Structural health monitoring emergency abnormal data diagnosis method based on hierarchical network

Abstract

The invention discloses a structural health monitoring emergency abnormal data diagnosis method based on a hierarchical network, belongs to the technical field of machine learning and health monitoring, and particularly relates to recognition of emergencies in structural health monitoring. The problems that in an existing sensor monitoring data analysis technology, emergency data and abnormal data cannot be accurately distinguished, intrinsic characteristics of a detected object are lost due to the fact that internal connection between sensor data is neglected, and then emergencies cannot be accurately recognized are solved. The method comprises the following steps: inputting an image sample set into an emergency and abnormal data classification and identification network to obtain an emergency and abnormal data classification and identification result. The structural health monitoring emergency abnormal data diagnosis method based on the hierarchical network is suitable for identifying emergency data mixed with various abnormal data types.

Description

Technical Field

[0001] The present invention relates to the fields of machine learning and health monitoring technology, and in particular to the identification of emergencies in structural health monitoring. Background Art

[0002] Over the past few decades, many bridges have been equipped with monitoring systems that continuously measure bridge responses and environmental conditions to continuously or periodically assess the integrity and safety of civil engineering infrastructure. As data-intensive systems, structural health monitoring systems generate a large amount of data continuously. The collected data sets are often contaminated by abnormal data due to factors such as system software and hardware failures. At the same time, unexpected events, such as earthquakes, ship collisions, and other excessive loads, may cause significant vibrations of the structure. The monitoring data generated by these events may have certain similarities with abnormal data in terms of characteristics, such as sudden changes in signals, abnormal amplitudes or frequency components, etc. Therefore, these data are often mixed and difficult to distinguish. The real unexpected event response may be misclassified as abnormal data, filtered out or ignored, resulting in an incorrect assessment of the structural health; conversely, abnormal data may also be mistaken for unexpected structural responses, triggering false alarms or unnecessary maintenance measures. Therefore, in practical applications, these abnormal situations must be carefully considered and handled.

[0003] Scholars at home and abroad have developed many methods for abnormal data diagnosis of health monitoring. In the context of massive and diverse abnormal data, traditional signal processing techniques require a lot of computation, and the uniqueness and robustness of the methods are still unsatisfactory. Larger structural health monitoring data sets require larger architectures to maximize their benefits. Machine learning has become an important method for processing large data sets in recent years. Methods based on reconstruction, classification, prediction, and clustering have shown good performance in various scenarios, but there are two problems. First, most methods either deal with sensor failures or emergencies, and rarely consider both types of data at the same time. How to accurately and quickly identify emergencies from monitoring data mixed with various types of abnormal data remains a challenge. Second, many of these methods fail to make full use of the complex interactions and correlations in the data, which are crucial to accurately understand and predict the health status of structures over time. The health monitoring system is a large distributed system with a complex structure. The measurement data usually exists in multiple forms. Relying on a single number of sensors cannot effectively and truly reflect the essential characteristics of the object being measured.

[0004] Attention mechanisms excel in processing complex spatiotemporal data and can capture key features and correlations in the data. These attention methods can play a role in vibration response data, but structural monitoring data has unique characteristics and is not suitable for direct application. In computer vision applications, researchers usually focus on the spatial correlation between image pixels or the spatiotemporal information in videos, which are usually strongly correlated and continuous. In health monitoring, data collected by sensors of different locations and types have large feature differences, and key information is reflected in the correlation and redundancy between sensors. In addition, the method of rearranging one-dimensional time series signals into two-dimensional mappings through matrix reconstruction has limited ability to learn from complex time series signals and is difficult to capture the deep patterns and time-varying characteristics of the signal. Therefore, it is necessary to design the structure in a targeted manner to make full use of the correlation between sensors and improve the learning ability and anomaly detection performance of the model. Summary of the invention

[0005] The present invention proposes a method for diagnosing abnormal data of sudden events in structural health monitoring based on a hierarchical network, which solves the problems existing in the existing sensor monitoring data analysis technology, that is, the inability to accurately distinguish between sudden event data and abnormal data, and the neglect of the internal connection between sensor data, which leads to the loss of the essential characteristics of the measured object, and thus the inability to accurately identify sudden events.

[0006] The method for diagnosing abnormal data of sudden incidents of structural health monitoring based on hierarchical network of the present invention has the following technical scheme: The method comprises the following steps: S1, acquiring original continuous data of vibration response of a structure collected from multiple acceleration sensors; S2, using a sliding window to segment the original continuous data to obtain multiple time series samples; Arrange all time series samples in the order of sampling time to form time series data; S3, based on the vibration characteristics of the structure, each time series sample is visualized in the time domain and the frequency domain, and then fused into a composite image, and the composite image is used as an image sample corresponding to each time series sample one by one; All image samples constitute an image sample set; S4, obtain the constructed emergency and abnormal data classification and recognition network; The emergency and abnormal data classification and recognition network includes an attention module and a hierarchical classification head; The attention module is divided into: sensor attention block, spatial attention block and channel attention block. The three attention blocks are connected through residual connections to form a hybrid attention block. The three attention blocks are used to calculate the attention parameters along the sensor direction, the spatial direction and the channel direction of each image sample using an alternating cyclic attention method, and obtain the attention parameters between tensors of different orders through a residual network; The hierarchical classification head is used to perform hierarchical detection tasks by combining the features of information at each level to classify and identify abnormal data and emergencies; S5, inputting the image sample set into the emergency and abnormal data classification and recognition network to obtain the emergency and abnormal data classification and recognition results; S6, using a sliding window to judge the emergency classification and recognition results, and obtaining the emergency classification and recognition results after removing interference.

[0007] Further, a preferred embodiment is provided, wherein step S2 is as follows: S2.1, aligning the original continuous data of the vibration response of the structure from multiple acceleration sensors in parallel according to the sampling time sequence; S2.2, using a sliding window according to a given step size, the original continuous data after side-by-side alignment is cut into multiple time series samples.

[0008] Furthermore, a preferred implementation is provided, in step S2.2, the width of the sliding window is 30 seconds, and the given step length is 5 seconds.

[0009] Further, a preferred embodiment is provided, in step S4, the three attention blocks are used to adopt an alternating cyclic attention method to calculate the attention parameters of each image sample along the sensor direction, the spatial direction and the channel direction, respectively, as follows: The sensor attention block is used to calculate the attention parameters along the sensor direction under the condition of fixed spatial and channel orders; The spatial attention block is used to calculate attention parameters along the spatial direction while fixing the order of sensors and channels; The channel attention block is used to calculate the attention parameters along the channel direction while fixing the order of the sensor and the space.

[0010] Further, a preferred embodiment is provided, wherein the sensor attention block is used to calculate the attention parameter along the sensor direction under the condition of fixed spatial and channel orders, as follows: ;

[0011] ; in, It is the LayerNorm layer; is the index of multiple attention heads; is the latent dimension of each attention head; is the dimension of pixel embedding; is the attention weight, calculated by dot product similarity; represents the key vector of the a-th attention head; Represents a sensor; Indicates channel; Indicates height; Indicates width; are the learnable parameters of the model regarding k, q, and v respectively; is the output of the previous network layer; is the query of the a-th attention head; is the value of the a-th attention head; is the attention weight of the a-th attention head; T represents transposition; is the index of the sensor; is the adjusted feature map.

[0012] Further, a preferred embodiment is provided, wherein the spatial attention block is used to calculate the attention parameters along the spatial direction while fixing the order of the sensor and the channel, as follows: ; ; ; in, is the key of a specific sensor s in the height h direction; represents a specific sensor s at height h; is the query of a specific sensor s in the height h direction; is the value of a specific sensor s in the direction of height h; Represents the tensor The dimension order is converted into W, H, C; is the key of a specific sensor s; S, C, H, W represent the dimensions of each tensor, S is the specific sensor, W is the width, H is the height, and C is the channel; s, c, h, w represent the specific dimensions of the feature map tensor in the attention block; They are the key, query, and value of a specific sensor s in the width h direction, respectively.

[0013] Further, a preferred embodiment is provided, in which the channel attention block is used to calculate the attention parameters along the channel direction under the condition of fixing the order of the sensor and the space, as follows: Use group convolution to achieve the difference between different channels. is the tensor obtained after compression by channel flattening and pooling; The weight is only considered by With The interactions between neighbors are calculated:

[0014] in, is the index of the channel to be calculated, yes The index of the first element of the support set of ; For the The weight of each channel; is the weight after local attention; for middle element; are the k interacting neighbors; represents a positive integer; Global attention is achieved through MLP and non-linear activation function: For each hidden layer, the channel weights are calculated as follows: ; in, is the weight after global attention; is the bias term; The first attention module of the sensor attention block is connected by setting a long-distance residual connection The input is connected to the last attention module of the channel attention block The output: ; in, is the result tensor.

[0015] Further, a preferred embodiment is provided, wherein step S6 is as follows: The emergency event classification and identification results are divided into a data point sequence arranged in the order of sampling time; the data point sequence includes two types of data points: normal data points and emergency event data points; Use a sliding window of a given width to slide gradually along the time axis of the data point sequence: If within a given width of the sliding window, there are a given number or more data points that are emergency data points, then this sliding window is an emergency window; Otherwise, this sliding window is a normal window; The emergency classification and recognition results after removing interference are obtained as follows: In a series of continuous sliding windows, the moment when the first emergency event data point appears in the first emergency event window is taken as the starting moment of the emergency event after interference is removed, and the moment when the last emergency event data point appears in the last emergency event window is taken as the ending moment of the emergency event after interference is removed.

[0016] The present invention also proposes a computer program product, and its technical solution is as follows: A computer program product includes a computer program / instruction, which, when executed by a processor, implements the steps of the above-mentioned hierarchical network-based structural health monitoring emergency abnormal data diagnosis method.

[0017] The present invention also proposes a computer storage medium, and its technical solution is as follows: A computer storage medium stores a computer program, and when the computer program is running, the above-mentioned structural health monitoring emergency abnormal data diagnosis method based on hierarchical network is executed.

[0018] The present invention has the following beneficial effects: 1. The hierarchical network-based structural health monitoring emergency abnormal data diagnosis method described in the present invention is based on the high redundancy and correlation of structural monitoring data. By fusing data from multiple sensors together for processing, emergency event data mixed with abnormal data can be effectively identified, thereby improving the fault tolerance and reliability of data analysis.

[0019] 2. The hierarchical network-based structural health monitoring sudden event abnormal data diagnosis method described in the present invention divides the attention module into three attention blocks: sensor attention block, spatial attention block and channel attention block. This division effectively reduces the computational burden of calculating the attention between each pixel point, significantly improves the computational efficiency, and minimizes the computational cost.

[0020] 3. The structural health monitoring emergency abnormal data diagnosis method based on hierarchical network described in the present invention identifies abnormal data and emergencies from two levels: single sensor image features and global aggregation features through a hierarchical classification head.

[0021] 4. Compared with the traditional image fusion method, the sensor fusion method based on hybrid attention mechanism of the structural health monitoring sudden event abnormal data diagnosis method based on hierarchical network described in the present invention can analyze images at a higher tensor dimension instead of just superimposing them, thereby obtaining the attention relationship between high-order tensors of different orders.

[0022] 5. The structural health monitoring emergency abnormal data diagnosis method based on hierarchical network described in the present invention can expand data and make up for the defect of insufficient data by cutting samples through sliding windows.

[0023] 6. The structural health monitoring emergency abnormal data diagnosis method based on hierarchical network described in the present invention eliminates the interference of individual erroneous data points through sliding window technology, thereby more accurately determining the emergency.

[0024] The structural health monitoring emergency abnormal data diagnosis method based on hierarchical network of the present invention is suitable for identifying emergency data mixed with various abnormal data types. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. 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 creative work.

[0026] Figure 1 A flowchart of a method for diagnosing abnormal data of structural health monitoring emergencies based on a hierarchical network in one embodiment of the present invention; Figure 2 A network architecture diagram of an emergency and abnormal data classification and identification network in one embodiment of the present invention; Figure 3 An example diagram of abnormal data in one embodiment of the present invention; Figure 4 In one embodiment of the present invention, an example diagram of a normal sample is shown in FIG. Figure 5 An example diagram of an earthquake sample and a ship-bridge collision sample in one embodiment of the present invention; Figure 6 An example diagram of a ship-bridge collision sample in one embodiment of the present invention; FIG7 is a schematic diagram of a confusion matrix of a network output result in one embodiment of the present invention; wherein FIG7(a) is an abnormal data recognition result, in which Predict represents a prediction result, Actual represents an actual result, 1 represents normal, 2 represents missing, 3 represents an outlier, 4 represents a mutation, 5 represents a sub-minimum, and 6 represents a drift; FIG7(b) is an emergency classification recognition result after removing interference; in which 1 represents a normal event, 2 represents an earthquake event, and 3 represents a ship-bridge collision event; FIG8 is a schematic diagram showing a comparison between the classification and identification results of emergencies and abnormal data and the distribution of abnormal data of actual emergencies in one embodiment of the present invention; wherein FIG8(a) is a distribution diagram of the classification and identification results of emergencies and abnormal data; and FIG8(b) is a distribution diagram of abnormal data of actual emergencies. DETAILED DESCRIPTION

[0027] In order to make the technical solutions and advantages of the present invention more clearly described, the specific implementation methods of the present invention will be further described in detail and completely in conjunction with the accompanying drawings. The various implementation methods described below are only part of the preferred solutions of the present invention, rather than all implementation methods; the various implementation methods described below are intended to explain the present invention and cannot be understood as limitations on the present invention; the reasonable combination of technical features defined in the various implementation methods of the present invention, and all other implementation methods obtained by ordinary technicians in the field without creative work based on the implementation methods of the present invention, are within the scope of protection of the present invention.

[0028] Implementation Method 1: Combination Figure 1 This embodiment describes an abnormal data diagnosis method for structural health monitoring emergencies based on a hierarchical network. The specific implementation contents are as follows: The method comprises the following steps: S1, acquiring original continuous data of vibration response of a structure collected from multiple acceleration sensors; S2, using a sliding window to segment the original continuous data to obtain multiple time series samples; Arrange all time series samples in the order of sampling time to form time series data; S3, based on the vibration characteristics of the structure, each time series sample is visualized in the time domain and the frequency domain, and then fused into a composite image, and the composite image is used as an image sample corresponding to each time series sample one by one; All image samples constitute an image sample set; S4, obtain the constructed emergency and abnormal data classification and recognition network; The emergency and abnormal data classification and recognition network includes an attention module and a hierarchical classification head; The attention module is divided into: sensor attention block, spatial attention block and channel attention block. The three attention blocks are connected through residual connections to form a hybrid attention block. The three attention blocks are used to calculate the attention parameters along the sensor direction, the spatial direction and the channel direction of each image sample using an alternating cyclic attention method, and obtain the attention parameters between tensors of different orders through a residual network; The hierarchical classification head is used to combine the features of each level of information (i.e., single image features and global aggregate features) to perform hierarchical detection tasks; S5, inputting the image sample set into the emergency and abnormal data classification and recognition network to obtain the network output result (i.e., the emergency and abnormal data classification and recognition result); S6, using a sliding window to judge the emergency classification and recognition results, and obtaining the emergency classification and recognition results after removing interference.

[0029] In this implementation, in step S3, visualization is performed in the time domain and the frequency domain so that the composite image includes the features of the time domain and the frequency domain, and there is no need to display the horizontal and vertical coordinates.

[0030] In this implementation, the emergency and abnormal data classification and identification network is named SSCHA-HNet.

[0031] In this implementation, a mixed attention block is used to enhance key features.

[0032] In this embodiment, visualization is performed in the time domain and the frequency domain, and the time domain and the frequency domain features are plotted on one image.

[0033] In this embodiment, the method divides the attention module into three attention blocks: sensor attention block, spatial attention block and channel attention block. This division effectively reduces the computational burden of calculating the attention between each pixel point, significantly improves the computational efficiency, and minimizes the computational cost.

[0034] In this implementation, a deep learning network (SSCHA-HNet) based on a hybrid attention mechanism is used to establish connections between sensor data instead of processing them separately. This can obtain information that a single sensor cannot collect, and can effectively identify emergency event data mixed with abnormal data, thereby enhancing the anti-interference and fault-tolerance capabilities of the monitoring system.

[0035] In this embodiment, compared with the traditional image fusion method, the sensor fusion method based on hybrid attention and hierarchical network can analyze images at higher tensor dimensions instead of just superimposing them, thereby obtaining the attention relationship between high-order tensors of different orders.

[0036] In this embodiment, the method is universal and can be applied to different types of structures, including bridges, reinforced concrete buildings, tunnels, and ships. In this embodiment, visualization results can be obtained through verification experiments, and the visualization results show that the sensor real-valued images with sudden event characteristics are the more concerned parts in these high-order attention maps, indicating that the method successfully captures the correlation between sensors.

[0037] In this embodiment, the emergency and abnormal data classification and recognition network adopts a hierarchical classification head, which belongs to a hierarchical network. The hierarchical classification head is composed of a shallow feedforward neural network (FNN).

[0038] The features of each level of information include single image features and global aggregation features; The hierarchical classification heads are connected to single image features and global aggregate features respectively; The hierarchical classification head is used to identify (sensor) abnormal data through single image features, and also to identify sudden events (types) through global aggregate features.

[0039] In this embodiment, in order to better train the emergency abnormal data diagnosis network, the image sample set can be divided into a training set, a verification set and a test set. The training set is used to train the emergency and abnormal data classification and recognition network, the verification set is used to verify the training results, and the test set is used to test the verified network. The generalization ability of the network can be further verified by training, verifying and testing the network with data sets of different contents.

[0040] It should be noted that the service environment of large bridges is harsh, which causes the monitoring system to generate a variety of abnormal data; at the same time, bridges may also be affected by emergencies (such as earthquakes or ships hitting bridges), and the data generated by them have certain similarities in characteristics with sensor error data, which are difficult to distinguish, affecting the accuracy of data analysis and early warning. The method described in this embodiment uses a sensor-space-channel hybrid attention hierarchical network (SSCHA-HNet) to analyze across multiple levels and dimensions. The sensor attention module builds the relationship between sensors, the space attention module captures long-distance dependencies, and the channel attention module dynamically adjusts the global and local feature representation capabilities within the feature map, and further uses the features of different levels of information to perform hierarchical detection tasks; it can effectively identify emergency event data mixed with various abnormal data types.

[0041] Implementation Method 2: Combination Figure 1 This embodiment is described. This embodiment further limits the abnormal data diagnosis method for structural health monitoring emergencies based on hierarchical networks described in the first embodiment. The specific implementation contents are as follows: The step S2 is as follows: S2.1, aligning the original continuous data of the vibration response of the structure from multiple acceleration sensors in parallel according to the sampling time sequence; S2.2, using a sliding window according to a given step size, the original continuous data after side-by-side alignment is cut into multiple time series samples.

[0042] Furthermore, a preferred implementation is provided, in step S2.2, the width of the sliding window is 30 seconds, and the given step length is 5 seconds.

[0043] At this time, the two adjacent time series samples have 25 seconds of overlapping data.

[0044] In this implementation, by cutting samples every 5 seconds through a sliding window, data can be expanded to make up for the defect of insufficient data.

[0045] Implementation Method 3: Combination Figure 1 This embodiment is described. This embodiment further limits the abnormal data diagnosis method for structural health monitoring emergencies based on hierarchical networks described in the first embodiment. The specific implementation contents are as follows: In step S4, the three attention blocks are used to use an alternating cyclic attention method to calculate the attention parameters of each image sample along the sensor direction, the spatial direction, and the channel direction, respectively, as follows: The sensor attention block is used to calculate the attention parameters along the sensor direction under the condition of fixed spatial and channel orders; The spatial attention block is used to calculate attention parameters along the spatial direction while fixing the order of sensors and channels; The channel attention block is used to calculate the attention parameters along the channel direction while fixing the order of the sensor and the space.

[0046] In this embodiment, the emergency and abnormal data classification and recognition network does not directly calculate the relationship between each pixel and other pixels in the image sample, but adopts the alternating cyclic attention method to construct a hybrid attention network, and independently refines the features of the sensor, space and channel dimensions by alternating the use of three attention blocks.

[0047] Implementation Method 4: Combination Figure 1 This embodiment is described as a further limitation of the abnormal data diagnosis method for structural health monitoring emergencies based on a hierarchical network described in the third embodiment. The specific implementation contents are as follows: The sensor attention block is used to calculate the attention parameters along the sensor direction under the condition of fixed spatial and channel orders, as follows: ;

[0048] ; in, It is the LayerNorm layer; is the index of multiple attention heads; is the latent dimension of each attention head; is the dimension of pixel embedding; is the attention weight, calculated by dot product similarity; represents the key vector of the a-th attention head; Represents a sensor; Indicates channel; Indicates height; Indicates width; are the learnable parameters of the model regarding k (key), q (query), and v (value); is the output of the previous network layer; is the query of the a-th attention head; is the value of the a-th attention head; is the attention weight of the a-th attention head; T represents transposition; is the index of the sensor; is the adjusted feature map.

[0049] In this implementation, the sensor attention block uses a multi-head attention method to calculate the relationship between different sensors while fixing the channel and spatial changes, which is called the "pixel-by-pixel" sensor attention method.

[0050] In this embodiment, the attention parameters include a query vector (consisting of query or Indicates, referred to as query), key vector (consisting of key or Representation, or the search vector, referred to as the key) and the content vector (by value or Representation, or value vector, or simply value).

[0051] In this embodiment, the attention mechanism achieves the distribution of attention weights by focusing the attention of the query vector and the key vector, so that the network focuses on the key parts. More specifically, the attention mechanism focuses the network's attention on a small part of the data by increasing the weight of a part of the data in the network input and reducing the weight of other parts. These weights are calculated through the query-key mechanism, where there is a one-to-one correspondence between the query vector and the key vector.

[0052] In this implementation, the LayerNorm layer is a layer normalization layer.

[0053] Implementation Method 5: Combination Figure 1 This embodiment is described as a further limitation of the abnormal data diagnosis method for structural health monitoring emergencies based on a hierarchical network described in the third embodiment. The specific implementation contents are as follows: The spatial attention block is used to calculate the attention parameters along the spatial direction while fixing the order of sensors and channels, as follows: ; ; ; in, is the key of a specific sensor s in the height h direction; represents a specific sensor s at height h; is the query for a specific sensor s at height h; is the value of a specific sensor s at height h; Represents the tensor The dimension order is converted into W, H, C; is the key of a specific sensor s; T[∙,∙,∙] means converting the tensor into a specified dimensional order; S, C, H, W represent the dimensions of each tensor, S is the specific sensor, W is the width, H is the height, and C is the channel; s, c, h, w represent the specific dimensions of the feature map tensor in the attention block; are the key, query, and value of a specific sensor s in the width h direction.

[0054] In this implementation, the spatial attention block generates an attention map between spatial pixels of a specific sensor to emphasize or suppress features at different spatial locations, which means fixing sensor variations and focusing only on spatial relationships.

[0055] Implementation Method 6: Combination Figure 1 This embodiment is described as a further limitation of the abnormal data diagnosis method for structural health monitoring emergencies based on a hierarchical network described in the third embodiment. The specific implementation contents are as follows: The channel attention block is used to calculate the attention parameters along the channel direction while fixing the order of the sensor and space, as follows: Use group convolution to achieve the difference between different channels. is the tensor obtained after compression by channel flattening and pooling; The weight is only considered by With The interactions between neighbors are calculated:

[0056] in, is the index of the channel to be calculated, yes The index of the first element of the support set of ; For the The weight of each channel; is the weight after local attention; for middle element; are the k interacting neighbors; represents a positive integer; Global attention is achieved through MLP and non-linear activation function: For each hidden layer, the channel weights are calculated as follows: ; in, is the weight after global attention; is the bias term; The first attention module of the sensor attention block is connected by setting a long-distance residual connection The input is connected to the last attention module of the channel attention block The output: ; in, is the result tensor.

[0057] In this implementation, the channel attention block obtains the attention relationship by setting local and global receptive fields.

[0058] In this implementation, a long-distance residual connection is set to prevent the gradient from disappearing.

[0059] Implementation Method 7: Combination Figure 1 This embodiment is described. This embodiment further limits the abnormal data diagnosis method for structural health monitoring emergencies based on hierarchical networks described in the first embodiment. The specific implementation contents are as follows: The step S6 is as follows: The emergency classification and identification result is divided into a data point sequence arranged in the order of sampling time; the data point sequence includes two types of data points: normal data points and emergency data points; for example, the normal data points can be represented by 0, and the emergency data points can be represented by 1; Use a sliding window of a given width to slide gradually along the time axis of the data point sequence: If within a given width of the sliding window, there are a given number or more data points with 1 representing an emergency data point, then this sliding window is an emergency window; Otherwise, this sliding window is a normal window; The emergency classification and recognition results after removing interference are obtained as follows: In a series of continuous sliding windows, the moment when the first 1 appears in the first sudden event window represents the sudden event data point as the starting moment of the sudden event after interference is removed, and the moment when the last 1 appears in the last sudden event window represents the sudden event data point as the ending moment of the sudden event after interference is removed.

[0060] Further, a preferred embodiment is provided, wherein the given width is the length of time occupied by 5 data points, and the given number is 3.

[0061] In this embodiment, the sliding window technology is used to eliminate the interference of individual erroneous data points, so as to more accurately determine the emergency event.

[0062] Implementation Method 8: Combination Figure 1 This embodiment is described as a specific example of the abnormal data diagnosis method for structural health monitoring emergencies based on hierarchical networks. The specific implementation contents are as follows: Step 1: Collect raw continuous data: First, actual monitoring vibration data of a long-span cable-stayed bridge is collected, including normal data and earthquake data (i.e., data in normal state and acceleration data during earthquake); Secondly, the finite element analysis tool ANSYS software was used to simulate the ship-bridge collision, and the acceleration data of each key position of the bridge (i.e. the location of the sensor arrangement) when the bridge and the ship collided was obtained. It should be noted that due to the lack of actual ship-bridge collision data, the ship-bridge collision data was obtained by using Ansys simulation.

[0063] Step 2, the original continuous data is segmented every 5 seconds using a 30-second sliding window and plotted in the time and frequency domains to expand the dataset to cope with the extremely rare and short-lived occurrence of emergencies. Step 3, multiple images from different sensors in the same time period are taken as an image sample, e.g. Figure 3 It should be noted that not all images collected by sensors show the characteristics of sudden events. Sometimes only a small number of images collected by sensors show subtle characteristics of sudden events, which makes the identification of sudden events particularly difficult.

[0064] Step 4: Input the image sample into the above-mentioned emergency and abnormal data classification and recognition network: The network utilizes convolutional layers to extract basic features, and then uses three attention blocks to refine and enrich the features in sensor, spatial, and channel dimensions, respectively.

[0065] The confusion matrix of network recognition results (emergency and abnormal data classification recognition results) is as follows: Figure 5The results show that the F1 scores of abnormal data and emergency diagnosis results on the test set are 90.4% and 91.4% respectively, indicating that this method can accurately identify emergencies.

[0066] Step 5: Figure 5 As shown, the sliding window technology is used to eliminate the interference of individual error points from the emergency classification and recognition results, so as to more accurately determine the emergency.

[0067] In summary, the above-mentioned emergency and abnormal data classification and recognition network can effectively identify emergency data mixed with abnormal data.

[0068] The technical solution provided by the present invention is further described in detail above through several specific implementation modes in order to highlight the advantages and benefits of the technical solution provided by the present invention. However, the several specific implementation modes described above are not intended to be used as limitations on the present invention. Any reasonable changes and improvements to the present invention, reasonable combinations of implementation modes and equivalent substitutions based on the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for diagnosing abnormal data of structural health monitoring emergencies based on hierarchical networks, characterized in that: The method comprises the following steps: S1, acquiring original continuous data of vibration response of a structure collected from multiple acceleration sensors; S2, using a sliding window to segment the original continuous data to obtain multiple time series samples; Arrange all time series samples in the order of sampling time to form time series data; S3, based on the vibration characteristics of the structure, each time series sample is visualized in the time domain and the frequency domain, and then fused into a composite image, and the composite image is used as an image sample corresponding to each time series sample one by one; All image samples constitute an image sample set; S4, obtain the constructed emergency and abnormal data classification and recognition network; The emergency and abnormal data classification and recognition network includes an attention module and a hierarchical classification head; The attention module is divided into: sensor attention block, spatial attention block and channel attention block. The three attention blocks are connected through residual connections to form a hybrid attention block. The three attention blocks are used to calculate the attention parameters along the sensor direction, the spatial direction and the channel direction of each image sample using an alternating cyclic attention method, and obtain the attention parameters between tensors of different orders through a residual network; The hierarchical classification head is used to perform hierarchical detection tasks by combining the features of information at each level to classify and identify abnormal data and emergencies; S5, inputting the image sample set into the emergency and abnormal data classification and recognition network to obtain the emergency and abnormal data classification and recognition results; S6, using a sliding window to judge the emergency classification and recognition results, and obtaining the emergency classification and recognition results after removing interference.

2. The method for diagnosing abnormal data of structural health monitoring emergencies based on hierarchical networks according to claim 1 is characterized in that: The step S2 is as follows: S2.1, aligning the original continuous data of the vibration response of the structure from multiple acceleration sensors in parallel according to the sampling time sequence; S2.2, using a sliding window according to a given step size, the original continuous data after side-by-side alignment is cut into multiple time series samples.

3. The method for diagnosing abnormal data of structural health monitoring emergencies based on hierarchical networks according to claim 2 is characterized in that: In step S2.2, the width of the sliding window is 30 seconds, and the given step length is 5 seconds.

4. The method for diagnosing abnormal data of structural health monitoring emergencies based on hierarchical networks according to claim 1 is characterized in that: In step S4, the three attention blocks are used to use an alternating cyclic attention method to calculate the attention parameters of each image sample along the sensor direction, the spatial direction, and the channel direction, respectively, as follows: The sensor attention block is used to calculate the attention parameters along the sensor direction under the condition of fixed spatial and channel orders; The spatial attention block is used to calculate attention parameters along the spatial direction while fixing the order of sensors and channels; The channel attention block is used to calculate the attention parameters along the channel direction while fixing the order of the sensor and the space.

5. The method for diagnosing abnormal data of structural health monitoring emergencies based on hierarchical networks according to claim 4 is characterized in that: The sensor attention block is used to calculate the attention parameters along the sensor direction under the condition of fixed spatial and channel orders, as follows: ； ； in, It is the LayerNorm layer; is the index of multiple attention heads; is the latent dimension of each attention head; is the dimension of pixel embedding; is the attention weight, calculated by dot product similarity; represents the key vector of the a-th attention head; Represents a sensor; Indicates channel; Indicates height; Indicates width; are the learnable parameters of the model regarding k, q, and v respectively; is the output of the previous network layer; is the query of the a-th attention head; is the value of the a-th attention head; is the attention weight of the a-th attention head; T represents transposition; is the index of the sensor; is the adjusted feature map.

6. The method for diagnosing abnormal data of structural health monitoring emergencies based on hierarchical networks according to claim 4 is characterized in that: The spatial attention block is used to calculate the attention parameters along the spatial direction while fixing the order of sensors and channels, as follows: ； ； ； in, is the key of a specific sensor s in the height h direction; represents a specific sensor s at height h; is the query of a specific sensor s in the height h direction; is the value of a specific sensor s in the direction of height h; Represents the tensor The dimension order is converted into W, H, C; is the key of a specific sensor s; S, C, H, W represent the dimensions of each tensor, S is the specific sensor, W is the width, H is the height, and C is the channel; s, c, h, w represent the specific dimensions of the feature map tensor in the attention block; They are the key, query, and value of a specific sensor s in the width h direction, respectively.

7. The method for diagnosing abnormal data of structural health monitoring emergencies based on hierarchical networks according to claim 4 is characterized in that: The channel attention block is used to calculate the attention parameters along the channel direction while fixing the order of the sensor and space, as follows: Use group convolution to achieve the difference between different channels. is the tensor obtained after compression by channel flattening and pooling; The weight is only considered by With The interactions between neighbors are calculated: in, is the index of the channel to be calculated, yes The index of the first element of the support set of ; For the The weight of each channel; is the weight after local attention; for middle element; are the k interacting neighbors; represents a positive integer; Global attention is achieved through MLP and non-linear activation function: For each hidden layer, the channel weights are calculated as follows: ； in, is the weight after global attention; is the bias term; The first attention module of the sensor attention block is connected by setting a long-distance residual connection The input is connected to the last attention module of the channel attention block The output: ； in, is the result tensor.

8. The method for diagnosing abnormal data of structural health monitoring emergencies based on hierarchical networks according to claim 1, characterized in that: The step S6 is as follows: The emergency event classification and identification results are divided into a data point sequence arranged in the order of sampling time; the data point sequence includes two types of data points: normal data points and emergency event data points; Use a sliding window of a given width to slide gradually along the time axis of the data point sequence: If within a given width of the sliding window, there are a given number or more data points that are emergency data points, then this sliding window is an emergency window; Otherwise, this sliding window is a normal window; The emergency classification and recognition results after removing interference are obtained as follows: In a series of continuous sliding windows, the moment when the first emergency event data point appears in the first emergency event window is taken as the starting moment of the emergency event after interference is removed, and the moment when the last emergency event data point appears in the last emergency event window is taken as the ending moment of the emergency event after interference is removed.

9. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the method for diagnosing abnormal data of sudden incidents of structural health monitoring based on a hierarchical network are implemented.

10. A computer storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is run, the method for diagnosing abnormal data of sudden incidents of structural health monitoring based on a hierarchical network as described in any one of claims 1 to 8 is executed.

Images