A regional ground motion field inversion method, device, equipment, medium and product

Abstract

The application discloses a regional seismic ground motion field inversion method, device, equipment, medium and product, relates to the technical field of earthquake engineering, and comprises the following steps: obtaining structural response monitoring data of a building structure in a real earthquake; pre-processing the structural response monitoring data to obtain matrix data; performing feature learning and nonlinear regression processing on the matrix data through a pre-trained neural network model to obtain seismic ground motion parameters, wherein the neural network model comprises a channel attention mechanism, a convolutional neural network and a long short-term memory network; and generating a continuously distributed regional seismic ground motion field through spatial interpolation or data fusion based on seismic ground motion parameters corresponding to multiple different positions. The application provides intuitive and reliable data support for rapid post-earthquake evaluation of spatial distribution of earthquake disasters and guidance of rescue force deployment.

Description

Technical Field

[0001] This application relates to the field of earthquake engineering technology, and in particular to a method, apparatus, equipment, medium and product for regional seismic ground motion field inversion. Background Technology

[0002] In earthquake-prone and densely populated areas, rapidly obtaining continuous and accurate seismic ground motion fields after an earthquake is crucial for emergency response, disaster assessment, and rescue decision-making.

[0003] Currently, methods for obtaining seismic ground motion fields are mainly divided into three categories: I. Monitoring methods based on dedicated seismic networks This method directly records ground motion using specialized equipment such as seismographs and strong-motion meters, and constructs the ground motion field through spatial interpolation. However, the construction and maintenance costs of dedicated seismic networks are high, and the distribution of stations is often sparse. In areas with sparse networks, the ground motion field obtained solely through interpolation of network data has significant blind spots and considerable uncertainty. Furthermore, this method cannot utilize the building structure monitoring equipment already widely installed in cities, resulting in a waste of data resources.

[0004] II. Numerical Simulation Methods Based on Physical Models This method simulates the propagation process of seismic motion using wave equations by setting source parameters and subsurface medium structure. Numerical simulation can fill gaps in seismic networks, but its results heavily depend on the subjective selection of model parameters (such as fault slip distribution) and simplified assumptions about the subsurface medium structure, leading to significant uncertainties in the simulation results. Furthermore, the computational demands of numerical simulation are enormous, making it difficult to complete within minutes to tens of minutes after an earthquake, thus failing to meet the timeliness requirements of emergency response.

[0005] III. Ground Motion Prediction Equation Method Based on Empirical Formulas This method rapidly calculates seismic intensity using empirical attenuation relationships based on parameters such as magnitude, epicentral distance, and site conditions. While simple and efficient, this method oversimplifies site effects and focal mechanisms, exhibiting significant regional dependence. Its prediction accuracy and reliability are difficult to guarantee, especially under complex geological conditions where errors are substantial.

[0006] In summary, existing methods struggle to simultaneously balance cost, coverage, accuracy, and efficiency. How to leverage existing structural health monitoring data from urban buildings to achieve low-cost, high-precision, near-real-time regional seismic field inversion is a pressing technical challenge in this field. Summary of the Invention

[0007] The purpose of this application is to provide a method, device, equipment, medium, and product for regional seismic field inversion, which enables direct utilization of monitoring equipment already installed on existing urban buildings, significantly reducing data acquisition costs and expanding data coverage; it achieves end-to-end, high-precision, near-real-time intelligent inversion from structural response to seismic parameters, fully utilizing the channel attention mechanism to enhance important features, the ability of CNN to extract local patterns, and the ability of LSTM to model global temporal dependencies; and it expands information from sparse monitoring points to continuous spatial fields, providing intuitive and reliable data support for rapid post-earthquake assessment of the spatial distribution of earthquake disasters and guidance for the deployment of rescue forces.

[0008] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for inverting regional seismic ground motion fields, including: Acquire structural response monitoring data of building structures during real earthquakes; The structural reaction monitoring data is preprocessed to obtain matrix data; Seismic motion parameters are obtained by performing feature learning and nonlinear regression on the matrix data using a pre-trained neural network model. The neural network model includes a channel attention mechanism, a convolutional neural network, and a long short-term memory network. Based on the ground motion parameters corresponding to multiple different locations, a continuously distributed regional ground motion field is generated through spatial interpolation or data fusion.

[0009] Optionally, the preprocessing of the structural response monitoring data to obtain matrix data specifically includes: The structural reaction monitoring data were resampled to ensure a consistent sampling frequency. The resampled data is truncated or padded with zeros to ensure consistent data length. Calculate the mean and standard deviation of the data after truncation or zero-padding, and then standardize them; The standardized time-series data is sliced ​​using a sliding window to obtain matrix data.

[0010] Optionally, the training process of the pre-trained neural network model includes: Obtain structural response monitoring data and corresponding seismic motion parameters to form sample data pairs; The structural response monitoring data in the sample data pair is preprocessed and converted into matrix data to obtain a training sample pair consisting of matrix data and corresponding ground motion parameters. The training sample pairs are divided into a training set, a validation set, and a test set; Construct a neural network model that integrates channel attention mechanism, convolutional neural network and long short-term memory network; The neural network model is trained by using matrix data from the training set as input and corresponding ground motion parameters as labels. The network hyperparameters are optimized using a validation set, and the inversion accuracy is evaluated using a test set to obtain the trained neural network model.

[0011] Optionally, the neural network model includes three convolutional layers with a kernel size of 6. The number of kernels in the three convolutional layers are 18, 36, and 64, respectively. Each convolutional layer is followed by a ReLU function for non-linear processing and a pooling layer with a pooling size of 2 for feature compression.

[0012] Optionally, the ground motion parameters include peak ground acceleration, peak ground velocity, and peak ground displacement.

[0013] Optionally, the step of generating a continuously distributed regional seismic field based on the seismic motion parameters corresponding to multiple different locations through spatial interpolation or data fusion specifically includes: Obtain the ground motion parameters corresponding to multiple different locations to form a discrete point dataset. Each discrete point contains the location coordinates and its corresponding ground motion parameter values. Based on the discrete point dataset, for any spatial location within the target area, the estimated value of the seismic motion parameters at that location is calculated using a spatial interpolation method or a data fusion algorithm. Iterate through all spatial locations within the target area to generate a continuously distributed regional seismic field.

[0014] Secondly, this application provides a regional seismic field inversion device, comprising: The data acquisition module is used to acquire structural response monitoring data of building structures during real earthquakes; The data preprocessing module is used to preprocess the structural reaction monitoring data to obtain matrix data; The seismic motion parameter inversion module is used to perform feature learning and nonlinear regression processing on the matrix data through a pre-trained neural network model to obtain seismic motion parameters. The neural network model includes a channel attention mechanism, a convolutional neural network, and a long short-term memory network. The seismic field generation module is used to generate a continuously distributed regional seismic field based on the seismic parameters corresponding to multiple different locations through spatial interpolation or data fusion.

[0015] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the regional seismic field inversion method described in any one of the above.

[0016] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the regional seismic field inversion method described in any one of the above.

[0017] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the regional seismic field inversion method described in any one of the above.

[0018] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method, apparatus, equipment, medium, and product for regional seismic ground motion field inversion. By acquiring structural response monitoring data of building structures during real earthquakes, it solves the problems of high data acquisition costs and sparse deployment caused by traditional methods relying on dedicated seismic networks or manually deployed sensors. It enables direct utilization of monitoring equipment already installed on existing urban buildings, significantly reducing data acquisition costs and expanding data coverage. Through preprocessing of the structural response monitoring data to obtain matrix data, and using a pre-trained neural network model to perform feature learning and nonlinear regression processing on the matrix data, seismic ground motion parameters are obtained. This addresses the problems of low accuracy and low efficiency in numerical simulation calculations of existing ground motion prediction equations. Addressing the issue of significant uncertainty in network interpolation results, this system achieves end-to-end, high-precision, near-real-time intelligent inversion from structural response to seismic ground motion parameters. It fully leverages the channel attention mechanism to enhance important features, the CNN's ability to extract local patterns, and the LSTM's ability to model global temporal dependencies. Based on seismic ground motion parameters corresponding to multiple different locations, it generates a continuously distributed regional seismic ground motion field through spatial interpolation or data fusion. This solves the problem that discrete point seismic ground motion parameters cannot directly serve regional earthquake damage assessment and emergency decision-making, realizing information expansion from sparse monitoring points to a continuous spatial field. It provides intuitive and reliable data support for rapidly assessing the spatial distribution of earthquake disasters after an earthquake and guiding the deployment of rescue forces. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is an application environment diagram of a regional seismic ground motion field inversion method according to an embodiment of this application; Figure 2 A schematic flowchart illustrating a regional seismic ground motion field inversion method provided in an embodiment of this application; Figure 3 This is a schematic diagram of the network structure of a neural network model in one embodiment of this application; Figure 4 This is a schematic diagram of a structural analysis model in one embodiment of this application; Figure 5 This is an acceleration response spectrum of a seismic motion record used to generate training data in one embodiment of this application; Figure 6 This is a comparison chart of the inversion results of the neural network model on the test set and the actual values ​​in one embodiment of this application; Figure 7 This is a schematic diagram of a seismic field generated by a numerical simulation method in one embodiment of this application; Figure 8 This is a comparison diagram of the seismic field inverted by the neural network model in one embodiment of this application and the actual value; Figure 9 A functional module diagram of a regional seismic field inversion device provided in another embodiment of this application; Figure 10 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] The regional seismic field inversion method provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on other servers. Terminal 102 can send the building structure monitoring data to be processed to server 104. After receiving the structural response monitoring data from a real earthquake, server 104 preprocesses the structural response monitoring data to obtain matrix data; then, it uses a pre-trained neural network model to perform feature learning and nonlinear regression processing on the matrix data to obtain the seismic motion parameters at the location of the monitoring data; based on the seismic motion parameters corresponding to multiple different locations, it generates a continuously distributed regional seismic field through spatial interpolation or data fusion. Server 104 can feed back the generated regional seismic field to terminal 102 to provide emergency command centers with information for earthquake damage assessment and rescue decisions. In addition, in some embodiments, the regional ground motion field inversion method can also be implemented by the server 104 or the terminal 102 separately. For example, the terminal 102 can directly perform regional ground motion field inversion processing on the acquired structural response monitoring data, or the server 104 can obtain pre-stored historical earthquake monitoring data from the data storage system and perform regional ground motion field inversion processing on the data.

[0024] The terminal 102 may be, but is not limited to, various structural health monitoring data acquisition devices, data aggregation nodes, desktop computers, laptops, smartphones, tablets, IoT devices, or emergency command terminals. The structural health monitoring data acquisition devices may include, but are not limited to, accelerometers and displacement sensors installed on the building structure. The server 104 may be implemented using a standalone server or a server cluster composed of multiple servers, or it may be a cloud server or an edge computing node.

[0025] In one exemplary embodiment, such as Figure 2 As shown, a method for inverting regional seismic ground motion fields is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 201 to 204. Wherein: Step 201: Obtain structural response monitoring data of the building structure during a real earthquake; Specifically, when a real earthquake occurs, monitoring equipment such as acceleration sensors deployed on the structures of various buildings within the target area record the acceleration response time history data of the structures during the vibration process. Server 104 can directly acquire this data from these monitoring devices or read it from the data storage system. The core advantage of this application is that it directly utilizes the monitoring equipment already installed on existing buildings, eliminating the need for an additional dedicated seismic network, thereby significantly reducing costs and increasing data coverage density.

[0026] Step 202: Preprocess the structural reaction monitoring data to obtain matrix data; Because raw monitoring data may have issues such as inconsistent sampling frequencies, inconsistent data lengths, and differences in units, preprocessing is required to adapt it to the neural network model. As an optional implementation method, preprocessing specifically includes: The structural response monitoring data were resampled to ensure a consistent sampling frequency. The resampled data is truncated or padded with zeros to ensure consistent data length. Calculate the mean and standard deviation of the data after truncation or zero-padding, and then standardize them; The standardized time-series data is sliced ​​using a sliding window to obtain matrix data.

[0027] Specifically, resampling unifies the sampling frequency of all data, such as 100Hz. Truncation or zero-padding is used to standardize the data length; for example, retaining 25 seconds of data before and after each sample's peak point. If the data length is insufficient, zeros are added; if it exceeds the limit, it is truncated, ensuring each sample contains 5000 sampling points. The standardized calculation formula is: ; In the formula, This represents standardized time-series data. This indicates data that has been truncated or padded with zeros. This represents the mean. This represents the standard deviation. Standardization makes the data conform to a standard normal distribution, which speeds up the convergence of neural networks.

[0028] The window size for slicing window sampling is s (which can be 10), the overlap ratio of adjacent windows is v (which can be 0.5), the number of rows of the generated matrix is ​​the total number of sampling points divided by (s×v) and then minus 1, that is, 5000÷(s×v)-1=999, and the number of columns is s×v=5.

[0029] Step 203: The matrix data is processed by feature learning and nonlinear regression through a pre-trained neural network model to obtain the seismic motion parameters. The neural network model includes a channel attention mechanism, a convolutional neural network, and a long short-term memory network. Specifically, the neural network model used in step 203 is pre-trained, and its network structure diagram is shown below. Figure 3 As shown, this neural network model integrates channel attention mechanism, convolutional neural network (CNN), and long short-term memory network (LSTM).

[0030] The CNN part contains three convolutional layers with a kernel size of 6 and the number of kernels being 18, 36, and 64 respectively. Each convolutional layer is followed by a ReLU activation function and a max pooling layer of size 2 to extract local features from the input matrix.

[0031] The channel attention mechanism performs channel recalibration on the feature maps extracted by CNN, automatically enhancing important feature channels.

[0032] LSTM layers are used to capture global temporal dependencies and remember long-term historical information.

[0033] Finally, regression is performed through multiple fully connected layers to output ground motion parameters, including peak ground acceleration (PGA), peak ground velocity (PGV), and peak ground displacement (PGD).

[0034] As an optional implementation method, the training process of a neural network model includes: Obtain structural response monitoring data and corresponding seismic motion parameters to form sample data pairs; The structural response monitoring data in the sample data pairs are preprocessed and converted into matrix data to obtain training sample pairs consisting of matrix data and corresponding ground motion parameters. The training sample pairs are divided into a training set, a validation set, and a test set; Construct a neural network model that integrates channel attention mechanism, convolutional neural network and long short-term memory network; The neural network model is trained by using matrix data from the training set as input and corresponding ground motion parameters as labels. The network hyperparameters are optimized using the validation set, and the inversion accuracy is evaluated using the test set to obtain the trained neural network model.

[0035] Specifically, since structural response data from real historical earthquakes are often limited, this embodiment uses numerical simulation methods to generate a large amount of sample data. Specifically, a multi-degree-of-freedom lumped mass shear (MDOF) model of a 5-story masonry structure is established, and the schematic diagram of the structural analysis model is shown below. Figure 4 As shown in Table 1, the mechanical parameters of the MDOF model are as follows.

[0036] Table 1 Mechanical parameters of the MDOF model

[0037] Six hundred ground motion records were selected, each amplitude-modulated five times, with a PGA range of 0.05g to 0.80g. The acceleration response spectra of the amplitude-modulated ground motion records used to generate training data are as follows: Figure 5 As shown, using amplitude-modulated ground motion as excitation, nonlinear dynamic analysis was performed on the MDOF model to calculate the acceleration response of each layer of the structure as structural response monitoring data. At the same time, the PGA, PGV, and PGD of the input ground motion were recorded as corresponding labels, thus forming a large number of sample data pairs.

[0038] Preprocessing (resampling, truncation / zero padding, standardization, windowing) is performed on the structural response monitoring data in the sample data pairs to obtain matrix data; the corresponding ground motion parameters (PGA, PGV, PGD) are not included in the preprocessing and are directly used as labels. This forms training sample pairs consisting of matrix data and corresponding ground motion parameters.

[0039] The training samples were randomly divided into training, validation, and test sets in a ratio of 8:1:1.

[0040] Build as Figure 3 The diagram shows a neural network model that integrates channel attention, CNN, and LSTM. The CNN part contains three convolutional layers (kernel size 6, number of kernels 18 / 36 / 64), each followed by a ReLU and a pooling layer (pooling size 2). The LSTM layer is followed by a fully connected layer for regression output.

[0041] Using a training set with matrix data as input and corresponding seismic motion parameters as labels, a neural network is trained using the Adam optimizer and the root mean square error (RMSE) as the loss function. The validation set is used to adjust hyperparameters (such as learning rate and batch size) to prevent overfitting. Finally, the inversion accuracy is evaluated on the test set using the Mean Arctangent Absolute Percentage Error (MAAPE) as the metric. The formula for calculating MAAPE is as follows: ; In the formula, This represents the true value of the i-th sample (i.e., the actual ground motion parameters, such as PGA, PGV, or PGD). This represents the predicted value of the i-th sample (i.e., the ground motion parameters obtained by inversion from the neural network model). This represents the total number of samples. A comparison of the neural network model's inversion results on the test set with the actual values ​​is shown below. Figure 6 As shown, the neural network model has high inversion accuracy.

[0042] Step 204: Based on the ground motion parameters corresponding to multiple different locations, a continuously distributed regional ground motion field is generated through spatial interpolation or data fusion.

[0043] Since step S203 obtains seismic motion parameters at discrete building locations, while emergency response requires continuous field information for the entire area, spatial expansion is necessary. As an optional implementation method, the process of generating a continuously distributed regional seismic field specifically includes: Obtain the ground motion parameters corresponding to multiple different locations to form a discrete point dataset. Each discrete point contains the location coordinates and its corresponding ground motion parameter values. Based on a discrete point dataset, for any spatial location within the target area, the estimated value of the seismic motion parameters at that location is calculated using spatial interpolation methods or data fusion algorithms. Iterate through all spatial locations within the target area to generate a continuously distributed regional seismic field.

[0044] Specifically, a discrete point dataset is constructed by acquiring the building coordinates and corresponding PGA, PGV, and PGD values ​​of all collected data within the target area. Then, a spatial interpolation method is selected, such as inverse distance weighted interpolation or kriging interpolation. Taking inverse distance weighted interpolation as an example, for any point to be estimated within the target area, its seismic motion parameter value is obtained by weighting the parameter values ​​of neighboring known points inversely proportional to the distance. Finally, the entire target area is traversed at a certain resolution (e.g., a 500m × 500m grid), and the estimated seismic motion parameters for each grid point are calculated, thereby generating a continuously distributed regional seismic field. The generated seismic field can visually display the spatial distribution of seismic intensity, providing crucial data support for post-earthquake emergency command, rescue force deployment, and disaster assessment.

[0045] Implementing steps 201 to 204 addresses the problems of high data acquisition costs and sparse deployment caused by traditional methods relying on dedicated seismic networks or manually deployed sensors. It enables direct utilization of monitoring equipment already installed on existing urban buildings, significantly reducing data acquisition costs and expanding data coverage. It also solves the problems of low accuracy in existing ground motion prediction equations, low efficiency in numerical simulation, and large uncertainty in network interpolation results. This achieves end-to-end, high-precision, near-real-time intelligent inversion from structural response to ground motion parameters, fully utilizing the channel attention mechanism to enhance important features, the CNN's ability to extract local patterns, and the LSTM's ability to model global temporal dependencies. Furthermore, it addresses the issue that discrete ground motion parameters cannot directly serve regional earthquake damage assessment and emergency decision-making, achieving information expansion from sparse monitoring points to a continuous spatial field. This provides intuitive and reliable data support for rapid post-earthquake assessment of the spatial distribution of earthquake disasters and guidance for the deployment of rescue forces.

[0046] To verify the generalization ability of the regional seismic field inversion method provided in this application, a virtual earthquake scenario (Tangshan area, magnitude 6.6) was generated using a hybrid simulation method. Specifically, the epicenter was set at 118°11′E, 39°38′N, with a fault strike of 55°, dip of 90°, and a grid resolution of 500m. The finite difference method was used for the low-frequency portion, and the random finite fault method was used for the high-frequency portion. The seismic field was generated by superimposing the data at 1Hz. A schematic diagram of the seismic field generated based on the numerical simulation method is shown below. Figure 7 As shown. From Figure 7 It can be seen that the simulated seismic field exhibits an elliptical attenuation characteristic centered on the epicenter, which is consistent with the typical seismic motion distribution pattern.

[0047] Then, based on the response data of multiple building structures in this scenario, a trained neural network model was used to invert the ground motion field. A comparison between the ground motion field inverted by the neural network model and the actual values ​​is shown below. Figure 8 As shown in the figure. The results show that the seismic field inverted by the regional seismic field inversion method provided in this application embodiment is highly consistent with the actual seismic field in terms of spatial distribution trend, with small error, proving that the regional seismic field inversion method provided in this application embodiment has good generalization ability and practical application value.

[0048] This application also provides an application scenario in which the aforementioned regional seismic ground motion field inversion method is applied. Specifically, the regional seismic ground motion field inversion method provided in this embodiment can be applied in post-earthquake emergency response scenarios. Monitoring data from building structures is collected from various monitoring nodes to a regional data center or cloud platform. After data cleaning and preprocessing, it enters the deep learning model inference chain to obtain the seismic ground motion parameters at each location, and finally generates a continuous seismic ground motion field, which is then used in the downstream earthquake damage assessment and emergency decision-making stages. The regional seismic ground motion field inversion method provided in this embodiment belongs to the intelligent inversion stage in the content processing of structural monitoring data. Specifically, in the process of processing post-earthquake monitoring data, seismic ground motion parameters can be automatically inverted based on a pre-trained neural network model, and then a regional seismic ground motion field can be generated through spatial interpolation, providing key reference information for the deployment of emergency rescue forces and the allocation of disaster relief resources.

[0049] Based on the same inventive concept, this application also provides a regional ground motion field inversion device for implementing the aforementioned regional ground motion field inversion method. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more regional ground motion field inversion device embodiments provided below can be found in the limitations of the regional ground motion field inversion method described above, and will not be repeated here.

[0050] In one exemplary embodiment, such as Figure 9As shown, a regional seismic ground motion field inversion device is provided, comprising: The data acquisition module is used to acquire structural response monitoring data of building structures during real earthquakes; The data preprocessing module is used to preprocess the structural response monitoring data to obtain matrix data; The seismic motion parameter inversion module is used to perform feature learning and nonlinear regression processing on matrix data through a pre-trained neural network model to obtain seismic motion parameters. The neural network model includes a channel attention mechanism, a convolutional neural network, and a long short-term memory network. The seismic field generation module is used to generate a continuously distributed regional seismic field based on seismic parameters corresponding to multiple different locations, through spatial interpolation or data fusion.

[0051] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 10 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs in the non-volatile storage media to run. The database stores structural response monitoring data, training sample data, seismic motion parameters, and pre-trained neural network model parameters. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a regional seismic field inversion method.

[0052] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0053] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0054] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0055] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0056] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0057] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0058] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0059] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0060] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for inverting regional seismic ground motion fields, characterized in that, The regional seismic field inversion method includes: Acquire structural response monitoring data of building structures during real earthquakes; The structural reaction monitoring data are preprocessed to obtain matrix data; Seismic motion parameters are obtained by performing feature learning and nonlinear regression on the matrix data using a pre-trained neural network model. The neural network model includes a channel attention mechanism, a convolutional neural network, and a long short-term memory network. Based on the ground motion parameters corresponding to multiple different locations, a continuously distributed regional ground motion field is generated through spatial interpolation or data fusion.

2. The regional seismic ground motion field inversion method according to claim 1, characterized in that, The preprocessing of the structural response monitoring data to obtain matrix data specifically includes: The structural reaction monitoring data were resampled to ensure a consistent sampling frequency. The resampled data is truncated or padded with zeros to ensure consistent data length. Calculate the mean and standard deviation of the data after truncation or zero-padding, and then standardize them; The standardized time-series data is sliced ​​using a sliding window to obtain matrix data.

3. The regional seismic ground motion field inversion method according to claim 1, characterized in that, The training process of the pre-trained neural network model includes: Obtain structural response monitoring data and corresponding seismic motion parameters to form sample data pairs; The structural response monitoring data in the sample data pair is preprocessed and converted into matrix data to obtain a training sample pair consisting of matrix data and corresponding ground motion parameters. The training sample pairs are divided into a training set, a validation set, and a test set; Construct a neural network model that integrates channel attention mechanism, convolutional neural network and long short-term memory network; The neural network model is trained by using matrix data from the training set as input and corresponding ground motion parameters as labels. The network hyperparameters are optimized using a validation set, and the inversion accuracy is evaluated using a test set to obtain the trained neural network model.

4. The regional seismic field inversion method according to claim 3, characterized in that, The neural network model contains three convolutional layers with a kernel size of 6. The number of kernels in the three convolutional layers are 18, 36, and 64, respectively. Each convolutional layer is followed by a ReLU function for non-linear processing and a pooling layer with a pooling size of 2 for feature compression.

5. The regional seismic ground motion field inversion method according to claim 1, characterized in that, The ground motion parameters include peak ground acceleration, peak ground velocity, and peak ground displacement.

6. The regional seismic ground motion field inversion method according to claim 1, characterized in that, The process of generating a continuously distributed regional seismic field based on the seismic motion parameters corresponding to multiple different locations through spatial interpolation or data fusion specifically includes: Obtain the ground motion parameters corresponding to multiple different locations to form a discrete point dataset. Each discrete point contains the location coordinates and its corresponding ground motion parameter values. Based on the discrete point dataset, for any spatial location within the target area, the estimated value of the seismic motion parameters at that location is calculated using a spatial interpolation method or a data fusion algorithm. Iterates through all spatial locations within the target area to generate a continuously distributed regional seismic field.

7. A regional seismic ground motion field inversion device, characterized in that, The regional seismic field inversion device includes: The data acquisition module is used to acquire structural response monitoring data of building structures during real earthquakes; The data preprocessing module is used to preprocess the structural reaction monitoring data to obtain matrix data; The seismic motion parameter inversion module is used to perform feature learning and nonlinear regression processing on the matrix data through a pre-trained neural network model to obtain seismic motion parameters. The neural network model includes a channel attention mechanism, a convolutional neural network, and a long short-term memory network. The seismic field generation module is used to generate a continuously distributed regional seismic field based on the seismic parameters corresponding to multiple different locations through spatial interpolation or data fusion.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the regional seismic field inversion method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the regional seismic field inversion method according to any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the regional seismic field inversion method according to any one of claims 1-6.

Images