Fire-fighting fire dynamic simulation method based on multi-physical field coupling

Abstract

The application relates to the technical field of fire-fighting simulation, and discloses a fire-fighting fire dynamic simulation method based on multi-physical field coupling. The method collects temperature field distribution data, smoke concentration gradient data, airflow velocity field data and building structure heat conduction parameters of a fire scene; a two-way coupling model of a temperature field and a smoke diffusion field is established, dynamic correction of a temperature gradient on a smoke convection effect is realized through real-time data exchange, a correlation model of an airflow velocity field and building structure heat conduction parameters is constructed, and a velocity field vortex coefficient is adjusted according to heat flow intensity; the two-way coupling model and the correlation model are integrated to form a multi-physical field joint calculation framework, a discrete time slice method is used to promote the simulation process; and a three-dimensional dynamic simulation atlas containing a temperature field evolution track, a smoke diffusion path and an airflow disturbance range is output. The method can accurately restore the multi-physical field synergistic process of a fire scene, and provides effective support for fire-fighting rescue decision-making and building fire prevention design.

Description

Technical Field

[0001] This invention relates to the field of fire simulation technology, specifically a dynamic fire simulation method based on multi-physics coupling. Background Technology

[0002] In the field of fire protection, dynamic fire simulation is an important tool for developing fire-fighting and rescue plans and optimizing building fire protection design. Traditional fire simulation systems often focus on analyzing a single physical field, such as predicting fire spread trends based solely on temperature field changes or planning evacuation routes based solely on smoke concentration distribution. These methods fail to fully reflect the complex characteristics of multi-physical field interactions at a fire scene. As building structures become increasingly complex and functional zones become more diverse, the limitations of single-physical-field simulation are becoming increasingly apparent.

[0003] In real-world fire scenarios, changes in the temperature field directly affect the convective diffusion trajectory of smoke. Hot air currents formed in high-temperature areas accelerate the movement of smoke in specific directions, while the accumulation of smoke alters the efficiency of heat radiation transfer in local spaces, thus having a counter-effect on the temperature field distribution. Simultaneously, the thermal conductivity of the building structure affects the rate of heat transfer in components such as walls and floors. This heat transfer, in turn, changes the temperature and density of the surrounding airflow, leading to vortex changes in the airflow velocity field. Traditional simulation systems do not consider these coupling relationships between multiple physics fields, relying solely on preset empirical formulas to calculate the parameters of each physics field in isolation. This results in simulation results that deviate significantly from the actual development of a fire.

[0004] Traditional simulation systems often employ a fixed time step during the simulation process, making it impossible to adjust the calculation accuracy according to the dynamic changes in fire development. During the rapid spread phase of a fire, a fixed time step struggles to capture instantaneous changes in the temperature and smoke fields, leading to lagging simulation results. Conversely, during the stable phase of a fire, a fixed time step wastes computational resources. Furthermore, traditional systems typically output simulation results in two-dimensional charts, failing to visually demonstrate the correlation between temperature field evolution, smoke diffusion paths, and airflow disturbances in three-dimensional space. This hinders firefighters from quickly understanding the fire situation and developing precise rescue strategies. These problems prevent traditional fire simulation systems from fully playing their role in decision support in practical applications and from meeting the needs of fire rescue and fire prevention design in complex building scenarios. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic simulation method for fire fighting based on multi-physics coupling, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a dynamic simulation method for fire fighting based on multiphysics coupling, the method comprising:

[0007] Collect temperature field distribution data, smoke concentration gradient data, airflow velocity field data, and building structure heat conduction parameters at the fire scene;

[0008] A two-way coupled model of temperature field and smoke diffusion field is established, and the dynamic correction of the smoke convection effect by temperature gradient is realized through real-time data exchange.

[0009] A correlation model between the airflow velocity field and the heat conduction parameters of the building structure is constructed, and the vortex coefficient of the velocity field is adjusted according to the heat flux intensity.

[0010] A multiphysics joint computation framework is formed by integrating bidirectional coupling models and correlation models, and the simulation process is advanced by using discrete time slicing.

[0011] The output includes a three-dimensional dynamic simulation map of temperature field evolution trajectory, smoke diffusion path, and airflow disturbance range.

[0012] Preferably, the specific process of establishing the two-way coupling model of the temperature field and the smoke diffusion field includes:

[0013] Extract the boundary coordinates of high-temperature regions from the temperature field distribution data;

[0014] Map the boundary coordinates of the high-temperature region to the spatial grid of the smoke concentration gradient data;

[0015] Calculate the migration acceleration of smoke particles under the influence of temperature gradient based on the mapping results;

[0016] An iterative feedback mechanism is used to input the migration acceleration in reverse to the transmission equation of the temperature field distribution data.

[0017] Preferably, the following operations are performed simultaneously when calculating the migration acceleration of smoke particles under the influence of a temperature gradient:

[0018] Detect the displacement of the boundary coordinates of the high-temperature region between adjacent time slices;

[0019] The sampling frequency of the smoke concentration gradient data is dynamically adjusted based on the displacement.

[0020] The adjusted sampling frequency is used as the time step constraint for the transmission equation.

[0021] Preferably, when constructing the correlation model between the airflow velocity field and the building structure heat conduction parameters:

[0022] Identify the distribution of material thermal resistivity in the thermal conduction parameters of building structures;

[0023] The velocity field computational domain is divided into multiple meshes based on the distribution of thermal resistivity.

[0024] A linear proportional relationship between heat flux intensity and vortex coefficient is established within each grid layer.

[0025] Preferably, the multi-layer mesh division process includes:

[0026] Mesh layering identifiers are generated based on abrupt changes in thermal resistivity.

[0027] An adaptive octree partitioning is used for the continuous thermal resistivity region;

[0028] A heat flux intensity buffer transition zone is set at the mesh interface.

[0029] Preferably, the operational logic of the multiphysics joint computation framework includes:

[0030] The bidirectional coupling model and the associated model are invoked synchronously at the start of each time slice.

[0031] The vortex coefficients output by the associated model are used as the initial boundary values ​​of the two-way coupled model.

[0032] The mesh parameters of the correlation model are corrected by the heat flux intensity generated by the two-way coupled model.

[0033] Preferably, the advancement control method of the time slice is as follows:

[0034] Monitor the maximum gradient change rate of the temperature field within each slice. When the change rate exceeds the critical threshold, shorten the time step of the next slice. Insert an auxiliary calculation slice when the smoke diffusion path bifurcates.

[0035] Preferably, the method for generating the three-dimensional dynamic simulation map includes:

[0036] Transform the temperature field evolution trajectory into an isothermal surface topology;

[0037] Embed smoke concentration chromaticity values ​​at the topology nodes;

[0038] The distribution layer of vortex vector arrows is generated based on the range of airflow disturbance.

[0039] Preferably, the process of constructing the isothermal surface topology includes:

[0040] Extract the extreme points of the temperature field at discrete time points as topological keyframes;

[0041] Transition surfaces between keyframes are generated using cubic spline interpolation;

[0042] Perform thermodynamic parameter-driven mesh subdivision on the transition surface.

[0043] Preferably, the method further includes a data verification mechanism:

[0044] Compare the timestamps of the multiphysics data before each simulation process proceeds;

[0045] Data resynchronization is triggered when the time deviation between the temperature field and the smoke diffusion field exceeds the tolerance.

[0046] Set a version number verification identifier for the thermal conduction parameters of the building structure.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] At the model construction level, the method establishes a two-way coupled model of the temperature field and the smoke diffusion field. Through real-time data exchange, it dynamically corrects the temperature gradient's effect on smoke convection, realistically reproducing the interaction between the temperature and smoke fields. Changes in the temperature field are fed back into the smoke diffusion field calculation in real time, adjusting the migration trajectory of smoke particles. Conversely, the heat transfer generated during smoke diffusion affects the temperature field distribution. This two-way coupling mechanism breaks the limitations of traditional single-physics-field simulations, making the simulation results of the temperature and smoke fields more consistent with the dynamic correlation characteristics of the two in actual fires. Simultaneously, the method constructs a correlation model between the airflow velocity field and the building structure's thermal conductivity parameters. By adjusting the vortex coefficient of the velocity field according to the heat flux intensity, it accurately reflects the influence of the building structure's thermal conductivity characteristics on airflow motion. Differences in thermal resistance coefficients in different areas of the building structure lead to different heat flux intensity distributions, thus altering the vortex state of the airflow velocity field. This correlation model, by establishing a correspondence between heat flux intensity and vortex coefficient, makes the simulation of the airflow velocity field more closely match the actual structural characteristics of the building, further improving the overall simulation accuracy.

[0049] At the computational framework level, the method integrates bidirectional coupling and correlation models to form a multiphysics joint computation framework, and employs a discrete-time slicing approach to advance the simulation process. This multiphysics joint computation framework enables coordinated calculation of temperature, smoke diffusion, airflow velocity, and building structure heat conduction. The calculation results of each physics field can be mutually fed back and corrected, avoiding the disconnect caused by isolated calculations of each physics field in traditional simulation systems. This makes the simulation process more closely reflect the actual situation of multiphysics interaction at a fire scene. The discrete-time slicing approach allows for flexible simulation advancement based on the dynamic process of fire development. Compared to traditional fixed time steps, it better balances simulation accuracy and computational efficiency. It can accurately capture instantaneous changes in the physics field during rapid fire phases and reasonably control computational resource consumption during stable fire phases, achieving a balance between simulation accuracy and computational efficiency.

[0050] At the output level, the method outputs a three-dimensional dynamic simulation map containing the temperature field evolution trajectory, smoke diffusion path, and airflow disturbance range. This three-dimensional dynamic format visually demonstrates the changes and relationships between various physical fields in three-dimensional space. Firefighters can clearly observe how the temperature field spreads within the building space over time, how smoke diffuses along specific paths, and how airflow forms vortex disturbances in different areas. This intuitive presentation helps rescuers quickly grasp the fire situation and understand the interactions between various physical fields, providing a clear reference for developing targeted firefighting and rescue plans and planning safe evacuation routes. Simultaneously, the three-dimensional dynamic simulation map can also be used for building fire protection design optimization. Designers can visually identify areas in the building structure prone to fire spread, smoke accumulation, or poor airflow through the map, thereby optimizing and adjusting the building's fire-resistant structure and ventilation system layout to improve the overall fire resistance performance of the building. Attached Figure Description

[0051] Figure 1 This is a timing diagram of the fire dynamic simulation method based on multi-physics coupling described in this invention.

[0052] Figure 2 A flowchart for establishing a two-way coupled model of the temperature field and the smoke diffusion field;

[0053] Figure 3 A flowchart for constructing a model relating airflow velocity field to building structure heat conduction parameters. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0055] Please see Figure 1 The present invention provides a fire dynamic simulation system based on multi-physics coupling. The system includes collecting temperature field distribution data, smoke concentration gradient data, airflow velocity field data and building structure heat conduction parameters through a sensor network deployed at the fire site. These data are stored in a central database in the form of point clouds.

[0056] Temperature field distribution data was acquired by an infrared thermal imager at a rate of 10 frames per second. Smoke concentration gradient data was measured using a laser scattering particulate sensor. Airflow velocity field data was recorded using an ultrasonic anemometer. Building structure thermal conductivity parameters, including the thermal conductivity, specific heat capacity, and density of the materials, were derived from the building information model. When establishing a two-way coupled model of the temperature field and the smoke diffusion field, the finite volume method was used to discretize the physical fields. A data exchange interface was set up within each computational unit to dynamically correct the temperature gradient for the smoke convection effect. Temperature field data was injected as a source term into the smoke transport equation, while smoke concentration feedback affected the radiative heat transfer calculation. When constructing a correlation model between the airflow velocity field and the building structure thermal conductivity parameters, an unstructured mesh was generated based on computational fluid dynamics. A functional relationship between heat flux intensity and vortex coefficient was established at the mesh nodes, and the velocity field parameters were updated by solving the energy equation. The multiphysics joint computation framework integrates the two-way coupled model and the correlation model, employing an explicit time integration scheme and advancing the simulation process with a 0.1-second time slice as the baseline. The control equations of each physical field are solved in parallel within each slice, and inter-process communication is achieved through a message passing interface. The output of the 3D dynamic simulation map uses an OpenGL visualization engine to render the temperature field evolution trajectory as an isosurface animation, the smoke diffusion path as particle trajectory lines, and the airflow disturbance range as streamline integrals, ultimately generating an interactive 3D scene.

[0057] Example 1: See Figure 2 The establishment of a two-way coupling model between the temperature field and the smoke diffusion field uses temperature field distribution data and smoke concentration gradient data collected at the fire scene as input. The extraction of the boundary coordinates of the high-temperature region employs a moving cube algorithm to process the temperature field distribution data. This algorithm scans the entire three-dimensional temperature data field, identifies voxel sets with temperature values ​​exceeding a preset threshold (e.g., 300 degrees Celsius), and generates continuous boundary surfaces by connecting the isosurfaces of these voxels. The boundary coordinates are recorded as a series of three-dimensional spatial point sets, each containing its spatial coordinates and corresponding temperature value. When mapping the boundary coordinates of the high-temperature region to the spatial grid of the smoke concentration gradient data, it is necessary to address the potential grid inconsistency between the two types of data. Smoke concentration data typically originates from uniformly divided regular grids, while the extracted boundary coordinates may be discrete points. Therefore, a spatial hash table is used to establish the coordinate correspondence, mapping each boundary point coordinate to the center of the nearest grid cell in the smoke grid. For boundary points falling within a grid cell, their precise weight coefficients in the smoke grid are calculated using trilinear interpolation. This mapping ensures the precise spatial alignment of the temperature field boundary and the smoke concentration field.

[0058] The migration acceleration of smoke particles under the influence of temperature gradient is calculated based on the mapping results. This step is based on the principle of buoyancy, where temperature differences cause changes in air density, resulting in a buoyancy effect. For each mapped grid cell, the temperature difference between its temperature and the average temperature of the surrounding environment is calculated. Combining the local gravitational acceleration and thermal expansion coefficient, the buoyancy force on the smoke particles is derived according to the Businesk approximation. Then, the migration acceleration vector is calculated based on the particle mass. This acceleration vector has magnitude and direction, with the direction perpendicular to the temperature isosurface and pointing from the low temperature region to the high temperature region. An iterative feedback mechanism is used to input the migration acceleration back into the conduction equation of the temperature field distribution data. Within each calculation time step, the heat conduction equation is first solved to obtain the preliminary temperature field distribution. Then, the calculated smoke particle migration acceleration is added as an additional source term to the energy equation. The impact of acceleration on the temperature field is reflected in the enhancement of its convective heat transfer effect. This feedback process requires multiple iterations to achieve coupling convergence between the two physical fields. The iteration termination condition is set to the temperature field change norm being less than a specified tolerance for two consecutive iterations.

[0059] When calculating the migration acceleration of smoke particles, the displacement of the boundary coordinates of the high-temperature region between adjacent time slices is detected simultaneously. The displacement is calculated by comparing the changes in the position of the boundary point set between the current time step and the previous time step. A point set registration algorithm (such as the iterative nearest point algorithm) is used to calculate the rigid body transformation parameters between the two point sets, thereby obtaining the overall displacement vector. The magnitude of the displacement reflects the speed of fire source spread or movement. The sampling frequency of the smoke concentration gradient data is dynamically adjusted according to the displacement. When the detected displacement exceeds a preset threshold (e.g., one-tenth of the grid size), it indicates a drastic change in the fire situation, requiring an increase in the data acquisition and simulation frequency. The system automatically increases the sampling frequency of the smoke concentration from the base value to a higher frequency to obtain denser time series data. The adjusted sampling frequency is used as a time step constraint for the propagation equation. In numerical simulation, the selection of the time step must meet the stability condition. The reciprocal of the sampling frequency directly determines the maximum time step that can be used in the simulation. The system dynamically adjusts the solver's time step parameters according to the current effective sampling frequency to ensure the numerical stability and accuracy of the simulation.

[0060] The bidirectional coupled model employs a divide-and-conquer strategy, breaking down the complex multiphysics problem into multiple sub-problems for sequential solution. Within each time step, the temperature field equations, considering heat conduction and radiation, are solved independently first, followed by the smoke transport equations, which incorporate temperature buoyancy effects. The two equations are coupled through alternating iterations, with the Jacobi iteration method used to handle nonlinear coupling terms during the iteration process. Regarding model parameter configuration, physical properties of smoke particles, such as diameter and density, are set to default values ​​based on the actual combustible material types in the fire scene. Material property parameters (such as thermal conductivity and specific heat capacity) in the temperature field conduction equations are dynamically obtained from a building structure database. Boundary conditions are set considering the actual scenario; for example, walls are set as adiabatic boundaries, and openings are set as pressure boundaries. The data exchange mechanism utilizes an efficient memory-sharing approach. The bidirectional coupled model runs on a multi-core computing platform, with temperature field and smoke field calculation tasks allocated to different processor cores. A semaphore mechanism controls the synchronization of data access, avoiding read-write conflicts.

[0061] Example 2: See Figure 3 The correlation model between airflow velocity field and building structure thermal conduction parameters aims to accurately reflect the influence of building structure thermal characteristics on airflow patterns. Identifying the distribution of material thermal resistance coefficients in the building structure's thermal conduction parameters is the starting point of the entire process. Material properties of all components such as walls, floors, doors, and windows are extracted from the building's digital model. These properties are usually stored in the building information model's material library and include basic parameters such as thermal conductivity, thickness, and density. The thermal resistance value at each location is obtained by calculating the ratio of thickness to thermal conductivity, generating a continuously varying thermal resistance scalar field within the building space. A multi-layered grid is created for the velocity field computational domain based on the thermal resistance coefficient distribution. The gradient of the thermal resistance field becomes the main basis for the grid density layout. Mesh refinement is performed in areas of abrupt change in thermal resistance, such as at the junctions of different materials or where the insulation layer is interrupted. An unstructured tetrahedral grid is generated using the leading-edge propagation method. The determination of the number of grid layers depends on the spectral analysis of the rate of change of the thermal resistance field, striving to achieve a balance between computational efficiency and accuracy.

[0062] A linear proportional relationship between heat flux intensity and vortex coefficient is established within each grid layer. Heat flux intensity is calculated based on Fourier's law, obtained by multiplying the temperature gradient and thermal conductivity. The vortex coefficient characterizes the rotational tendency of airflow driven by heat. The proportionality between the two needs to be calibrated through benchmark tests in standard fire scenarios. This relationship allows the heat transfer effect to directly influence the turbulence characteristics of the airflow. The multi-layer grid generation process uses abrupt changes in thermal resistance coefficient as boundaries to generate grid layer markers. When the thermal resistance difference between adjacent grid cells exceeds a preset threshold, the system automatically sets a layer marker at that interface. These markers become control points for subsequent adaptive grid refinement. The marker generation employs an edge detection algorithm to ensure accurate capture of the material interface location. An adaptive octree partitioning method is used for the continuous thermal resistivity region. The initial computational mesh is relatively coarse to improve computational efficiency. During the solution process, the mesh refinement is determined based on the estimated error of the second derivative of the thermal resistivity field. Regions with large errors are subdivided into progressively smaller mesh cells. The maximum refinement depth is limited by computational resources. The goal of refinement is to maintain the linearity of the thermal resistivity field within a single mesh cell. A heat flux intensity buffer transition zone is set at the mesh interface. The width of the transition zone is dynamically determined based on the size of adjacent meshes and the difference in thermal resistivity. A high-order interpolation method is used within the zone to smooth the heat flux intensity values, avoiding numerical discontinuities caused by abrupt changes in material properties. This approach effectively improves computational stability.

[0063] The solution of the associated model is based on the fundamental governing equations of computational fluid dynamics, including mass conservation, momentum conservation, and energy conservation equations. The Reynolds stress model, suitable for fire simulation, is selected for the turbulence model. Heat flux intensity is added as a source term in the energy equation to the momentum equation, reflecting the influence of heat-driven airflow. The vortex coefficient is updated at each time step. First, the velocity distribution is obtained by solving the flow field; then, the vortex coefficient is corrected based on the local heat flux intensity value; finally, the turbulent viscosity is updated to complete the constitutive relation closure. Mesh data is managed using a hierarchical data structure. Each mesh cell stores its geometric information, physical parameters, and computational state variables. During parallel computation, a space-filling curve is used for domain decomposition to ensure load balance. Boundary condition treatment considers the specific characteristics of building fires. No-slip boundary conditions and the standard wall function method are used to handle near-wall flow on solid walls. Pressure boundary conditions are set for open boundaries based on the indoor-outdoor pressure difference. The calculation of heat flux intensity considers the coupling effect of radiation and convection. The initialization of model parameters relies on a physical property database of building materials, which contains the thermal performance parameters of common building materials at different temperatures. During the simulation, parameter values ​​are dynamically updated according to the local temperature. The calculation results were validated by comparing them with standard fire experimental data, focusing on the degree of matching between the velocity field distribution and the temperature field. Error analysis employed statistical methods to evaluate the reliability of the simulation results. The entire correlated model and the bidirectional coupled model maintained synchronous data exchange, together forming an important component of the multiphysics joint calculation framework.

[0064] Example 3: The multiphysics joint computation framework serves as the core scheduling engine of the system, responsible for coordinating data interaction and computational synchronization between the temperature field-smoke diffusion field bidirectional coupling model and the airflow velocity field-building structure heat conduction parameter correlation model. The framework adopts a master-slave parallel architecture, with the master process acting as the scheduler and slave processes each carrying a physical field model instance. Cross-process communication is achieved through a message passing interface. At the start of each time slice, the master process broadcasts a synchronization signal to all slave processes, triggering parallel calls to the bidirectional coupling model and the correlation model. The bidirectional coupling model prioritizes initializing the temperature field solution routine, while the correlation model simultaneously starts the airflow field calculation task. The model call process employs a non-blocking communication mode, allowing overlapping execution of computation and data exchange to improve efficiency. The vortex coefficients output by the correlation model are used as the initial boundary values ​​of the bidirectional coupling model. The vortex coefficient matrix is ​​read from the memory buffer of the correlation model and mapped to the mesh boundary nodes of the bidirectional coupling model using a spatial interpolation method. The interpolation process employs an inverse distance weighted algorithm to ensure the continuity of physical quantities at the interface. The boundary values ​​are set considering the local mesh size and physical characteristics to avoid numerical singularities. The heat flux intensity generated by the bidirectional coupling model corrects the mesh parameters of the associated model. The heat flux intensity field is passed to the associated model as a scalar input, driving the mesh adaptive reconstruction process. The reconstruction judgment is based on the local error estimate of the heat flux intensity gradient. When the error exceeds the limit, mesh refinement or coarsening is triggered.

[0065] The time-slice advancement control method monitors the maximum gradient rate of change of the temperature field within the previous slice. This rate of change is obtained by calculating the difference quotient of the temperature gradient amplitudes between consecutive time slices. The monitoring range covers the entire computational domain, and a parallel reduction algorithm is used to quickly locate the maximum value. When the rate of change exceeds a critical threshold, the time step of the next slice is shortened. The threshold is dynamically set according to the simulation accuracy requirements, and the step size adjustment strategy adopts a linear decay mode. The new step size is the weighted average of the original step size and the reciprocal of the rate of change. When a bifurcation occurs in the smoke diffusion path, an auxiliary computational slice is inserted. Bifurcation detection is based on topological analysis of the smoke concentration field, determining connectivity changes by calculating the Euler eigenvalues ​​of the concentration isosurfaces. When isosurface splitting or merging is detected, a micro-step auxiliary slice is inserted after the current slice. The duration of the auxiliary slice is set to a fraction of the main step size to ensure the accuracy of capturing transient phenomena. The insertion mechanism adopts an event-driven approach to avoid unnecessary computational overhead.

[0066] During framework operation, a data synchronization mechanism ensures temporal consistency between physical fields. At the start of each time slice, the data timestamps of each model are verified; if deviations exceed limits, interpolation or extrapolation compensation is triggered. The transfer of vortex coefficients uses shared memory to reduce data copy overhead. A relaxation factor is introduced into the feedback loop for heat flux intensity to control the iteration convergence speed. The mesh parameter correction process involves dynamic load balancing; when mesh reconstruction leads to uneven computational load, a domain decomposition and repartitioning algorithm is initiated to optimize task allocation among processors. Adaptive adjustment of the time step follows numerical stability criteria, constraining the maximum allowable step size based on CFL conditions. A sliding window average filter is introduced to monitor the rate of change, smoothing out misjudgments caused by instantaneous fluctuations. The bifurcation detection algorithm integrates machine learning-assisted identification; training data comes from historical fire cases to improve the accuracy of bifurcation judgment. Auxiliary slice calculations use a low-precision mode for rapid advancement, with the results serving as initial guesses for the main slice, accelerating overall convergence. The joint framework's fault tolerance includes computational anomaly detection and recovery; when the physical field solution diverges, it reverts to the previous slice state, reducing the step size and retrying. The framework interface design supports modular expansion, allowing the integration of additional physics models, such as toxic gas diffusion or structural deformation simulations. The visualization module is asynchronously coupled with the computation engine, rendering the simulation progress in real time and supporting interactive parameter adjustments by the user.

[0067] The formula is used to quantify the calculation of the maximum rate of change of the temperature field gradient:

[0068] ;

[0069] in: This represents the maximum rate of change of the temperature gradient. Represents the spatial extent of the computational domain. It is the temperature gradient vector. At the current time point, It is the time step of all previous pieces. This represents the Euclidean norm. The formula captures the intensity of fire development by calculating the maximum rate of change of the temperature gradient norm between adjacent time slices. Exceeding the preset threshold At that time, the time step adjustment mechanism is triggered.

[0070] The rate of change is calculated based on discrete temperature field data; the gradient is solved using the central difference method; the norm calculation covers all grid cells; and the maximum value retrieval utilizes a parallel scanning algorithm to optimize performance. Critical value. The framework is designed with reference to the ignition point characteristics and spatial scale of the reference materials, and dynamically adjusts the strategy to account for the simulation stage and error accumulation effects. Formulas are integrated into the framework monitoring module and are automatically executed at the end of each time slice, with the results used to determine the next parameter input. Mesh parameter correction and heat flux intensity feedback form a closed loop; heat flux intensity values ​​drive the update of turbulence parameters in the associated model, and vortex coefficients are recalculated and fed back to the bidirectional coupled model, with iterations continuing until residual convergence. Time slice management employs a hierarchical strategy: the main slice handles macroscopic evolution, auxiliary slices capture local mutations, and data transfer between slices ensures the consistency of physical field coupling. The framework is implemented based on C++ and the MPI library, with standardized model interfaces facilitating the integration of third-party solvers. The computational state is periodically saved to support breakpoint continuation. An exception handling mechanism monitors memory usage and computational accuracy, and an automatic degradation strategy addresses hardware failures. Simulation results are output in multiple formats, supporting post-analysis and visualization.

[0071] Taking a fire in a large underground parking lot as an example, the parking lot is a reinforced concrete structure with an area of ​​approximately 8,000 square meters and a floor height of 4.5 meters. It is equipped with a mechanical smoke extraction system and an automatic sprinkler system. The cause of the fire was a short circuit in an electric vehicle charging station, with the initial fire source located in the central area of ​​the parking lot. The operating logic of the multiphysics joint computing framework in this scenario is as follows: During the initialization of the simulation system, a three-dimensional model of the parking lot is loaded, and the space is discretized into a hexahedral mesh with a side length of 0.5 meters, with a total of approximately 600,000 meshes. At the start of each time slice, the master process distributes computing tasks to the computing cluster nodes. Node A is responsible for the bidirectional coupling model of the temperature field and the smoke diffusion field, while node B simultaneously starts the correlation model of the airflow velocity field and the thermal conduction parameters of the building structure. The bidirectional coupling model first reads the temperature distribution data from the previous time step, and the correlation model reads the airflow field monitoring data in parallel. The two models exchange data through MPI non-blocking communication.

[0072] The vortex coefficient matrix output by the associated model after calculation is passed to the boundary condition processing module of the two-way coupled model through a spatial mapping relationship. This matrix contains the turbulence intensity parameters of each grid cell, and the mapping process uses a conservation interpolation method to maintain the flux balance of physical quantities. In the third time slice (simulation time 0.3 seconds), the temperature field monitoring module detected a drastic change in the fire source area: the rate of change of the temperature gradient of the grid cells located in the charging pile area suddenly increased, and the system automatically triggered the time step adjustment mechanism. The original time step of 0.1 seconds was shortened to 0.05 seconds, and a 0.02-second auxiliary calculation slice was inserted after the current slice. The auxiliary slice is specifically designed to handle the local flow details near the fire source, and the smoke transport process is recalculated using a refined grid (0.2-meter resolution).

[0073] When the simulation reached the 12th time slice (1.2 seconds), the smoke diffusion path bifurcated in the parking lot pillar area. The topology analysis algorithm detected the tearing of the concentration isosurface and immediately inserted an auxiliary slice into the regular calculation sequence. This slice focused on calculating the flow effect around the pillar, using large eddy simulation to analyze the vortex shedding process. The calculation results were fed back to the main simulation loop to correct the global flow field. The heat flux intensity correction mechanism continued to play a role in the simulation. The two-way coupled model output a heat flux intensity distribution map at each time step, and this data was passed to the grid management system of the associated model through interpolation. When the heat flux intensity in a certain area was detected to exceed the threshold, the associated model automatically adaptively refined the grid in that area, increasing the grid resolution to 0.3 meters at the ceiling above the fire source, accurately capturing the ceiling jet phenomenon.

[0074] The time-slice advancement control module employs a dynamic monitoring strategy, continuously tracking changes in five physical quantities: maximum temperature gradient, smoke concentration variance, peak airflow vorticity, extreme heat flux intensity, and pressure fluctuation amplitude. When any two of these parameters simultaneously exceed a threshold, the system automatically halves the time step. At 45 seconds into the simulation, the sprinkler system activated, causing a sudden temperature drop; the time step was then adjusted from 0.1 seconds to 0.2 seconds to improve computational efficiency. The data synchronization mechanism faces practical challenges: the temperature sensors in the parking lot sample at 10Hz, the smoke detectors at 5Hz, and the anemometers at 20Hz. The system coordinates multi-source data using a timestamp alignment algorithm, reconstructing the time series from low-frequency data using cubic spline interpolation, and downsampling high-frequency data. When a sensor's data delay exceeds 200 milliseconds, the system activates a prediction model based on physical laws for data compensation.

[0075] The anomaly handling module recorded a network interruption event: communication with the sensor group located in the northwest corner of the parking lot was interrupted for 1.2 seconds. The system immediately switched to backup data processing mode, using data assimilation technology combined with readings from nearby sensors for spatial interpolation. After communication was restored, data resynchronization was automatically performed, filling in the missing time period data through model extrapolation. Throughout the simulation process, the time synchronization status of each physical field was continuously recorded. The generated timestamp deviation statistics table showed that the average deviation between the temperature field and the smoke field was controlled within 80 milliseconds, meeting the simulation accuracy requirements. In the simulation results visualization stage, the system successfully reproduced key phenomena in fire development: the stratification phenomenon formed by the spread of hot smoke along the ceiling, the influence of vortices generated behind the columns on smoke transport, and the interaction between the mechanical smoke exhaust system and the buoyancy drive of the fire source. The time slice control log showed that 17 time step adjustments were triggered during the entire 120-second simulation, and 9 auxiliary calculation slices were inserted, effectively capturing the abrupt changes in fire development.

[0076] Example 4: Generation of 3D Dynamic Simulation Maps. This transforms numerical data generated from multiphysics coupled simulations into intuitive visualization output. Taking a standard office floor fire scenario as an example, the floor is 30 meters long, 20 meters wide, and 3 meters high, containing typical structures such as office partitions, meeting rooms, and corridors. The fire source is set as the burning of a wastepaper basket in the corner of the meeting room. When transforming the temperature field evolution trajectory into an isothermal surface topology, three characteristic isothermal surfaces of 200℃, 400℃, and 600℃ are first extracted from the simulation data. The moving cube algorithm traverses the entire 3D scalar field. This algorithm calculates the isotropic points on the edges of each grid cell through linear interpolation, connecting these points to form a triangular patch network. For areas with insufficient resolution, adaptive subdivision technology is used to increase details. The generated isothermal surface mesh includes vertex coordinates, normal vectors, and temperature attributes. Smoke concentration chromaticity values ​​are embedded at the nodes of the topological structure. The smoke concentration data comes from the output of the bidirectional coupled model. The concentration values ​​are mapped to each vertex of the isothermal surface mesh using a nearest neighbor interpolation algorithm. The chromaticity encoding adopts the HSV color space, where hue represents the concentration level, saturation is related to the temperature value, and brightness is fixed at 80% to ensure visibility. A vortex vector arrow distribution layer is generated based on the airflow disturbance range. The rotation center is obtained by calculating the vorticity from the airflow velocity field data. The arrow starting point is placed at the vortex core position, the arrow direction is determined by the local velocity direction, and the length is proportional to the vorticity magnitude. The color encoding uses a red-blue gradient to represent the velocity direction.

[0077] The construction process of the isothermal surface topology extracts discrete time-domain temperature field extrema as topological keyframes. The simulation duration is 120 seconds, with keyframes extracted at 5-second intervals, totaling 24 key time points. Each keyframe uses a region growing algorithm to identify temperature extrema regions, recording their centroid coordinates, volume, and highest temperature value. Transition surfaces between keyframes are generated using cubic spline interpolation. Control points are taken from feature points on the isothermal surface boundary, and the tension parameter is set to 0.7 to ensure surface smoothness. The interpolation step size is 0.1 seconds, generating a total of 1200 transition frames. Thermodynamic parameter-driven mesh subdivision is performed on the transition surfaces. The subdivision criteria are based on the local temperature gradient and heat flux intensity, with higher gradient regions receiving higher subdivision levels. The maximum subdivision is 4 times, and the average side length of the subdivided mesh does not exceed 0.1 meters. The visualization pipeline uses GPU-based real-time rendering technology. Isothermal surface data undergoes frustum clipping and geometric transformation through vertex shaders, and fragment shaders implement Phong lighting model calculations. The smoke concentration visualization employs volume rendering technology, transforming the concentration field into a semi-transparent medium, and generating the smoke effect through a ray casting algorithm. Airflow vector arrows are rendered using instantiation; each arrow, acting as a Billboard sprite, always faces the camera, dynamically updating its position and direction parameters. The time-based animation system achieves smooth transitions through keyframe interpolation, with interpolation weights changing linearly over time, supporting adjustable playback speeds from 0.5x to 2x.

[0078] The interactive features allow users to switch display modes, viewing temperature, smoke, or airflow fields individually, or overlaying multiple fields. View control offers first-person panning and a global overview mode, supporting zoom, rotation, and panning. The data output module generates sequence frame images for creating animated videos and exports STL-formatted 3D meshes for 3D printing. The color mapping scheme follows fire visualization standards, using a red-yellow-blue gradient for the temperature field (from high to low temperature), grayscale to represent smoke concentration, and red-green encoding for airflow direction (inflow-outflow). See Table 1 for the isothermal surface topology keyframe data.

[0079] Table 1: Isothermal Surface Topology Keyframe Data

[0080]

[0081] The scene lighting setup takes into account the characteristics of a fire environment, employing a multi-light source system to simulate fire effects. The main light source is located at the center of the fire source to simulate flame illumination, while auxiliary light sources simulate ambient lighting. Shadow rendering utilizes variance shadow mapping technology, supporting shadow calculations for semi-transparent objects. In post-processing, glow effects are added to simulate flame glare, motion blur represents airflow dynamics, and depth-of-field effects enhance the immersive 3D experience.

[0082] For performance optimization, layered detail technology is employed to dynamically adjust mesh complexity based on viewpoint distance, while a simplified model is used for long-distance displays. Memory management utilizes virtual texture technology to stream large-scale data, automatically downgrading to CPU-based software rendering when video memory is insufficient. The compatibility design supports various display devices, from desktop workstations to mobile VR headsets, with output resolution automatically adjusted according to device capabilities. The user interface provides a simulation parameter adjustment panel, allowing real-time modification of isothermal surface thresholds, color mapping schemes, and display effects. A history function saves user operation paths, supporting retrospective comparison of fire scene states at different times. The help system includes built-in visual legends to guide users in correctly interpreting the meaning of various visual elements in the 3D atlas.

[0083] Example 5: The data verification mechanism is embedded in each time step of the multiphysics coupled simulation system. Taking a fire simulation case of a large shopping mall as an example, the building has three floors, an atrium structure, and multiple fire compartments. Data sources include temperature sensors, smoke detectors, and anemometers placed on the ceiling, walls, and vents. Simultaneously, the building information model provides structural thermal parameters. At the start of the simulation, the system loads the initial dataset, and the timestamps are uniformly calibrated to the simulation reference time of zero seconds. Before each simulation progress, the timestamps of the multiphysics data are compared. The timestamp data comes from the hardware clocks of each sensor, which are synchronized to millisecond precision via a network time protocol. The system maintains a global timeline, and each physics data packet is accompanied by a generated timestamp. The comparison process employs an optimistic concurrency control strategy, first checking the order of the timestamps, and then calculating the relative deviation. When the time deviation between the temperature field and the smoke diffusion field exceeds the tolerance limit, data resynchronization is triggered. The tolerance value is set to 100 milliseconds according to the simulation accuracy requirements. The deviation detection algorithm traverses the time scale of all data points to find the maximum positive and negative offsets. During the resynchronization process, the current calculation thread is paused, and the data interpolation program is started to extrapolate the data of the lagging field forward, while the data of the leading field is truncated or cached. A version number verification identifier is set for the heat conduction parameters of the building structure. The version number is generated using a hash value based on SHA-256. The hash is recalculated after each parameter modification. During loading, the current hash is checked to see if it matches the stored master version. If they do not match, the parameter rollback mechanism is triggered.

[0084] The timestamp comparison operation occurs at the beginning of the simulation cycle. The system reads the latest data blocks of temperature, smoke, and airflow fields from shared memory, parses the header information of each data block to obtain the acquisition time, and uses Unix timestamps in milliseconds. The comparison algorithm calculates the absolute value of the difference between adjacent physical field timestamps, while simultaneously checking the continuity of the time series to prevent data packet loss. The tolerance threshold is dynamically adjusted according to the simulation scenario. For stages of rapid fire development, the tolerance is tightened to 50 milliseconds, and relaxed to 200 milliseconds during stable stages. The threshold adjustment strategy is automatically optimized based on historical deviation statistics. The resynchronization process includes data alignment and consistency repair. Temperature field data is usually acquired at a higher frequency. When smoke field data lags, cubic spline interpolation is used to resample the temperature data to the smoke field time point. When there is a reverse lag, linear extrapolation is used to compensate for missing data. Correction amounts are recorded during the synchronization process for error analysis.

[0085] Version number verification is implemented for building structure heat conduction parameter files. These files are stored in JSON format and include attributes such as material type, thickness, and thermal conductivity. The version number is stored in the metadata segment of the file header. During verification, the hash value of the current file content is calculated and compared with the baseline hash registered in the simulation configuration. When a version change occurs, a difference report is generated listing the modified parameter items. Verification failures are handled with tiered responses: minor inconsistencies allow simulation to continue but are marked with a warning log; major changes interrupt the simulation and prompt the user to confirm parameter validity. The system retains the five most recent valid versions for emergency rollback. Data integrity checks are implemented using cyclic redundancy check (CRC) codes. Each data packet is appended with a 32-bit CRC checksum. Data corrupted during transmission or storage is discarded, triggering a re-acquisition request. The timestamp synchronization mechanism considers network latency; fluctuations in the arrival time of sensor data at the simulation server are smoothed using a sliding window averaging filter, and outliers are removed using a median filter. Version management is integrated into the parameter editing interface; any modification to heat conduction parameters automatically triggers a version number update, and a change comment function records the reason for the modification.

[0086] The error handling module is designed with a multi-layered structure. When a timestamp deviation exceeds the limit for the first time, it attempts automatic resynchronization. After three consecutive failures, it escalates to a manual intervention event, and the system sends an alert to the administrator. For parameter version conflicts, a merging tool is provided to assist in resolving the conflict, supporting partial rollback or selective application of parameter values. All verification events are recorded in the audit log, including timestamps, deviation amounts, processing actions, and result status. The log file is cyclically overwritten and saves records from the most recent 30 days. Verification points are set at the starting barrier of each time step before the simulation process progresses. All computation threads synchronize at this point. If the verification passes, the barrier is released and the simulation continues; if the verification fails, the thread waits until the problem is resolved. Adaptive adjustment of tolerance values ​​is based on a predictive model, analyzing early deviation trends to predict subsequent allowable ranges. The model training uses gradient descent to optimize threshold parameters. The interpolation method for the resynchronization algorithm considers data characteristics. For temperature field data, piecewise cubic Hermitian interpolation is used to maintain monotonicity; for smoke concentration data, conservative linear interpolation is used to avoid negative values. The building structure parameter version control system supports branch management, allowing experimental parameter modifications to be tested on independent branches and merged into the main branch after maturation. The version comparison feature visualizes the impact of parameter changes on simulation results, helping users assess the necessity of modifications. Hash calculation optimization employs an incremental update method, recalculating hashes only for modified parts to improve efficiency.

[0087] During the system integration testing phase, historical fire data is used to verify the reliability of the verification mechanism, simulating various abnormal scenarios such as sensor clock drift, network interruption, and incorrect parameter modification to ensure the robustness of the mechanism under various fault conditions. During implementation and deployment, the verification module runs as an independent service, interacting with the simulation core via remote procedure calls and supporting hot-swappable updates to the verification strategy. The entire data verification mechanism is deeply integrated with the simulation process, and the overhead of verification operations is optimized through asynchronous processing and caching techniques to ensure that it does not affect the real-time performance of the main simulation thread. The user interface provides real-time monitoring of the verification status, graphically displaying the time synchronization status and parameter version health of each physical field, with color coding intuitively indicating the anomaly level. The simulation results output includes a verification report appendix, detailing the verification results and intervention measures before each step, providing a transparent record for result credibility assessment.

[0088] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0089] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dynamic simulation method for fire fighting based on multiphysics coupling, characterized in that, Includes the following steps: Collect temperature field distribution data, smoke concentration gradient data, airflow velocity field data, and building structure heat conduction parameters at the fire scene; A two-way coupled model of temperature field and smoke diffusion field is established, and the dynamic correction of the smoke convection effect by temperature gradient is realized through real-time data exchange. A correlation model between the airflow velocity field and the heat conduction parameters of the building structure is constructed, and the vortex coefficient of the velocity field is adjusted according to the heat flux intensity. A multiphysics joint computation framework is formed by integrating bidirectional coupling models and correlation models, and the simulation process is advanced by using discrete time slicing. The output includes a three-dimensional dynamic simulation map of temperature field evolution trajectory, smoke diffusion path, and airflow disturbance range; When constructing the correlation model between the airflow velocity field and the building structure heat conduction parameters: Identify the distribution of material thermal resistivity in the thermal conduction parameters of building structures; The velocity field computational domain is divided into multiple meshes based on the distribution of thermal resistivity. A linear proportional relationship between heat flux intensity and vortex coefficient is established within each grid layer; The process of dividing the multi-layer mesh includes: Mesh layering identifiers are generated based on abrupt changes in thermal resistivity. An adaptive octree partitioning is used for the continuous thermal resistivity region; A heat flux intensity buffer transition zone is set at the mesh interface.

2. The fire dynamic simulation method based on multiphysics coupling according to claim 1, characterized in that, The specific process of establishing the two-way coupling model of the temperature field and the smoke diffusion field includes: Extract the boundary coordinates of high-temperature regions from the temperature field distribution data; Map the boundary coordinates of the high-temperature region to the spatial grid of the smoke concentration gradient data; Calculate the migration acceleration of smoke particles under the influence of temperature gradient based on the mapping results; An iterative feedback mechanism is used to input the migration acceleration in reverse to the transmission equation of the temperature field distribution data.

3. The fire dynamic simulation method based on multiphysics coupling according to claim 2, characterized in that, The following operations are performed simultaneously when calculating the migration acceleration of smoke particles under the influence of a temperature gradient: Detect the displacement of the boundary coordinates of the high-temperature region between adjacent time slices; The sampling frequency of the smoke concentration gradient data is dynamically adjusted based on the displacement. The adjusted sampling frequency is used as the time step constraint for the transmission equation.

4. The fire dynamic simulation method based on multiphysics coupling according to claim 1, characterized in that, The operational logic of the multiphysics joint computation framework includes: The bidirectional coupling model and the associated model are invoked synchronously at the start of each time slice. The vortex coefficients output by the associated model are used as the initial boundary values ​​of the two-way coupled model. The mesh parameters of the correlation model are corrected by the heat flux intensity generated by the two-way coupled model.

5. The fire dynamic simulation method based on multiphysics coupling according to claim 4, characterized in that, The time slice advancement control method is as follows: Monitor the maximum gradient change rate of the temperature field within each slice. When the change rate exceeds the critical threshold, shorten the time step of the next slice. Insert an auxiliary calculation slice when the smoke diffusion path bifurcates.

6. The fire dynamic simulation method based on multiphysics coupling according to claim 1, characterized in that, The method for generating the three-dimensional dynamic simulation map includes: Transform the temperature field evolution trajectory into an isothermal surface topology; Embed smoke concentration chromaticity values ​​at the topology nodes; The distribution layer of vortex vector arrows is generated based on the range of airflow disturbance.

7. The fire dynamic simulation method based on multiphysics coupling according to claim 6, characterized in that, The process of constructing the isothermal surface topology includes: Extract the extreme points of the temperature field at discrete time points as topological keyframes; Transition surfaces between keyframes are generated using cubic spline interpolation; Perform thermodynamic parameter-driven mesh subdivision on the transition surface.

8. The fire dynamic simulation method based on multiphysics coupling according to claim 1, characterized in that, It also includes a data verification mechanism: Compare the timestamps of the multiphysics data before each simulation process proceeds; Data resynchronization is triggered when the time deviation between the temperature field and the smoke diffusion field exceeds the tolerance. Set a version number verification identifier for the thermal conduction parameters of the building structure.

Images