Building fire perception-recognition-inference integrated method and system

Abstract

The present application relates to the technical field of building fire safety, and particularly relates to a building fire sensing-identification-inference integrated method and system, which solves the technical problem that the fire detection system in the prior art is difficult to determine the position of the fire source and the release rate of the fire, and is difficult to support fire situation analysis and rescue decision-making. The method provided by the present application inverses the fire source position and the heat release rate of the fire source from sparse temperature points by using an STGCN model, and then couples the fire source position and the heat release rate to a fire situation inference model based on GAN as conditions to output a temperature field distribution and a smoke visibility distribution, so that real-time dynamic reconstruction from "point data" to "full field cloud map" is realized. In addition, a spatial adjacency matrix between multiple temperature sensors is adopted, and the accuracy of the predicted fire source position and heat release rate can still be maintained under the condition that the temperature sensors are partially failed by relying on a neighborhood feature aggregation mechanism and reasoning through the spatio-temporal correlation of effective nodes.

Description

Technical Field

[0001] This invention relates to the field of building fire safety technology, and more specifically, to an integrated method and system for building fire perception, identification, and deduction. Background Technology

[0002] Building fires are one of the major hazards threatening lives and property, especially in complex buildings such as hospitals, shopping malls, and residences. Due to their enclosed spaces, limited ventilation, and dense populations, fires can easily cause serious casualties and property damage. Effective fire prediction can provide early warnings and crucial information for emergency responders and trapped individuals.

[0003] Traditional fire detection systems, such as smoke temperature sensors and temperature sensors, can only trigger alarms based on thresholds. Furthermore, temperature sensors have a limited detection range and cannot provide the spatial location and intensity of the fire source. As a result, fire detection systems can only determine that a fire has occurred in a building, but cannot determine the location of the fire source or the rate at which the fire is released, making it difficult to support fire situation analysis and rescue decisions. Summary of the Invention

[0004] The purpose of this invention is to provide an integrated method and system for building fire perception, identification and simulation, which solves the technical problems in the prior art where fire detection systems are unable to determine the location of the fire source and the rate of fire release, and are unable to support fire situation analysis and rescue decision-making.

[0005] As a first aspect of the present invention, the present invention provides an integrated method for building fire perception-identification-inference with topology adaptation and disaster feedback functions, comprising: A fire situation analysis model is constructed, which includes a fire source identification model constructed from a spatiotemporal graph convolutional network and a fire situation inference model constructed from a multi-scale conditional generative adversarial network. Acquire real-time temperature data detected by multiple temperature sensors inside the building space, the real-time temperature data including the real-time temperature detected by each temperature sensor at the detection time; A spatial adjacency matrix is ​​constructed between multiple temperature sensors based on their spatial locations within the building space. The real-time temperature data and the spatial adjacency matrix between multiple temperature sensors are input into the fire source identification model for identification, so as to generate the real-time fire source location and real-time heat release rate of the fire source in the building space. The real-time fire source location and the real-time heat release rate are input into the fire situation simulation model for calculation to generate the real-time environmental parameter distribution within the building space. The real-time environmental parameter distribution includes: real-time temperature field distribution and real-time smoke visibility distribution.

[0006] In one embodiment of the present invention, the method further includes: The spatial adjacency matrix is ​​updated based on the real-time temperature field distribution and the real-time temperatures detected by multiple temperature sensors.

[0007] In one embodiment of the present invention, a spatial adjacency matrix between multiple temperature sensors is constructed based on their spatial locations within the building space, including: Acquire the spatial locations of multiple temperature sensors within the building space; Based on the A-star search algorithm, the shortest path between any two temperature sensors is calculated according to the wall structure information of the building space and the spatial location of multiple temperature sensors in the building space. Calculate the spatial adjacency weight of two adjacent temperature sensors based on the shortest path between any two temperature sensors; A spatial adjacency matrix of multiple temperature sensors is constructed based on the spatial adjacency weights of two adjacent temperature sensors.

[0008] In one embodiment of the present invention, updating the spatial adjacency matrix based on the real-time temperature field distribution and the real-time temperatures detected by multiple temperature sensors includes: Based on the real-time temperature field distribution and the real-time temperatures detected by multiple temperature sensors, the prediction error of each temperature sensor is calculated. Calculate the real-time temperature gradient between two adjacent temperature sensors based on the real-time temperature field distribution. The real-time average temperature within the building space is calculated based on the real-time temperatures detected by multiple temperature sensors. The spatial adjacency weight between two adjacent temperature sensors is updated based on the real-time temperature gradient between the two adjacent temperature sensors, the reference temperature gradient, the prediction error between the two adjacent temperature sensors, and the real-time average temperature within the building space. The spatial adjacency matrix is ​​updated based on the updated spatial adjacency matrix between the two temperature sensors.

[0009] In one embodiment of the present invention, a fire situation analysis model is constructed, including: A training database is constructed based on fire dynamics simulation software. The training database includes multiple training data. Each training data includes the building space structure, the spatial location of temperature sensors in the building space structure, the historical temperature data detected by each temperature sensor, the simulated fire source location, and the simulated heat release rate of the fire source. A spatial adjacency matrix is ​​constructed between multiple temperature sensors based on their spatial locations within the building space. A fire source identification model is constructed using a spatiotemporal graph convolutional network architecture, and the fire source identification model is trained based on the spatial adjacency matrix and multiple training data to generate the completed fire source identification model. A fire situation simulation model is constructed based on a multi-scale conditional generative adversarial network, and the fire situation simulation model is trained based on multiple training data to generate the completed fire situation simulation model.

[0010] In one embodiment of the present invention, the fire source identification model includes a first temporal convolutional layer, a spatial graph convolutional layer, and a second temporal convolutional layer; The process of training a fire source identification model based on the spatial adjacency matrix and multiple training data includes: Based on the first time convolutional layer, extract the time-dependent features of historical temperature data in the training data; The spatial graph convolutional layer is used to determine the spatial correlation between multiple temperature sensors based on the spatial adjacency matrix; Based on the second temporal convolutional layer, spatiotemporal feature fusion and normalization are performed according to the temporal dependence features in the temperature data and the spatial correlation between multiple temperature sensors.

[0011] In one embodiment of the present invention, the fire source identification model is constructed using a spatiotemporal graph convolutional network with thermal energy conservation, temperature gradient, and mean square error as loss functions.

[0012] As a second aspect of the present invention, the present invention also provides an integrated building fire perception-identification-deduction system with topology adaptation and disaster feedback functions, comprising: The data acquisition module is used to acquire real-time temperature data detected by multiple temperature sensors inside the building space; A topology building module is used to construct a spatial adjacency matrix between multiple temperature sensors based on their spatial locations within the building space. A fire source identification model is used to calculate the real-time fire source location and real-time heat release rate within the building space based on the real-time temperature data and the spatial adjacency matrix between multiple temperature sensors. The fire source identification model is constructed using a spatiotemporal graph convolutional network. The fire situation simulation model is used to calculate the real-time fire source location and the real-time heat release rate to generate the real-time environmental parameter distribution within the building space. The real-time environmental parameter distribution includes: real-time temperature field distribution and real-time smoke visibility distribution. The fire situation simulation model is constructed using a multi-scale conditional generative adversarial network.

[0013] In one embodiment of the present invention, the system further includes: An update module is used to update the spatial adjacency matrix based on the real-time temperature field distribution and the real-time temperatures detected by multiple temperature sensors.

[0014] In one embodiment of the present invention, the system further includes: The early warning module is used to determine fire alarm information based on the real-time location of the fire source, the real-time heat release rate, and the real-time distribution of environmental parameters within the building space.

[0015] This invention provides an integrated method for building fire perception, identification, and inference with topology adaptation and disaster feedback functions. First, it uses an STGCN model to infer the fire source location and heat release rate from sparse temperature points. Then, the fire source location and heat release rate are coupled as conditional inputs to a GAN-based fire situation inference model to output temperature field distribution and smoke visibility distribution, achieving real-time dynamic reconstruction from "point data" to "full-field cloud map (temperature / smoke)". Furthermore, when the fire source identification model identifies the fire source, it employs a spatial adjacency matrix between multiple temperature sensors. Relying on a neighborhood feature aggregation mechanism, it performs inference through the spatiotemporal correlation of effective nodes. Even when some temperature sensors fail, it can maintain the accuracy of the predicted fire source location and the stability of the fire source's heat release rate. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 The diagram shown is a flowchart of an integrated building fire perception-identification-inference method with topology adaptation and disaster feedback functions provided by an embodiment of the present invention.

[0018] Figure 2 The figure shown is a comparison between the fire source location and heat release rate output by the fire source identification model when all temperature sensors are normal, according to an embodiment of the present invention, and the actual fire source location and actual heat release rate.

[0019] Figure 3 The figure shown is a result of an embodiment of the present invention, which shows the effect of the number of failed temperature sensors on the fire source location and heat release rate output by the fire source identification model when some temperature sensors fail.

[0020] Figure 4The figure shown is a result of how the duration of temperature data from the temperature sensor used in an embodiment of the present invention affects the fire source location and heat release rate output by the fire source identification model.

[0021] Figure 5 The figure shown is a comparison between the temperature field generated by the fire simulation model provided in an embodiment of the present invention and the actual temperature field.

[0022] Figure 6 The image shown is a comparison between the smoke visibility field generated by the fire simulation model provided in an embodiment of the present invention and the actual smoke visibility field.

[0023] Figure 7 The diagram shown is a flowchart of an integrated building fire perception-identification-inference method with topology adaptation and disaster feedback functions provided by another embodiment of the present invention.

[0024] Figure 8 The diagram shown is a flowchart of an integrated building fire perception-identification-inference method with topology adaptation and disaster feedback functions provided by another embodiment of the present invention.

[0025] Figure 9 The diagram shown is a working block diagram of an integrated building fire perception-identification-inference system with topology adaptation and disaster feedback functions provided by an embodiment of the present invention.

[0026] Figure 10 The diagram shown is a working block diagram of an integrated building fire perception-identification-deduction system with topology adaptation and disaster feedback functions, provided by another embodiment of the present invention.

[0027] Figure 11 The diagram shown is a working block diagram of an integrated building fire perception-identification-deduction system with topology adaptation and disaster feedback functions, provided by another embodiment of the present invention.

[0028] Figure 12 The diagram shown is a block diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0029] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0030] In the description of this invention, it should be noted that the terms "upper", "lower", "front", "horizontal", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0031] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the term "installation" should be interpreted broadly. For example, it can refer to a fixed connection, a detachable connection, or an integral connection; it can refer to a mechanical connection or an electrical connection; it can refer to a direct connection or an indirect connection through an intermediate medium; it can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0032] Application Overview In the critical initial stage of a fire, rapidly and accurately determining the coordinates of the fire source and its heat release rate can significantly improve early warning quality, evacuation efficiency, and the scientific basis of firefighting commands. Therefore, with the development of artificial intelligence technology, data-driven fire source parameter inversion technology has become a research hotspot. Specifically: (1) Vision-based detection technology: using video surveillance combined with convolutional neural networks (CNN) to identify flame or smoke features. Although the response speed is fast, it is limited by the field of view of the camera and occlusion problems, and it is difficult to accurately quantify the heat release rate.

[0033] (2) Intelligent algorithms based on physical constraints: such as Bayesian inversion, artificial fish swarm algorithm, particle swarm optimization algorithm, etc., combined with physical models for fire source localization. However, they usually have high computational complexity.

[0034] (3) Deep learning model based on temperature sensor data: An end-to-end neural network model is constructed using multi-point temperature and smoke temperature sensor data for inversion. For example, a convolutional neural network (CNN) is used to process the spatial matrix of temperature sensor data, or a long short-term memory network (LSTM) or gated recurrent unit (GRU) is used to process the time series features of temperature sensor data, thereby establishing a nonlinear mapping relationship between temperature sensor readings and fire source location / HRR. However, existing CNN or RNN / LSTM methods usually treat the readings of multiple temperature sensors as independent input channels or simple grid matrices, ignoring the complex physical structure inside the building (such as wall partitions, door and window locations). In real fires, the spread of temperature is not a simple Euclidean distance diffusion, but is limited by wall obstruction and connection paths, making it difficult to accurately predict the location of heat sources and the heat release rate under complex building layouts. In addition, at fire scenes, temperature sensors are easily damaged by high temperatures, communication interruptions, or power supply failures. Once some temperature sensor data is missing, the feature extraction capability of the model will drop significantly, resulting in serious distortion of the prediction results.

[0035] (4) Most existing technologies stop at the “identification” stage, that is, only output fire source parameters and fail to use the identification results for continuous dynamic simulation of fire development. In other words, they cannot predict the future temperature field and smoke visibility field distribution, and it is difficult to support the full-process situation assessment in the digital twin system.

[0036] Therefore, this invention provides an integrated method for building fire perception, identification, and inference with topology adaptation and disaster feedback functions. First, the STGCN model is used to inversely deduce the fire source location and heat release rate from sparse temperature points. Then, the fire source location and heat release rate are coupled as conditional inputs to a GAN-based fire situation inference model to output the temperature field distribution and smoke visibility distribution, achieving real-time dynamic reconstruction from "point data" to "full-field cloud map (temperature / smoke)". Furthermore, when the fire source identification model identifies the fire source, a spatial adjacency matrix between multiple temperature sensors is used. Relying on a neighborhood feature aggregation mechanism, inference is performed through the spatiotemporal correlation of effective nodes. Even when some temperature sensors fail, the stability of the predicted fire source location and heat release rate can still be maintained.

[0037] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0038] Exemplary methods As a first aspect of the present invention, the present invention provides an integrated method for building fire perception, identification, and inference with topology adaptation and disaster feedback functions. Figure 1The diagram shown is a flowchart illustrating an integrated building fire perception-identification-inference method with topology adaptation and disaster feedback functions according to an embodiment of the present invention. Figure 1 As shown, an integrated method for building fire perception, identification, and inference with topology adaptation and disaster feedback functions includes the following steps: S1: Construct a fire situation analysis model, which includes a fire source identification model constructed from a spatiotemporal graph convolutional network and a fire situation inference model constructed from a multi-scale conditional generative adversarial network. Specifically, the input to the fire source identification model is the temperature detected by multiple temperature sensors and the spatial adjacency matrix of the multiple temperature sensors, and the output is the location of the fire source and the heat release rate of the fire source.

[0039] The inputs to the fire situation simulation model are the location of the fire source and the heat release rate of the fire source, and the outputs are the temperature field distribution and smoke visibility field distribution within the building space.

[0040] Specifically, smoke visibility refers to the farthest distance at which the human eye can identify a target object in a smoky environment.

[0041] Temperature field distribution refers to the overall state of temperature at every point within a building space at a given moment. It reflects the continuous changes in temperature within the building space.

[0042] Specifically, both the fire source identification model and the fire situation simulation model were pre-built.

[0043] S2: Acquire real-time temperature data detected by multiple temperature sensors inside the building space. The real-time temperature data includes the real-time temperature detected by each temperature sensor at the detection time. Specifically, in practical applications, the analysis model is deployed within a building digital twin platform. When monitoring building spaces, in the event of a fire, real-time temperature data from multiple temperature sensors within the building space can be acquired. This real-time temperature data includes the actual temperature detected by each temperature sensor at the time of the event.

[0044] S3: Construct a spatial adjacency matrix between multiple temperature sensors based on their spatial locations within the building space; Temperature sensors are installed at different locations within the building space. Therefore, a spatial adjacency matrix between multiple temperature sensors can be constructed based on their spatial locations within the building space.

[0045] S4: Input the real-time temperature data and the spatial adjacency matrix between multiple temperature sensors into the fire source identification model for identification, so as to generate the real-time fire source location and real-time heat release rate in the building space. The spatial adjacency matrix and the real-time temperature detected by each temperature sensor are input into the fire source identification model for identification, so as to generate the real-time fire source location and the real-time heat release rate of the fire source in the building space.

[0046] When identifying fire sources, the fire source identification model uses a spatial adjacency matrix between multiple temperature sensors. Relying on the neighborhood feature aggregation mechanism, it performs inference through the spatiotemporal correlation of effective nodes. Even when some temperature sensors fail, the predicted fire source location and the heat release rate of the fire source can still be kept stable.

[0047] S5: Input the real-time fire source location and real-time heat release rate into the fire situation simulation model for calculation to generate the real-time environmental parameter distribution within the building space. The real-time environmental parameter distribution includes: real-time temperature field distribution and real-time smoke visibility distribution.

[0048] The real-time fire source location and real-time heat release rate output by the fire source identification model are input into the fire situation simulation model for calculation, so as to obtain the real-time temperature field distribution and real-time smoke visibility distribution in the building space, so as to realize dynamic fire reconstruction and situation simulation.

[0049] This invention provides an integrated method for building fire perception, identification, and inference with topology adaptation and disaster feedback functions. First, it uses an STGCN model to infer the fire source location and heat release rate from sparse temperature points. Then, it couples the fire source location and heat release rate as conditional inputs to a GAN-based fire situation inference model to output temperature field distribution and smoke visibility distribution, achieving real-time dynamic reconstruction from "point data" to "full-field cloud map (temperature / smoke)". Furthermore, when the fire source identification model identifies the fire source, it employs a spatial adjacency matrix between multiple temperature sensors. Relying on a neighborhood feature aggregation mechanism, it performs inference through the spatiotemporal correlation of effective nodes, maintaining the stability of the predicted fire source location and heat release rate even when some temperature sensors fail.

[0050] In one embodiment of the present invention, the analysis model constructed in S1 may include a fire source identification model and a fire situation simulation model. The fire source identification model and the fire situation simulation model are pre-constructed. Specifically, the construction method of the fire source identification model and the fire situation simulation model, i.e., S1 (constructing the fire situation analysis model), specifically includes the following steps: S110: A training database is constructed based on fire dynamics simulation software. The training database includes multiple training data. Each training data includes the building space structure, the spatial location of the temperature sensor in the building space structure, the simulated temperature data corresponding to each temperature sensor, the simulated fire source location or the simulated heat release rate of the fire source. Specifically, the simulated fire source locations and simulated heat release rates within the building space are input into fire dynamics simulation software, such as the Fire Dynamics Simulator (FDS), for simulation calculations to obtain simulated temperature data corresponding to multiple temperature sensors within the building space. A corresponding set of training data includes: the building space structure, the spatial locations of the temperature sensors within the building space structure, the simulated temperature data for each temperature sensor, and the simulated fire source location. Alternatively, a training database may include: the building space structure, the spatial locations of the temperature sensors within the building space structure, the simulated temperature data for each temperature sensor, and the simulated heat release rate of the fire source. For example, the training database in this invention may include 192 training data points, covering 16 simulated fire source locations and 12 simulated heat release rates for different types of fire sources.

[0051] S120: Construct a spatial adjacency matrix between multiple temperature sensors based on their spatial locations within the building space; Since temperature sensors are installed at different locations within the building space, a spatial adjacency matrix between multiple temperature sensors can be constructed based on their different locations within the building space.

[0052] Specifically, the construction of the spatial adjacency matrix among multiple temperature sensors may include the following steps: (1) Based on the A-star search algorithm, calculate the shortest path between any two temperature sensors according to the wall structure information of the building space and the spatial location of multiple temperature sensors in the building space; Specifically, the locations of multiple temperature sensors within the building space can be determined based on the design drawings.

[0053] Once the locations of the different temperature sensors are determined, the shortest path between any two temperature sensors can be calculated based on the A* search algorithm, taking into account the wall structure information of the building space and the spatial locations of multiple temperature sensors within the building space.

[0054] (2) Calculate the spatial adjacency weight of two adjacent temperature sensors based on the shortest path between any two temperature sensors; (3) Construct a spatial adjacency matrix of multiple temperature sensors based on the spatial adjacency weights of two adjacent temperature sensors.

[0055] By constructing a spatial adjacency matrix of multiple temperature sensors within the building space, a corresponding spatial adjacency matrix can be constructed for different installations of the temperature sensors, thus explicitly representing the non-Euclidean spatial dependencies between nodes.

[0056] This invention utilizes the A* algorithm to calculate the "shortest accessible path" (rather than straight-line distance) between sensor nodes. This is used to construct a spatial adjacency matrix, explicitly encoding the physical constraints of building wall layout and heat propagation into the input structure of the neural network, significantly improving the model's ability to understand complex built environments.

[0057] S130: Construct a fire source identification model using Spatio-Temporal Graph Convolutional Networks (STGCN), and train the fire source identification model based on the spatial adjacency matrix and multiple training data to generate the completed fire source identification model; Specifically, the historical temperature data from each training dataset and the spatial adjacency matrix between the corresponding multiple temperature sensors are input into the fire source identification model for training, in order to construct the fire source identification model.

[0058] When training the fire source identification model, heat energy conservation, temperature gradient, and mean square error are used as loss functions. The simulated fire source location and simulated heat release rate of the fire source in each training data are used to optimize the fire source identification model through backpropagation.

[0059] Specifically, adding heat energy conservation and temperature gradient constraints to the loss function of the fire source identification model can achieve physical consistency and improve model generalization.

[0060] Optionally, the fire source identification model includes a first temporal convolutional layer, a spatial graph convolutional layer, and a second temporal convolutional layer; the corresponding construction of the fire source identification model, namely S130 (constructing the fire source identification model), specifically includes the following steps: S121: Based on the first time convolutional layer, extract the time-dependent features of historical temperature data in the training data; S122: Spatial graph convolutional layers are used to determine the spatial correlation between multiple temperature sensors based on the spatial adjacency matrix; Specifically, the spatial graph convolutional layer employs spectral domain convolution based on the Chebyshev polynomial approximation to approximate the graph Laplacian operator with low computational cost. This invention uses spectral domain graph convolution with the Chebyshev polynomial approximation to efficiently capture the spatial thermal diffusion characteristics between nodes while reducing computational complexity.

[0061] S123: Based on the second temporal convolutional layer, spatiotemporal feature fusion and normalization are performed according to the temporal dependence features in the temperature data and the spatial correlation between multiple temperature sensors.

[0062] The first and second temporal convolutional layers employ a causal temporal convolutional structure, using zero-padding at the left end of the time series to ensure that the output depends only on historical information. In the time dimension, causal convolution with left-side zero-padding ensures that the fire source identification model relies only on current and historical data, guaranteeing the physical compliance of real-time predictions.

[0063] S140: A fire situation simulation model is constructed based on a multi-scale conditional generative adversarial network, and the fire situation simulation model is trained based on multiple training data to generate the completed fire situation simulation model.

[0064] Specifically, the model architecture of the fire situation simulation model is: Multi-Scale conditional Generative Adversarial Network (MS cGAN).

[0065] During the training of the fire situation simulation model, the inputs are: the location distribution of multiple temperature sensors in the building space, the simulated location of the fire source, and the simulated heat release rate of the fire source. The outputs of the fire situation simulation model are: the temperature field distribution and the smoke visibility distribution. The fire situation simulation model achieves simultaneous modeling of temperature and smoke through a multi-scale coupled generator, and uses structural similarity loss at the discriminator end to ensure the continuity of thermal features.

[0066] Once the analysis model is built, the fire source identification model and fire simulation model within the analysis model are tested to obtain the test results, as follows: (1) When all temperature sensors in the building space are functioning normally: Multiple test data are input into the fire source identification model for identification, so as to generate multiple test fire source locations and multiple test heat release rates respectively.

[0067] A comparison graph showing the test fire source location (x, y) and the test heat release rate (HRR), and the simulated fire source location and simulated heat release rate in the test data, is shown below. Figure 2 As shown. From Figure 2 From this, we can derive the prediction determination coefficient R for both the prediction of the fire source location and the prediction of the heat release rate. 2 All values ​​are greater than 0.99. The mean square error (MSE) is less than 0.008. This indicates that the fire source identification model can accurately reconstruct the core parameters of a fire scene.

[0068] (2) When a failed temperature sensor is present among multiple temperature sensors in the building space: Test data from varying numbers of failed temperature sensors were input into the fire source identification model for identification, generating multiple test fire source locations and multiple test heat release rates. The effect of the number of failed temperature sensors on the fire source location and heat release rate output by the fire source identification model is shown in the figure below. Figure 3 As shown. From Figure 3 The results show that the proportion of failed sensors reaches approximately 1 / 6 (about 17%), yet the predictive performance of the fire source identification model remains stable. Even with this missing sensor ratio, the mean squared error (MSE) of the fire source identification model remains below 0.0153, and R0... 2 It remains above 0.80. This means that even if the fire source identification model constructed in this invention is burned or communication is interrupted, it still has a high prediction accuracy.

[0069] (3) The duration of the temperature data used in the test: The prediction results of the fire source identification model were analyzed when the temperature data collected by the temperature sensor used in the test data were of different collection durations. The analysis yielded a graph showing the impact of the duration of the temperature sensor data on the fire source location and heat release rate output by the fire source identification model. Figure 4 As shown, requiring only 10 seconds of temperature data after a fire starts, the fire source identification model can predict the location of the fire source using R0. 2 The accuracy rate can approach 90%; when the input time reaches 20 seconds, the prediction of the heat release rate (HRR) also tends to be accurate and stable. The fire source identification model of this invention can quickly locate the fire source within the "golden time" of fire development, without waiting for the fire to fully spread.

[0070] (4) When inputting test data into the fire simulation model: The image shows a comparison between the temperature field generated when the test data is input into the fire simulation model and the actual temperature field. The actual temperature field is the simulated temperature field obtained by inputting the temperature sensor data from the test data into the Fire Dynamics Simulator (FDS) for simulation calculation.

[0071] Figure 5 The figure shown is a comparison between the temperature field generated by the fire simulation model provided in an embodiment of the present invention and the actual temperature field. Figure 6 The image shown is a comparison between the smoke visibility field generated by the fire simulation model provided in an embodiment of the present invention and the actual smoke visibility field. From... Figure 5 as well as Figure 6The results show that the temperature field and smoke visibility field output by the fire simulation model are in high agreement with the simulation values ​​of the fire dynamics simulator. The structural similarity index (SSIM) is greater than 0.73, and the normalized root mean square error (NRMSE) is less than 0.03. This further verifies the effectiveness of the "perception-identification-inference" link, which can provide fire command with an intuitive dynamic situation map. That is, it can infer the future temperature field distribution and smoke visibility field distribution based on the current temperature data, so that the digital twin system can perform full-process situation analysis of the fire situation in the building environment.

[0072] In another embodiment of the invention, such as Figure 7 As shown, the integrated building fire perception-identification-inference method with topology adaptation and disaster feedback functions provided by the present invention further includes the following steps: S6: Update the spatial adjacency matrix based on the real-time temperature field distribution and the real-time temperature detected by multiple temperature sensors.

[0073] This invention uses the output results (temperature field distribution and smoke visibility distribution) of the fire situation simulation model to back-calculate and update the spatial adjacency matrix between multiple sensors, thereby achieving topological adaptive evolution and further improving the prediction accuracy of the analysis models (fire source identification model and fire situation simulation model).

[0074] Optional, such as Figure 8 As shown, S3 (constructing a spatial adjacency matrix between multiple temperature sensors based on their spatial locations within the building space) specifically includes the following steps: S31: Acquire the spatial locations of multiple temperature sensors within the building space; Specifically, the spatial locations of multiple temperature sensors within a building space can be obtained from the building space design information. The specific spatial location refers to their planar position within the building space; for example, the spatial coordinates of the temperature sensors are (x, y).

[0075] S32: Based on the A-star search algorithm, calculate the shortest path between any two temperature sensors according to the wall structure information of the building space and the spatial location of multiple temperature sensors in the building space. S33: Calculate the spatial adjacency weight of two adjacent temperature sensors based on the shortest path between any two temperature sensors; Specifically, the formula for calculating the spatial adjacency weight of two adjacent temperature sensors is as follows: Formula 1 In Formula 1, For the first i The sensor and the first j The spatial adjacency weights between sensors, Lij For the first i The sensor and the first j The shortest path between sensors. α This is the distance attenuation coefficient.

[0076] Formula 1 can be used to calculate the spatial adjacency weight between any two adjacent temperature sensors.

[0077] S34: Construct a spatial adjacency matrix for multiple temperature sensors based on the spatial adjacency weights of two adjacent temperature sensors.

[0078] Once the spatial adjacency weights between two adjacent temperature sensors are determined, a spatial adjacency matrix between multiple temperature sensors can be constructed.

[0079] Based on building plan topology information and the obstruction features of walls, doors and windows, this invention uses the A* algorithm to calculate the shortest passable path distance between any two sensor nodes; it constructs a spatial adjacency matrix that can be updated according to changes in environmental conditions to explicitly represent the non-Euclidean spatial dependencies between nodes.

[0080] Correspondingly, the specific update method for the spatial adjacency matrix, namely S6 (updating the spatial adjacency matrix based on the real-time temperature field distribution and the real-time temperature detected by multiple temperature sensors), includes the following steps: S61: Calculate the prediction error of each temperature sensor based on the real-time temperature field distribution and the real-time temperature detected by multiple temperature sensors. S62: Calculate the real-time temperature gradient between two adjacent temperature sensors based on the real-time temperature field distribution; S63: Calculate the real-time average temperature within the building space based on the real-time temperatures detected by multiple temperature sensors; S64: Update the spatial adjacency weight between two temperature sensors based on the real-time temperature gradient between two adjacent temperature sensors, the reference temperature gradient, the prediction error between two adjacent temperature sensors, and the real-time average temperature in the building space. Specifically, the updated formula for calculating spatial adjacency weights is as follows: Formula 2: Formula 2 In Formula 2, The updated spatial adjacency weights, The spatial adjacency weights before the update. γ For learning rate, The weights are the values ​​corresponding to the temperature gradient. Let be the prediction error of the i-th sensor. Let be the prediction error of the j-th temperature sensor, where the i-th and j-th temperature sensors are two adjacent temperature sensors. The weights are the values ​​corresponding to the prediction error. As a reference temperature gradient, This represents the real-time average temperature within the building space. For the first i The sensor and the first j Real-time temperature gradient between sensors.

[0081] S65: Update the spatial adjacency matrix based on the updated spatial adjacency matrix between the two temperature sensors.

[0082] Once the spatial adjacency weights between any two temperature sensors are updated, the spatial adjacency matrix between multiple temperature sensors can be updated.

[0083] Exemplary System As a second aspect of the present invention, the present invention provides an integrated system for building fire perception, identification, and simulation with topology adaptation and disaster feedback functions. Figure 9 The diagram shown is a working block diagram of an integrated building fire perception-identification-inference system with topology adaptation and disaster feedback functions provided in an embodiment of the present invention. Figure 9 As shown, a building fire perception-identification-deduction integrated system 10 with topology adaptation and disaster feedback functions includes: The data acquisition module 100 is used to acquire real-time temperature data detected by multiple temperature sensors inside the building space; Specifically, the data acquisition module 100 is communicatively connected to multiple temperature sensors and retrieves the temperature data detected by each temperature sensor. That is, the data acquisition module 100 executes S2 in the above-described integrated method for building fire perception-identification-inference with topology adaptation and disaster feedback functions.

[0084] Topology building module 200 is used to build a spatial adjacency matrix between multiple temperature sensors based on their spatial locations within the building space. Specifically, the topology building module 200 can connect to design software to obtain design information of the building space, including architectural design information and the spatial locations of multiple temperature sensors within the building space. That is, the topology building module 200 executes S3 in the aforementioned integrated building fire perception-identification-inference method with topology adaptation and disaster feedback functions.

[0085] The fire source identification model 300 is used to calculate the real-time fire source location and real-time heat release rate within the building space based on real-time temperature data and the spatial adjacency matrix between multiple temperature sensors. The fire source identification model is constructed using a spatiotemporal graph convolutional network.

[0086] The fire source identification model 300 is pre-built, and the specific construction method is as described above. The fire source identification model 300 is used to execute S4 in the above-described integrated method for building fire perception-identification-inference with topology adaptation and disaster feedback functions.

[0087] Fire situation simulation model 400 is used to calculate the real-time fire source location and real-time heat release rate to generate the real-time environmental parameter distribution within the building space. The real-time environmental parameter distribution includes: real-time temperature field distribution and real-time smoke visibility distribution. The fire situation simulation model is constructed using a multi-scale conditional generative adversarial network.

[0088] Specifically, the fire situation simulation model 400 is pre-built, and the specific construction method is as described above. The fire situation simulation model 400 is used to execute S5 in the above-described integrated building fire perception-identification-simulation method with topology adaptation and disaster feedback functions.

[0089] This invention provides an integrated building fire perception-identification-inference system with topology adaptation and disaster feedback functions. First, it uses the STGCN model to infer the fire source location and heat release rate from sparse temperature points. Then, it couples the fire source location and heat release rate as conditional inputs to a GAN-based fire situation inference model to output temperature field distribution and smoke visibility distribution, achieving real-time dynamic reconstruction from "point data" to "full-field cloud map (temperature / smoke)". Furthermore, when the fire source identification model identifies the fire source, it employs a spatial adjacency matrix between multiple temperature sensors. Relying on a neighborhood feature aggregation mechanism, it performs inference through the spatiotemporal correlation of effective nodes, maintaining the stability of the predicted fire source location and heat release rate even when some temperature sensors fail.

[0090] Optional, such as Figure 10 As shown, the integrated building fire perception-identification-deduction system 10 with topology adaptation and disaster feedback functions also includes: The update module 500 is used to update the spatial adjacency matrix based on the real-time temperature field distribution and the real-time temperature detected by multiple temperature sensors.

[0091] Specifically, the update module is used to update the spatial adjacency matrix.

[0092] Optionally, such as Figure 11As shown, the integrated building fire perception-identification-deduction system 10 with topology adaptation and disaster feedback functions also includes: The early warning module 600 is used to determine fire alarm information based on the real-time location of the fire source, the real-time heat release rate, and the real-time distribution of environmental parameters within the building space.

[0093] Specifically, the early warning module 600 interfaces with the building management system and outputs fire alarm information and evacuation information based on the fire situation results (such as fire source location, heat release rate, distribution field, and smoke visibility distribution). The fire alarm information and evacuation information are then transmitted to the building management system so that staff can evacuate people in a timely manner based on the fire alarm information and evacuation information to reduce losses.

[0094] Exemplary electronic devices As a fourth aspect of the present invention, the present invention also provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, it executes the above-described integrated method for building fire perception-identification-inference with topology adaptation and disaster feedback functions.

[0095] Specifically, the electronic device includes a processor, memory, network interface, and input device connected via a device bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores operating devices and computer programs. The internal memory provides an environment for the operation of the operating devices and computer programs in the non-volatile storage medium. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it follows the steps of an integrated building fire perception-identification-inference method with topology adaptation and disaster feedback functions according to various embodiments of this specification, as described in the above embodiments.

[0096] The processor may include the main processor, as well as baseband chips, modems, etc.

[0097] The memory stores a program that executes the technical solution of this invention, and may also store operating devices and other key business functions. Specifically, the program may include program code, which includes computer operation instructions. More specifically, the memory may include read-only memory (ROM), other types of static storage devices capable of storing static information and instructions, random access memory (RAM), other types of dynamic storage devices capable of storing information and instructions, disk storage, flash memory, etc.

[0098] The processor can be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits used to control the execution of the program of the present invention. It can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0099] Input devices may include devices that receive data and information input by the user, such as keyboards, mice, cameras, scanners, light pens, voice input devices, touch screens, pedometers, or gravity sensors.

[0100] Output devices may include devices that allow information to be output to the user, such as displays, printers, speakers, etc.

[0101] The communication interface may include any transceiver-like device for communicating with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0102] The processor executes the program stored in the memory and calls other devices, which can be used to implement any of the steps of the integrated building fire perception-identification-inference method with topology adaptation and disaster feedback functions provided in the above embodiments of this specification.

[0103] The electronic device may also include a display component and a voice component. The display component may be a liquid crystal display or an e-ink display. The input device of the controller may be a touch layer covering the display component, or a button, trackball, or touchpad set on the controller housing, or an external keyboard, touchpad, or mouse, etc.

[0104] Those skilled in the art will understand that Figure 12 The structures shown are merely block diagrams of a portion of the structure related to the scheme described in this specification, and do not constitute a limitation on the electronic devices to which the scheme described in this specification is applied. Specific electronic devices may include more or fewer components than those shown in the figures, or may combine certain components, or may have different component arrangements.

[0105] Exemplary computer program products and storage media In addition to the methods and devices described above, the ion implantation control method provided in the embodiments of this specification can also be a computer program product, which includes computer program instructions. When the computer program instructions are run by a processor, the processor performs the steps of an integrated building fire perception-identification-inference method with topology adaptation and disaster feedback function according to various embodiments of this specification, as described in the "Exemplary Methods" section above.

[0106] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0107] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this specification. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages.

[0108] Furthermore, embodiments of this specification also provide a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor of the steps in an integrated building fire perception-identification-inference method with topology adaptation and disaster feedback functions according to various embodiments of this specification, as described in the "Exemplary Methods" section above.

[0109] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this specification can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0110] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0111] The embodiments described above are merely illustrative of several implementation methods outlined in this specification. While the descriptions are specific and detailed, they should not be construed as limiting the scope of the solutions provided in this specification. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this specification, and these all fall within the scope of protection of this specification. Therefore, the scope of protection for this patent should be determined by the appended claims.

Claims

1. A building fire perception-identification-inference integrated method with topology adaptation and disaster feedback functions, characterized in that, include: A fire situation analysis model is constructed, which includes a fire source identification model constructed from a spatiotemporal graph convolutional network and a fire situation inference model constructed from a multi-scale conditional generative adversarial network. Acquire real-time temperature data detected by multiple temperature sensors inside the building space, the real-time temperature data including the real-time temperature detected by each temperature sensor at the detection time; A spatial adjacency matrix is ​​constructed between multiple temperature sensors based on their spatial locations within the building space. The real-time temperature data and the spatial adjacency matrix between multiple temperature sensors are input into the fire source identification model for identification, so as to generate the real-time fire source location and real-time heat release rate of the fire source in the building space. The real-time fire source location and the real-time heat release rate are input into the fire situation simulation model for calculation to generate the real-time environmental parameter distribution within the building space. The real-time environmental parameter distribution includes: real-time temperature field distribution and real-time smoke visibility distribution.

2. The method according to claim 1, characterized in that, Also includes: The spatial adjacency matrix is ​​updated based on the real-time temperature field distribution and the real-time temperatures detected by multiple temperature sensors.

3. The method according to claim 2, characterized in that, A spatial adjacency matrix is ​​constructed among multiple temperature sensors based on their spatial locations within the building space, including: Acquire the spatial locations of multiple temperature sensors within the building space; Based on the A-star search algorithm, the shortest path between any two temperature sensors is calculated according to the wall structure information of the building space and the spatial location of multiple temperature sensors in the building space. Calculate the spatial adjacency weight of two adjacent temperature sensors based on the shortest path between any two temperature sensors; A spatial adjacency matrix of multiple temperature sensors is constructed based on the spatial adjacency weights of two adjacent temperature sensors.

4. The method according to claim 3, characterized in that, The spatial adjacency matrix is ​​updated based on the real-time temperature field distribution and the real-time temperatures detected by multiple temperature sensors, including: Based on the real-time temperature field distribution and the real-time temperatures detected by multiple temperature sensors, the prediction error of each temperature sensor is calculated. Calculate the real-time temperature gradient between two adjacent temperature sensors based on the real-time temperature field distribution. The real-time average temperature within the building space is calculated based on the real-time temperatures detected by multiple temperature sensors. The spatial adjacency weight between two adjacent temperature sensors is updated based on the real-time temperature gradient between two adjacent temperature sensors, the reference temperature gradient, the prediction error corresponding to each of the two adjacent temperature sensors, and the real-time average temperature in the building space. The spatial adjacency matrix is ​​updated based on the updated spatial adjacency matrix between the two temperature sensors.

5. The method according to claim 1, characterized in that, The fire situation analysis model includes: A training database is constructed based on fire dynamics simulation software. The training database includes multiple training data. Each training data includes the building space structure, the spatial location of the temperature sensor in the building space structure, the simulated temperature data corresponding to each temperature sensor, and the simulated fire source location or simulated heat release rate of the fire source. A spatial adjacency matrix is ​​constructed between multiple temperature sensors based on their spatial locations within the building space. A fire source identification model is constructed using a spatiotemporal graph convolutional network architecture, and the fire source identification model is trained based on the spatial adjacency matrix and multiple training data to generate the completed fire source identification model. A fire situation simulation model is constructed based on a multi-scale conditional generative adversarial network, and the fire situation simulation model is trained based on multiple training data to generate the completed fire situation simulation model.

6. The method according to claim 5, characterized in that, The fire source identification model includes a first temporal convolutional layer, a spatial graph convolutional layer, and a second temporal convolutional layer; The process of training a fire source identification model based on the spatial adjacency matrix and multiple training data includes: Based on the first time convolutional layer, extract the time-dependent features of historical temperature data in the training data; The spatial graph convolutional layer is used to determine the spatial correlation between multiple temperature sensors based on the spatial adjacency matrix; Based on the second temporal convolutional layer, spatiotemporal feature fusion and normalization are performed according to the temporal dependence features in the temperature data and the spatial correlation between multiple temperature sensors.

7. The method according to claim 5, characterized in that, The fire source identification model is constructed using a spatiotemporal graph convolutional network, with thermal energy conservation, temperature gradient, and mean square error as loss functions.

8. An integrated building fire perception-identification-deduction system with topology adaptation and disaster feedback functions, characterized in that, include: The data acquisition module is used to acquire real-time temperature data detected by multiple temperature sensors inside the building space; A topology building module is used to construct a spatial adjacency matrix between multiple temperature sensors based on their spatial locations within the building space. A fire source identification model is used to calculate the real-time fire source location and real-time heat release rate within the building space based on the real-time temperature data and the spatial adjacency matrix between multiple temperature sensors. The fire source identification model is constructed using a spatiotemporal graph convolutional network. The fire situation simulation model is used to calculate the real-time fire source location and the real-time heat release rate to generate the real-time environmental parameter distribution within the building space. The real-time environmental parameter distribution includes: real-time temperature field distribution and real-time smoke visibility distribution. The fire situation simulation model is constructed using a multi-scale conditional generative adversarial network.

9. The system according to claim 8, characterized in that, Also includes: An update module is used to update the spatial adjacency matrix based on the real-time temperature field distribution and the real-time temperatures detected by multiple temperature sensors.

10. The system according to claim 8, characterized in that, Also includes: The early warning module is used to determine fire alarm information based on the real-time location of the fire source, the real-time heat release rate, and the real-time distribution of environmental parameters within the building space.

Images