Cross-domain structure damage identification method based on digital twinning and time-space feature fusion

Abstract

The invention discloses a cross-domain structural damage identification method based on digital twinning and time-space feature fusion, belongs to the technical field of structural health monitoring, and provides a cross-domain transfer learning damage identification technical scheme fusing a digital twinning model and structural multi-sensor data time-space features. A measurement point connection relation of a digital twinborn model and an actual physical structure is abstracted into a graph structure, and a spatial-temporal feature learning model for simultaneously processing multi-sensor time domain data and spatial feature information is designed based on classic time domain convolution and graph convolution network. A specific model architecture is provided for deep fusion of a structural digital twin model and spatial-temporal features of multi-sensor data of an actual structure to realize structural damage detection in a cross-data-domain scene, so that the problems of insufficient actual damage data, insufficient fusion of structural multi-dimensional feature data and the like are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of structural health monitoring, and specifically relates to a cross-domain structural damage identification method based on digital twins and time-space feature fusion. Background Art

[0002] With the development of deep learning technology, the research on structural damage identification using various neural network models has gradually attracted widespread attention. Existing research mainly trains neural network models by collecting a large amount of damage data under different damage scenarios, and identifies different types of damage based on damage data feature mapping. However, in actual engineering measurements, structural damage monitoring data is very scarce, and neural network models often have insufficient generalization ability due to the small amount of training data, thereby reducing recognition accuracy.

[0003] In order to make up for the scarcity of measured data, existing technologies use batch simulation data generated by structural finite element models or digital twin models in multiple damage scenarios as training data to train a neural network model, and use a small amount of measured data to fine-tune the neural network model to achieve cross-domain data-driven damage task transfer identification. In the current mainstream research methods of integrating cross-domain data and transfer learning, it is believed that structural vibration data is the most promising data for structural damage state identification. Most studies are based on structural time domain vibration data, with the goal of minimizing the difference between cross-domain data, using time domain data or variant data (such as frequency domain, time-frequency domain data) for deep learning model training and feature learning, and finally achieve the mapping relationship fitting from structural vibration data to damage state detection. However, these research methods ignore the node spatial characteristics corresponding to the structural vibration data, and there are few research results that integrate the spatial relationship between structural vibration data and the location of its sensors to carry out damage identification.

[0004] According to the mechanism of sensor optimization layout, the sensor layout measurement points are the most critical set of nodes obtained by comprehensive optimization by combining structural dynamic characteristics and modal analysis. The data set or single training sample data constructed by the vibration data on these nodes only reflects the time-course response information of each node, but does not reflect the spatial layout of each node. According to the structural dynamic characteristics, the dynamic responses of different nodes of the structure must have some correlation that obeys the structural dynamic characteristics. This spatial correlation may have a potential impact on the time domain response signal between multiple sensor measurement points. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention relies on the dominant position of structural vibration data and considers the key issues ignored by existing research results. It proposes a cross-domain structural damage identification method based on digital twins and time-space feature fusion to solve the problem that the current cross-domain structural damage identification research methods are mainly based on the time domain, frequency domain, time-frequency domain or other variant data of structural vibration data, and few studies consider the problem of integrating the spatial characteristics of multiple sensor measurement points of the structure for damage judgment.

[0006] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a cross-domain structural damage identification method based on digital twin and time-space feature fusion, comprising the following steps: S1. Build a digital twin model corresponding to the actual engineering structure and optimize the sensor layout based on the digital twin model; S2, construct a two-domain spatiotemporal dataset of actual engineering structures under various damage scenarios; The two-domain spatiotemporal data sets include a measured domain data set and a simulated domain data set; the measured domain data set includes measured time-domain vibration data of the actual engineering structure under various damage scenarios and its corresponding adjacency matrix; the simulated domain data set includes simulated time-domain vibration data generated by the corresponding digital twin model of the actual engineering structure under various damage scenarios and its corresponding adjacency matrix; the adjacency matrix represents the spatial position association of the sensors after the optimized arrangement; S3, construct a spatiotemporal feature learning network model, and train it using the spatiotemporal dataset to obtain a cross-domain structural damage recognition model; The cross-domain structural damage identification model includes an input layer, a time domain feature extraction layer, a spatial feature extraction layer and a damage prediction layer; S4, processing the time domain vibration data of the two structures to be identified for damage by using a time domain adaptive model to obtain adaptive time domain data; S5. Input the adaptive time domain data and its corresponding adjacency matrix into the cross-domain structural damage identification model for processing to obtain a cross-domain structural damage identification result.

[0007] Furthermore, in step S3, the method for the cross-domain structural damage identification model to process input data is specifically as follows: S31, receiving time domain vibration data and an adjacency matrix through an input layer, and constructing them into a spatiotemporal feature matrix; S32, performing a separate convolution on the time domain vibration data of each sensor in the spatiotemporal feature matrix through a time domain feature extraction layer to extract corresponding time domain features; S33, based on the extracted time domain features, extracting corresponding multi-dimensional spatial features through a spatial feature extraction layer; S34, flatten the multi-dimensional spatial features into one-dimensional damage information through the damage prediction layer, and output the damage identification result.

[0008] Furthermore, in the cross-domain structural damage identification model, the time domain feature extraction layer is a one-dimensional convolutional neural network, which extracts corresponding time domain features by performing a separate convolution operation on the time domain vibration data of each sensor.

[0009] Furthermore, in the cross-domain structural damage identification model, the spatial feature extraction layer extracts corresponding multi-dimensional spatial features from the input graph structure; In the graph structure, the graph nodes correspond to the positions of the sensors, the features of the graph nodes are the extracted time domain features, and the connection relationship of the graph nodes is the spatial position association of the sensors represented by the adjacency matrix; The spatial feature extraction layer includes a first graph convolution layer, a first ReLU activation function, a second graph convolution layer, a second ReLU activation function and a global average pooling layer connected in sequence; Graph attention mechanisms are embedded in both the first graph convolutional layer and the second graph convolutional layer.

[0010] Furthermore, in the cross-domain structural damage identification model, the damage prediction layer includes a fully connected layer and a Softmax function layer, which respectively flattens the input multi-dimensional spatial features in one dimension and maps the damage prediction score value to the [0,1] interval to obtain a damage identification result.

[0011] Furthermore, the training loss function of the cross-domain structural damage identification model is for:

[0012]

[0013] In the formula, and Respectively represent samples belong to the true damage label The score and predicted probability values ​​of Indicates The true damage label of the sample is , represents the total number of samples, Represents the total number of true damage labels.

[0014] Furthermore, in step S3, the method of using the spatiotemporal data set to train the spatiotemporal feature learning network model to obtain the cross-domain structural damage recognition model is: The spatiotemporal feature learning network model is trained by using a simulated domain data set to obtain a pre-trained spatiotemporal feature learning network model; The pre-trained spatiotemporal feature learning network model is fine-tuned through the measured domain dataset to obtain a cross-domain structural damage recognition model.

[0015] Further, in step S4, the time domain adaptive model includes a first fast Fourier transform, a one-dimensional deep convolutional neural network and a second Fourier transform; The step S4 comprises the following sub-steps: S41, performing a first fast Fourier transform on the time domain vibration data of the two structures respectively to obtain corresponding frequency domain data; S42, inputting the frequency domain data of the two structures into the corresponding one-dimensional deep convolutional neural network for data alignment; S43, performing a second Fourier transform on the aligned frequency domain data to restore it to the time domain to obtain adaptive time domain data.

[0016] Furthermore, in step S4, the loss function of the time domain adaptive model is an MMD loss function.

[0017] The beneficial effects of the present invention are: Existing studies on using two-domain data for transfer learning to improve the accuracy of structural damage identification have not taken into account the spatial characteristics between the two-domain structures. The present invention proposes the SfLN model, which integrates the spatial characteristic information between the time domain and key measurement points of the two-domain data, thereby greatly improving the accuracy of cross-domain structural damage identification and providing a new solution for cross-structure damage identification. The specific beneficial effects are as follows: (1) A cross-domain transfer learning damage identification technology solution is proposed that integrates the digital twin model and the spatiotemporal characteristics of structural multi-sensor data to solve the problem of incomplete sensor data information (without considering the spatial feature information of the sensor) in the field of structural damage identification.

[0018] (2) A spatiotemporal feature learning model is proposed that can simultaneously process the temporal domain data and spatial feature information of multiple sensors, providing a specific model architecture for the deep fusion of the structural digital twin model and the spatiotemporal features of multi-sensor data of the actual structure.

[0019] (3) Based on the structure digital twin model and the spatiotemporal feature learning model trained with measured data, a technical architecture that can be used for cross-structure damage identification is proposed to solve problems such as the scarcity of damage data in actual structures and the cross-structure damage identification task with the same sensor measurement point arrangement. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Flowchart of the cross-domain structural damage identification method based on digital twins and spatiotemporal feature fusion provided by the present invention.

[0021] Figure 2 This is a structural diagram of the spatiotemporal feature learning network model (SfLN) provided by the present invention. DETAILED DESCRIPTION

[0022] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0023] The embodiment of the present invention provides a cross-domain structural damage identification method based on digital twin and spatiotemporal feature fusion, such as Figure 1 As shown, the following steps are included: S1. Build a digital twin model corresponding to the actual engineering structure and optimize the sensor layout based on the digital twin model; S2, construct a two-domain spatiotemporal dataset of actual engineering structures under various damage scenarios; S3, construct a spatiotemporal feature learning network model, and train it using the spatiotemporal dataset to obtain a cross-domain structural damage recognition model; Among them, the cross-domain structural damage identification model includes an input layer, a time domain feature extraction layer, a spatial feature extraction layer, and a damage prediction layer; S4, processing the time domain vibration data of the two structures to be identified for damage by using a time domain adaptive model to obtain adaptive time domain data; S5. Input the adaptive time domain data and its corresponding adjacency matrix into the cross-domain structural damage identification model for processing to obtain a cross-structure damage identification result between the two structures.

[0024] In step S1 of the embodiment of the present invention, a digital twin model of the actual engineering structure is constructed based on its engineering data, size, material, environment and other parameters, and the sensors are optimally arranged using information such as the modal vibration shape data of the digital twin model to obtain a reasonable sensor arrangement plan.

[0025] In step S2 of an embodiment of the present invention, a small amount of measured data under various damage scenarios in actual engineering structures and a large amount of simulated data under various damage scenarios generated by digital twin models are collected, and a spatiotemporal dataset is constructed based on the spatial position information of each sensor; specifically, the two-domain spatiotemporal dataset in this embodiment includes a measured domain dataset and a simulated domain dataset; the measured domain dataset includes the measured time-domain vibration data of the actual engineering structure under various damage scenarios and its corresponding adjacency matrix; the simulated domain dataset includes the simulated time-domain vibration data of the actual engineering structure under various damage scenarios generated by the corresponding digital twin model and its corresponding adjacency matrix; wherein the adjacency matrix represents the spatial position association of the sensors after the optimized arrangement.

[0026] In step S3 of the embodiment of the present invention, the spatiotemporal feature learning network model (SfLN) is as follows: Figure 2 As shown in the figure, the cross-domain structural damage recognition model obtained by training includes an input layer, a time domain feature extraction layer, a spatial feature extraction layer and a damage prediction layer; the spatial position of the sensor is regarded as a "graph" node, and the spatial distribution association between the sensors is regarded as the edge of the "graph". The local time domain vibration data and the global node "graph" information are processed synchronously through a one-dimensional convolutional network, a graph convolutional network and a graph attention mechanism, and the common characteristics of the time domain data of the two domain structures and the spatial characteristics of the key nodes of the structure are extracted from the local to the whole, thereby realizing the structural damage feature learning and task transfer that integrates the multi-dimensional feature information of the structure in time and space, which is expected to improve the accuracy of structural damage recognition and reliability analysis driven by cross-domain data.

[0027] Specifically, in this embodiment, the method for the cross-domain structural damage identification model to process input data is as follows: S31, receiving time domain vibration data and an adjacency matrix through an input layer, and constructing them into a spatiotemporal feature matrix; S32, performing a separate convolution on the time domain vibration data of each sensor in the spatiotemporal feature matrix through a time domain feature extraction layer to extract corresponding time domain features; S33, based on the extracted time domain features, extracting corresponding multi-dimensional spatial features through a spatial feature extraction layer; S34, flatten the multi-dimensional spatial features into one-dimensional damage information through the damage prediction layer, and output the damage identification result.

[0028] In step S31 of this embodiment, the input layer is an important layer for collecting structural spatiotemporal feature data in the cross-domain structural damage identification model, and mainly inputs the time domain vibration data matrix collected under different damage scenarios of the structure. and an adjacency matrix reflecting the spatial position association of the sensor, wherein the time domain vibration data matrix only reflects the time-course response information on the key nodes of the structure, and does not include the spatial attributes of the sensor, while the adjacency matrix only reflects the spatial relationship of the key nodes of the structure, and does not include any sensor time-course data. To facilitate the processing of subsequent modules, the two are combined in step S31 to construct a spatiotemporal feature matrix that contains both structural time-course data and node spatial feature information, which is expressed as:

[0029] in, yes dimensional time domain data matrix ( represents the number of sensors, represents the number of points for collecting data), yes dimensional adjacency matrix (the adjacency matrix is ​​an important step in graph convolution and an important step in integrating spatial information. The adjacency matrix converts the physical connection relationship between sensors into a matrix. For example, if sensor 1 and sensor 2 are physically connected, the adjacency matrix constructs A 12 =1), It contains complete spatiotemporal feature information under different damage scenarios of the structure, which is then input into the subsequent structure of SfLN, which is conducive to mining the constraint relationship between the time domain response of key nodes of the structure and its spatial characteristics, and deeply integrating the spatiotemporal feature information of the structure for damage feature extraction and identification.

[0030] In step S32 of this embodiment, in the cross-domain structural damage identification model, the time domain feature extraction layer is a one-dimensional convolutional neural network (1DCNN), which extracts the corresponding time domain features by performing a separate convolution operation on the time domain vibration data of each sensor.

[0031] Specifically, taking the sensor data of a channel under a certain structural damage scenario as an example, the calculation method is as follows:

[0032] Among them, * represents the convolution operation, Represents the acceleration vibration response data of a certain sensor, represents the convolution kernel, represents the bias, Represents the output time domain characteristic data of the sensor.

[0033] In step S33 of this embodiment, in the cross-domain structural damage identification model, the spatial feature extraction layer extracts corresponding multi-dimensional spatial features from the input graph structure; in the graph structure, the graph nodes correspond to the positions of the sensors, the features of the graph nodes are the extracted time domain features, and the connection relationship of the graph nodes is the spatial position association of the sensors represented by the adjacency matrix; The spatial feature extraction layer includes the first graph convolution layer (GCN1), the first ReLU activation function (ReLU1), the second graph convolution layer (GCN2), the second ReLU activation function (ReLU2) and the global average pooling layer (GAP) connected in sequence; at the same time, the first graph convolution layer (GCN1) and the second graph convolution layer (GCN2) are embedded with the graph attention mechanism.

[0034] Among them, by embedding the graph attention mechanism, we can better pay attention to the importance of each sensor measurement point and further assist the graph convolutional network to better extract the spatial information of the structure.

[0035] In step S34 of the present embodiment, in the cross-domain structural damage identification model, the damage prediction layer includes a fully connected layer and a Softmax function layer, which respectively perform one-dimensional flattening on the input multi-dimensional spatial features and map the damage prediction score value to the [0,1] interval to obtain a damage identification result.

[0036] Specifically, the structural damage identification using deep learning in this embodiment can essentially be attributed to the multi-classification problem of mapping the damage location or degree from the structural response data. In the present invention, after the time domain feature matrix is ​​processed by the spatial feature extraction module, the multi-dimensional spatial features are obtained, which need to be flattened into one-dimensional damage information to complete the damage prediction. Therefore, the damage prediction layer mainly includes two parts: the fully connected layer and the Softmax function layer, which respectively complete the information flattening and map the damage prediction score value to the [0,1] interval to obtain the damage identification result.

[0037] In the embodiment of the present invention, the calibration method of the real damage label of the structure under consideration always satisfies the probability distribution. The damage prediction score value output by the fully connected layer may not meet this condition, resulting in the incomparability between the damage prediction value and the real value. To this end, the Softmax function is first used to map the damage prediction score value to the probability distribution in the [0,1] interval, and then the cross entropy loss function is used to measure the difference between the probability distribution of the damage prediction and the real damage, and the SfLN model performance is reversely optimized with the minimum loss value as the goal. Therefore, the training loss function of the cross-domain structural damage recognition model in this embodiment is for:

[0038]

[0039] In the formula, and Respectively represent samples belong to the true damage label The score and predicted probability values ​​of Indicates The true damage label of the sample is , represents the total number of samples, Represents the total number of true damage labels.

[0040] After the above Softmax function layer and the calculation of the loss function, the SfLN model's ability to distinguish different damage data can be measured, and the loss function value can directly reflect the parameter update process of the SfLN model. When a large loss value is obtained, it indicates that the model has a poor prediction effect on individual damage categories, prompting the model to further strengthen the learning of such damage features. In addition, during the training process, the Softmax cross entropy loss can provide a stable gradient, helping the SfLN model maintain good stability and robustness when facing different damage features.

[0041] In step S3 of the embodiment of the present invention, a transfer learning strategy is adopted to train a spatiotemporal feature learning network model using a spatiotemporal data set to obtain a cross-domain structural damage recognition model: The spatiotemporal feature learning network model is trained by using a simulated domain data set to obtain a pre-trained spatiotemporal feature learning network model; The pre-trained spatiotemporal feature learning network model is fine-tuned through the measured domain dataset to obtain a cross-domain structural damage recognition model.

[0042] In step S4 of the embodiment of the present invention, when completing the cross-structure loss identification task, the time domain acceleration data of the two structures need to be first adapted to the time domain data domain. The specific method of time domain adaptation is: the time domain vibration signals collected in the healthy state of the two structures are subjected to fast Fourier transformation (Fast Fourier Transformation, FFT) to obtain frequency domain data, and then the frequency domain data is sent to a one-dimensional deep convolutional neural network model (One-Dimensional DeepConvolutional Neural Network, 1DDCNN) to bring the data distribution characteristics between the two structures closer. The brought-closed data is then restored to the time domain through an inverse fast Fourier transform (Inverse Fast Fourier Transform, IFFT).

[0043] Therefore, in step S4 of the embodiment of the present invention, a time domain adaptive model is designed, which includes a first fast Fourier transform, a one-dimensional deep convolutional neural network and a second Fourier transform; specifically, the method for obtaining adaptive time domain data in step S4 is specifically: S41, performing a first fast Fourier transform on the time domain vibration data of the two structures respectively to obtain corresponding frequency domain data; S42, inputting the frequency domain data of the two structures into the corresponding one-dimensional deep convolutional neural network for data alignment; S43, performing a second Fourier transform on the aligned frequency domain data to restore it to the time domain to obtain adaptive time domain data.

[0044] In this embodiment, the loss function of the time domain adaptive model is an MMD loss function, which measures the degree of data alignment by using the Maximum Mean Discrepancy (MMD) method.

[0045] In this embodiment, MMD is used as the loss function of the time domain adaptive model to guide its training, and the trained model is saved. The data of the two structures in all states collected subsequently are all passed through this time domain adaptive model to reduce the gap between the time domain data of the two structures. Finally, the spatial feature data and the processed time domain data are sent to SfLN again, and cross-structure damage identification between the two structures can be realized.

[0046] It should be noted that the cross-domain structural damage identification method provided in the present invention can be used for structural damage identification across measured domains and simulation domains (cross-data domain), and can also be used for cross-structural damage identification between two similar structures (cross-structural domain).

[0047] In an embodiment of the present invention, an effect verification example of the above-mentioned cross-domain structural damage identification method is provided. By comparing a two-dimensional convolutional neural network model (2DCNN) that does not consider spatial features and a SfLN model using the same data, a damage identification accuracy comparison table is shown in Table 1.

[0048] Table 1 Comparison of accuracy, recall and F1 value

[0049] In the actual experiment, we set up four specific damage scenarios, including bolt loosening and bolt falling off, and used hammer excitation to simulate the natural excitation of the actual structure. Seven sensors were arranged on the structure to collect vibration acceleration signals at different positions. It can be seen from Table 1 that the recognition accuracy and various indicators of the SfLN in the present invention are significantly higher than those of the two-dimensional convolutional neural network model (2DCNN) without considering spatial features.

[0050] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

[0051] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A cross-domain structural damage identification method based on digital twins and spatiotemporal feature fusion, characterized in that: The following steps are involved: S1. Build a digital twin model corresponding to the actual engineering structure and optimize the sensor layout based on the digital twin model; S2, construct a two-domain spatiotemporal dataset of actual engineering structures under various damage scenarios; The two-domain spatiotemporal data sets include a measured domain data set and a simulated domain data set; the measured domain data set includes measured time-domain vibration data of the actual engineering structure under various damage scenarios and its corresponding adjacency matrix; the simulated domain data set includes simulated time-domain vibration data generated by the corresponding digital twin model of the actual engineering structure under various damage scenarios and its corresponding adjacency matrix; the adjacency matrix represents the spatial position association of the sensors after the optimized arrangement; S3, construct a spatiotemporal feature learning network model, and train it using the spatiotemporal dataset to obtain a cross-domain structural damage recognition model; The cross-domain structural damage identification model includes an input layer, a time domain feature extraction layer, a spatial feature extraction layer and a damage prediction layer; S4, processing the time domain vibration data of the two structures to be identified for damage by using a time domain adaptive model to obtain adaptive time domain data; S5. Input the adaptive time domain data and its corresponding adjacency matrix into the cross-domain structural damage identification model for processing to obtain a cross-domain structural damage identification result.

2. The cross-domain structural damage identification method based on digital twin and spatiotemporal feature fusion according to claim 1 is characterized in that: In step S3, the method for the cross-domain structural damage identification model to process input data is specifically as follows: S31, receiving time domain vibration data and an adjacency matrix through an input layer, and constructing them into a spatiotemporal feature matrix; S32, performing a separate convolution on the time domain vibration data of each sensor in the spatiotemporal feature matrix through a time domain feature extraction layer to extract corresponding time domain features; S33, based on the extracted time domain features, extracting corresponding multi-dimensional spatial features through a spatial feature extraction layer; S34, flatten the multi-dimensional spatial features into one-dimensional damage information through the damage prediction layer, and output the damage identification result.

3. The cross-domain structural damage identification method based on digital twin and spatiotemporal feature fusion according to claim 1 is characterized in that: In the cross-domain structural damage identification model, the time domain feature extraction layer is a one-dimensional convolutional neural network, which extracts the corresponding time domain features by performing a separate convolution operation on the time domain vibration data of each sensor.

4. The cross-domain structural damage identification method based on digital twin and spatiotemporal feature fusion according to claim 1 is characterized in that: In the cross-domain structural damage identification model, the spatial feature extraction layer extracts corresponding multi-dimensional spatial features from the input graph structure; In the graph structure, the graph nodes correspond to the positions of the sensors, the features of the graph nodes are the extracted time domain features, and the connection relationship of the graph nodes is the spatial position association of the sensors represented by the adjacency matrix; The spatial feature extraction layer includes a first graph convolution layer, a first ReLU activation function, a second graph convolution layer, a second ReLU activation function and a global average pooling layer connected in sequence; Graph attention mechanisms are embedded in both the first graph convolutional layer and the second graph convolutional layer.

5. The cross-domain structural damage identification method based on digital twin and spatiotemporal feature fusion according to claim 1 is characterized in that: In the cross-domain structural damage identification model, the damage prediction layer includes a fully connected layer and a Softmax function layer, which respectively flattens the input multi-dimensional spatial features in one dimension and maps the damage prediction score value to the [0,1] interval to obtain a damage identification result.

6. The cross-domain structural damage identification method based on digital twin and spatiotemporal feature fusion according to claim 1 is characterized in that: The training loss function of the cross-domain structural damage identification model is for: In the formula, and Respectively represent samples belong to the true damage label The score and predicted probability values ​​of Indicates The true damage label of the sample is , represents the total number of samples, Represents the total number of true damage labels.

7. The cross-domain structural damage identification method based on digital twin and spatiotemporal feature fusion according to claim 1 is characterized in that: In step S3, the method of using the spatiotemporal data set to train the spatiotemporal feature learning network model to obtain the cross-domain structural damage identification model is: The spatiotemporal feature learning network model is trained by using a simulated domain data set to obtain a pre-trained spatiotemporal feature learning network model; The pre-trained spatiotemporal feature learning network model is fine-tuned through the measured domain dataset to obtain a cross-domain structural damage recognition model.

8. The cross-domain structural damage identification method based on digital twin and spatiotemporal feature fusion according to claim 1 is characterized in that: In step S4, the time-domain adaptive model includes a first fast Fourier transform, a one-dimensional deep convolutional neural network, and a second Fourier transform; The step S4 comprises the following sub-steps: S41, performing a first fast Fourier transform on the time domain vibration data of the two structures respectively to obtain corresponding frequency domain data; S42, inputting the frequency domain data of the two structures into the corresponding one-dimensional deep convolutional neural network for data alignment; S43, performing a second Fourier transform on the aligned frequency domain data to restore it to the time domain to obtain adaptive time domain data.

9. The cross-domain structural damage identification method based on digital twin and spatiotemporal feature fusion according to claim 8 is characterized in that: In step S4, the loss function of the time domain adaptive model is an MMD loss function.

Images