CN-121997233-A - Bridge pulling sling response self-adaptive prediction and anomaly identification method and system based on graph neural network

CN121997233ACN 121997233 ACN121997233 ACN 121997233ACN-121997233-A

Abstract

The invention provides a bridge pulling sling response self-adaptive prediction and anomaly identification method and system based on a graph neural network, wherein the method comprises the following steps: and obtaining node characteristic vectors of each monitoring point, wherein the node characteristic vectors are used as graph nodes, and all the node characteristic vectors form a graph node set. And calculating edge weights among the nodes of each graph according to the node feature vectors, and judging whether the nodes of each graph are connected or not according to a preset physical connection matrix. If the judgment result is yes and the edge weight is greater than the weight threshold, adding the corresponding edge weight into the edge set, and constructing a graph structure according to the graph node set and the edge set. The graph structure is input into a preset graph neural network to be predicted, predicted response data and side note force weights are obtained, an abnormal recognition result is obtained according to the predicted response data, preset actual response data and the side note force weights, and accuracy of bridge pulling and sling prediction and abnormal recognition is improved.

Inventors

MA JUNHAI
ZHANG BIN
WANG YANG
LIU TIANCHENG
CHENG QIAN
WANG XIAONING
SHI XIAOPENG
Ren qingda
LI CHUANBO
RU CHAO

Assignees

中交公路长大桥建设国家工程研究中心有限公司

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (10)

1. The bridge pull sling response self-adaptive prediction and anomaly identification method based on the graph neural network is characterized by comprising the following steps of: the node characteristic vector of each monitoring point is obtained, wherein the node characteristic vector is used as a graph node, and all the node characteristic vectors form a graph node set; calculating edge weights among the graph nodes according to the node feature vectors, and judging whether the graph nodes are connected or not according to a preset physical connection matrix; if the judgment result is yes and the edge weight is greater than the weight threshold, adding the corresponding edge weight into an edge set, and constructing a graph structure according to the graph node set and the edge set; And inputting the graph structure into a preset graph neural network for prediction to obtain predicted response data and side note force weights, and obtaining an abnormal recognition result according to the predicted response data, the preset actual response data and the side note force weights.
2. The adaptive prediction and anomaly identification method for bridge pull sling response based on the graph neural network according to claim 1, wherein the obtaining the node feature vector of each monitoring point comprises: Acquiring the original data of each monitoring point; Calculating corresponding mean value, standard deviation, effective data proportion and time embedded vector according to the original data; and splicing the mean value, the standard deviation, the effective data proportion and the time embedded vector to obtain the node characteristic vector of each monitoring point.
3. The adaptive prediction and anomaly identification method for bridge pull sling response based on graph neural network according to claim 1, wherein the node feature vector comprises a mean value, a standard deviation and an effective data proportion, the calculating the edge weight between each graph node according to the node feature vector comprises: calculating the similarity between each graph node according to the mean value and the standard deviation; Calculating the data integrity between each two graph nodes according to the effective data proportion and a preset adjustable coefficient; and adding the similarity and the data integrity to obtain the edge weight.
4. The adaptive prediction and anomaly identification method for bridge pull sling response based on a graph neural network according to claim 1, wherein the obtaining the anomaly identification result according to the predicted response data, the preset actual response data and the side note force weight comprises: calculating a cable residual index according to the predicted response data and the actual response data; Calculating a path change index according to the side note force weight; And searching in a preset index result mapping table according to the cable residual index and the path change index to obtain the abnormal identification result.
5. The adaptive prediction and anomaly identification method for bridge pull-sling response based on a graph neural network according to claim 4, wherein the preset response data comprises a predicted rope force and a predicted acceleration root mean square, the actual response data comprises an actual rope force and an actual acceleration root mean square, and the calculating the cable residual index according to the predicted response data and the actual response data comprises: subtracting the predicted cable force from the actually measured cable force and taking an absolute value to obtain first data; subtracting the root mean square of the measured acceleration from the root mean square of the predicted acceleration and taking an absolute value to obtain second data; and adding the first data and the second data to obtain the cable residual index.
6. The adaptive prediction and anomaly identification method for bridge pull sling response based on a graph neural network according to claim 4, wherein the calculating the path change index according to the side note force weight comprises: And subtracting the side note meaning weight of the current time window from the side note meaning weight of the previous time window to obtain third data, and adding the third data of the cable node and all neighbor nodes to obtain the path change index.
7. The bridge pull sling response self-adaptive prediction and abnormality identification method based on a graph neural network according to claim 1, wherein the graph neural network comprises a graph attention module and a time modeling module, the graph structure is input into a preset graph neural network for prediction to obtain prediction response data and side attention weight, and the method comprises the following steps: Inputting the graph structure to the graph annotation meaning module for weighted aggregation to obtain node feature codes; And inputting the node characteristic codes into the time modeling module for prediction to obtain the prediction response data and the side note force weight.
8. Bridge pull sling response self-adaptive prediction and anomaly identification system based on graph neural network, which is characterized by comprising: the system comprises an acquisition unit, a graph node acquisition unit and a graph node acquisition unit, wherein the acquisition unit is used for acquiring a node characteristic vector of each monitoring point, and the node characteristic vector is used as a graph node, and all the node characteristic vectors form a graph node set; The computing unit is used for computing the edge weight between each graph node according to the node characteristic vector and judging whether the graph nodes are connected or not according to a preset physical connection matrix; The construction unit is used for adding the corresponding edge weight into an edge set and constructing a graph structure according to the graph node set and the edge set if the judgment result is yes and the edge weight is greater than a weight threshold; the prediction unit is used for inputting the graph structure into a preset graph neural network to perform prediction to obtain prediction response data and side note force weight, and obtaining an abnormal recognition result according to the prediction response data, the preset actual response data and the side note force weight.
9. An electronic device comprising a memory and a processor, the memory storing a computer program, the processor implementing the adaptive prediction and anomaly identification method of bridge pull-sling response based on a graph neural network of any one of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the adaptive prediction and anomaly identification method for bridge pull-sling response based on a graph neural network as defined in any one of claims 1 to 7.

Description

Bridge pulling sling response self-adaptive prediction and anomaly identification method and system based on graph neural network Technical Field The invention belongs to the field of bridge anomaly recognition, and particularly relates to a bridge pulling sling response self-adaptive prediction and anomaly recognition method and system based on a graph neural network. Background Along with the continuous increase of the number of the large-span bridges and the increasing complexity of service environments thereof, the bridge structure shows remarkable multi-factor coupling characteristics in the long-term operation process, particularly in a cable-stayed bridge and a suspension bridge, a cable-stayed system is simultaneously influenced by various factors such as temperature change, wind load, structural deformation, vehicle action and the like, and the response behavior has obvious nonlinearity, time variability and structural dependence. Current bridge structural health monitoring systems typically collect environmental parameters and structural response data by deploying multiple types of sensors, and perform anomaly analysis and assessment of structural state based on these data. However, the existing methods depend on finite element models, simplified mechanical analysis or empirical rules for inference, and the methods have high dependence on structural modeling precision and parameter calibration, are difficult to continuously apply under actual operation conditions, and have low accuracy in bridge sling prediction and anomaly identification. Therefore, how to improve the accuracy of bridge pulling sling prediction and abnormality identification becomes a technical problem to be solved urgently. Disclosure of Invention The invention aims to design a bridge pull sling response self-adaptive prediction and anomaly identification method and system based on a graph neural network, and can improve the accuracy of bridge pull sling prediction and anomaly identification. In order to achieve the above object, in a first aspect of the present invention, there is provided a bridge sling response adaptive prediction and anomaly identification method based on a graph neural network, the method comprising: the node characteristic vector of each monitoring point is obtained, wherein the node characteristic vector is used as a graph node, and all the node characteristic vectors form a graph node set; calculating edge weights among the graph nodes according to the node feature vectors, and judging whether the graph nodes are connected or not according to a preset physical connection matrix; if the judgment result is yes and the edge weight is greater than the weight threshold, adding the corresponding edge weight into an edge set, and constructing a graph structure according to the graph node set and the edge set; And inputting the graph structure into a preset graph neural network for prediction to obtain predicted response data and side note force weights, and obtaining an abnormal recognition result according to the predicted response data, the preset actual response data and the side note force weights. Further, the obtaining the node feature vector of each monitoring point includes: Acquiring the original data of each monitoring point; Calculating corresponding mean value, standard deviation, effective data proportion and time embedded vector according to the original data; and splicing the mean value, the standard deviation, the effective data proportion and the time embedded vector to obtain the node characteristic vector of each monitoring point. Further, the node feature vector includes a mean value, a standard deviation, and a valid data proportion, and the calculating the edge weight between each of the graph nodes according to the node feature vector includes: calculating the similarity between each graph node according to the mean value and the standard deviation; Calculating the data integrity between each two graph nodes according to the effective data proportion and a preset adjustable coefficient; and adding the similarity and the data integrity to obtain the edge weight. Further, the obtaining the abnormal recognition result according to the predicted response data, the preset actual response data and the side note force weight includes: calculating a cable residual index according to the predicted response data and the actual response data; Calculating a path change index according to the side note force weight; And searching in a preset index result mapping table according to the cable residual index and the path change index to obtain the abnormal identification result. Further, the preset response data includes a predicted cable force and a predicted acceleration root mean square, the actual response data includes an actual cable force and an actual acceleration root mean square, and the calculating the cable residual index according to the predicted response data and the actual response data i