Method for predicting displacement response of walls of adjacent rooms under building implosion
By establishing a sample library of wall displacement responses and training it with a neural network, and employing a multi-part loss function method, the problem of predicting the wall displacement response of adjacent rooms under blasting was solved, achieving rapid and accurate damage assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XIAN MODERN CHEM RES INST
- Filing Date
- 2026-01-19
- Publication Date
- 2026-06-05
AI Technical Summary
Existing technologies struggle to accurately predict the wall displacement response of adjacent rooms in a building under blast attack, especially due to the complex and difficult-to-assess damage effects caused by the reflection, diffraction, and structural influence of the shock wave.
A sample library of wall displacement responses in adjacent rooms under building implosion was established. The wall displacement was predicted by training a neural network. A multi-part loss function, including data loss and physical loss, was adopted to ensure that the prediction results meet the physical conditions.
It enables rapid and accurate prediction of the displacement response of adjacent room walls under building implosion, meets physical constraints, and improves the accuracy and efficiency of damage assessment.
Smart Images
Figure CN122154789A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of damage assessment, and more specifically, to a method for predicting the displacement response of adjacent room walls under a building implosion. Background Technology
[0002] In recent years, frequent high-tech warfare and workplace accidents in some regions have severely damaged buildings, primarily reinforced concrete structures, causing enormous property losses and casualties. With the continuous improvement of precision-guided systems, anti-tank munitions equipped with such systems can directly strike specific areas inside buildings, causing even more severe damage to personnel and equipment in specific rooms and adjacent rooms. Therefore, the analysis of the destructive effects and assessment of the damage to the internal structures of buildings under anti-tank attacks has become an important area of research urgently needed in the military field.
[0003] Regarding the problem of building implosion under blast attack, due to the influence of structures such as floor slabs, shear walls, masonry walls, and doors and windows, the shock wave has problems such as reflection, diffraction, and wall penetration. The internal explosive load is very complex, and the damage effect on rooms near the blast center is difficult to predict. Summary of the Invention
[0004] To overcome at least one deficiency in the prior art, this application provides a method for predicting the displacement response of adjacent room walls under building implosion.
[0005] Firstly, a method for predicting the displacement response of adjacent room walls under a building implosion is provided, including: Extract the charge amount and the corresponding wall displacement response matrix under multiple working conditions, and establish a sample library of wall displacements in adjacent rooms under building implosion; each element in the wall displacement response matrix represents the wall displacement at different sampling points; Data from the sample library of adjacent room wall displacements under building implosion are preprocessed to obtain a preprocessed sample library. Construct a neural network and train it based on the preprocessed sample library to obtain the trained neural network; The target dosage is input into the trained neural network to obtain the wall displacement response matrix corresponding to the target dosage.
[0006] In one embodiment, preprocessing includes sample augmentation, partitioning of the training and validation sets, and data format conversion.
[0007] In one embodiment, the neural network is a fully connected neural network, and the activation function is the tanh function.
[0008] In one embodiment, the total loss function used during training is:
[0009] in, For the total loss function, λ , μ , ν For the weights of the loss function, For data loss, For boundary loss, This refers to the physical loss at the interior point.
[0010] In one embodiment, data loss for:
[0011] in, The number of training samples. Let be the vector dimension of the flattened wall displacement response matrix. For the first i The neural network prediction value for each sample. For the first i The true value of each sample.
[0012] In one embodiment, boundary loss for:
[0013] in, The number of training samples. For the set of boundary locations, The number of boundary points. For the first i Each sample is located at (j,k) The predicted value.
[0014] In one embodiment, interior point physical loss for:
[0015]
[0016]
[0017]
[0018] in, The number of training samples. For the first i The amplitude-constrained residuals of each sample For the first i Smoothness constraint residuals for each sample For the first i The non-negative constrained residuals of each sample , , These are the weighting coefficients. Indicates the first i Neural network prediction values for each sample The root mean square value, For the first i The input drug dosage for each sample. Let be the vector dimension of the flattened wall displacement response matrix. for The discrete Laplace operator, It is the Frobenius norm. For the first i Each sample is located at (j,k) The predicted value, For the first i The maximum position coordinates of each sample along the length of the wall. For the first i The maximum position coordinates of each sample along the width of the wall.
[0019] Secondly, a device for predicting the displacement response of adjacent room walls under a building implosion is provided, comprising: The sample library construction module is used to extract the charge amount and the corresponding wall displacement response matrix under multiple working conditions, and to establish a sample library of wall displacements in adjacent rooms under building implosion; each element in the wall displacement response matrix is the wall displacement at different sampling points; The preprocessing module is used to preprocess the data in the sample library of displacement of adjacent room walls under building implosion to obtain the preprocessed sample library; The network construction and training module is used to construct neural networks and train them based on a preprocessed sample library to obtain the trained neural network. The prediction module is used to input the target drug amount into the trained neural network to obtain the wall displacement response matrix corresponding to the target drug amount.
[0020] Thirdly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the aforementioned method for predicting the displacement response of adjacent room walls under building implosion.
[0021] Fourthly, a computer program product is provided, including a computer program / instruction, which, when executed by a processor, implements the aforementioned method for predicting the displacement response of adjacent room walls under building implosion.
[0022] Compared with existing technologies, this application has the following advantages: The method for predicting the displacement response of adjacent room walls under building implosion in this application establishes a neural network for the displacement response of adjacent room walls based on a sample library of munition and adjacent room wall displacements. It employs a loss function containing multiple components. Data loss is mainly learned through data supervision of the neural network in the database, while physical loss mainly includes fixed boundary conditions and interior point physical losses, ensuring that the prediction results meet physical conditions. This application can achieve rapid prediction of the displacement response of adjacent room walls under building implosion. Attached Figure Description
[0023] This application can be better understood by referring to the description given below in conjunction with the accompanying drawings, which, together with the detailed description below, are incorporated in and form part of this specification. In the drawings: Figure 1 A flowchart illustrating a method for predicting the displacement response of adjacent room walls under a building implosion is shown. Figure 2 The diagram shows the sampling method of the wall displacement response; Figure 3 It shows W =5kg wall displacement response prediction results diagram Detailed Implementation Exemplary embodiments of the present application will be described below with reference to the accompanying drawings. For clarity and brevity, not all features of the actual embodiments are described in the specification. However, it should be understood that many embodiment-specific decisions can be made in the development of any such actual embodiment to achieve the developer’s specific objectives, and these decisions may vary as the embodiments differ.
[0024] It should also be noted that, in order to avoid obscuring this application with unnecessary details, only the device structure closely related to the solution of this application is shown in the accompanying drawings, while other details that are not closely related to this application are omitted.
[0025] It should be understood that this application is not limited to the described embodiments by virtue of the following description with reference to the accompanying drawings. In this document, embodiments may be combined with each other, features may be substituted or borrowed between different embodiments, and one or more features may be omitted in one embodiment, where feasible.
[0026] Neural networks, due to their end-to-end advantage, can directly establish the correlation between input and output parameters, making them crucial in numerous fields. Addressing the issues in predicting damage from building implosions, this application proposes a method for predicting the displacement response of walls in adjacent rooms under implosion, using the blast center parameters as input and the damage characteristic parameters of rooms near the blast center as output, thus establishing a correlation between the two. Therefore, to establish a predictive model for the internal structural response of rooms near a building implosion and improve prediction efficiency, this application proposes a method for predicting the displacement response of walls in adjacent rooms under a building implosion, taking building implosion under blast attack as a typical scenario. This method provides a way to predict the internal structural response of buildings, enabling the establishment of a neural network for the displacement response of walls in adjacent rooms based on a sample library of munition and adjacent room wall displacements, achieving rapid prediction of the displacement response of walls in adjacent rooms under building implosion.
[0027] This application provides a method for predicting the displacement response of adjacent room walls under a building implosion. Figure 1 A flowchart illustrating a method for predicting the displacement response of adjacent room walls under a building implosion is shown. (See attached diagram) Figure 1 The method mainly includes the following steps: Step S1: Extract the charge amount and the corresponding wall displacement response matrix under multiple working conditions, and establish a sample library of wall displacements in adjacent rooms under building implosion; each element in the wall displacement response matrix represents the wall displacement at different sampling points.
[0028] Typical operating conditions (dosage) are obtained through experiments or simulations. W Displacement response data of adjacent room walls. U It is a two-dimensional matrix with a size of M×N, where M is the number of sampling points along the length of the wall and N is the number of sampling points along the width of the wall. The sampling method is uniform sampling, and the matrix element value is the displacement value at the corresponding sampling point.
[0029] For the sidewall displacement response results under a certain working condition, assuming that... x , y Ten sampling points were set in each direction. Figure 2 The diagram shows the sampling method for wall displacement response and the recorded dosage. W Establish the displacement response matrix based on the displacement at each sample point. U , W When =1kg:
[0030] Extract the dosage of the drug under all operating conditions W and displacement response matrix U Establish a sample library of wall displacements in adjacent rooms under building implosion.
[0031] Step S2: Preprocess the data in the sample library of displacement of adjacent room walls under building implosion to obtain the preprocessed sample library.
[0032] Sample augmentation: Using spline interpolation, more samples are generated based on known data. W Use the training data provided to ensure data continuity; Training and validation set partitioning: The total dataset is randomly divided into training and validation sets in an 8:2 ratio. A random seed is set to ensure reproducibility of results. The training and validation sets are recorded. W distributed; Data format conversion: M × N The U matrix flattened as Q dimensional vector ( Q 维数 = M × N This converts the input data into the dlarray format required by deep learning frameworks. Specifically, it can convert 10×10 arrays... U The matrix is flattened into a 100-dimensional vector.
[0033] Step S3: Construct a neural network and train it based on the preprocessed sample library to obtain the trained neural network.
[0034] Construct 1 input ( W value), Q (100) Output (flattened) U A fully connected neural network (matrix) with the activation function defined as the tanh function to ensure output smoothness.
[0035] The training process includes: Training parameter initialization: Set the maximum number of training epochs, learning rate, and initialize the momentum parameters of the Adam optimizer, etc. Here, you can set the maximum number of training epochs to 5000 and the learning rate to 0.001.
[0036] Forward propagation: This involves passing the training set... W The values are input into the neural network, and the predictions are obtained through calculations by each layer of the network. U matrix( Q Dimensional output, Q =100); Loss function calculation: Establish a loss function that includes multiple parts of loss. Data loss is mainly learned by supervising the neural network through data in the displacement sample library, and physical loss mainly includes fixed boundary conditions and interior point physical loss. The total loss function used is:
[0037] in, For the total loss function, λ , μ , νFor the weights of the loss function, For data loss, For boundary loss, This refers to the physical loss at the interior point.
[0038] Data loss for:
[0039] in, The number of training samples. Let be the vector dimension of the flattened wall displacement response matrix. For the first i The neural network prediction values of each sample (two-dimensional matrix flattened into vectors). For the first i The true value of each sample.
[0040] Boundary loss for:
[0041] in, The number of training samples. For the set of boundary locations, The number of boundary points. For the first i Each sample is located at (j,k) The predicted value.
[0042] Interior point physical loss for:
[0043]
[0044]
[0045]
[0046] in, The number of training samples. For the first i The amplitude-constrained residuals of each sample For the first i Smoothness constraint residuals for each sample For the first i The non-negative constrained residuals of each sample , , Here, are the weighting coefficients. , , , Indicates the first iNeural network prediction values for each sample The root mean square value, , For the first i The input drug dosage for each sample, with 0.5 as an empirical proportionality coefficient. Let be the vector dimension of the flattened wall displacement response matrix. for The discrete Laplace operator, It is the Frobenius norm. For the first i Each sample is located at (j,k) The predicted value, For the first i The maximum position coordinates of each sample along the length of the wall. For the first i The maximum position coordinates of each sample along the width of the wall.
[0047] Backpropagation and parameter update: Calculate the gradient of the total loss with respect to each parameter of the network, use the Adam optimizer to update the network weights and biases, and record the loss values for monitoring the training process; Validation and early stopping mechanism: Calculate the validation set loss after each epoch. If the validation loss does not improve for 500 consecutive epochs, trigger early stopping and save the network parameters with the minimum validation loss as the optimal model.
[0048] Real-time monitoring and visualization: Real-time plotting of loss curves and dynamic updates of prediction results for training and validation sets.
[0049] Model evaluation: Calculate the final training loss and validation loss, draw a comparison map of the real field and the predicted field, and an error distribution heatmap, etc. Model saving: Saves trained network parameters, training history, and other relevant information.
[0050] Step S4: Input the target dosage into the trained neural network to obtain the wall displacement response matrix corresponding to the target dosage.
[0051] Invoke the pre-trained neural network and input the drug dosage. W When =5kg, the output is obtained U This enables displacement field prediction.
[0052]
[0053] Figure 3 It shows W =5kg wall displacement response prediction results diagram.
[0054] In this embodiment, a neural network for the displacement response of adjacent room walls is established based on a sample database of ammunition and adjacent room wall displacements. A loss function containing multiple components is employed. Data loss is primarily learned through data supervision of the neural network from the database, while physical loss mainly includes fixed boundary conditions and in-point physical losses, ensuring that the prediction results meet the physical conditions. This application enables rapid prediction of the displacement response of adjacent room walls under building implosion.
[0055] Based on the same inventive concept as the method for predicting the displacement response of adjacent room walls under a building implosion, this embodiment also provides a corresponding device for predicting the displacement response of adjacent room walls under a building implosion, including: The sample library construction module is used to extract the charge amount and the corresponding wall displacement response matrix under multiple working conditions, and to establish a sample library of wall displacements in adjacent rooms under building implosion; each element in the wall displacement response matrix is the wall displacement at different sampling points; The preprocessing module is used to preprocess the data in the sample library of displacement of adjacent room walls under building implosion to obtain the preprocessed sample library; The network construction and training module is used to construct neural networks and train them based on a preprocessed sample library to obtain the trained neural network. The prediction module is used to input the target drug amount into the trained neural network to obtain the wall displacement response matrix corresponding to the target drug amount.
[0056] The device for predicting the displacement response of adjacent room walls under building implosion in this embodiment has the same inventive concept as the method for predicting the displacement response of adjacent room walls under building implosion described above. Therefore, the specific implementation of the device can be found in the embodiment section of the method for predicting the displacement response of adjacent room walls under building implosion described above, and its technical effects correspond to the technical effects of the above method, so it will not be repeated here.
[0057] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the above-described method for predicting the displacement response of adjacent room walls under building implosion.
[0058] This application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the above-described method for predicting the displacement response of adjacent room walls under building implosion.
[0059] The above descriptions are merely various embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for predicting the displacement response of adjacent room walls under a building implosion, characterized in that, include: Extract the charge amount and the corresponding wall displacement response matrix under multiple working conditions to establish a sample library of wall displacements in adjacent rooms under building implosion; each element in the wall displacement response matrix represents the wall displacement at different sampling points; The data in the sample library of adjacent room wall displacements under the building implosion are preprocessed to obtain the preprocessed sample library. A neural network is constructed, and the neural network is trained based on the preprocessed sample library to obtain a trained neural network; The target dosage is input into the trained neural network to obtain the wall displacement response matrix corresponding to the target dosage.
2. The method as described in claim 1, characterized in that, The preprocessing includes sample augmentation, division of training and validation sets, and data format conversion.
3. The method as described in claim 1, characterized in that, The neural network is a fully connected neural network with the tanh function as the activation function.
4. The method as described in claim 1, characterized in that, The total loss function used during training is: in, For the total loss function, λ , μ , ν For the weights of the loss function, For data loss, For boundary loss, This refers to the physical loss at the interior point.
5. The method as described in claim 4, characterized in that, The data loss for: in, The number of training samples. Let be the vector dimension of the flattened wall displacement response matrix. For the first i The neural network prediction value for each sample. For the first i The true value of each sample.
6. The method as described in claim 4, characterized in that, The boundary loss for: in, The number of training samples. For the set of boundary locations, The number of boundary points. For the first i Each sample is located at (j,k) The predicted value.
7. The method as described in claim 4, characterized in that, The physical loss at the interior point for: in, The number of training samples. For the first i The amplitude-constrained residuals of each sample For the first i Smoothness constraint residuals for each sample For the first i The non-negative constrained residuals of each sample , , These are the weighting coefficients. Indicates the first i Neural network prediction values for each sample The root mean square value, For the first i The input drug dosage for each sample. Let be the vector dimension of the flattened wall displacement response matrix. for The discrete Laplace operator, It is the Frobenius norm. For the first i Each sample is located at (j,k) The predicted value, For the first i The maximum position coordinates of each sample along the length of the wall. For the first i The maximum position coordinates of each sample along the width of the wall.
8. A device for predicting the displacement response of adjacent room walls under a building implosion, characterized in that, include: The sample library construction module is used to extract the amount of explosive under multiple working conditions and the wall displacement response matrix corresponding to the amount of explosive, and to establish a sample library of wall displacement of adjacent rooms under building implosion; each element in the wall displacement response matrix is the wall displacement at different sampling points; The preprocessing module is used to preprocess the data in the sample library of displacement of adjacent room walls under the building implosion to obtain the preprocessed sample library; The network construction and training module is used to construct a neural network and train the neural network based on the preprocessed sample library to obtain the trained neural network. The prediction module is used to input the target dosage into the trained neural network to obtain the wall displacement response matrix corresponding to the target dosage.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method for predicting the displacement response of adjacent room walls under building implosion as described in any one of claims 1-7.
10. A computer program product, characterized in that, Includes a computer program / instruction, which, when executed by a processor, implements the method for predicting the displacement response of adjacent room walls under building implosion as described in any one of claims 1-7.