Structure global dynamic response prediction method and related equipment thereof

Abstract

The invention provides a structure global dynamic response prediction method and related equipment thereof. The method comprises the following steps: acquiring a reality data set of a structure in a current state; based on a preset finite element model, the multiple pieces of sparse sensing data are processed, a simulation data set is output, and the simulation data set comprises multiple paired sparse input-global output data pairs; based on the simulation data set, pre-training the blank neural network model to obtain a virtual model; performing physical correction processing on the virtual model to obtain a real model; processing the sensing data of the current structure through a reality model, and determining global dynamic response field data of the current structure; based on the global dynamic response field data, risk deformation of the current structure is predicted, and corresponding risk early warning operation is executed. Through the steps of the method, global dynamic response prediction of a large-scale complex structure can be realized under the condition that only a small amount of sensing data is depended on, the sensor layout cost is reduced, and the structural risk monitoring precision and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent testing, and in particular to a method, device, electronic device and storage medium for predicting the global dynamic response of a structure. Background Art

[0002] Currently, health assessment of large-scale infrastructure relies primarily on two independent technical paradigms: direct monitoring based on physical sensors and simulation analysis based on numerical models. However, both paradigms have fundamental and insurmountable limitations, making it impossible to achieve a comprehensive, accurate, and real-time assessment of the structural condition.

[0003] Existing technologies typically deploy large-scale sensor arrays on structures to capture response data under different operating conditions for damage identification and risk assessment. However, in complex structures such as bridges, tunnels, and large buildings, the number of sensors deployed is limited by cost, energy consumption, and maintenance requirements. Consequently, only a limited amount of sparse sensor data can be obtained in practical applications. This makes it difficult for monitoring data to fully reflect the global response of the structure, severely limiting the accuracy and reliability of structural risk prediction and early warning.

[0004] On the other hand, finite element models, as a high-fidelity numerical simulation tool, can theoretically output the full-domain dynamic response field. However, these models rely heavily on idealized input parameters and boundary conditions, resulting in a "simulation-reality gap" between their results and the actual structure. Traditional finite element model updating methods often require large amounts of high-precision measured data for reverse correction, which conflicts with the reality of sparse sensor data and makes it difficult to meet the dual requirements of real-time performance and accuracy in practical engineering.

[0005] Therefore, the existing methods for predicting the global dynamic response of structures cannot accurately reconstruct the global dynamic response of structures and achieve real-time risk warnings by relying solely on sparse sensor data. Summary of the Invention

[0006] The present invention provides a method for predicting the global dynamic response of a structure to solve the problem that existing methods for predicting the global dynamic response of a structure cannot accurately reconstruct the global dynamic response of the structure and achieve real-time risk warning when relying solely on sparse sensor data.

[0007] In a first aspect, the present invention provides a method for predicting the global dynamic response of a structure, the method comprising the following steps: Acquire a real data set of the structure in a current state, wherein the real data set includes a plurality of sparse sensor data corresponding to the current structure; Based on a preset finite element model, the plurality of sparse sensor data are processed to output a simulation data set, wherein the simulation data set includes a plurality of paired sparse input-global output data pairs; Pre-training a blank neural network model based on the simulation data set to obtain a virtual model; Performing physical correction processing on the virtual model to obtain a real model; Process the sensor data of the current structure through the realistic model to determine the global dynamic response field data of the current structure; Based on the global dynamic response field data, the risk deformation of the current structure is predicted, and corresponding risk warning operations are performed.

[0008] Optionally, obtaining a real data set of the structure in the current state includes: Arranging the preset sensor array at a vulnerable position in the structure, the preset sensor array comprising an acceleration sensor, a strain sensor, a displacement sensor and an environmental sensor; The preset sensor array is used to collect sparse sensing data of vulnerable positions of the structure in the current state to obtain a real data set of the structure in the current state. The sparse sensing data corresponding to the vulnerable positions are used to maximize the information reflecting the overall state of the structure.

[0009] Optionally, processing the plurality of sparse sensing data based on a preset finite element model and outputting a simulation data set includes: Determine the virtual load condition data required for the current simulation; Loading the virtual load condition data into the preset finite element model, and processing the plurality of sparse sensor data using a finite element algorithm to determine a plurality of corresponding finite element response outputs; Pairing multiple sparse sensor data with corresponding multiple finite element response outputs to obtain multiple sparse input-global output data pairs; A simulation dataset is constructed based on multiple sparse input-global output data pairs.

[0010] Optionally, pre-training a blank neural network model based on the simulation data set to obtain a virtual model includes: Inputting the sparse sensor data in the simulation data set into the blank neural network model to obtain a training full domain output; The finite element response output corresponding to the sparse sensing data is used as a supervision label, and the training global output is compared and calculated with the finite element response output according to a time-frequency domain contrast regularization fidelity loss function to obtain a calculation result, wherein the time-frequency domain contrast regularization fidelity loss function is constructed based on the time-domain morphological fidelity data, frequency-domain robustness feature data, and multi-scale physical consistency data corresponding to the sparse sensing data of the current structure; Taking minimization of calculation results as optimization goal, the blank neural network model is iteratively trained, and after the iterative training is completed, a virtual model is obtained.

[0011] Optionally, performing physical correction processing on the virtual model to obtain a real model includes: Obtain sparse sensor data of the current structure; Adjust and train the output layer and the correction layer of the virtual model according to the sparse sensing data of the current structure, and determine correction parameters corresponding to the output layer and the correction layer of the virtual model; Based on the correction parameters, the virtual model is corrected and trained, and a real model is obtained after the training converges. The real model is used to output physical property data reflecting the current structure.

[0012] Optionally, processing the sensor data of the current structure using the reality model to determine the global dynamic response field data of the current structure includes: The sparse sensing data of the current real-time structure is input into the reality model for nonlinear mapping reasoning to determine the global dynamic response field data corresponding to the real-time current structure. The global dynamic response field data is used to reflect the dynamic response distribution of the current structure at different positions and times.

[0013] Optionally, the predicting of risk deformation of the current structure based on the global dynamic response field data and performing corresponding risk warning operations include: Performing temporal and spatial feature calculations on the global dynamic response field data to determine deformation trends of the current structure at different locations; Comparing the deformation trend with a preset risk trend to identify potential deformation areas; Based on the potential deformation area, corresponding risk warning information is generated, and a corresponding risk warning operation is triggered.

[0014] In a second aspect, the present invention further provides a device for predicting the global dynamic response of a structure, the device comprising: A first acquisition module is used to acquire a real data set of the structure in a current state, wherein the real data set includes a plurality of sparse sensor data corresponding to the current structure; A first processing module is configured to process the plurality of sparse sensor data based on a preset finite element model and output a simulation data set, wherein the simulation data set includes a plurality of paired sparse input-global output data pairs; A first training module is used to pre-train a blank neural network model based on the simulation data set to obtain a virtual model; A second processing module is used to perform physical correction processing on the virtual model to obtain a real model; A first determination module is used to process the sensor data of the current structure through the reality model to determine the global dynamic response field data of the current structure; The first execution module is used to predict the risk deformation of the current structure based on the global dynamic response field data and perform corresponding risk warning operations.

[0015] In a third aspect, the present invention provides an electronic device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method for predicting the global dynamic response of a structure provided by the present invention are implemented.

[0016] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for predicting the global dynamic response of a structure provided by the invention.

[0017] The present invention obtains a real data set of the structure in its current state; based on a preset finite element model, multiple sparse sensor data are processed to output a simulation data set, the simulation data set including multiple paired sparse input-global output data pairs; based on the simulation data set, a blank neural network model is pre-trained to obtain a virtual model; the virtual model is physically corrected to obtain a real model; the sensor data of the current structure is processed by the real model to determine the global dynamic response field data of the current structure; based on the global dynamic response field data, the risk deformation of the current structure is predicted, and corresponding risk warning operations are performed. Through the above method steps, the global dynamic response prediction of large and complex structures can be achieved while relying only on a small amount of sensor data, reducing the cost of sensor deployment and improving the precision and accuracy of structural risk monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 is a flow chart of a structure global dynamic response prediction method provided by an embodiment of the present application; Figure 2 is a structural schematic diagram of another structure global dynamic response prediction device provided in an embodiment of the present application; Figure 3 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the protection scope of the present application.

[0021] As shown in Figure 1 , Figure 1 is a flow chart of a structure global dynamic response prediction method provided by an embodiment of the present application. The structure global dynamic response prediction method comprises the following steps: 101, acquiring a real data set of a structure in a current state.

[0022] In the embodiments of the present application, the above-mentioned structure global dynamic response prediction method can be applied to a structure global dynamic response prediction platform. The above-mentioned structure global dynamic response prediction platform has functions of global dynamic response prediction data processing, global dynamic response prediction data transceiving, global dynamic response prediction data memory storage and the like, and can be constructed based on a server or a server cluster. The above-mentioned server or server cluster can be an electronic device with global dynamic response prediction data processing capability.

[0023] The above-mentioned current state can refer to a physical state of a structure at a specific time or under a specific working condition. It can be understood that the above-mentioned state can be jointly determined according to external environmental loads of the structure, such as vibration, wind blowing, and internal service conditions, such as aging, temperature influence, and the like. For example, the stress condition of a bridge under traffic load in a peak period, and the dynamic response of a building under strong wind can be regarded as the current state of a certain structure.

[0024] The above-mentioned structure can refer to an engineering object that needs to be monitored and predicted for global dynamic response, and can include a bridge, a tunnel, a large building, a sea platform and the like typical civil engineering infrastructure.

[0025] The real data set can include, but is not limited to, sparse sensing data corresponding to the current structure, and generally, can be a set of actual monitoring data collected by sensors deployed on the structure, including outputs of various sensors, such as vibration signals output by acceleration sensors, local strain values collected by strain gauges, structural displacement values collected by displacement sensors, and environmental parameters such as temperature, humidity, and wind speed. It can be understood that the data in the real data set is derived from specific physical measurement quantities and can reflect the dynamic response of the structure and external environmental conditions.

[0026] The sparse sensing data can be data extracted from the real data set and obtained by deploying a limited number of sensors. The data collected after deployment at key positions of the entire structure only covers part of the nodes of the structure. For example, when acceleration sensors and strain gauges are deployed at the midspan, quarter points, and tower top of a large-span bridge, the obtained data is sparse sensing data. Therefore, although the amount of the sparse sensing data is limited, the data content can be used to represent key information of the overall dynamic characteristics of the structure.

[0027] In a possible embodiment, the structure global dynamic response prediction platform collects the real data set by deploying a limited number of sensors at key positions when the structure is in a specific working condition.

[0028] 102. Process the plurality of sparse sensing data based on the preset finite element model, and output a simulation data set.

[0029] In the embodiment of the application, the preset finite element model can be a high-fidelity numerical model established by the structure global dynamic response prediction platform according to design drawings, material parameters, and actual boundary conditions of the current structure. The preset finite element model can simulate the dynamic response of the structure under virtual working conditions such as traffic load, wind load, or earthquake action through numerical calculation, that is, a small amount of sparse sensing data is calculated to output a large amount of corresponding dynamic response data as simulation data. This process can be used as a basis for generating simulation data.

[0030] In a possible embodiment, the structure global dynamic response prediction platform loads the set virtual load working condition to the preset finite element model, extracts response data at nodes corresponding to the sparse sensing data positions, and generates dynamic response output in the global range. In this way, the data of a limited number of sensor points can be associated with the global response results to form a pair of sparse input and global output.

[0031] The above-mentioned simulation data set may include but is not limited to multiple paired sparse input-global output data pairs. Specifically, each set of data samples contains the time-series response of the sparse sensing node and its corresponding global dynamic response field. By repeatedly generating under various virtual load conditions, multiple sets of samples covering different states can be constructed, thereby providing a training data source for the pre-training of the neural network model.

[0032] In another possible embodiment, the above-mentioned structural global dynamic response prediction platform applies virtual load conditions to the finite element model, extracts response results at nodes corresponding to sensor layout positions, and outputs a dynamic response field covering the entire domain. By pairing sparse input with global output, multiple groups of data samples are formed, and then a simulation data set is constructed to provide training data for the subsequent pre-training of the neural network model.

[0033] 103. Based on the simulation data set, the blank neural network model is pre-trained to obtain a virtual model.

[0034] In an embodiment of the present invention, the above-mentioned blank neural network model may refer to a neural network structure that has not undergone any data training before pre-training, and its parameters are all in an initial state. Specifically, the above-mentioned blank neural network model may adopt a spatiotemporal neural network structure (such as spatiotemporal U-Net) and have the ability to perform feature extraction and nonlinear mapping on input sparse data. It can be understood that in the initial state, that is, under the premise that no training has been performed, the above-mentioned blank neural network model does not have an effective structural global response prediction capability.

[0035] In this embodiment, the blank neural network model can be initially trained based on the simulation data set generated by the finite element model, using the sparse input as the input layer data of the model and the global response output as the supervision label. By iteratively optimizing the loss function, the blank neural network model can gradually master the mapping relationship between the sparse input and the global output. It can be understood that the purpose of the above-mentioned pre-training process can be to allow the above-mentioned blank neural network model to learn universal physical laws in a virtual environment, thereby laying the foundation for subsequent physical correction based on real data.

[0036] The above-mentioned supervision labels can also be complete global dynamic response outputs, which are used to adjust the model to learn universal physical laws.

[0037] The above-mentioned virtual model may refer to a neural network model obtained through pre-training. The above-mentioned virtual model is already able to predict the corresponding global dynamic response field based on the above-mentioned sparse sensor data under virtual working conditions, and has certain predictive ability and generalization. However, the above-mentioned virtual model reflects the physical laws contained in the finite element simulation data, and there is still a simulation-reality gap with the actual structure. It needs to be further adjusted to a real model through physical correction to make up for the difference in global response with the actual structure.

[0038] In a possible embodiment, the above-mentioned structural global dynamic response prediction platform uses the sparse input in the simulation data set as the input of the model, and the corresponding global dynamic response output as the supervision label, and iteratively trains the model using the set loss function. During this training process, the model gradually learns the complex nonlinear mapping relationship between the sparse input and the global output until it converges and has generalization ability, and finally forms a virtual model that can characterize the dynamic characteristics of the structure.

[0039] 104. Perform physical correction processing on the virtual model to obtain a real model.

[0040] In an embodiment of the present invention, the virtual model can be corrected using sparse sensor data collected in reality. Specifically, by introducing a small amount of real data into the output layer and correction layer of the virtual model for fine-tuning training, the model can gradually eliminate the difference between the finite element simulation and the actual structure, bridge the "simulation-reality gap", avoid the large-scale data requirements required for training the neural network from scratch, and complete the model correction by relying only on limited measured data.

[0041] The aforementioned real-world model can be a neural network model formed after physical calibration. While maintaining the universal physical laws learned in the virtual model, this real-world model also incorporates the actual response characteristics of the structure, enabling accurate prediction of the structure's global dynamic response field under real-world operating conditions. It is understood that this real-world model can output predictions consistent with the actual structural dynamic behavior when fed with real-time sparse sensor data, providing a reliable basis for structural health monitoring and risk warning.

[0042] In one possible embodiment, the aforementioned global dynamic response prediction platform collects a small amount of real-world sparse sensor data from the structure and uses this data as input to adjust and train the output and correction layers of the virtual model. By iteratively updating some network parameters, the model can compensate for the differences between the virtual model and the real structure. After training converges, a realistic model that reflects the target structure's true physical properties is obtained. This model, when receiving real-time sparse sensor data, can output a global dynamic response field that aligns with the actual situation.

[0043] 105. The sensor data of the current structure is processed through the reality model to determine the global dynamic response field data of the current structure.

[0044] In an embodiment of the present invention, the above-mentioned global dynamic response field data may refer to the dynamic response results of the structure in the entire spatial range in the current state, including but not limited to the time-series distribution of displacement field, strain field, stress field or acceleration field. It should be noted that the above-mentioned global dynamic response field data is different from the sparse sensing data that can only reflect local monitoring points. The above-mentioned global dynamic response field data can display the overall stress and deformation characteristics of the structure, thereby comprehensively reflecting the dynamic behavior of the structure under specific working conditions.

[0045] For example, during the operation of a long-span bridge, the global dynamic response field data generated by the realistic model can demonstrate the following structural dynamic behavior characteristics under specific operating conditions: Stress characteristics: When heavy vehicles pass through the mid-span of a bridge, the displacement and stress in the mid-span area increase significantly, while the tops of the pylons and the anchorage areas exhibit corresponding force transfer paths. This global distribution reflects the overall transfer of vehicle loads in the structure. Deformation characteristics: In strong wind environments, global dynamic response field data can depict the overall vibration modes of the bridge along the span, such as lateral swing or vertical vibration, and show the deformation trends of the bridge deck and pylons at different time points, thereby reflecting the global dynamic behavior of the structure under wind loads.

[0046] Through these data, managers can not only observe the response of a single sensor point, but also obtain the stress distribution and dynamic deformation process of the entire bridge under specific working conditions, thereby making a more comprehensive judgment on potential risk areas.

[0047] In one possible embodiment, the above-mentioned global dynamic response prediction platform of the structure inputs the sparse sensor data collected in real time into the reality model. The model performs inference calculations on the input data based on the calibrated network parameters to generate global dynamic response field data covering the entire structure range, reflecting the dynamic response distribution of the structure at different locations and times, which is used for subsequent risk identification and early warning operations.

[0048] 106. Based on the global dynamic response field data, the risk deformation of the current structure is predicted and corresponding risk warning operations are performed.

[0049] In an embodiment of the present invention, the global dynamic response prediction platform analyzes and determines the potential for abnormal deformation of a structure over a period of time or under specific operating conditions based on global dynamic response field data. Specifically, it models the evolution of displacement, strain, or stress at different monitoring points and, combined with preset risk thresholds, identifies deformation quantities or growth rates that could lead to structural safety hazards. For example, if the deflection in a mid-span area of ​​a bridge increases rapidly over time and approaches the design limit, a potential risk is predicted for that area.

[0050] The aforementioned risk warning operation can refer to the automatic generation and issuance of response measures after predicting a potential risk deformation. This operation can include generating visual warning information (such as highlighting risk areas on a 3D structural model), triggering audible and visual alarms to alert on-site personnel, or transmitting warning instructions to the structural health management system via a data interface, thereby achieving automated coordinated control.

[0051] In one possible embodiment, the aforementioned global structural dynamic response prediction platform analyzes the temporal and spatial characteristics of the global dynamic response field data, extracting deformation trends at different locations. This information is then compared with preset risk reference values ​​to identify areas where deformation may exceed specified limits. After identifying potential risky deformation areas, risk warning information is generated and executed through audio-visual alarms, visual prompts, or linkage with the management system, enabling operations and maintenance personnel to take timely action.

[0052] In an embodiment of the present invention, a real data set of the structure in the current state is obtained; based on a preset finite element model, multiple sparse sensor data are processed to output a simulation data set, which includes multiple paired sparse input-global output data pairs; based on the simulation data set, a blank neural network model is pre-trained to obtain a virtual model; the virtual model is physically corrected to obtain a real model; the sensor data of the current structure is processed by the real model to determine the global dynamic response field data of the current structure; based on the global dynamic response field data, the risk deformation of the current structure is predicted, and corresponding risk warning operations are performed. The above method steps can realize the global dynamic response prediction of large and complex structures while relying only on a small amount of sensor data, reduce the cost of sensor deployment, and improve the precision and accuracy of structural risk monitoring.

[0053] Optionally, in the step of obtaining a realistic data set of the structure in the current state, a preset sensor array may be arranged at a vulnerable position in the structure; sparse sensing data of the vulnerable position of the structure in the current state may be collected by the preset sensor array to obtain a realistic data set of the structure in the current state.

[0054] In an embodiment of the present invention, the above-mentioned preset sensor array may include but is not limited to acceleration sensors, strain sensors, displacement sensors, and environmental sensors. For example, it may be an acceleration sensor that captures high-frequency vibrations, a strain gauge that monitors local stress, a GNSS displacement sensor that obtains macroscopic displacements, and a sensor that collects environmental parameters such as wind speed and temperature, etc., which are sensor devices that can obtain the physical state of the structure. Specifically, the sensors in the above-mentioned preset sensor array can be arranged according to structural mechanics analysis and optimal sensor layout methods, so as to achieve the purpose of maximizing the amount of information with a limited number of sensors.

[0055] The above-mentioned vulnerable locations may refer to the critical areas that are most prone to damage or abnormalities, determined based on the structural stress characteristics and long-term operating experience. For example, in large bridges, typical vulnerable locations include the mid-span area, the connection between the beam end and the support, and the top of the bridge tower. These locations often bear large dynamic loads or environmental loads and are identified as potential weak points in finite element simulation analysis and engineering experience. Therefore, by deploying sensors at these locations, the possible risk responses of the structure can be captured earlier and more directly, thereby improving the accuracy of reconstructing the global dynamic response field based on sparse data. That is, the sparse sensing data corresponding to the above-mentioned vulnerable locations can be used to maximize the information reflecting the overall state of the structure.

[0056] In a possible embodiment, the above-mentioned structural global dynamic response prediction platform collects sparse sensing data of the above-mentioned vulnerable positions under current working conditions through the above-mentioned preset sensor array to form a realistic data set. It can be understood that since the sensor layout positions correspond to the potential weak points of the structure, the data obtained can more directly reflect the key response characteristics of the structure.

[0057] Optionally, in the step of processing multiple sparse sensor data based on a preset finite element model and outputting a simulation data set, it also includes determining the virtual load condition data required for the current simulation; loading the virtual load condition data into the preset finite element model, and processing the multiple sparse sensor data through a finite element algorithm to determine multiple corresponding finite element response outputs; pairing the multiple sparse sensor data with the corresponding multiple finite element response outputs to obtain multiple sparse input-global output data pairs; and constructing a simulation data set based on the multiple sparse input-global output data pairs.

[0058] In an embodiment of the present invention, the above-mentioned virtual load condition data may refer to a combination of loads and boundary conditions artificially set in the finite element simulation, including traffic loads, wind loads, seismic motions, temperature changes and other working condition parameters. It can be understood that the above-mentioned virtual load condition data is applied to the finite element model in a virtual manner to simulate the dynamic response of the structure under different environments and actions.

[0059] The above-mentioned finite element response output can be the structural response result obtained by solving the above-mentioned finite element algorithm, including but not limited to time series or spatial distribution data such as node displacement, strain, stress, acceleration, etc.

[0060] In this embodiment, a correspondence can be established between the above-mentioned sparse sensing data and the finite element global response results obtained under the same working conditions. Specifically, a small amount of sparse input can be matched one-to-one with the output results covering the entire structure to form a training sample that meets the requirements of supervised learning.

[0061] In a possible embodiment, the above-mentioned structural global dynamic response prediction platform loads the above-mentioned virtual load condition data into a preset finite element model, and extracts the response output at the position corresponding to the sparse sensor data through the finite element algorithm to obtain multiple finite element response results, and pairs the sparse sensor data with the corresponding finite element response output to form multiple sparse input-global output data pairs, and further constructs a simulation data set based on these data pairs to provide training samples for pre-training of the neural network model.

[0062] Optionally, in the step of pre-training a blank neural network model based on a simulation data set to obtain a virtual model, the sparse sensor data in the simulation data set is also input into the blank neural network model to obtain a training global output; the finite element response output corresponding to the sparse sensor data is used as a supervision label, and the training global output and the finite element response output are compared and calculated based on the time-frequency domain contrast regularization fidelity loss function to obtain a calculation result; the blank neural network model is iteratively trained with minimizing the calculation result as the optimization goal, and a virtual model is obtained after the iterative training is completed.

[0063] In an embodiment of the present invention, the above-mentioned training global output may refer to a global dynamic response prediction result covering the entire structure generated by network inference after a blank neural network model receives sparse sensor data as input during the pre-training process.

[0064] The above-mentioned supervision label can refer to the global dynamic response output generated by the finite element simulation, which serves as the reference truth value in the neural network training.

[0065] The above-mentioned time-frequency domain contrast regularization fidelity loss function may refer to the simultaneous introduction of time domain and frequency domain contrast constraints in the loss function, combined with dynamic time warping (soft-DTW) to ensure the temporal morphological fidelity of the predicted sequence, combined with frequency domain feature contrast to enhance the robustness of the model in the spectral space, and maintain physical rationality through power spectral density consistency constraints. Specifically, the above-mentioned time-frequency domain contrast regularization fidelity loss function can also be constructed based on the time domain morphological fidelity data, frequency domain robustness feature data and multi-scale physical consistency data corresponding to the sparse sensing data of the current structure.

[0066] In this embodiment, the model can be guided to conduct deeper and physically intuitive learning by combining three complementary sub-loss terms. The specific overall structure is as follows: ; in, For fidelity loss, is the frequency domain contrast loss, is the multi-scale PSD consistency loss, and It is used as a hyperparameter to balance the weights of various loss terms.

[0067] More specifically, the above-mentioned loss of fidelity It can be used to pursue the accuracy of the structure in morphology and timing. It uses differentiable dynamic time warping (soft-DTW) as the core metric. Unlike the Euclidean distance, which is extremely sensitive to time offsets, DTW can find the optimal nonlinear alignment path between two time series, thereby measuring their morphological similarity. Specifically, the above-mentioned fidelity loss can be constructed using the following formula:

[0068]

[0069]

[0070] Among them, the morphological loss Directly calculated by soft-DTW, timing loss Excessive distortion of the alignment path is penalized. and Corresponding to the real sequence and the predicted sequence respectively, is the best alignment path found by the DTW algorithm, Is the path The index pairs in For real sequence The time index, To predict the sequence Zhongyu Aligned time index, is a hyperparameter used to balance the importance of morphology and timing.

[0071] The above frequency domain contrast loss It can be used to learn robust physical representations. It can adopt the strategy of "pulling similar ones closer and pushing dissimilar ones apart" to learn robust representations that are invariant to data enhancement. By using the frequency domain representation to deal with noise and slight time offsets, the above frequency domain contrast loss is constructed through the following steps: 1. First, Convert to the frequency domain via real fast Fourier transform (RFFT) ; 2. Construct positive and negative samples: The positive sample can be represented as two slightly different enhanced samples formed by applying two independent data augmentation operations (for example, adding Gaussian noise, applying random time domain masks, or performing small-scale amplitude scaling and jittering on the signal) to the same original input in a batch. These two enhanced samples are fed into the model separately to obtain their frequency domain representations. and , forming a positive sample pair; Negative samples can be represented as predictions from different original inputs in a batch, and their frequency domain representation is are all regarded as negative samples.

[0072] 3. Using the NT-Xent (Normalized Temperature-scaled Cross-Entropy) loss function, we maximize the similarity of positive sample pairs in the frequency domain representation space while minimizing the similarity with negative samples. This enables the model to learn stable frequency domain features that can distinguish different physical events, greatly enhancing the model's robustness to real-world noise. Specifically, this is expressed as:

[0073] in, represents the batch size, is the cosine similarity, is the temperature hyperparameter, which is used to adjust the sharpness of the similarity distribution. This will widen the similarity score differences between different samples, making the model more focused on distinguishing difficult negative samples. 、 and Respectively represent the frequency domain representation vector of the sample after rfft transformation. Among them, and Come from the same original input, forming a positive sample pair, and It represents the batch Non-homologous negative samples.

[0074] The above multi-scale power spectral density consistency loss It can be used to ensure the physical rationality of multiple scales. By keeping the PSD consistent with the PSD of the real signal at the original resolution and the low resolution after downsampling, it can not only capture the high-frequency vibration details (modal characteristics), but also accurately reproduce the long-term macro trend (thermal effect, etc.). Specifically, the above multi-scale power spectral density consistency loss can be constructed by the following formula :

[0075] in,

[0076]

[0077] For downsampling operation, the sequence length is halved by average pooling. is a hyperparameter used to balance the importance of different scales. and Corresponding sequences and Middle The value at a time point.

[0078] By combining the above three losses to construct the final time-frequency domain contrast regularization fidelity loss function, we can solve the physical inductive bias problem of ill-posed structures under specific working conditions based on their synergistic effect. Through the organic synergy of the three losses, we provide a set of "physical inductive bias" for solving the structure under specific working conditions. Specifically, through Ensure that the solution has a consistent shape at the known data points; by The frequency domain information of these sparse points provides global physical constraints for the reconstruction of the entire unknown area, forcing the intrinsic physical properties of the solution to be robust and consistent. This ensures that the energy distribution of the solution in the frequency domain conforms to physical laws at both macro and micro scales. The combination of these three factors significantly compresses the space of possible solutions, guiding the model to learn a physically consistent and reasonable solution with high generalization capabilities for real-world data.

[0079] In this embodiment, the difference between the training global output and the supervision label in the time domain and frequency domain can be calculated by using the loss function, thereby reflecting the gap between the model prediction value and the target true value.

[0080] The above calculation result may refer to the error or difference value obtained after the comparison calculation, which is usually expressed in numerical form and is the basis for training optimization. It can be understood that the smaller the above calculation result, the closer the model prediction output is to the true global response.

[0081] The parameters of the neural network can be continuously updated using the calculation results in multiple training cycles. In each iteration, the network parameters will be adjusted according to the error feedback, so that the model performance will be gradually improved to achieve the purpose of iterative training.

[0082] In a possible embodiment, the above-mentioned structural global dynamic response prediction platform inputs the sparse sensor data in the simulation data set into a blank neural network model to generate a training global output, and uses the finite element response output corresponding to the sparse sensor data as a supervisory label. According to the time-frequency domain contrast regularization fidelity loss function, the training global output and the supervisory label are compared and calculated, and the blank neural network model is iteratively trained with the minimization of the calculation result as the optimization goal. After the training converges, a virtual model that can characterize the mapping relationship between the sparse input and the global output is obtained.

[0083] Optionally, in the step of performing physical correction processing on the virtual model to obtain the real model, it also includes obtaining sparse sensing data of the current structure; adjusting and training the output layer and correction layer of the virtual model based on the sparse sensing data of the current structure, and determining the correction parameters corresponding to the output layer and correction layer of the virtual model; based on the correction parameters, performing correction training on the virtual model, and obtaining the real model after the training converges.

[0084] In an embodiment of the present invention, the above-mentioned output layer may refer to a layer in the above-mentioned neural network structure used to generate the final prediction result. It can be understood that the above-mentioned output layer is used to convert the intermediate feature map into global dynamic response field data. Therefore, adjusting the above-mentioned output layer can directly correct the final prediction accuracy of the model, making its output closer to the physical performance of the real structure.

[0085] The correction layer may refer to a layer provided in the virtual model for subsequent physical correction, which may generally be provided before the output layer. It is understandable that a small amount of parameter updates are used to compensate for the difference between the virtual model and the actual structure, thereby achieving a transition from simulation to reality, and reducing the need for large-scale parameter updates to improve training efficiency and reduce dependence on real data.

[0086] In this embodiment, the above-mentioned structural global dynamic response prediction platform can use the actual sparse sensor data collected as input, and use supervised learning to update only the parameters of the output layer and the correction layer, thereby achieving the purpose of adjustment training. It should be noted that, unlike full model training, adjustment training only optimizes limited parameters, so the training efficiency is faster and can avoid destroying the universal physical laws that the virtual model has learned.

[0087] The above-mentioned correction parameters may refer to the model parameters such as weights and biases that are updated in the output layer and the correction layer during the adjustment training process. The correction values ​​of the corresponding layers of the above-mentioned correction parameters can be determined through continuous iterative optimization, and are used to correct the prediction deviation of the virtual model so that the model output can be closer to the physical characteristics corresponding to the real sensor data.

[0088] In a possible embodiment, the above-mentioned structural global dynamic response prediction platform keeps the main parameters of the virtual model unchanged and only optimizes the output layer and the correction layer, so that the model gradually eliminates the difference between the virtual simulation and the real data.

[0089] In another possible embodiment, the above-mentioned structural global dynamic response prediction platform obtains sparse sensor data of the current structure and inputs it into the virtual model. Based on these real collected data, the output layer and correction layer of the virtual model are adjusted and trained, and the correction parameters related to the layer are determined. The virtual model is corrected and trained according to the above-mentioned correction parameters so that its prediction results gradually approach the dynamic response of the real structure. After the training converges, the real model is obtained. The above-mentioned real model can be used to output physical property data reflecting the current structure, as well as subsequent global dynamic response prediction and risk warning.

[0090] Optionally, in the step of processing the sensor data of the current structure through the reality model to determine the global dynamic response field data of the current structure, the sparse sensor data of the current real-time structure is also input into the reality model for nonlinear mapping inference to determine the global dynamic response field data corresponding to the real-time current structure.

[0091] In an embodiment of the present invention, after receiving sparse sensor data from the global dynamic response prediction platform, the above-mentioned reality model utilizes the nonlinear activation function, convolution operation, and spatiotemporal feature extraction mechanism in the model network to map the sparse sensor data into a global dynamic response field. In this process, the above-mentioned reality model no longer relies on simple linear interpolation, but instead uses learned complex nonlinear functional relationships to spatially expand and reconstruct the temporal evolution of the input limited node information. Through this reasoning method, the above-mentioned reality model can bridge the gap between sparse observations and global responses, ensuring that the prediction results are both consistent with physical laws and have high temporal fidelity and frequency domain robustness. For example, when a bridge is equipped with only limited sensors at the mid-span and tower top, nonlinear mapping reasoning can reconstruct the global vibration modes and displacement distribution of the entire bridge at different time points.

[0092] The above global dynamic response field data can also be used to reflect the dynamic response distribution of the current structure at different locations and at different times.

[0093] In one possible embodiment, the above-mentioned structural global dynamic response prediction platform inputs the sparse sensor data of the current real-time structure into a physically corrected real-world model. The model performs nonlinear mapping inference based on its internally trained parameters to generate global dynamic response field data covering the entire structural range from limited sparse inputs.

[0094] Optionally, in the step of predicting the risk deformation of the current structure based on the global dynamic response field data and executing the corresponding risk warning operation, it also includes performing time and space feature calculations on the global dynamic response field data to determine the deformation trend of the current structure at different positions; comparing the deformation trend with the preset risk trend to identify potential deformation areas; based on the potential deformation area, generating corresponding risk warning information, and triggering the execution of corresponding risk warning operations.

[0095] In an embodiment of the present invention, the above-mentioned structural global dynamic response prediction platform performs temporal and spatial feature calculations on the global dynamic response field data, and extracts the deformation trend of the structure changing with time at different monitoring positions; the obtained deformation trend is compared with the preset risk trend, and when it is found that the deformation amount or evolution speed of a certain position exceeds the risk reference range, the position is identified as a potential deformation area, and corresponding risk warning information is generated according to the potential deformation area, and the risk warning operation is triggered through sound and light alarms, visual interface prompts or linkage with the management system, so that operation and maintenance personnel can obtain structural risk information in a timely manner.

[0096] For example, during the operation of a long-span suspension bridge, the global dynamic response prediction platform of the structure, based on the global dynamic response field data generated by the real model, conducted a time-series and spatial analysis of the stress and deformation characteristics of the bridge under traffic loads, and found that the displacement trend in the mid-span area of ​​the bridge increased rapidly in a short period of time, and compared with the preset risk trend, the growth rate was close to the safety threshold. By comparing the results, the platform identified the mid-span area as a potential deformation area and immediately generated risk warning information. After the warning operation is triggered, the platform sends the risk warning to the operation and maintenance management terminal, and highlights the area in the visual interface of the bridge monitoring center, prompting maintenance personnel to take load reduction or traffic control measures, thereby effectively avoiding possible structural safety accidents.

[0097] like Figure 2 As shown, an embodiment of the present invention further provides a device 200 for predicting the global dynamic response of a structure. The device 200 includes: A first acquisition module 201 is configured to acquire a real data set of a structure in a current state, wherein the real data set includes a plurality of sparse sensor data corresponding to the current structure; A first processing module 202 is configured to process the plurality of sparse sensor data based on a preset finite element model and output a simulation data set, wherein the simulation data set includes a plurality of paired sparse input-global output data pairs; A first training module 203 is configured to pre-train a blank neural network model based on the simulation data set to obtain a virtual model; The second processing module 204 is used to perform physical correction processing on the virtual model to obtain a real model; A first determination module 205 is configured to process the sensor data of the current structure using a reality model to determine global dynamic response field data of the current structure; The first execution module 206 is configured to predict risk deformation of the current structure based on the global dynamic response field data and execute corresponding risk warning operations.

[0098] Optionally, the first obtaining module 201 includes: A first placement submodule is configured to place the preset sensor array at a vulnerable position in the structure, wherein the preset sensor array includes an acceleration sensor, a strain sensor, a displacement sensor, and an environmental sensor; The first acquisition submodule is used to collect sparse sensing data of the vulnerable positions of the structure in the current state through the preset sensor array to obtain a real data set of the structure in the current state. The sparse sensing data corresponding to the vulnerable positions are used to maximize the information reflecting the overall state of the structure.

[0099] Optionally, the first processing module 202 includes: The first determination submodule is used to determine the virtual load condition data required for the current simulation; A second determination submodule is configured to load the virtual load condition data into the preset finite element model, and process the plurality of sparse sensing data using a finite element algorithm to determine a plurality of corresponding finite element response outputs; A second acquisition submodule is used to pair the plurality of sparse sensing data with the corresponding plurality of finite element response outputs to obtain a plurality of sparse input-global output data pairs; The first construction submodule is used to construct a simulation data set based on multiple sparse input-global output data pairs.

[0100] Optionally, the first training module 203 includes: A third acquisition submodule is used to input the sparse sensor data in the simulation data set into the blank neural network model to obtain a training full domain output; A second construction submodule is configured to use the finite element response output corresponding to the sparse sensing data as a supervision label, and compare and calculate the training global output with the finite element response output according to a time-frequency domain contrast regularization fidelity loss function to obtain a calculation result, wherein the time-frequency domain contrast regularization fidelity loss function is constructed based on the time-domain morphological fidelity data, frequency-domain robustness feature data, and multi-scale physical consistency data corresponding to the sparse sensing data of the current structure; The first training submodule is used to iteratively train the blank neural network model with minimization of calculation results as the optimization goal, and obtain a virtual model after the iterative training is completed.

[0101] Optionally, the second processing module 204 includes: The fourth acquisition submodule is used to obtain sparse sensing data of the current structure; A second training submodule is configured to adjust and train the output layer and the correction layer of the virtual model according to the sparse sensing data of the current structure, and determine correction parameters corresponding to the output layer and the correction layer of the virtual model; The third training submodule is used to perform correction training on the virtual model based on the correction parameters, and obtain a real model after the training converges. The real model is used to output physical property data reflecting the current structure.

[0102] Optionally, the first determining module 205 includes: The third determination submodule is used to input the sparse sensing data of the current real-time structure into the reality model for nonlinear mapping inference, and determine the global dynamic response field data corresponding to the real-time current structure. The global dynamic response field data is used to reflect the dynamic response distribution of the current structure at different positions and at different times.

[0103] Optionally, the first execution module 206 includes: a fourth determination submodule, configured to perform temporal and spatial feature calculations on the global dynamic response field data to determine deformation trends of the current structure at different locations; an identification submodule, configured to compare the deformation trend with a preset risk trend to identify potential deformation areas; The operation submodule is used to generate corresponding risk warning information based on the potential deformation area and trigger the execution of corresponding risk warning operations.

[0104] like Figure 3 As shown, an embodiment of the present invention further provides an electronic device 300, including a processor, and the processor can execute any of the above-mentioned methods for predicting the global dynamic response of a structure.

[0105] Specifically, the system includes a processor 301, a memory 302, and a computer program for executing a method for predicting the global dynamic response of a structure, which is stored in the memory 302 and can be run on the processor 301, wherein: The processor 301 runs the computer program of the structural global dynamic response prediction method stored in the memory 302 and performs the following steps: Acquire a real data set of the structure in a current state, wherein the real data set includes a plurality of sparse sensor data corresponding to the current structure; Based on a preset finite element model, the plurality of sparse sensor data are processed to output a simulation data set, wherein the simulation data set includes a plurality of paired sparse input-global output data pairs; Pre-training a blank neural network model based on the simulation data set to obtain a virtual model; Performing physical correction processing on the virtual model to obtain a real model; Process the sensor data of the current structure through the realistic model to determine the global dynamic response field data of the current structure; Based on the global dynamic response field data, the risk deformation of the current structure is predicted, and corresponding risk warning operations are performed.

[0106] Optionally, the processor 301 executes the step of obtaining a real data set of the structure in the current state, including: Arranging the preset sensor array at a vulnerable position in the structure, the preset sensor array comprising an acceleration sensor, a strain sensor, a displacement sensor and an environmental sensor; The preset sensor array is used to collect sparse sensing data of vulnerable positions of the structure in the current state to obtain a real data set of the structure in the current state. The sparse sensing data corresponding to the vulnerable positions are used to maximize the information reflecting the overall state of the structure.

[0107] Optionally, the processor 301 executes the processing based on the preset finite element model to process the plurality of sparse sensor data and output a simulation data set, including: Determine the virtual load condition data required for the current simulation; Loading the virtual load condition data into the preset finite element model, and processing the plurality of sparse sensor data using a finite element algorithm to determine a plurality of corresponding finite element response outputs; Pairing multiple sparse sensor data with corresponding multiple finite element response outputs to obtain multiple sparse input-global output data pairs; A simulation dataset is constructed based on multiple sparse input-global output data pairs.

[0108] Optionally, the processor 301 performs pre-training of the blank neural network model based on the simulation data set to obtain a virtual model, including: inputting sparse sensing data in the simulation data set into the blank neural network model to obtain a training global output; taking a finite element response output corresponding to the sparse sensing data as a supervision label, and comparing and calculating the training global output with the finite element response output according to a time-frequency domain comparison regularization fidelity loss function to obtain a calculation result, the time-frequency domain comparison regularization fidelity loss function being constructed according to time domain form fidelity data, frequency domain robustness feature data and multi-scale physical consistency data corresponding to the sparse sensing data of the current structure; taking minimization of the calculation result as an optimization target, iteratively training the blank neural network model, and obtaining the virtual model after the iterative training is completed.

[0109] Optionally, the processor 301 performs physical correction processing on the virtual model to obtain a real model, including: obtaining sparse sensing data of a current structure; adjusting and training an output layer and a correction layer of the virtual model according to the sparse sensing data of the current structure to determine correction parameters corresponding to the output layer and the correction layer of the virtual model; based on the correction parameters, performing correction training on the virtual model to obtain a real model after training convergence, the real model being used to output physical property data reflecting the current structure.

[0110] Optionally, the processor 301 performs processing on sensing data of a current structure by a real model to determine global dynamic response field data of the current structure, including: inputting sparse sensing data of the current real-time structure into the real model for non-linear mapping inference to determine global dynamic response field data corresponding to the real-time current structure, the global dynamic response field data being used to reflect dynamic response distribution of the current structure at different positions and different times.

[0111] Optionally, the processor 301 performs prediction of risk deformation of the current structure based on the global dynamic response field data and performs corresponding risk warning operations, including: performing time sequence and space feature calculation on the global dynamic response field data to determine deformation trends of the current structure at different positions; comparing the deformation trends with preset risk trends to identify potential deformation areas; based on the potential deformation areas, generating corresponding risk warning information and triggering execution of corresponding risk warning operations.

[0112] The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement each process of the structural global dynamic response prediction method provided by the embodiment of the application or the application end structural global dynamic response prediction method, and the same technical effects can be achieved. To avoid repetition, details are not described herein.

[0113] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware, and can be stored in a computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiment methods. The storage medium can be a magnetic disc, an optical disc, a read-only memory (ROM), or a random access memory (RAM), etc.

[0114] The above only discloses preferred embodiments of the application, and of course cannot limit the scope of the right of the application, so equivalent changes made according to the claims of the application are still within the scope of the application.

Claims

1. A method for predicting the global dynamic response of a structure, characterized in that: include: Acquire a real data set of the structure in a current state, wherein the real data set includes a plurality of sparse sensor data corresponding to the current structure; Based on a preset finite element model, the plurality of sparse sensor data are processed to output a simulation data set, wherein the simulation data set includes a plurality of paired sparse input-global output data pairs; Pre-training a blank neural network model based on the simulation data set to obtain a virtual model; Performing physical correction processing on the virtual model to obtain a real model; Process the sensor data of the current structure through the realistic model to determine the global dynamic response field data of the current structure; Based on the global dynamic response field data, the risk deformation of the current structure is predicted, and corresponding risk warning operations are performed.

2. The method for predicting the global dynamic response of a structure according to claim 1, wherein: The obtaining of a real data set of the structure in the current state includes: Arranging a preset sensor array at a vulnerable position in the structure, the preset sensor array comprising an acceleration sensor, a strain sensor, a displacement sensor, and an environmental sensor; The preset sensor array is used to collect sparse sensing data of vulnerable positions of the structure in the current state to obtain a real data set of the structure in the current state. The sparse sensing data corresponding to the vulnerable positions are used to maximize the information reflecting the overall state of the structure.

3. The method for predicting the global dynamic response of a structure according to claim 1, wherein: The processing of the plurality of sparse sensor data based on a preset finite element model and outputting a simulation data set includes: Determine the virtual load condition data required for the current simulation; Loading the virtual load condition data into the preset finite element model, and processing the plurality of sparse sensor data using a finite element algorithm to determine a plurality of corresponding finite element response outputs; Pairing multiple sparse sensor data with corresponding multiple finite element response outputs to obtain multiple sparse input-global output data pairs; A simulation dataset is constructed based on multiple sparse input-global output data pairs.

4. The method for predicting the global dynamic response of a structure according to claim 3, wherein: The method of pre-training a blank neural network model based on the simulation data set to obtain a virtual model includes: Inputting the sparse sensor data in the simulation data set into the blank neural network model to obtain a training full domain output; The finite element response output corresponding to the sparse sensing data is used as a supervision label, and the training global output is compared and calculated with the finite element response output according to a time-frequency domain contrast regularization fidelity loss function to obtain a calculation result, wherein the time-frequency domain contrast regularization fidelity loss function is constructed based on the time-domain morphological fidelity data, frequency-domain robustness feature data, and multi-scale physical consistency data corresponding to the sparse sensing data of the current structure; Taking minimization of calculation results as optimization goal, the blank neural network model is iteratively trained, and after the iterative training is completed, a virtual model is obtained.

5. The method for predicting the global dynamic response of a structure according to claim 1, wherein: The performing physical correction processing on the virtual model to obtain a real model includes: Obtain sparse sensor data of the current structure; Adjust and train the output layer and the correction layer of the virtual model according to the sparse sensing data of the current structure, and determine correction parameters corresponding to the output layer and the correction layer of the virtual model; Based on the correction parameters, the virtual model is corrected and trained, and a real model is obtained after the training converges. The real model is used to output physical property data reflecting the current structure.

6. The method for predicting the global dynamic response of a structure according to claim 1, wherein: The processing of the sensor data of the current structure by the reality model to determine the global dynamic response field data of the current structure includes: The sparse sensing data of the current real-time structure is input into the reality model for nonlinear mapping reasoning to determine the global dynamic response field data corresponding to the real-time current structure. The global dynamic response field data is used to reflect the dynamic response distribution of the current structure at different positions and times.

7. The method for predicting the global dynamic response of a structure according to claim 1, wherein: The predicting of risk deformation of the current structure based on the global dynamic response field data and performing corresponding risk warning operations include: Performing temporal and spatial feature calculations on the global dynamic response field data to determine deformation trends of the current structure at different locations; Comparing the deformation trend with a preset risk trend to identify potential deformation areas; Based on the potential deformation area, corresponding risk warning information is generated, and a corresponding risk warning operation is triggered.

8. A device for predicting the global dynamic response of a structure, characterized in that: include: A first acquisition module is used to acquire a real data set of the structure in a current state, wherein the real data set includes a plurality of sparse sensor data corresponding to the current structure; A first processing module is configured to process the plurality of sparse sensor data based on a preset finite element model and output a simulation data set, wherein the simulation data set includes a plurality of paired sparse input-global output data pairs; A first training module is used to pre-train a blank neural network model based on the simulation data set to obtain a virtual model; A second processing module is used to perform physical correction processing on the virtual model to obtain a real model; A first determination module is used to process the sensor data of the current structure through the reality model to determine the global dynamic response field data of the current structure; The first execution module is used to predict the risk deformation of the current structure based on the global dynamic response field data and perform corresponding risk warning operations.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps in the method for predicting the global dynamic response of a structure as claimed in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the method for predicting the global dynamic response of a structure according to any one of claims 1 to 7.

Images