoke concentration at key locations; Extract the measured temperature and measured smoke concentration at the same physical location from the newly collected real-time sensing data; Calculate the absolute difference between the measured value and the expected value at the same location; The average and variance of the absolute differences at all monitoring locations are calculated as a measure of overall data deviation.

9. The construction project fire emergency response method based on digital twins according to claim 7, characterized in that, When the data deviation exceeds a preset threshold, the model parameter calibration process of the digital twin is triggered, using new real-time sensing data to correct key parameters of the simulation model, including: Set the average and variance thresholds for the data deviation; When the calculated data deviation exceeds any threshold, the calibration process is initiated. Identify the spatial locations with the largest data deviations and use them as the main calibration areas; Use new real-time sensing data within the main calibration area as the calibration target value; Inversely adjust the key parameters in the computational fluid dynamics simulation model that affect the simulation results of the main calibration region. These key parameters include the rate of increase of the heat release rate of the fire source and the pyrolysis characteristics of the wall material. The key parameters are iteratively adjusted so that the simulation output of the adjusted model in the main calibration area is close to the new real-time sensing data. The corrected key parameters are then used as the output of the model parameter calibration process.

10. The construction project fire emergency response method based on digital twins according to claim 7, characterized in that, The step of feeding back the corrected simulation model's key parameters to the data assimilation technology stage as model parameters during the digital twin update process includes: When the data assimilation technology dynamically updates the digital twin, the corrected key parameters of the simulation model are used as the process model parameters of the Kalman filter algorithm. Using the updated process model parameters, predict the state of the digital twin at the next time step; By fusing new real-time sensing data with the state predicted using new parameters, the digital twin is updated.

Images