A method, system, medium, and computer program for ancient bridge damage identification based on an improved directed graph convolutional network.

Abstract

This invention relates to a method, system, medium, and computer program for identifying damage to ancient bridges based on an improved directed graph convolutional network. The damage identification method includes: constructing a training dataset based on ancient bridge monitoring and detection information; deriving a dynamic equation considering damage based on possible damage conditions of the ancient bridge, and solving for the target solution set by integration; calculating a damage sensitivity matrix based on the target solution set, reflecting the sensitivity relationship between damage and test response; correcting the sensitivity matrix according to sensor type, and using the corrected sensitivity matrix as a weighting factor for the directed graph; then, performing time-domain differencing on the training data to expand the features in the time domain dimension of the data; finally, establishing a graph neural network for training and using it for ancient bridge damage identification. Compared with existing technologies, this invention can effectively identify damage patterns of ancient bridges, with high accuracy, precision, recall, and F1 score, providing scientific support for the preventive protection of bridge cultural relics.

Description

Technical Field

[0001] This invention relates to the field of civil engineering, and in particular to a method for identifying damage to ancient bridges based on an improved directed graph convolutional network. Background Technology

[0002] Ancient bridges, as witnesses to history and inheritors of culture, carry rich historical information and artistic value. However, due to long-term exposure to the natural environment, including sun and rain, these cultural relics suffer from damage such as cracks and deformation. Therefore, identifying this damage is crucial for the protection and preservation of these precious cultural heritages.

[0003] However, the current damage identification based on structural health monitoring mainly has the following problems: (1) Some damage characteristic tests of ancient bridges, such as natural vibration frequency, may not be sensitive to local or small damage, and more sensitive feature quantities need to be extracted from dynamic response signals. In fact, many monitoring projects of ancient bridges often only have low-frequency static data. (2) The measured data are greatly affected by environmental noise, and a lot of effort needs to be invested in the decoupling of test information and environmental factors. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a method, system, medium, and computer program for identifying ancient bridge damage based on an improved directed graph convolutional network. Based on data from various types of sensors, it can achieve bridge damage identification that takes into account environmental impacts under conditions of low sampling frequency and long monitoring period.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] The first aspect of this invention provides a method for identifying damage to ancient bridges based on an improved directed graph convolutional network, comprising the following steps:

[0007] S1: Based on the information obtained from continuous monitoring and on-site detection of the ancient bridge, a training dataset is constructed. In the training dataset, the input data is data collected from different types of detection sensors, and the output data is a binary array of labels.

[0008] S2: Based on the linear elastic dynamic equation, and considering the possible damage in the ancient bridge structure, the corresponding dynamic equation considering damage is derived. Then, by integrating the dynamic equation, the target solution set is obtained.

[0009] S3: Based on the obtained target solution set, further calculate the damage sensitivity matrix, which is used to reflect the sensitivity relationship between damage and monitoring response;

[0010] S4: Based on the actual test content of different types of sensors, correct the sensitivity matrix. Find the maximum / minimum value of the updated sensitivity matrix in the time dimension. Then, calculate the Pearson coefficient matrix of the sensitivity matrix in the damage dimension. The calculated Pearson coefficient matrix is ​​the weight factor matrix of the directed graph.

[0011] S5: Establish a graph neural network, train the graph neural network, and use the graph neural network to identify damage to ancient bridges.

[0012] Furthermore, in S1, in the training dataset, the input of the training set is the monitoring data of different types of sensors within a preset time period, and the output label data is a series of binary arrays with values ​​of 0 or 1, and the dimension of the label array is the number of damage categories detected.

[0013] Furthermore, in S2, the linear elastic dynamic equation is shown in equation (1):

[0014]

[0015] Where M, C, and K represent the mass matrix, damping matrix, and stiffness matrix, respectively;

[0016] x represents the acceleration, velocity, and displacement response arrays, respectively;

[0017] F represents the input stimulus.

[0018] Furthermore, in S2, when there are n e In the finite element method with 1 element, the parameter reduction matrix is ​​denoted as {γ}. k}, where k∈[1,n e ], and has γ k ∈[0,1],γ k The damage to element k is represented by the damage dynamics equation (2):

[0019]

[0020] Among them, M k,0 and K k,0 Let represent the mass and stiffness matrices of element k when it is undamaged, respectively.

[0021] α and β are Rayleigh damping constants;

[0022] Current damage target solution set By integrating equation (2), we can obtain the result, where A represents the selection matrix and represents the degree of freedom of the corresponding sensor placement location.

[0023] Furthermore, in S3, the damage sensitivity matrix is ​​Y with respect to damage γ. kThe sensitivity matrix is ​​calculated using equation (3):

[0024]

[0025] Among them, S l (γ) represents the response Y of degree of freedom l. l , responding to Y l There are a total of t τ A discrete value;

[0026] The result is obtained using the central difference method shown in equation (4):

[0027]

[0028] Furthermore, in S4, the specific process includes:

[0029] Based on the sensitivity matrix, the solution set is determined according to the sensor type. Transform into corresponding test content The updated sensitivity matrix (5) is now:

[0030]

[0031] in This indicates finding the maximum value of the array over time, where Ns represents the number of sensor channels;

[0032] The coefficients of the Pearson matrix in the γ dimension of equation (5) are obtained using equation (6). This indicates that the correlation between sensors is determined based on the strength of the correlation between the damage sensitivity matrices of each sensor.

[0033]

[0034] Furthermore, in S5, the graph neural network is an improved graph neural network using a time-domain difference operator. The improvement process includes:

[0035] Add a difference layer to the graph neural network, treating the sensor-sampled data as a graph. The difference layer differs the sensor data along the time axis to expand the features of the graph in the channel dimension. The difference method used is shown in equation (7):

[0036]

[0037] in Indicates that sensor q at t i Test data at any given time, This represents the m-th order difference, where zeros are padded in empty positions in the array.

[0038] A second aspect of the present invention provides a system for identifying damage to ancient bridges based on an improved directed graph convolutional network, comprising:

[0039] Data acquisition and labeling module: Based on the information obtained from continuous monitoring and on-site detection of the ancient bridge, a training dataset is constructed. In the training dataset, the input data is data collected from different types of detection sensors, and the output data is a binary array label.

[0040] Damage dynamics equation solving module: Based on the linear elastic dynamics equation, and considering the possible damage in the ancient bridge structure, the module derives the corresponding dynamic equation that takes damage into account, and then solves the target solution set by integrating the dynamic equation.

[0041] Damage sensitivity calculation module: Based on the obtained target solution set, further calculate the damage sensitivity matrix, which is used to reflect the sensitivity relationship between damage and other relevant factors;

[0042] The directed graph matrix solving module corrects the sensitivity matrix according to the actual test content of different types of sensors. It then finds the maximum value of the updated sensitivity matrix in the time dimension. Next, it calculates the Pearson coefficient matrix of the sensitivity matrix in the damage dimension. The resulting Pearson coefficient matrix is ​​the weight factor matrix of the directed graph.

[0043] The ancient bridge damage identification module uses a dataset to train a graph neural network, which is then used to identify damage to the ancient bridge.

[0044] A third aspect of 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 above-described method for identifying ancient bridge damage based on an improved directed graph convolutional network.

[0045] A fourth aspect of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the above-described ancient bridge damage identification method based on an improved directed graph convolutional network.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] 1) The present invention proposes an ancient bridge damage identification method based on an improved directed graph convolutional network. It adopts the Pearson coefficient matrix with the maximum damage sensitivity in the time domain to replace the Pearson coefficient matrix of the original sensor data, which can effectively eliminate the influence of environmental factors on the construction of the directed graph matrix.

[0048] 2) High-dimensional features in the temporal dimension of the sensor were constructed using differential layering, expanding its temporal characteristics. High-dimensional features in the spatial and mechanical dimensions of the sensor were constructed using directed graphs and graph convolutional networks, enabling the network to fully learn the potential patterns in the data. Attached Figure Description

[0049] Figure 1 This is a flowchart of the ancient bridge damage identification method based on an improved directed graph convolutional network in this invention. Detailed Implementation

[0050] Overall, this invention discloses a method for identifying damage to ancient bridges based on an improved directed graph convolutional network. First, an improved method for constructing directed graph matrices is established to avoid environmental influences encountered during the construction process. Then, a temporal difference-improved graph convolutional neural network is built to identify damage, expanding the features in the temporal dimension of the data. The proposed method can effectively identify damage patterns in ancient bridges, exhibiting high accuracy, precision, recall, and F1 score. This invention can provide scientific support for the preventative protection of bridge cultural relics.

[0051] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Component models, material names, connection structures, control methods, algorithms, and other features not explicitly described in this technical solution are considered common technical features disclosed in the prior art.

[0052] Example 1

[0053] This embodiment discloses a method for identifying damage to ancient bridges based on an improved directed graph convolutional network. See [link to relevant documentation]. Figure 1 This includes the following steps:

[0054] Step 1: Establish a structural health monitoring system for the ancient bridge, conduct long-term monitoring, and carry out on-site damage detection every month in the early stages of the monitoring project. Based on the monitoring and detection results, construct a training dataset. The input to the training set consists of different types of sensor data over a certain period of time, and the output label data is a series of binary arrays with values ​​of 0 or 1, where the dimension of the label array is the number of damage categories detected.

[0055] Step 2: The linear elastic dynamic equation is shown in equation (1):

[0056]

[0057] Where M, C, and K represent the mass matrix, damping matrix, and stiffness matrix, respectively. x represents the acceleration, velocity, and displacement response arrays, respectively. F represents the input excitation. For n... e For a finite element method with elements, the parameter reduction matrix can be denoted as {γ}.k}, where k∈[1,n e ], and has γ k ∈[0,1]. γ k Let k represent the damage to element k. Then, the dynamic equation considering the damage can be transformed into equation (2).

[0058]

[0059] Among them, M k,0 and K k,0 Let represent the mass and stiffness matrices of element k without damage, respectively. α and β are Rayleigh damping constants. At this point, the target solution set for the current damage is... It can be obtained by integrating equation (2). Where A represents the selection matrix, and represents the degree of freedom of the corresponding sensor placement location. Step 3: Y with respect to damage γ k The sensitivity matrix can be calculated by equation (3).

[0060]

[0061] Among them, S l (γ) represents the response Y of degree of freedom l. l (This response has a total of t) τ (discrete values), regarding damage γ k The sensitivity value. It can be obtained by the central difference method shown in equation (4).

[0062]

[0063] Step 4: Calculate the sensitivity matrix obtained from equations (3) and (4), and then determine the solution set according to the sensor type. Transform into corresponding test content For example, the displacement is removed to calculate the height, and the inverse trigonometric angle is obtained. Taking the maximum value of the updated sensitivity matrix in the time dimension, we obtain equation (5).

[0064]

[0065] in This indicates finding the maximum value of the array in the time dimension, where Ns represents the number of sensor channels. At this point, equation (6) is used to calculate the coefficients of the Pearson matrix in the γ dimension of equation (5). This indicates that the correlation between sensors is determined based on the correlation strength of the damage sensitivity matrices of each sensor. Compared with directly calculating the Pearson matrix between the original sensor data, it has two advantages: (1) It is calculated based on finite element theory, which can avoid the confusion of the correlation between signals by environmental factors; (2) The correlation coefficient obtained through the damage sensitivity value can better reflect the correlation behavior between different sensors when the structural damage evolves.

[0066]

[0067] Step 5: The established graph neural network can be improved using a temporal difference operator. The sensor-sampled data is regarded as a graph, and the purpose of the difference layer is to expand the features of the graph in the channel dimension by differentiating the sensor data on the time axis. The difference method used is shown in Equation (7).

[0068]

[0069] in This represents the test data of sensor q at time ti. This represents m-order differences. Because differences can lead to data misalignment, zeros are padded in the missing positions. Thus, the original single-channel sensor image data can be transformed into m-channel image data. Training is completed on the training set until the maximum number of training epochs is reached. Then, the network is tested on the test set using the network with the optimal number of epochs to verify its performance.

[0070] This embodiment also provides an ancient bridge damage identification system based on an improved directed graph convolutional network, including:

[0071] Data acquisition and labeling module: Based on the information obtained from continuous monitoring and on-site detection of the ancient bridge, a training dataset is constructed. In the training dataset, the input data is data collected from different types of detection sensors, and the output data is a binary array label.

[0072] Damage dynamics equation solving module: Based on the linear elastic dynamics equation, and considering the possible damage in the ancient bridge structure, the module derives the corresponding dynamic equation that takes damage into account, and then solves the target solution set by integrating the dynamic equation.

[0073] Damage sensitivity calculation module: Based on the obtained target solution set, further calculate the damage sensitivity matrix, which is used to reflect the sensitivity relationship between damage and other relevant factors;

[0074] The directed graph matrix solving module corrects the sensitivity matrix according to the actual test content of different types of sensors. It then finds the maximum value of the updated sensitivity matrix in the time dimension. Next, it calculates the Pearson coefficient matrix of the sensitivity matrix in the damage dimension. The resulting Pearson coefficient matrix is ​​the weight factor matrix of the directed graph.

[0075] The ancient bridge damage identification module uses a dataset to train a graph neural network, which is then used to identify damage to the ancient bridge.

[0076] In the various embodiments of the present invention, the functional modules can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional units.

[0077] This embodiment also provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the steps of the above-described method for identifying ancient bridge damage based on an improved directed graph convolutional network. The storage medium can be an electronic medium, magnetic medium, optical medium, electromagnetic medium, infrared medium, or a semiconductor system or propagation medium. The storage medium may also include semiconductor or solid-state memory, magnetic tape, removable computer disk, random access memory (RAM), read-only memory (ROM), hard disk, and optical disc. Optical discs may include optical disc-read-only memory (CD-ROM), optical disc-read / write (CD-RW), and DVD.

[0078] This embodiment also provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned ancient bridge damage identification method based on an improved directed graph convolutional network. When the computer program runs on the processor, it efficiently implements the aforementioned ancient bridge damage identification method based on an improved directed graph convolutional network. Starting with collecting continuous monitoring and on-site inspection information of the ancient bridge to construct a training dataset, it derives and solves a dynamic equation considering damage based on the linear elastic dynamic equation, then calculates the damage sensitivity matrix, utilizes the solved directed graph matrix, and finally uses the dataset to train a graph neural network to achieve accurate identification of ancient bridge damage.

[0079] Verification Example 1

[0080] To evaluate the benefits of this invention, the following verification example is used. In this case, a structural health monitoring system for an ancient bridge is constructed, including 2 dual-channel accelerometers, 2 dual-channel tilt sensors, 2 settlement sensors, 4 displacement sensors, 10 strain sensors, 1 temperature sensor, and 1 humidity sensor, totaling 26 sensor channels. During testing, samples are taken once per hour, 24 times per day, for a total of 32 months. To expand the training data, the original data is first linearly interpolated, and then, using 100 data points as a baseline, the interpolated data is sampled with a 60% overlap rate. Each 100 consecutive data points are then considered as one sample, resulting in 3488 training data points. These 3488 training data points are sequentially divided into 10 subsets for 10-fold cross-validation. A spatiotemporal difference graph-based convolutional neural network is used for training. The Gated Conv module includes two 64-channel output layers (the last Gated Conv outputs 16 channels), with a 1×3 two-dimensional convolutional kernel. The output of the Gated Conv module is fed into a graph convolutional layer. After passing through a ReLU activation function, the output is fed back into the Gated Conv module, and then applied to a residual module (1×1 convolutional kernel). This process is repeated once. The final output, after being processed once by the Gated Conv module, is then subjected to adaptive average pooling and passed through two fully connected layers (FC layers, 256 hidden neurons). The result is then Sigmoid-encoded and multiple-hot encoded with a threshold of 0.5, converting the output into a binary matrix of 0s and 1s. The learning rate is 0.001, the training epochs are 200, the total difference order is 4, and the batch size is 64.

[0081] The calculation results show that on the test set, the complete accuracy prediction rate can reach 92.26%, the average precision is 93.5%, the average recall rate is 90.4%, the average false alarm rate is 3.0%, and the average F1 score is 0.917, indicating that the obtained model can identify damage relatively accurately on the test set.

[0082] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0083] The above description of the embodiments is provided to enable those skilled in the art to understand and use the invention. It will be apparent to those skilled in the art that various modifications can be made to these embodiments, and the general principles described herein can be applied to other embodiments without inventive effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the invention should be within the protection scope of the present invention.

Claims

1. A method for identifying damage to ancient bridges based on an improved directed graph convolutional network, characterized in that, Includes the following steps: S1: Based on the information obtained from continuous monitoring and on-site detection of the ancient bridge, a training dataset is constructed. In the training dataset, the input data is data collected from different types of detection sensors, and the output data is a binary array of labels. S2: Based on the linear elastic dynamic equation, and considering the possible damage in the ancient bridge structure, the corresponding dynamic equation considering damage is derived. Then, by integrating the dynamic equation, the target solution set is obtained. S3: Based on the obtained target solution set, further calculate the damage sensitivity matrix, which is used to reflect the sensitivity relationship between damage and monitoring response; S4: Based on the actual test content of different types of sensors, correct the sensitivity matrix, take the maximum value of the updated sensitivity matrix in the time dimension, and then calculate the Pearson matrix coefficient matrix of the sensitivity matrix in the damage dimension. The calculated Pearson matrix coefficient matrix is ​​the weight factor matrix of the directed graph. S5: Establish a graph neural network, train the graph neural network, and use the graph neural network to identify damage to ancient bridges.

2. The ancient bridge damage identification method based on an improved directed graph convolutional network according to claim 1, characterized in that, In S1, the input to the training dataset is monitoring data from different types of sensors within a preset time period, and the output label data is a series of binary arrays with values ​​of 0 or 1. The dimension of the label array is the number of damage categories detected.

3. The ancient bridge damage identification method based on an improved directed graph convolutional network according to claim 1, characterized in that, In S2, the linear elastic dynamic equation is shown in equation (1): Where M, C, and K represent the mass matrix, damping matrix, and stiffness matrix, respectively; x represents the acceleration, velocity, and displacement response arrays, respectively; F represents the input stimulus.

4. The ancient bridge damage identification method based on an improved directed graph convolutional network according to claim 3, characterized in that, In S2, when there are ne elements in the finite element method, the parameter reduction matrix is ​​denoted as {γ}. k }, where k∈[1,n e ], and has γ k ∈[0,1],γ k The damage to element k is represented by the damage dynamics equation (2): Among them, M k,0 and K k,0 These represent the mass and stiffness matrices of element k when it is undamaged, respectively. α and β are Rayleigh damping constants; Current damage target solution set By integrating equation (2), we can obtain the result, where A represents the selection matrix and represents the degree of freedom of the corresponding sensor placement location.

5. The ancient bridge damage identification method based on an improved directed graph convolutional network according to claim 4, characterized in that, In S3, the damage sensitivity matrix is ​​Y with respect to damage γ k The sensitivity matrix is ​​calculated using equation (3): Among them, S l (γ) represents the response Y of degree of freedom l. l , responding to Y l There are a total of t τ A discrete value; The result is obtained using the central difference method shown in equation (4):

6. The ancient bridge damage identification method based on an improved directed graph convolutional network according to claim 5, characterized in that, In S4, the specific process includes: Based on the sensitivity matrix, the solution set is determined according to the sensor type. Transform into corresponding test content The updated sensitivity matrix (5) is now: in This indicates finding the maximum value of the array over time, where Ns represents the number of sensor channels; The coefficients of the Pearson matrix in the γ dimension of equation (5) are obtained using equation (6). This indicates that the correlation between sensors is determined based on the strength of the correlation between the damage sensitivity matrices of each sensor.

7. The ancient bridge damage identification method based on an improved directed graph convolutional network according to claim 5, characterized in that, In S5, the graph neural network is an improved graph neural network using a time-domain difference operator. The improvement process includes: Add a difference layer to the graph neural network, treating the sensor-sampled data as a graph. The difference layer differs the sensor data along the time axis to expand the features of the graph in the channel dimension. The difference method used is shown in equation (7): in Indicates that sensor q at t i Test data at any given time, This represents the m-th order difference, where zeros are padded in empty positions in the array.

8. A damage identification system for ancient bridges based on an improved directed graph convolutional network, comprising: Data acquisition and labeling module: Based on the information obtained from continuous monitoring and on-site detection of the ancient bridge, a training dataset is constructed. In the training dataset, the input data is data collected from different types of detection sensors, and the output data is a binary array label. Damage dynamics equation solving module: Based on the linear elastic dynamics equation, and considering the possible damage in the ancient bridge structure, the module derives the corresponding dynamic equation that takes damage into account, and then solves the target solution set by integrating the dynamic equation. Damage sensitivity calculation module: Based on the obtained target solution set, further calculate the damage sensitivity matrix, which is used to reflect the sensitivity relationship between damage and other relevant factors; The directed graph matrix solving module corrects the sensitivity matrix according to the actual test content of different types of sensors, takes the maximum value of the updated sensitivity matrix in the time dimension, and then calculates the Pearson matrix coefficient matrix in the damage dimension of the sensitivity matrix. The calculated Pearson matrix coefficient matrix is ​​the weight factor matrix of the directed graph. The ancient bridge damage identification module uses a dataset to train a graph neural network, which is then used to identify damage to the ancient bridge.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the ancient bridge damage identification method based on an improved directed graph convolutional network as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the ancient bridge damage identification method based on an improved directed graph convolutional network as described in any one of claims 1 to 7.

Images