Physical Enhancement and Reconstruction Method and System for Explosive Overpressure Fields in Sparse Observations

Abstract

This invention discloses a physical enhancement and reconstruction method and system for explosion overpressure fields oriented towards sparse observations. It utilizes a pre-trained and optimized explosion overpressure sparse reconstruction model to physically enhance and reconstruct the on-site explosion overpressure field. The explosion overpressure sparse reconstruction model includes a sparse enhancement module, a feature identification module, a denoising module, and an output module. The sparse reconstruction method of this invention offers high reconstruction accuracy, strong generalization ability, and high efficiency, enabling high-precision reconstruction of the entire flow field of sparse overpressure data from explosion impacts in complex, multi-dimensional, and multi-scale scenarios.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular to a physical enhancement and reconstruction method and system for explosive overpressure fields oriented towards sparse observation. Background Technology

[0002] Explosive overpressure refers to the energy released by the explosion source during an explosion, which rapidly pushes the surrounding medium to produce compressive fluctuations, forming a propagating overpressure field closely related to the explosion yield, the distance from the explosion center, environmental conditions, and the properties of the medium. Explosive overpressure sparse reconstruction refers to the process of recovering the propagation characteristics of the explosion shock wave in space from limited, sparse observational data using sparse reconstruction techniques. This process not only improves data utilization but also ensures that the reconstruction results match physical laws. Explosive overpressure sparse reconstruction is fundamental for assessing explosion effects, optimizing protective designs, and developing emergency response plans; it is also a prerequisite for providing more comprehensive and accurate predictions of the explosion shock wave overpressure field.

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

[0004] For example, regarding sparse reconstruction of explosion overpressure, 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.) discloses a shock wave signal acquisition technique based on sparse Fourier transform, including: utilizing the sparsity characteristics of the frequency domain to downsample the transient shock wave signal, reducing the amount of data acquired; and comparing the simulation results of the above experiment with the simulation results of compressed sensing low-sampling observations. However, while the above method achieves sparse sampling of the signal by constructing a shock wave signal acquisition technique based on sparse Fourier transform, it cannot fully utilize the physical laws of the explosion shock wave for more accurate reconstruction. This makes it difficult to compensate for the lack of data from a physical perspective under sparse data conditions, thus affecting the accuracy of the results.

[0005] In summary, the problems with existing technologies are: insufficient accuracy of explosive overpressure sparse reconstruction, poor versatility, and low efficiency. Summary of the Invention

[0006] The purpose of this invention is to provide a physical enhancement and reconstruction method and system for sparse observation of explosion overpressure fields, which has high reconstruction accuracy, strong generalization ability and high efficiency.

[0007] The technical solution to achieve the purpose of this invention is as follows:

[0008] Firstly, a physical enhancement and reconstruction method for explosive overpressure fields oriented towards sparse observations is provided. This method utilizes a pre-trained and optimized explosive overpressure sparse reconstruction model to physically enhance and reconstruct the on-site explosive overpressure field. The explosive overpressure sparse reconstruction model includes a sparse enhancement module, a feature identification module, a denoising module, and an output module.

[0009] The sparse augmentation module achieves upsampling and augmentation of the input sparse observation data by setting a fully connected layer;

[0010] The feature recognition module is constructed by stacking several graph convolutional layers and a channel attention layer. The channel attention layer includes a pooling layer, a first convolutional layer, a ReLU activation layer and a second convolutional layer. The input of the pooling layer is the result of upsampling augmentation. The output of the second convolutional layer is multiplied by the result of upsampling augmentation after a sigmoid operation. The noisy information obtained by the multiplication is used as the output of the channel attention layer.

[0011] 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 convolutional layer, a fourth convolutional layer, an LSTM layer, a fifth convolutional layer, and a sixth convolutional layer in series. The hybrid feature extraction and recognition layer includes a seventh to twelfth convolutional layer in series. The input of the third convolutional layer is the noisy information. The noisy information is concatenated with the output of the sixth convolutional layer and used as the input of the seventh convolutional layer. The twelfth convolutional layer outputs the denoised information.

[0012] The output module includes a series of deconvolutional layers and ReLU activation layers. The input to the deconvolutional layer is the denoised information, and the output of the ReLU activation layer is the reconstructed result.

[0013] Preferably, the fully connected layer consists of an input layer, an intermediate layer, an activation layer, and an output layer. The input layer has 20 neurons, the intermediate layer consists of 6 layers of dense networks with 64, 128, 256, 512, 1024, and 2048 neurons per layer, respectively, and the output layer consists of a single-layer neural network with 2048 neurons. Adjacent layers in the fully connected layer are connected using the Tanh activation function.

[0014] Preferably, the following physical information constraints are embedded in the explosion overpressure sparse reconstruction model:

[0015] ,

[0016] ,

[0017] Where ρ represents air density and p is shock wave pressure. Let u1, u2, and u3 represent the velocity components of the air medium in the lateral, longitudinal, and vertical directions, respectively, and let E be the total energy contained per unit volume. δ ij It is the Kronecker function. γ=1.4 is the adiabatic index.

[0018] Preferably, the pre-training and optimization of the explosion overpressure sparse reconstruction model are achieved using an explosion overpressure field database containing training and validation datasets;

[0019] The steps for constructing an explosion overpressure field database include:

[0020] Based on real-world working conditions across multiple scenarios and dimensions, we use Abaqus numerical simulation to solve the overpressure distribution under explosion impact and obtain a sparse dataset of the explosion overpressure field.

[0021] After data processing and transformation, an explosion overpressure field database containing training and validation datasets is obtained from the sparse dataset of the explosion overpressure field.

[0022] Preferably, the acquisition of the sparse dataset of the explosion overpressure field includes:

[0023] Acquisition of real-world working conditions in multiple scenarios and dimensions: By setting different explosion source characteristics, different building structures, different node types, and different scale distances, real-world working conditions in multiple scenarios and dimensions can be obtained;

[0024] Acquisition of sparse dataset of explosion overpressure field: Based on real working conditions in multiple scenarios and dimensions, the overpressure distribution under explosion impact is solved through Abaqus numerical simulation to obtain a sparse dataset of explosion overpressure field.

[0025] Preferably, the construction of the explosion overpressure field database includes:

[0026] Data processing: Data preprocessing, feature extraction, and data standardization are performed on the sparse dataset of the explosion overpressure field to obtain standardized data;

[0027] Data transformation: The adjacency matrix is ​​used to represent the connection relationship between standardized data nodes, and the features of nodes and edges are combined into a graph. Graph enhancement is achieved through random sampling to obtain a standardized graph storage structure.

[0028] Overpressure field database construction: The standardized graphical storage structure is hierarchically sampled to construct an explosion overpressure field database containing training and validation datasets.

[0029] Preferably, the loss function of the explosion overpressure sparse reconstruction model The design is as follows:

[0030] ,

[0031] in, The data error that measures how well the model fits the training dataset; The physical constraint loss term characterizes and measures the deviation of the model from obeying physical laws. The weighting coefficients characterize the physical constraint loss term.

[0032] Secondly, a physical enhancement and reconstruction system for explosion overpressure fields oriented towards sparse observations is also provided, including:

[0033] Model unit, used for pre-training and optimization of the explosive overpressure sparse reconstruction model;

[0034] The reconstruction unit is used to physically enhance and reconstruct the on-site explosion overpressure field using the explosion overpressure sparse reconstruction model.

[0035] Thirdly, a computer-readable storage medium is also provided for storing one or more programs, said one or more programs including instructions that, when executed by a computing device, cause the computing device to perform the method described above.

[0036] Fourthly, an electronic device is also provided, 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 configured to be executed by the one or more processors, and the one or more programs include instructions for performing the methods described above.

[0037] Compared with the prior art, the significant advantages of this invention are:

[0038] 1. Higher reconstruction accuracy: Based on multi-scenario and multi-dimensional actual working conditions, this invention constructs an overpressure field database through numerical simulation to express the overpressure distribution characteristics under explosion impact. By constructing a physical information-constrained multi-module graph neural network, the learning and recognition capabilities of the overpressure propagation characteristics of explosion impact are enhanced. By training the neural network, an explosion overpressure sparse reconstruction model is obtained, realizing high-precision reconstruction of the entire flow field of sparse overpressure data of explosion impact in complex multi-dimensional and multi-scale scenarios.

[0039] 2. Strong generalization ability and high efficiency: This invention establishes an overpressure field database and combines effective sparse data with physical information constraints to train a multi-module graph neural network to realize an explosion overpressure sparse reconstruction model. Through learning from various 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, thus achieving efficient reconstruction of multi-scenario and multi-dimensional explosion overpressure distribution. Attached Figure Description

[0040] Figure 1 This is a flowchart of a method according to an embodiment of the present invention.

[0041] Figure 2 This is a schematic diagram of the sparse enhancement module.

[0042] Figure 3 This is a structural diagram of the feature recognition module.

[0043] Figure 4 This is a schematic diagram of the channel attention layer.

[0044] Figure 5 This is a structural diagram of the noise reduction module.

[0045] Figure 6 This is a schematic diagram of the output module. Detailed Implementation

[0046] like Figure 1 As shown, the present invention provides a physical enhancement and reconstruction method for explosive overpressure fields for sparse observations, comprising the following steps:

[0047] S10. Construction of Overpressure Field Database: Based on multi-scenario and multi-dimensional actual working conditions, an explosion overpressure field sparse dataset is obtained through Abaqus numerical simulation. The explosion overpressure field sparse dataset is then processed and transformed to obtain an explosion overpressure field database containing training and validation datasets.

[0048] The steps of S10 and the construction of the overpressure field database include:

[0049] S101. Acquisition of Sparse Dataset of Explosion Overpressure Field: Based on multi-scenario and multi-dimensional actual working conditions, the overpressure distribution under explosion impact is solved through Abaqus numerical simulation to obtain the sparse dataset of explosion overpressure field.

[0050] The acquisition of the S101 and the sparse dataset of the explosion overpressure field includes:

[0051] S1011. Acquisition of real-world working conditions in multiple scenarios and dimensions: By setting different explosion source characteristics, different building structures, different node types, and different proportional distances, real-world working conditions in multiple scenarios and dimensions can be acquired.

[0052] The S1011 multi-scenario, multi-dimensional acquisition of actual working conditions includes:

[0053] 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, and combine the physicochemical properties of the explosives to determine the initial conditions of the explosion center point, including the initial pressure, temperature, and energy of the explosion center.

[0054] S10112. Building Structure Type Setting: Based on the actual scenario, set various building structures (such as high-rise buildings, industrial facilities, residences, underground structures, etc.) and set their corresponding material properties and geometry.

[0055] S10113. Node type setting: Set the corresponding positions and structural nodes according to the different building structures;

[0056] S10114. Proportional Distance Acquisition: Based on the actual scenario, different distance ranges are set, and the ratio of the blast center distance to the explosive yield, i.e., the proportional distance, is expressed using a normalized distance formula:

[0057] ,

[0058] Where Z is the proportional distance, R is the distance from the blast center, and W is the explosive yield (TNT equivalent mass).

[0059] S1012. Acquisition of Sparse Dataset of Explosion Overpressure Field: Based on the multi-scenario and multi-dimensional actual working conditions obtained in S1011, the overpressure distribution under explosion impact is solved through Abaqus numerical simulation to obtain the sparse dataset of explosion overpressure field.

[0060] S102. Construction of the explosion overpressure field database: Perform data processing and data transformation on the sparse dataset of the explosion overpressure field to construct an explosion overpressure field database containing training datasets and validation datasets.

[0061] The construction of the S102 explosion overpressure field database includes:

[0062] S1021. Data processing: The sparse dataset of the explosion overpressure field is preprocessed, features are extracted, and data is standardized to obtain standardized data. It should be noted that the data processing used in this embodiment of the invention is existing technology.

[0063] S1021, data processing construction includes:

[0064] S10211. Data preprocessing: The dataset data in the overpressure field database is processed in batches and transformed into a graph structure that can be learned.

[0065] The learnable graph structure is as follows:

[0066] ,

[0067] ,

[0068] ,

[0069] Where G represents the graph structure, V represents the set of nodes, and node v n E represents a spatial location, building unit, or measurement point, and E represents a set of edges, with edges representing physical connections or spatial relationships.

[0070] S10212, Feature Extraction: Vectorize the features of nodes and edges in the graph structure to construct feature vectors from multi-source data.

[0071] S10213. Data standardization: Normalize the feature vectors to obtain standardized data.

[0072] The feature vector normalization process is as follows:

[0073] ,

[0074] Where Z represents the normalized data, and X represents the feature vector.

[0075] The standardized data is as follows:

[0076] ,

[0077] Where M represents standardized data, Zn Normalized data representing the multi-scenario, multi-dimensional actual working conditions (e.g., explosive equivalent, proportional distance, building structure geometric parameters, etc.).

[0078] S1022. Data Transformation: The adjacency matrix is ​​used to represent the connection relationship between standardized data nodes, and the features of nodes and edges are combined into a graph. Graph enhancement is achieved through random sampling to obtain a standardized graph storage structure.

[0079] The adjacency matrix is ​​expressed as follows:

[0080] ,

[0081] ,

[0082] Where A represents the adjacency matrix, ω ij For v i and v j The weights between them.

[0083] S1023. Construction of the overpressure field database: The standardized graphical storage structure is hierarchically sampled to construct an explosion overpressure field database containing training datasets and validation datasets.

[0084] S20. Establishment of the Explosion Overpressure Sparse Reconstruction Model: By constructing a sparse enhancement module, a feature identification module, a denoising module, and an output module, physical information is embedded, and the explosion overpressure field database containing training and validation datasets is used for training and optimization to obtain a physically constrained explosion overpressure sparse reconstruction model.

[0085] S201. By setting a fully connected layer, a sparse enhancement module is constructed. The fully connected layer interpolates and supplements the input sparse data to obtain the distribution data of the entire flow field, thereby realizing the upsampling and enhancement of the sparse data.

[0086] like Figure 2 As shown, the fully connected layer consists of an input layer, an intermediate layer, an activation layer, and an output layer. The input layer has 20 neurons, the intermediate layer consists of 6 layers of dense networks with 64, 128, 256, 512, 1024, and 2048 neurons respectively, and the layers are connected by the Tanh activation function. The output layer has 2048 neurons.

[0087] S202. By setting up graph convolutional layers, a graph neural network is built. The graph neural network message passing mechanism is used to realize the feature aggregation of deep features by the feature recognition module, thereby constructing the feature recognition module.

[0088] like Figure 3As shown, the graph neural network message passing mechanism consists of five graph convolutional layers stacked together and then connected in series with a channel attention layer to achieve message passing and node feature updates.

[0089] The message passing format is represented as follows:

[0090] ,

[0091] Among them, h i (k+1) Characterizes the features of node i at layer k+1; h i (k) Characterizes the features of node i in layer k; M ij (k) Represents the message from node j to node i; The set of neighboring nodes of node i is represented; Aggregate(•) represents the aggregation function; Update(•) represents the node feature update function.

[0092] like Figure 4 As shown, the channel attention layer is composed of pooling compression and a Sigmoid excitation network connected in sequence to enhance the deep features of the network and restore the details and textures.

[0093] S203. A denoising module is constructed using a convolutional recurrent network as the core architecture. For example... Figure 5 As shown, this module mainly consists of two parts: noise feature recognition and hybrid feature extraction and recognition. The noise feature recognition is composed of four 1×1 convolutional networks and one LSTM recurrent network connected in series, while the hybrid feature extraction and recognition is composed of six 1×1 convolutional networks. This module realizes the denoising of deep aggregated feature information.

[0094] S204, such as Figure 6 As shown, an output module is constructed by concatenating deconvolutional layers and ReLU activation layers, enabling the output module to decode the deep denoising feature information.

[0095] S205. Physical Constraint Setting: Physical information is extracted from the mathematical modeling of the explosion impact phenomenon. Physical constraints are implemented using the Euler equations without viscosity terms. The basic forms of the conservation equations for mass, momentum, and energy of inviscid compressible flow are as follows:

[0096] ,

[0097] ,

[0098] Where ρ represents air density. It is shock wave pressure. Let u1, u2, and u3 represent the velocity components of the air medium in the lateral, longitudinal, and vertical directions, respectively, and let E be the total energy contained per unit volume. δ ij The Kronecker function takes two integers as parameters; if they are equal, the output is 1; otherwise, the output is 0. γ=1.4 is the adiabatic index.

[0099] .

[0100] By combining physical constraints with graph neural networks and through iterative training, the model's understanding of physical phenomena can be enhanced, accurately capturing the interactions between sparse three-dimensional nodes, thus improving the accuracy and robustness of predictions.

[0101] S206. Loss Function Design: The loss function is designed using a composite loss function that combines standard data error and physical constraint error. The basic form is as follows:

[0102] ,

[0103] in, Characterizes data error and measures how well the model fits the training data; Characterizes the physical constraint loss, measuring the deviation of the model from obeying physical laws; The weighting coefficients characterize the physical constraint terms, balancing the effects of data errors and physical constraints.

[0104] S207. Training of the explosion overpressure sparse reconstruction model: The explosion overpressure sparse reconstruction model is trained using the training dataset.

[0105] S208. Optimization of the explosion overpressure sparse reconstruction model: Using the validation dataset, test and evaluate the explosion overpressure sparse reconstruction model, and optimize the model based on the evaluation results to obtain the optimized explosion overpressure sparse reconstruction model.

[0106] The optimization steps for the S208 explosion overpressure sparse reconstruction model include:

[0107] S2081. Error Assessment: The explosion overpressure sparse reconstruction model is verified using the aforementioned verification dataset. By comparing the explosion overpressure distribution results reconstructed by the reconstruction model with simulation data or experimental data, the error of the explosion overpressure distribution results is assessed using an error metric.

[0108] The error assessment step, S2081, is specifically as follows:

[0109] The mean square error (MSE) is obtained by calculating the average of the squared differences between the reconstructed explosion overpressure distribution results from the model and the simulation or experimental data, thus measuring the overall error. The coefficient of determination of the model is also calculated. This is used to measure the reconstruction of the model and to evaluate the overall fitting effect of the model.

[0110] S2082. Optimization of the explosion overpressure sparse reconstruction model: Based on the error evaluation results, an optimized explosion overpressure sparse reconstruction model based on data-matter hybrid driving is obtained through parameter tuning, structure optimization, data structure enhancement and balancing.

[0111] The optimization steps of the S2082 explosion overpressure sparse reconstruction model include:

[0112] S20821. Hyperparameter tuning: Based on the error assessment results, perform parameter tuning.

[0113] S20822, Structural Optimization: The model structure is optimized by adjusting the number of layers, node aggregation method, and convolution method of the graph neural network. The reconstruction results of the explosion overpressure sparse reconstruction model based on data-matter hybrid driving after structural optimization are then evaluated for error.

[0114] S20823. Data Structure Enhancement and Balancing: Data with large errors is balanced by undersampling and oversampling methods. Data enhancement and balancing are achieved through scene data expansion, data perturbation, and change graph structure.

[0115] S20824, Model Optimization: Repeat steps S20821 to S20823 until the error evaluation criteria are met to obtain the optimized explosion overpressure sparse reconstruction model.

[0116] S30. Explosion Overpressure Sparse Reconstruction: Based on the on-site environment and explosion data, construct an on-site explosion input dataset, input the on-site explosion input dataset into the optimized explosion shock wave sparse reconstruction model, and obtain on-site explosion overpressure distribution reconstruction data.

[0117] The S30 step, explosive overpressure sparse reconstruction step, includes:

[0118] S301. Input Dataset Construction: Obtain the reconstruction site environment and explosion data, and construct the reconstruction site input dataset.

[0119] The specific steps for constructing the input dataset for S301 are as follows:

[0120] By using sensors to measure environmental parameters such as explosion source characteristics, building structure, node type, and scale distance in the required reconstructed scenario, a multi-source input dataset is obtained through data preprocessing and feature extraction. The input dataset is then constructed through data standardization and data transformation.

[0121] S302, Explosion Overpressure Sparse Reconstruction: Input the reconstruction site input dataset into the optimized explosion overpressure sparse reconstruction model, and its output is the site explosion overpressure distribution reconstruction data.

[0122] This invention, based on multi-scenario and multi-dimensional actual working conditions, obtains a sparse dataset of explosion overpressure fields through numerical simulation. Through data processing and transformation, an overpressure field database is constructed, and an explosion overpressure sparse reconstruction model is established. The model is then trained based on a physical information-constrained multi-module graph neural network, resulting in an explosion overpressure sparse reconstruction model driven by a hybrid data and physics approach. This model possesses high-precision sparse data reconstruction capabilities under noise and strong generalization ability. Furthermore, error evaluation and parameter tuning are performed on the model to improve its reconstruction capability, thus solving the problem of sparse reconstruction of overpressure distribution data under explosion impacts in complex scenarios and at multiple scales.

[0123] This invention also designs a physical enhancement and reconstruction system for explosive overpressure fields oriented towards sparse observations, comprising:

[0124] Model unit, used for pre-training and optimization of the explosive overpressure sparse reconstruction model;

[0125] The reconstruction unit is used to physically enhance and reconstruct the on-site explosion overpressure field using the explosion overpressure sparse reconstruction model.

[0126] The technical solution of the above-mentioned precise analysis system for the distribution pattern of explosion shock waves is similar to the aforementioned precise analysis method for the distribution pattern of explosion shock waves, and will not be repeated here.

[0127] Based on the same technical solution, the present invention also discloses a computer-readable storage medium storing one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform the above-described method for enhancing and reconstructing the physics of an explosion overpressure field for sparse observations.

[0128] 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 configured to be executed by the one or more processors, and the one or more programs include instructions for executing the above-described explosive overpressure field physics enhancement and reconstruction method for sparse observations.

[0129] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied 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.

[0130] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0131] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work 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 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0132] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0133] The above embodiments are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solutions based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. A physical enhancement and reconstruction method for explosion overpressure fields oriented towards sparse observations, characterized in that, This method utilizes a pre-trained and optimized sparse reconstruction model of explosion overpressure 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 identification module, a denoising module, and an output module; The sparse augmentation module achieves upsampling and augmentation of the input sparse observation data by setting a fully connected layer; The feature recognition module is constructed by stacking several graph convolutional layers and a channel attention layer. The channel attention layer includes a pooling layer, a first convolutional layer, a ReLU activation layer and a second convolutional layer. The input of the pooling layer is the result of upsampling augmentation. The output of the second convolutional layer is multiplied by the result of upsampling augmentation after a sigmoid operation. The noisy information obtained by the 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 convolutional layer, a fourth convolutional layer, an LSTM layer, a fifth convolutional layer, and a sixth convolutional layer in series. The hybrid feature extraction and recognition layer includes a seventh to twelfth convolutional layer in series. The input of the third convolutional layer is the noisy information. The noisy information is concatenated with the output of the sixth convolutional layer and used as the input of the seventh convolutional layer. The twelfth convolutional layer outputs the denoised information. The output module includes a series of deconvolutional layers and ReLU activation layers. The input to the deconvolutional layer is the denoised information, and the output of the ReLU activation layer is the reconstructed result.

2. The method according to claim 1, characterized in that, The fully connected layer consists of an input layer, intermediate layers, activation layers, and an output layer. The input layer has 20 neurons, the intermediate layers consist of 6 dense networks with 64, 128, 256, 512, 1024, and 2048 neurons per layer, respectively, and the output layer consists of a single-layer neural network with 2048 neurons. 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 Indicates air density, p It represents the shock wave pressure, and u1, u2, and u3 represent the air velocity components in the transverse, longitudinal, and vertical directions, respectively. δ ij It is the Kronecker function, and E is the total energy contained in a unit volume.

4. The method according to claim 1, characterized in that, Pre-training and optimization of an explosion overpressure sparse reconstruction model are achieved using an explosion overpressure field database containing training and validation datasets; The steps for constructing an explosion overpressure field database include: Based on real-world working conditions across multiple scenarios and dimensions, we use Abaqus numerical simulation to solve the overpressure distribution under explosion impact and obtain a sparse dataset of the explosion overpressure field. After data processing and transformation, an explosion overpressure field database containing training and validation datasets is obtained from the sparse dataset of the explosion overpressure field.

5. The method according to claim 4, characterized in that, The acquisition of the sparse dataset of the explosion overpressure field includes: Acquisition of real-world working conditions in multiple scenarios and dimensions: By setting different explosion source characteristics, different building structures, different node types, and different scale distances, real-world working conditions in multiple scenarios and dimensions can be obtained; Acquisition of sparse dataset of explosion overpressure field: Based on real working conditions in multiple scenarios and dimensions, the overpressure distribution under explosion impact is solved through Abaqus numerical simulation to obtain a sparse dataset of explosion overpressure field.

6. The method according to claim 4, characterized in that, The construction of the explosion overpressure field database includes: Data processing: Data preprocessing, feature extraction, and data standardization are performed on the sparse dataset of the explosion overpressure field to obtain standardized data; Data transformation: The adjacency matrix is ​​used to represent the connection relationship between standardized data nodes, and the features of nodes and edges are combined into a graph. Graph enhancement is achieved through random sampling to obtain a standardized graph storage structure. Overpressure field database construction: The standardized graphical storage structure is hierarchically sampled to construct an explosion overpressure field database containing training and validation datasets.

7. The method according to claim 3, characterized in that, Loss function of the explosion overpressure sparse reconstruction model The design is as follows: ， in, The data error that characterizes and measures the model’s fit to the training dataset; The physical constraint loss term characterizes and measures the deviation of the model from obeying physical laws. The weighting coefficients characterize the physical constraint loss term.

8. A system applying the physical enhancement and reconstruction method for sparse observations of an explosion overpressure field as described in any one of claims 1 to 7, characterized in that, include: Model unit, used for pre-training and optimization of the explosive overpressure sparse reconstruction model; The reconstruction unit is used to physically enhance and reconstruct the on-site explosion overpressure field using the explosion overpressure sparse reconstruction model.

9. A computer-readable storage medium storing one or more programs, said one or more programs comprising instructions, characterized in that, When the instruction is executed by the computing device, it causes the computing device to perform the method as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes 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 configured to be executed by the one or more processors, and the one or more programs include instructions for performing the method as described in any one of claims 1 to 7.

Images