Explosion overpressure field physical enhancement reconstruction method and system oriented to sparse observation

Abstract

The invention discloses a sparse observation-oriented explosion overpressure field physical enhancement reconstruction method and system. A pre-trained and optimized explosion overpressure sparse reconstruction model is utilized to perform physical enhancement reconstruction on a field explosion overpressure field; wherein the explosion overpressure sparse reconstruction model comprises a sparse enhancement module, a feature identification module, a denoising module and an output module. The sparse reconstruction method disclosed by the invention is high in reconstruction precision, strong in generalization ability and high in efficiency, and can realize full-flow-field high-precision reconstruction of sparse overpressure data of explosion impact in a complex multi-dimensional and multi-scale scene.

Description

Technical Field

[0001] The invention belongs to the technical field of data processing, and in particular to a method and system for physically enhancing reconstruction of explosion overpressure fields oriented to sparse observations. Background Art

[0002] Explosion overpressure refers to the energy released by the explosion source when an explosion occurs, which quickly drives the surrounding medium to produce compression fluctuations, forming a propagating overpressure field that is closely related to the explosion equivalent, the distance from the explosion center, environmental conditions and the properties of the medium. Explosion overpressure sparse reconstruction refers to the process of using sparse reconstruction technology to restore the propagation characteristics of the explosion shock wave in space from limited and sparse observation data. This process not only improves data utilization, but also ensures that the reconstruction results are consistent with physical laws. Explosion overpressure sparse reconstruction is the basis for evaluating explosion effects, optimizing protection designs and formulating emergency response plans. It is also a prerequisite for providing more comprehensive and accurate predictions of explosion shock wave overpressure fields.

[0003] At present, for the sparse reconstruction of explosion overpressure, as described in the paper "Modeling of the whole process of shock wave overpressure of free-field air explosion" (Xue, Zai-qing, et al. Defense Technology 15.5 (2019): 815-820.), a calculation formula for the free-field shock wave pressure peak based on scaling parameters is proposed, including: reconstructing the free-field shock wave pressure time history curve during the immune explosion; introducing scaling parameters according to the explosion similarity law, and constructing a calculation formula for the free-field shock wave pressure peak based on scaling parameters; using finite element numerical simulation for comparison and correction. However, the above method relies on the traditional calculation formula for the free-field shock wave pressure peak and specific scenario conditions. Faced with the complexity of explosion shock wave propagation, it lacks the ability to cope with the sparse reconstruction of explosion shock waves in complex environments, multiple obstacles, inhomogeneous media and other scenarios, thus affecting the universality of the method.

[0004] For another example, regarding the sparse reconstruction of explosion overpressure, as described in the paper “Shockwave Signal Downsampling Rate Acquisition Based on Sparse Fourier Transform” (B. Xu et al., in IEEE Access, vol. 10, pp. 44076-44087, 2022, doi: 10.1109 / ACCESS.2022.3169603.), it discloses a shock wave signal capture technology based on sparse Fourier transform, including: using the sparse characteristics of the frequency domain to downsample and collect transient shock wave signals to reduce the amount of collected data; comparing the simulation results of the above experiment with the simulation results of Compressed Sensing low sampling observation. However, the above method realizes sparse sampling of the signal by constructing a shock wave signal capture technology based on sparse Fourier transform. However, the above method cannot make full use of the physical laws of the explosion shock wave for more accurate reconstruction, resulting in difficulty in making up for the lack of data from a physical perspective under sparse data, thereby affecting the accuracy of the results.

[0005] In summary, the problems existing in the prior art are: the accuracy of explosion overpressure sparse reconstruction is not high enough, the versatility is poor, and the efficiency is relatively low. Contents of invention

[0006] The purpose of the present invention is to provide a method and system for physically enhanced reconstruction of explosion overpressure fields for sparse observations, which have high reconstruction accuracy, strong generalization ability and high efficiency.

[0007] The technical solution to achieve the purpose of the present invention is: In the first aspect, a method for physically enhancing and reconstructing explosion overpressure fields for sparse observation is provided. The method uses a pre-trained and optimized explosion overpressure sparse reconstruction model to physically enhance and reconstruct the on-site explosion overpressure field; wherein the explosion overpressure sparse reconstruction model includes a sparse enhancement module, a feature recognition module, a denoising module, and an output module; The sparse enhancement module achieves upsampling and expansion enhancement of the input sparse observation data by setting a fully connected layer; The feature recognition module is constructed by stacking several graph convolution layers and a channel attention layer. The channel attention layer includes a pooling layer, a first convolution layer, a ReLU activation layer, and a second convolution layer in series. The input of the pooling layer is the result of upsampling expansion and enhancement. The output of the second convolution layer is subjected to a Sigmoid operation and then dot-multiplied with the result of upsampling expansion and enhancement. The noisy information obtained by the dot multiplication is used as the output of the channel attention layer. The denoising module includes a noise feature recognition layer and a hybrid feature extraction and recognition layer. The noise feature recognition layer includes a third convolution layer, a fourth convolution layer, an LSTM layer, a fifth convolution layer, and a sixth convolution layer connected in series. The hybrid feature extraction and recognition layer includes seventh to twelfth convolution layers connected in series. The input of the third convolution layer is noisy information, which is concatenated with the output of the sixth convolution layer as the input of the seventh convolution layer. The twelfth convolution layer outputs denoised information. The output module includes a series of deconvolution layers and ReLU activation layers. The input of the deconvolution layer is the denoising information, and the output of the ReLU activation layer is the reconstruction result.

[0008] Preferably, the fully connected layer consists of an input layer, an intermediate layer, an activation layer and an output layer, wherein the number of neurons in the input layer is 20, the intermediate layer consists of 6 layers, each layer consists of a dense network of 64, 128, 256, 512, 1024, and 2048 neurons, respectively, the output layer consists of a single-layer neural network of 2048 neurons, and the adjacent layers in the fully connected layer are connected using the Tanh activation function.

[0009] Preferably, the following physical information constraints are embedded in the explosion overpressure sparse reconstruction model: , , Where ρ represents the air density, p is the shock wave pressure, is the total velocity of the air medium, u1, u2 and u3 represent the air medium velocity components in the lateral, longitudinal and vertical directions, E is the total energy contained in the unit capacity, δ ij It's the Cronek function, ;γ=1.4 is the adiabatic index.

[0010] Preferably, the explosion overpressure field database including a training data set and a validation data set is used to implement pre-training and optimization of the explosion overpressure sparse reconstruction model; The steps to build an explosion overpressure field database include: Based on multi-scenario and multi-dimensional actual working conditions, the overpressure distribution under explosion impact is solved through Abaqus numerical simulation, and a sparse data set of explosion overpressure field is obtained; After data processing and data conversion of the explosion overpressure field sparse data set, an explosion overpressure field database containing a training data set and a verification data set is obtained.

[0011] Preferably, the acquisition of the explosion overpressure field sparse data set includes: Acquisition of actual working conditions in multiple scenarios and dimensions: Acquisition of actual working conditions in multiple scenarios and dimensions by setting different explosion source characteristics, different building structures, different node types, and different proportional distances; Acquisition of sparse datasets of explosion overpressure fields: Based on multi-scenario and multi-dimensional actual working conditions, the overpressure distribution under explosion impact is solved through Abaqus numerical simulation to obtain sparse datasets of explosion overpressure fields. Preferably, the explosion overpressure field database construction includes: Data processing: perform data preprocessing, feature extraction and data standardization on the sparse data set of the explosion overpressure field to obtain standardized data; Data conversion: Use the adjacency matrix to represent the connection relationship between standardized data nodes, and combine the features of nodes and edges into a graph. Through random sampling, the graph is enhanced to obtain a standardized graph storage structure. Construction of overpressure field database: Stratified sampling is performed on the standardized graph storage structure to construct an explosion overpressure field database containing training data sets and verification data sets.

[0012] Preferably, the loss function of the explosion overpressure sparse reconstruction model is The design is as follows: , in, Represents the data error that measures how well the model fits the training data set; A physical constraint loss term that represents the deviation of the model from complying with physical laws; Characterizes the weight coefficient of the physical constraint loss term.

[0013] In the second aspect, a physical enhanced reconstruction system of explosion overpressure field for sparse observation is also provided, including: Model unit, used for pre-training and optimizing explosion overpressure sparse reconstruction model; The reconstruction unit is used to perform physical enhanced reconstruction of the on-site explosion overpressure field using the explosion overpressure sparse reconstruction model.

[0014] According to a third aspect, a computer-readable storage medium storing one or more programs is provided. The one or more programs include instructions. When the instructions are executed by a computing device, the computing device executes the method as described above.

[0015] In a fourth aspect, an electronic device is also provided, comprising one or more processors, one or more memories and one or more programs, wherein the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing the above method.

[0016] Compared with the prior art, the present invention has the following significant advantages: 1. Higher reconstruction accuracy: Based on multi-scenario and multi-dimensional actual working conditions, the present invention constructs an overpressure field database through numerical simulation to express the overpressure distribution characteristics under explosion impact, and constructs a multi-module graph neural network constrained by physical information to enhance the learning and recognition capabilities of the overpressure propagation characteristics of explosion impact. By training the neural network, a sparse reconstruction model of explosion overpressure is obtained, which realizes high-precision reconstruction of the full flow field of sparse overpressure data of explosion impact in complex multi-dimensional and multi-scale scenarios; 2. Strong generalization ability and high efficiency: The present invention establishes an overpressure field database, combines effective sparse data with physical information to constrain multi-module graph neural network training, and realizes a sparse reconstruction model of explosion overpressure. By learning from multiple types of sparse data, the model has the ability to generalize and reconstruct the explosion overpressure field distribution under different explosion source characteristics, different building structures, different node types, and different proportional distance conditions, thereby realizing efficient reconstruction of multi-scenario and multi-dimensional explosion overpressure distribution. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a method flow chart of an embodiment of the present invention.

[0018] Figure 2 It is a structural diagram of the sparse enhancement module.

[0019] Figure 3 It is a structural diagram of the feature identification module.

[0020] Figure 4 It is a structural diagram of the channel attention layer.

[0021] Figure 5 It is a structural diagram of the noise denoising module.

[0022] Figure 6 It is a structural diagram of the output module. DETAILED DESCRIPTION

[0023] like Figure 1 As shown, the physical enhancement reconstruction method of explosion overpressure field for sparse observation of the present invention comprises the following steps: S10. Construction of overpressure field database: Based on multi-scenario and multi-dimensional actual working conditions, a sparse data set of explosion overpressure field is obtained through Abaqus numerical simulation. The sparse data set of explosion overpressure field is processed and converted to obtain an explosion overpressure field database containing a training data set and a verification data set. The step of constructing the overpressure field database in S10 includes: S101. Acquisition of sparse data set of explosion overpressure field: Based on multi-scenario and multi-dimensional actual working conditions, the overpressure distribution under the explosion impact is solved through Abaqus numerical simulation to obtain a sparse data set of explosion overpressure field.

[0024] The S101, acquisition of explosion overpressure field sparse data set includes: S1011. Acquisition of actual working conditions in multiple scenarios and dimensions: Acquisition of actual working conditions in multiple scenarios and dimensions by setting different explosion source characteristics, different building structures, different node types, and different proportional distances.

[0025] The S1011 multi-scenario multi-dimensional actual working condition acquisition includes: S10111. Explosion source characteristic setting: Based on the principle of explosion mechanics, determine the equivalent, density, and energy release rate parameters of different types of explosives. Combined with the physical and chemical properties of the explosives, determine the initial conditions of the explosion center, including the initial pressure, temperature, and energy of the explosion center. S10112. Building structure type setting: According to the actual scene, set a variety of building structures (such as high-rise buildings, industrial facilities, residential buildings, underground structures, etc.), and set their corresponding material properties and geometric shapes; S10113. Node type setting: setting corresponding positions and structural nodes according to the different building structures; S10114. Proportional distance acquisition: According to the actual scenario, different distance ranges are set, and the ratio of the explosion center distance to the explosion equivalent, i.e. the proportional distance, is expressed by the normalized distance formula: , Where Z is the proportional distance, R is the distance from the center of the explosion, and W is the explosion equivalent (TNT equivalent mass).

[0026] S1012. Acquisition of sparse data set of explosion overpressure field: Based on the multi-scenario and multi-dimensional actual working conditions obtained in S1011, the overpressure distribution under the explosion impact is solved through Abaqus numerical simulation to obtain a sparse data set of explosion overpressure field.

[0027] S102, constructing an explosion overpressure field database: performing data processing and data conversion on the explosion overpressure field sparse data set to construct an explosion overpressure field database including a training data set and a verification data set.

[0028] The said S102, construction of explosion overpressure field database comprises: S1021, data processing: performing data preprocessing, feature extraction and data standardization on the explosion overpressure field sparse data set to obtain standardized data; it should be noted here that the data processing adopted in the embodiment of the present invention is a prior art.

[0029] The S1021, data processing construction includes: S10211, data preprocessing: batch processing the data set data in the overpressure field database and converting them into a graph structure that can be used for learning.

[0030] The graph structure for learning is as follows: , , , Where G represents the graph structure, V represents the node set, and node v n It represents a spatial location, building unit or measurement point, E represents a set of edges, and edges represent physical connections or spatial relationships.

[0031] S10212. Feature extraction: Vectorize the features of nodes and edges in the graph structure and construct feature vectors of multi-source data.

[0032] S10213. Data standardization: normalize the feature vector to obtain standardized data.

[0033] The eigenvector normalization process is as follows: , Wherein, Z represents the normalized data, and X represents the feature vector.

[0034] The standardized data are as follows: , Among them, M characterizes the standardized data, Z n Characterize the normalized data under the actual working conditions of multiple scenarios and multiple dimensions (for example, explosion equivalent, proportional distance, building structure geometric parameters, etc.).

[0035] S1022. Data conversion: Use the adjacency matrix to represent the connection relationship between standardized data nodes, and combine the features of nodes and edges into a graph. Through random sampling, the graph is enhanced to obtain a standardized graph storage structure.

[0036] The adjacency matrix is ​​expressed as follows: , , where A represents the adjacency matrix, ω ij v i and v j weight between.

[0037] S1023, overpressure field database construction: performing stratified sampling on the standardized graphic storage structure to construct an explosion overpressure field database including a training data set and a verification data set.

[0038] S20. Establishment of explosion overpressure sparse reconstruction model: By constructing a sparse enhancement module, a feature recognition module, a denoising module and an output module, embedding physical information, and using the explosion overpressure field database containing a training data set and a verification data set for training and optimization, a sparse reconstruction model of explosion overpressure constrained by physical information is obtained.

[0039] S201. A sparse enhancement module is constructed by setting a fully connected layer. The fully connected layer obtains the distribution data of the entire flow field by interpolating and supplementing the input sparse data, thereby achieving upsampling, expansion and enhancement of the sparse data.

[0040] like Figure 2 As shown in the figure, the fully connected layer consists of an input layer, an intermediate layer, an activation layer, and an output layer. The number of neurons in the input layer is 20, and the intermediate layer consists of 6 layers, each with 64, 128, 256, 512, 1024, and 2048 neurons. Each layer is connected using the Tanh activation function, and the output layer has 2048 neurons.

[0041] S202. By setting a graph convolution layer, building a graph neural network, and utilizing the graph neural network message passing mechanism, the feature recognition module can aggregate the deep features, thereby constructing a feature recognition module.

[0042] like Figure 3 As shown in the figure, the graph neural network message passing mechanism consists of 5 stacked graph convolutional layers connected in series with a channel attention layer to achieve message passing and update of node features.

[0043] The message delivery form is expressed as follows: , Among them, h i (k+1) Characterize the characteristics of node i in layer k+1; h i (k) Characterize the characteristics of node i in the k-layer; M ij (k) Characterizes the message from node j to node i; Represents the set of neighbor nodes of node i; Aggregate(•) represents the aggregation function; Update(•) represents the node feature update function.

[0044] like Figure 4 As shown, the channel attention layer is composed of a pooling compression network and a Sigmoid excitation network connected in sequence to enhance the deep features of the network and repair the detailed texture.

[0045] S203, using the convolutional recurrent network as the core architecture to build a denoising module. Figure 5As shown in the figure, the module mainly consists of two parts: noise feature recognition and mixed feature extraction and recognition. The noise feature recognition is composed of four 1×1 convolutional networks and an LSTM recurrent network in series, and the mixed feature extraction and recognition is composed of six layers of 1×1 convolutional networks. This module realizes the denoising of deep aggregation feature information.

[0046] S204, Figure 6 As shown in FIG. 1 , by connecting the deconvolution layer and the ReLU activation layer in series, an output module is constructed to realize decoding of the deep denoising feature information by the output module.

[0047] S205, physical constraint setting: physical information is extracted from the mathematical modeling of the explosion impact phenomenon, and physical constraints are realized with the help of the Euler equation without viscosity terms. The basic form of the conservation equations of mass, momentum and energy of inviscid compressible flow is as follows: , , where ρ represents the air density, It's shock wave pressure, is the total velocity of the air medium, u1, u2 and u3 represent the air medium velocity components in the lateral, longitudinal and vertical directions, E is the total energy contained in the unit capacity, δ ij is the Kronecker function: The Kronecker function takes two integers as parameters, and if they are equal, the output is 1; otherwise, the output is 0; γ = 1.4 is the adiabatic index; .

[0048] By combining physical constraints with graph neural networks and undergoing iterative training, the model’s understanding of physical phenomena can be enhanced, the interactions between three-dimensional sparse nodes can be accurately captured, and the accuracy and robustness of predictions can be improved.

[0049] S206. Loss function design: A composite loss function combining standard data error and physical constraint error is used to implement loss function design. The basic form is as follows: , in, Characterize data error and measure the model's fit to the training data; Characterizes the physical constraint loss, measuring the deviation of the model in complying with physical laws; The weight coefficient that represents the physical constraint term balances the impact of data error and physical constraints.

[0050] S207, explosion overpressure sparse reconstruction model training: using the training data set to train the explosion overpressure sparse reconstruction model.

[0051] S208, explosion overpressure sparse reconstruction model optimization: using the validation data set, testing and evaluating the explosion overpressure sparse reconstruction model, optimizing the model according to the evaluation results, and obtaining an optimized explosion overpressure sparse reconstruction model. The step S208, optimization of the explosion overpressure sparse reconstruction model, includes: S2081. Error assessment: The explosion overpressure sparse reconstruction model is verified using the verification data set, and the explosion overpressure distribution result reconstructed by the reconstruction model is compared with the simulation data or experimental data, and the error measurement index is used to perform error assessment on the explosion overpressure distribution result.

[0052] The S2081 and error evaluation steps are specifically: The mean square error (MSE) is obtained by calculating the average of the square differences between the explosion overpressure distribution results reconstructed by the calculation model and the simulation data or experimental data, which measures the overall error. To measure the reconstruction of the model and evaluate the overall fitting effect of the model.

[0053] S2082. Optimization of explosion overpressure sparse reconstruction model: Based on the error evaluation results, an optimized explosion overpressure sparse reconstruction model based on hybrid digital-physical drive is obtained through parameter tuning, structural optimization, data structure enhancement and balance.

[0054] The step S2082, explosion overpressure sparse reconstruction model optimization step comprises: S20821. Hyperparameter tuning: Perform parameter tuning based on error evaluation results.

[0055] S20822. Structural optimization: Optimize the model structure by adjusting the number of layers, node aggregation method and convolution method of the graph neural network, and conduct error evaluation on the reconstruction results of the explosion overpressure sparse reconstruction model based on digital-physical hybrid drive after structural optimization.

[0056] S20823. Data structure enhancement and balance: Balance data with large errors through undersampling and oversampling methods, and achieve data enhancement and balance through scene data expansion, data perturbation, and changing graph structure.

[0057] S20824, model optimization: repeat steps S20821 to S20823 until the error evaluation standard is met, and an optimized explosion overpressure sparse reconstruction model is obtained; S30, sparse reconstruction of explosion overpressure: constructing an on-site explosion input data set according to the on-site environment and explosion data, inputting the data of the on-site explosion input data set into the optimized explosion shock wave sparse reconstruction model to obtain on-site explosion overpressure distribution reconstructed data.

[0058] The step S30, explosion overpressure sparse reconstruction, comprises: S301, input data set construction: obtain the reconstruction scene environment and explosion data, and construct the reconstruction scene input data set. The step of constructing the input data set in S301 is specifically as follows: With the help of actual sensor measurements, the environmental parameter conditions such as explosion source characteristics, building structure, node type, proportional distance, etc. in the required reconstruction scenario are obtained. Through data preprocessing and feature extraction, a multi-source input data set is obtained. Through data standardization processing and data conversion, an input data set is constructed.

[0059] S302, explosion overpressure sparse reconstruction: inputting the reconstructed on-site input data set data into the optimized explosion overpressure sparse reconstruction model, and the output thereof is the on-site explosion overpressure distribution reconstructed data.

[0060] The present invention is based on multi-scenario and multi-dimensional actual working conditions, and obtains a sparse data set of explosion overpressure field through numerical simulation. Through data processing and conversion, an overpressure field database is constructed, and a sparse reconstruction model of explosion overpressure is established. The sparse reconstruction model is trained based on a multi-module graph neural network constrained by physical information, thereby obtaining a sparse reconstruction model of explosion overpressure based on digital-physical hybrid drive with high-precision sparse data reconstruction capability under the influence of noise and extremely strong generalization capability. The model is also error evaluated and parameter tuned to improve the reconstruction capability of the sparse reconstruction model of explosion overpressure, and to realize the sparse reconstruction of overpressure distribution data under explosion impact in complex scenarios and at multiple scales.

[0061] The present invention also designs a physical enhancement reconstruction system of explosion overpressure field for sparse observation, including: Model unit, used for pre-training and optimizing explosion overpressure sparse reconstruction model; The reconstruction unit is used to perform physical enhanced reconstruction of the on-site explosion overpressure field using the explosion overpressure sparse reconstruction model.

[0062] The technical solution of the above-mentioned precise analysis system of the explosion shock wave distribution pattern is similar to the aforementioned precise analysis method of the explosion shock wave distribution pattern, which will not be repeated here.

[0063] Based on the same technical solution, the present invention also discloses a computer-readable storage medium storing one or more programs, wherein the one or more programs include instructions, which, when executed by a computing device, enable the computing device to execute the above-mentioned method for physically enhanced reconstruction of explosion overpressure fields for sparse observations.

[0064] Based on the same technical solution, the present invention also discloses a computing device, including one or more processors, one or more memories and one or more programs, wherein the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing the above-mentioned method for physically enhanced reconstruction of explosion overpressure fields for sparse observations.

[0065] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0066] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0067] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0068] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0069] The above embodiments are only for illustrating the technical idea of ​​the present invention, and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for physically enhanced reconstruction of explosion overpressure fields based on sparse observations, characterized in that: The method uses a pre-trained and optimized explosion overpressure sparse reconstruction model to physically enhance the on-site explosion overpressure field reconstruction; wherein the explosion overpressure sparse reconstruction model includes a sparse enhancement module, a feature recognition module, a denoising module and an output module; The sparse enhancement module achieves upsampling and expansion enhancement of the input sparse observation data by setting a fully connected layer; The feature recognition module is constructed by stacking several graph convolution layers and a channel attention layer. The channel attention layer includes a pooling layer, a first convolution layer, a ReLU activation layer, and a second convolution layer in series. The input of the pooling layer is the result of upsampling expansion and enhancement. The output of the second convolution layer is subjected to a Sigmoid operation and then dot-multiplied with the result of upsampling expansion and enhancement. The noisy information obtained by the dot multiplication is used as the output of the channel attention layer. The denoising module includes a noise feature recognition layer and a hybrid feature extraction and recognition layer. The noise feature recognition layer includes a third convolution layer, a fourth convolution layer, an LSTM layer, a fifth convolution layer, and a sixth convolution layer connected in series. The hybrid feature extraction and recognition layer includes seventh to twelfth convolution layers connected in series. The input of the third convolution layer is noisy information, which is concatenated with the output of the sixth convolution layer as the input of the seventh convolution layer. The twelfth convolution layer outputs denoised information. The output module includes a series of deconvolution layers and ReLU activation layers. The input of the deconvolution layer is the denoising information, and the output of the ReLU activation layer is the reconstruction result.

2. The method according to claim 1, characterized in that The fully connected layer consists of an input layer, an intermediate layer, an activation layer, and an output layer. The number of neurons in the input layer is 20, the intermediate layer consists of 6 layers, each layer consists of a dense network of 64, 128, 256, 512, 1024, and 2048 neurons, and the output layer consists of a single-layer neural network with 2048 neurons. The adjacent layers in the fully connected layer are connected using the Tanh activation function.

3. The method according to claim 1, characterized in that The following physical information constraints are embedded in the explosion overpressure sparse reconstruction model: ， in, ρ represents the air density, p is the shock wave pressure, u1, u2 and u3 represent the air medium velocity components in the lateral, longitudinal and vertical directions, δ ij is the Kronecker function, .

4. The method according to claim 1, characterized in that The explosion overpressure field database containing training data sets and validation data sets is used to realize the pre-training and optimization of the explosion overpressure sparse reconstruction model; The steps for building the explosion overpressure field database include: Based on multi-scenario and multi-dimensional actual working conditions, the overpressure distribution under explosion impact is solved through Abaqus numerical simulation, and a sparse data set of explosion overpressure field is obtained; After data processing and data conversion of the explosion overpressure field sparse data set, an explosion overpressure field database containing a training data set and a verification data set is obtained.

5. The method according to claim 4, characterized in that The acquisition of explosion overpressure field sparse data set includes: Acquisition of actual working conditions in multiple scenarios and dimensions: Acquisition of actual working conditions in multiple scenarios and dimensions by setting different explosion source characteristics, different building structures, different node types, and different proportional distances; Acquisition of sparse data sets of explosion overpressure field: Based on multi-scenario and multi-dimensional actual working conditions, Abaqus numerical simulation is used to solve the overpressure distribution under explosion impact and obtain a sparse data set of explosion overpressure field.

6. The method according to claim 4, characterized in that The construction of explosion overpressure field database includes: Data processing: perform data preprocessing, feature extraction and data standardization on the sparse data set of the explosion overpressure field to obtain standardized data; Data conversion: Use the adjacency matrix to represent the connection relationship between standardized data nodes, and combine the features of nodes and edges into a graph. Through random sampling, the graph is enhanced to obtain a standardized graph storage structure. Construction of overpressure field database: Stratified sampling is performed on the standardized graph storage structure to construct an explosion overpressure field database containing training data sets and verification data sets.

7. The method according to claim 3, characterized in that Loss function of explosion overpressure sparse reconstruction model The design is as follows: ， in, Represents the data error that measures how well the model fits the training data set; A physical constraint loss term that represents the deviation of the model from complying with physical laws; The weight coefficient that represents the physical constraint loss term.

8. A physical enhanced reconstruction system of explosion overpressure field for sparse observation, characterized in that: include: Model unit, used for pre-training and optimizing explosion overpressure sparse reconstruction model; The reconstruction unit is used to perform physical enhanced reconstruction of the on-site explosion overpressure field using the explosion overpressure sparse reconstruction model.

9. A computer-readable storage medium storing one or more programs, wherein the one or more programs include instructions, characterized in that: When the instructions are executed by a computing device, the computing device is caused to perform the method according to claims 1 to 7.

10. An electronic device, characterized in that: The method comprises one or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing the method according to claims 1 to 7.

Images