A structural damage identification method and system based on improved channel attention mechanism and one-dimensional convolutional neural network

Abstract

This invention belongs to the field of structural damage identification and discloses a structural damage identification method and system based on an improved channel attention mechanism and a one-dimensional convolutional neural network. The method specifically includes: establishing experimental models of intact structures and structures with different damage locations; applying excitation to the structure to be tested, measuring acceleration time-history data at each data acquisition point, and performing data preprocessing; constructing a data set and dividing the data set into a training set and a test set; constructing a neural network model based on the improved channel attention mechanism and a one-dimensional convolutional neural network, and training it using the training set; inputting the data from the test set into the trained convolutional neural network model for calculation to obtain structural damage identification results. The present invention can accurately identify the location of structural damage and quantify the importance of different sensor channels.

Description

Technical Field

[0001] The present invention belongs to the field of structural damage identification, and specifically relates to a structural damage identification method and system based on an improved channel attention mechanism and a one-dimensional convolutional neural network. Background Art

[0002] In recent years, research has focused on structural health monitoring. Compared to traditional static nondestructive testing methods (such as acoustic or ultrasonic waves, thermal imaging, X-rays, pulse-echo methods, and Lamb waves), vibration-based structural health monitoring offers significant advantages. This method eliminates the need for prior knowledge of the approximate location of structural damage, offers low testing costs, and does not affect the normal operation of the structure, making it a research hotspot in the field of structural damage identification.

[0003] Structural damage is primarily manifested through changes in its physical parameters (such as stiffness, mass, and damping), which are directly related to its measured dynamic response. After damage occurs, changes in the structural physical parameters lead to changes in its dynamic response. Vibration-based identification methods exploit this characteristic, analyzing the structural dynamic response to determine the presence, location, and severity of structural damage.

[0004] Deep learning algorithms, with their efficient computing power and strong generalization performance, are playing an increasingly important role in the field of structural health monitoring. Compared to traditional machine learning methods, deep learning methods can automatically learn abstract features from input signals, eliminating the need for manual feature selection through iterative training. Vibration signals measured in real-world projects contain significant noise, which can prevent traditional damage identification methods from detecting damage. Deep learning algorithms also exhibit high robustness and significant noise immunity, making them more suitable for practical engineering applications.

[0005] Structural health monitoring response data, such as displacement, acceleration, stress, and strain, are mostly one-dimensional time series. One-dimensional convolutional neural networks (1D-CNNs) can directly process these raw signals without requiring additional dimensionality conversion, making them particularly well-suited for analyzing one-dimensional time series data collected by sensors. Compared to two-dimensional convolutional neural networks (2D-CNNs), 1D-CNNs significantly reduce model complexity and improve computational efficiency.

[0006] Because damage information at different locations on the structure varies in importance, the training process requires a strong focus on extracting features from key locations. By introducing a channel-attention mechanism (SENet), the convolutional neural network (CNN) can focus on the sensor channels most important for damage identification, thereby enhancing the network's ability to extract damage information and further improving recognition efficiency and accuracy.

[0007] At present, research on structural damage identification based on CNN has made some progress, but there are still limitations that restrict its widespread application in the field of structural damage identification. First, in multi-sensor signals, there are significant differences in the contribution of different sensors to damage information, but existing methods use consistent processing for all channels, resulting in the model failing to effectively focus on the characteristics of key sensors. Secondly, the embedding position of traditional SENet modules is usually biased towards the back, making it difficult to fully capture the importance differences between channels. Its feature extraction method uses global average pooling, which is insensitive to positive and negative characteristics, and the Sigmoid activation function easily leads to excessive concentration of weights, weakening the ability to distinguish channel weights. In addition, the computational efficiency of 2D-CNN in time series data processing is low, which also limits application performance. Summary of the Invention

[0008] To solve the problems existing in the prior art, the present invention provides a structural damage identification method and system based on an improved channel attention mechanism and a one-dimensional convolutional neural network, which can accurately identify the location of structural damage and quantify the importance of different sensor channels.

[0009] To achieve the above object, the present invention provides the following solutions:

[0010] A structural damage identification method based on an improved channel attention mechanism and a one-dimensional convolutional neural network, the method comprising:

[0011] Establish experimental models of intact structures and structures with different damage locations;

[0012] Apply excitation to the structure to be measured, measure the acceleration time history data of each data acquisition point and transmit it to the multi-channel sensor;

[0013] Divide multi-channel sensor data into acceleration time series segments of fixed length;

[0014] Construct a dataset based on acceleration time series segments and divide the dataset into training and test sets;

[0015] A convolutional neural network model is constructed based on the improved channel attention mechanism and one-dimensional convolutional neural network, and trained using the training set;

[0016] The data of the test set is input into the trained convolutional neural network model to obtain the structural damage identification results.

[0017] Preferably, the acceleration time history data received by the sensor is ,in Represents the acceleration time history data received by the cth sensor, where the total number of acceleration sensors is C.

[0018] Preferably, the improved channel attention mechanism includes:

[0019] S41: The insertion position of the channel attention mechanism module is improved from after the convolution operation to before the convolution operation, and the channel attention mechanism is used to weight the C sensor data to determine the importance of the data at different measurement points;

[0020] S42: Improve global average pooling to global maximum pooling;

[0021] S43: Improve the compression of the number of neurons in the middle layer to appropriate expansion;

[0022] S44: Improve the activation function to PReLU function.

[0023] Preferably, the convolutional neural network model includes: an improved channel attention mechanism module, a first convolutional layer, a first maximum pooling layer, a second convolutional layer, a second maximum pooling layer, a global average pooling layer and an output layer connected in sequence.

[0024] Preferably, the loss function used in the convolutional neural network model is the cross entropy loss function, and the formula is:

[0025]

[0026] Where, represents the loss function, represents the number of categories, Represents a sign function, if the sample The true category is equal to Take 1, otherwise take 0; Represents an observation sample Belong to category The predicted probability of is the number of samples.

[0027] The present invention also provides a structural damage identification system based on an improved channel attention mechanism and a one-dimensional convolutional neural network, the system being used to implement any one of the methods described above, the system comprising: an experimental model construction module, a data measurement module, a data segmentation module, a data set construction module, a network model construction module, and a damage calculation module;

[0028] The experimental model building module is used to establish experimental models of intact structures and structures with different damage positions;

[0029] The data measurement module is used to apply excitation to the structure to be measured, measure the acceleration time history data of each data collection point and transmit it to the multi-channel sensor;

[0030] The data segmentation module is used to segment the multi-channel sensor data into acceleration time series segments of fixed length;

[0031] The data set construction module is used to construct a data set based on the acceleration time series segments and divide the data set into a training set and a test set;

[0032] The network model building module is used to build a convolutional neural network model based on an improved channel attention mechanism and a one-dimensional convolutional neural network, and train it using a training set;

[0033] The damage calculation module is used to input the data of the test set into the trained convolutional neural network model for calculation to obtain the structural damage identification result.

[0034] Preferably, the acceleration time history data received by the sensor is ,in Represents the acceleration time history data received by the cth sensor, where the total number of acceleration sensors is C.

[0035] Preferably, the improved channel attention mechanism includes:

[0036] S41: The insertion position of the channel attention mechanism module is improved from after the convolution operation to before the convolution operation, and the channel attention mechanism is used to weight the C sensor data to determine the importance of the data at different measurement points;

[0037] S42: Improve global average pooling to global maximum pooling;

[0038] S43 improves the compression of the number of neurons in the middle layer to an appropriate expansion;

[0039] S44: Improve the activation function to PReLU function.

[0040] Preferably, the convolutional neural network model includes: an improved channel attention mechanism module, a first convolutional layer, a first maximum pooling layer, a second convolutional layer, a second maximum pooling layer, a global average pooling layer and an output layer connected in sequence.

[0041] Preferably, the loss function used in the convolutional neural network model is the cross entropy loss function, and the formula is:

[0042]

[0043] Where, represents the loss function, represents the number of categories, Represents a sign function, if the sample The true category is equal to Take 1, otherwise take 0; Represents an observation sample Belong to category The predicted probability of is the number of samples.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] By optimizing the embedding position of the channel attention mechanism module, the present invention significantly improves the model's ability to focus on key sensor channels and enhances CNN's ability to extract structural damage information, achieving higher recognition efficiency and accuracy.

[0046] The present invention can better extract the significant features of the acceleration signal through global maximum pooling, and use the PReLU activation function to enhance the distinguishing ability of channel weight distribution;

[0047] The present invention weights multi-channel sensors by improving the channel attention mechanism module and visually analyzes the weights of sensor channels, providing a reliable basis for sensor layout optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. 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.

[0049] Figure 1 This is a flowchart of an exemplary method for structural damage identification based on an improved channel attention mechanism and a one-dimensional convolutional neural network in an embodiment of the present invention;

[0050] Figure 2 This is a diagram showing the arrangement of measuring points for a steel frame structure in an embodiment of the present invention;

[0051] Figure 3 This is a graph showing the accuracy of the channel attention mechanism before improvement in an embodiment of the present invention;

[0052] Figure 4 This is a loss curve diagram before the improvement of the channel attention mechanism in an embodiment of the present invention;

[0053] Figure 5 This is a graph showing the accuracy of the improved channel attention mechanism in an embodiment of the present invention;

[0054] Figure 6 This is a loss curve diagram after the channel attention mechanism is improved in an embodiment of the present invention;

[0055] Figure 7 This is a visualization diagram of the 13-channel sensor weights in an embodiment of the present invention;

[0056] Figure 8This is a graph showing the results of selecting the CH.1 sensor data training model in an embodiment of the present invention;

[0057] Figure 9 This is a graph showing the results of selecting the CH.1 and 3 sensor data training model in an embodiment of the present invention;

[0058] Figure 10 This is a graph showing the results of selecting the CH.5, 8, 11, and 13 sensor data training model in an embodiment of the present invention;

[0059] Figure 11 This is a graph showing the results of selecting the CH.1, 3, 6, and 9 sensor data training model in an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0061] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0062] Example 1

[0063] The present invention discloses a structural damage identification method based on an improved channel attention mechanism and a one-dimensional convolutional neural network, including the following contents:

[0064] S1 Model establishment: Establish experimental models of intact structures and structures with different damage locations;

[0065] S2 data sampling: Excitation is applied to the structure to be tested, and the acceleration time history data of each data collection point is measured;

[0066] S3 data preprocessing: divide multi-channel sensor data into acceleration time series segments of fixed length;

[0067] S4 creates a data set: constructs a data set based on the acceleration data samples obtained in S3, and divides the data set into a training set and a test set;

[0068] S5 improves the channel attention mechanism: To efficiently extract structural damage information, the channel attention mechanism module is improved;

[0069] S6 builds a neural network model: embeds the improved channel attention mechanism module into the one-dimensional CNN model, and trains the neural network model using the training set in S4;

[0070] S7 Recognition result verification: Substitute the test set data in S4 into the trained neural network model in S6 to obtain the prediction result; if the damage recognition accuracy of the test set is higher than the set threshold, save the training parameters to obtain a neural network model that can be applied to structural damage recognition; otherwise, adjust the neural network model parameters and continue training;

[0071] S8 Structural damage identification: Structural damage identification results are obtained based on the neural network model in S7.

[0072] Furthermore, in step S1, the lossy structure is a lossless structure and its damage is simplified by using an added mass method.

[0073] Furthermore, in step S2, the acceleration time history data received by the sensor is ,in Represents the acceleration time history data received by the cth sensor, where the total number of acceleration sensors is C.

[0074] Furthermore, in step S3, the acceleration data collected by each sensor is processed, invalid data is removed, and valid acceleration data is cut into data sequences of equal length.

[0075] Furthermore, in step S4, the channel attention mechanism module is improved as follows:

[0076] S41 Embedding Position: The channel attention mechanism module is typically embedded after the convolution operation. This approach combines all sensor channels and fails to consider the importance of sensor channels at different measurement points. However, the importance of damage information contained at different locations in the structure often varies. Therefore, this invention improves the insertion position of the channel attention mechanism module from after the convolution operation to before the convolution operation. The channel attention mechanism is used to weight the C sensor data and thus determine the importance of the data at different measurement points.

[0077] S42 Feature Extraction Method: The original channel attention mechanism uses global average pooling to process the data. However, the acceleration data has both positive and negative values, so global average pooling cannot effectively extract the data features. Therefore, global average pooling is improved to global maximum pooling.

[0078] S43 Channel Expansion: Improve the compression of the number of neurons in the middle layer to appropriate expansion. In computer vision applications, SENet is located after the convolution operation, which often faces hundreds or even more than 1,000 feature channels. Therefore, the number of neurons in the first fully connected layer is relatively small, which plays a role in controlling the complexity of the model. However, this may also lead to the loss of some effective information. In structural damage identification applications, the number of sensor channels is often small. Therefore, it is proposed to replace the operation of compressing the number of neurons in SENet with expanding to the nearest power of 2. This can both ensure the integrity of information and control the complexity of the model.

[0079] S44 Activation Function: The activation function used in the original channel attention mechanism is Sigmoid. However, when the Sigmoid function is used in this problem, the neural network model will converge most of the channel weights to 1 after training. This is not conducive to comparing the weights of different sensor channels. Therefore, the activation function is improved to the PReLU function, which is more conducive to linear expansion.

[0080] The formula for improving the attention mechanism is derived as follows:

[0081] Input tensor: (Batch size is B, number of sensor channels is C, and sample time length is L)

[0082] 1. Global Max Pooling

[0083] Take the maximum value of the time dimension for each sensor channel:

[0084]

[0085] Output tensor:

[0086] 2. Neuron Number Expansion Strategy

[0087] Neuron dimension conversion formula:

[0088]

[0089] Where E is the nearest power of 2 that is larger than C.

[0090] 3. Multi-layer Perceptron (MLP) weight generation

[0091] Generate channel weights through fully connected layers and nonlinear activation:

[0092]

[0093]

[0094]

[0095] in, is the first layer weight matrix (channel expansion), is the first layer weight matrix (channel restoration).

[0096] Output weight tensor:

[0097] 4. Feature Recalibration

[0098] Broadcast the weights to the time dimension and multiply them channel-wise with the input:

[0099]

[0100] Output tensor:

[0101] Furthermore, in step S5, the neural network model includes an improved channel attention mechanism module, a first convolutional layer, a first maximum pooling layer, a second convolutional layer, a second maximum pooling layer, a global average pooling layer and an output layer connected in sequence.

[0102] The model structure formula is as follows:

[0103] Enhanced concentration:

[0104] First layer of convolution and pooling:

[0105] Second layer of convolution and pooling:

[0106] Global average pooling layer:

[0107] Output layer:

[0108] The overall forward propagation formula of the model is:

[0109]

[0110] Furthermore, the loss function used in the neural network model is the cross entropy loss function, and the formula is as follows:

[0111]

[0112] Where, represents the loss function, represents the number of categories, Represents a sign function, if the sample The true category is equal to Take 1, otherwise take 0; Represents an observation sample Belong to category The predicted probability of is the number of samples.

[0113] The technical solution of the present invention optimizes the embedding position, feature extraction method and activation function of the channel attention mechanism by designing an improved channel attention mechanism. First, the embedding position of the channel attention mechanism is adjusted so that it can generate independent channel weights for each sensor, thereby improving the ability to focus on important channels; second, the global average pooling is improved to the global maximum pooling to better extract the significant features of the acceleration signal; finally, the PReLU activation function is used instead of the Sigmoid to avoid the weights converging to the same value, thereby enhancing the weight distribution differences between sensors. Multi-channel acceleration data is input into the improved channel attention mechanism module, so that it can perform weighted processing on the multi-channel sensor data, allowing the convolutional neural network to focus on the sensor channels it needs to pay most attention to, which can enhance the convolutional neural network's ability to extract structural damage information, thereby improving recognition efficiency and accuracy. The present invention explores and visualizes the sensor channel weights within the model, which can provide a reliable reference for the optimal arrangement of sensors.

[0114] Example 2

[0115] like Figure 1 As shown, the present invention proposes a structural damage identification method based on a channel attention mechanism and a one-dimensional convolutional neural network. The process of structural damage identification of an asymmetric steel frame structure based on this method is as follows:

[0116] Introduction to experimental conditions:

[0117] The experimental model of this embodiment is a 3-story 2-span asymmetric steel frame structure of the laboratory. The asymmetric steel frame structure is composed of 13 identical rods, each of which is 500 mm long and has a cross-section of The initial elastic modulus of the steel used in the structural members is 200GPa, and the density is measured to be 7700 .

[0118] The working condition settings include one non-destructive condition and 23 damaged conditions. The non-destructive condition corresponds to the intact state of the structure, while the damaged condition is achieved by adding a mass block in the middle of the structural member to simulate the situation at different damage locations. The specific working condition settings are shown in Table 1:

[0119] Table 1 Working condition setting details

[0120]

[0121] Data sampling:

[0122] An acceleration sensor is placed in the middle of each rod of the steel frame to collect the dynamic response signal of the structure under excitation. The data is collected using Donghua DH5922D data acquisition instrument with a sampling frequency of 1000 Hz. The arrangement of structural measurement points and the location of the hammer striking point are as follows: Figure 2 In the experiment, a hammer was used to apply excitation at specific locations, and the acceleration response of each measuring point was measured in real time by an accelerometer. The acceleration signal collected by the sensor was transmitted to the data acquisition instrument via a data transmission line for recording.

[0123] Data preprocessing:

[0124] To ensure data quality and the validity of model input, the raw acceleration data collected in the experiment must undergo the following processing steps:

[0125] Abnormal data removal: Check the integrity and validity of the collected acceleration data, and remove invalid data with noise interference, missing data, or abnormal fluctuations to ensure that the data used for analysis and training is of high quality.

[0126] Data segmentation: After removing abnormal data, the acceleration time history signal is segmented according to a fixed length. The 13-channel effective acceleration data is divided into data segments of length 1024, each data segment contains the acceleration response signal of 13 channels. The size of each data sample is 13 , where 13 represents the number of sensor channels (one sensor per rod), and 1024 represents the time step of each channel's data, reflecting the dynamic response characteristics within a specific time window. This segmentation method generates multiple data samples from the original acceleration signal for each operating condition.

[0127] Create a dataset:

[0128] The steel frame is divided into 13 units, each corresponding to an accelerometer, numbered 1 to 13. The damage location label is defined as a vector of length 13. When a unit number 𝑖 is damaged, the 𝑖th position of the label vector is set to 1, and the remaining positions are set to 0. For example, when unit 5 is damaged, the label is [0,0,0,0,1,0,0,0,0,0,0,0,0]; in the intact state, the label is [0,0,0,0,0,0,0,0,0,0,0,0,0].

[0129] Add corresponding labels to all data samples of each working condition, and finally generate a complete data set containing sample data and labels, where the format of each data sample is: (data sample, label vector) All data samples are randomly shuffled and divided into training set and test set in the ratio of 80% and 20%.

[0130] Improved channel attention mechanism:

[0131] 1. Because the importance of damage information contained in different structural locations often varies, the channel attention mechanism module is inserted before the convolution operation instead of after it. The channel attention mechanism is used to weight the data from 13 sensors and thus determine the importance of data at different measurement points.

[0132] 2. Because acceleration data has positive and negative fluctuations, directly performing global average pooling can cause positive and negative values ​​to cancel each other out, failing to fully capture the data's characteristics. Therefore, we improved global average pooling to global maximum pooling, avoiding information loss caused by the data's positive and negative characteristics and improving the model's ability to capture the key features of acceleration signals.

[0133] 3. When using Sigmoid as the activation function, most channel weights converge to 1 after network training, which is not conducive to comparing the weights of different sensor channels. Therefore, the activation function is improved to the prelu function, which is more conducive to linear expansion.

[0134] Neural network model construction and training:

[0135] The neural network model consists of an improved channel attention mechanism layer, a 1D-CNN layer, a global average pooling layer, and a linear output layer. The prepared dataset is fed into the constructed neural network model for training. The model parameters are considered trained by minimizing the difference between the model output and the data labels until the number of training cycles reaches a set threshold or the loss function reaches a set range and reaches equilibrium. The model results are then saved. During training, the cross-entropy loss function is used as the loss function.

[0136] The accuracy and loss curves before and after the improvement of the channel attention mechanism are as follows Figure 3 、 Figure 4 、 Figure 5 、 Figure 6 As shown, the accuracy and loss values ​​are shown in Table 2:

[0137] Table 2 Comparison of damage identification accuracy and loss before and after improvement

[0138]

[0139] Comparing the accuracy and loss curves shows that the improved model converges significantly faster and achieves higher accuracy more quickly. Experimental results demonstrate that the proposed method can effectively extract features related to structural damage from acceleration data, thereby accurately identifying the damage locations of steel frame structures. The improved model's loss on the test set decreased from 0.1130 to 0.0024, while the accuracy increased from 97.2% to 100%. This result validates the effectiveness of the model improvements, demonstrating that the improved model can more efficiently extract structural damage information while significantly improving identification accuracy.

[0140] By visually analyzing the sensor channel weights of the trained model (such as Figure 7 As shown in the figure, we can clearly see the difference in the contribution of different sensor channels in damage identification. Based on the weight analysis results, we selected several sensor channel data with larger weights to retrain the model and predict the results. Figure 8 、 Figure 9 The results show that 99.98% accuracy can be achieved by using only the data from the four sensor channels with larger weights.

[0141] In order to fully verify the rationality of this method, the four sensors with the smallest weight and the four sensors with the largest weight are used for training and prediction respectively. The results are as follows: Figure 10 、 Figure 11 The model using the data from the four most heavily weighted sensors achieves faster convergence and higher accuracy, demonstrating that the proposed method can effectively identify structural damage information using a small number of sensors. This method can also provide a reference for optimizing sensor placement when sensors are insufficient for large structures.

[0142] Example 3

[0143] The present invention also provides a structural damage identification system based on an improved channel attention mechanism and a one-dimensional convolutional neural network, the system being used to implement any one of the methods described above, the system comprising: an experimental model construction module, a data measurement module, a data segmentation module, a data set construction module, a network model construction module, and a damage calculation module;

[0144] The experimental model building module is used to build experimental models of intact structures and structures with different damage locations;

[0145] The data measurement module is used to apply excitation to the structure to be measured, measure the acceleration time history data of each data collection point and transmit it to the multi-channel sensor;

[0146] The data segmentation module is used to segment the multi-channel sensor data into acceleration time series segments of fixed length;

[0147] The dataset construction module is used to construct a dataset based on acceleration time series segments and divide the dataset into a training set and a test set;

[0148] The network model building module is used to build a convolutional neural network model based on the improved channel attention mechanism and one-dimensional convolutional neural network, and train it using the training set;

[0149] The damage calculation module is used to input the test set data into the trained convolutional neural network model to calculate and obtain the structural damage identification results.

[0150] In this embodiment, the acceleration time history data received by the sensor is ,in Represents the acceleration time history data received by the cth sensor, where the total number of acceleration sensors is C.

[0151] In this embodiment, the improved channel attention mechanism includes:

[0152] S41: The insertion position of the channel attention mechanism module is improved from after the convolution operation to before the convolution operation, and the channel attention mechanism is used to weight the C sensor data to determine the importance of the data at different measurement points;

[0153] S42: Improve global average pooling to global maximum pooling;

[0154] S43: Channel expansion: Improve the compression operation of the number of neurons in the middle layer to expand to the nearest power of 2;

[0155] S44: Improve the activation function to PReLU function.

[0156] In this embodiment, the convolutional neural network model includes: an improved channel attention mechanism module, a first convolutional layer, a first maximum pooling layer, a second convolutional layer, a second maximum pooling layer, a global average pooling layer and an output layer connected in sequence.

[0157] In this embodiment, the loss function used in the convolutional neural network model is the cross entropy loss function, and the formula is:

[0158]

[0159] Where, represents the loss function, represents the number of categories, Represents a sign function, if the sample The true category is equal to Take 1, otherwise take 0; Represents an observation sample Belong to category The predicted probability of is the number of samples.

[0160] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by persons skilled in the art should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A structural damage identification method based on an improved channel attention mechanism and a one-dimensional convolutional neural network, characterized in that: The method comprises: Establish experimental models of intact structures and structures with different damage locations; Apply excitation to the structure to be measured, measure the acceleration time history data of each data acquisition point and transmit it to the multi-channel sensor; Divide multi-channel sensor data into acceleration time series segments of fixed length; Construct a dataset based on acceleration time series segments and divide the dataset into training and test sets; A convolutional neural network model is constructed based on the improved channel attention mechanism and one-dimensional convolutional neural network, and trained using the training set; The test set data is input into the trained convolutional neural network model to obtain the structural damage identification results; The improved channel attention mechanism includes: S41: The insertion position of the channel attention mechanism module is improved from after the convolution operation to before the convolution operation, and the channel attention mechanism is used to weight the C sensor data to determine the importance of the data at different measurement points; S42: Improve global average pooling to global maximum pooling; S43: Improve the compression of the number of neurons in the middle layer to expansion; S44: Improve the activation function to PReLU function.

2. The method according to claim 1, characterized in that The acceleration time history data received by the sensor is ,in Represents the acceleration time history data received by the cth sensor, where the total number of acceleration sensors is C.

3. The method according to claim 1, characterized in that The convolutional neural network model includes: an improved channel attention mechanism module, a first convolutional layer, a first maximum pooling layer, a second convolutional layer, a second maximum pooling layer, a global average pooling layer and an output layer connected in sequence.

4. The method according to claim 1, wherein The loss function used in the convolutional neural network model is the cross entropy loss function, and the formula is: Where, represents the loss function, represents the number of categories, Represents a sign function, if the sample The true category is equal to Take 1, otherwise take 0; Represents an observation sample Belong to category The predicted probability of is the number of samples.

5. A structural damage identification system based on an improved channel attention mechanism and a one-dimensional convolutional neural network, the system being used to implement the method according to any one of claims 1 to 4, characterized in that: The system includes: an experimental model construction module, a data measurement module, a data segmentation module, a data set construction module, a network model construction module and a damage calculation module; The experimental model building module is used to establish experimental models of intact structures and structures with different damage positions; The data measurement module is used to apply excitation to the structure to be measured, measure the acceleration time history data of each data collection point and transmit it to the multi-channel sensor; The data segmentation module is used to segment the multi-channel sensor data into acceleration time series segments of fixed length; The data set construction module is used to construct a data set based on the acceleration time series segments and divide the data set into a training set and a test set; The network model building module is used to build a convolutional neural network model based on an improved channel attention mechanism and a one-dimensional convolutional neural network, and train it using a training set; The damage calculation module is used to input the data of the test set into the trained convolutional neural network model for calculation to obtain the structural damage identification result.

6. The system according to claim 5, characterized in that The acceleration time history data received by the sensor is ,in Represents the acceleration time history data received by the cth sensor, where the total number of acceleration sensors is C.

7. The system according to claim 5, characterized in that The improved channel attention mechanism includes: S41: The insertion position of the channel attention mechanism module is improved from after the convolution operation to before the convolution operation, and the channel attention mechanism is used to weight the C sensor data to determine the importance of the data at different measurement points; S42: Improve global average pooling to global maximum pooling; S43: Improve the compression of the number of neurons in the middle layer to expansion; S44: Improve the activation function to PReLU function.

8. The system according to claim 5, wherein: The convolutional neural network model includes: an improved channel attention mechanism module, a first convolutional layer, a first maximum pooling layer, a second convolutional layer, a second maximum pooling layer, a global average pooling layer and an output layer connected in sequence.

9. The system according to claim 5, characterized in that The loss function used in the convolutional neural network model is the cross entropy loss function, and the formula is: Where, represents the loss function, represents the number of categories, Represents a sign function, if the sample The true category is equal to Take 1, otherwise take 0; Represents an observation sample Belong to category The predicted probability of is the number of samples.

Images