CN114969934A - Stay cable damage degree identification method and model construction method - Google Patents
Stay cable damage degree identification method and model construction method Download PDFInfo
- Publication number
- CN114969934A CN114969934A CN202210608594.9A CN202210608594A CN114969934A CN 114969934 A CN114969934 A CN 114969934A CN 202210608594 A CN202210608594 A CN 202210608594A CN 114969934 A CN114969934 A CN 114969934A
- Authority
- CN
- China
- Prior art keywords
- cable
- actual
- matrix
- stay
- stayed bridge
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computer Hardware Design (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Mathematics (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Analysis (AREA)
- Molecular Biology (AREA)
- Structural Engineering (AREA)
- Civil Engineering (AREA)
- Architecture (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Bridges Or Land Bridges (AREA)
Abstract
The embodiment of the invention provides a method for identifying damage degree of a stay cable and a method for constructing a model, wherein the method comprises the steps of obtaining position information of each stay cable in a target cable-stayed bridge and condition data of the target cable-stayed bridge; the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data; normalizing the acceleration data to generate a corresponding characteristic matrix; establishing an adjacency matrix according to the position information of each cable-stayed bridge; and constructing a neural network model combining a graph neural network and a gate control cycle unit according to the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix. The neural network model provided by the application can make full use of the relevance between the stay cables, well retains the information of the nodes, and realizes perfect combination of global and local.
Description
Technical Field
The invention relates to the field of damage identification, in particular to a stay cable damage degree identification method and a model construction method.
Background
The stay cable is a key stressed component of the cable-stayed bridge, and the corrosion of the stay cable caused by environmental factors or the fatigue of the stay cable caused by long-term load can form serious threats and hidden dangers to the operation safety and the use durability of the whole structure, and is directly related to the smooth operation of national economy and the life and property safety of people. Therefore, in order to ensure the safe service of the cable-stayed bridge and the smooth road network, the damage position and the damage degree of the stay cable are rapidly identified, and the method has important scientific research value and engineering application value. With the development of artificial intelligence technology, machine learning and deep learning methods are widely applied to identifying damages of stay cables.
Stay cable damage identification is an important modeling task, but at present, many researches and applications ignore the relevance between stay cables, and the existing identification model does not consider the spatial correlation. Since all the stay cables form a whole and have relevance, the damage of one stay cable may affect the vibration signal of the adjacent stay cable beside, so that the vibration signal of the adjacent cable is changed, but the change is theoretically opposite to the change trend of the damaged cable. If the space factor is not considered, the recognition result is deviated, the stay cable damages the inherent nonlinear characteristic of the recognition task, and a simple linear model is often insufficient to generate reliable prediction. Common non-linear models such as artificial neural networks, recurrent neural networks and long-term memory networks (LSTM) do not take into account the correlation between different stay cables. A Convolutional Neural Network (CNN) which is excellent in many fields can depend on the position of an existing stay cable as a space factor, but the traditional CNN cannot process irregular data. Although the distance between each stayed cable on the bridge surface is equal, the angle between each stayed cable and the beam is changed, so the graph structure determined by the position of the stayed cable is irregular, in the topological graph, the distance relation between each node and the characteristic relation are different, the graph has more complex characteristics and lacks of translation invariance, and therefore the CNN processing graph data structure is not preferable.
However, the multi-layer GCN (graph convolutional neural network) can cause excessive interaction of node information, and the characteristics of each node tend to be consistent and cannot be distinguished obviously.
Disclosure of Invention
In view of the above problems, the present application is directed to a method for identifying damage level of a stay cable and a method for constructing a model, which overcome or at least partially solve the above problems, and includes:
a construction method of a model for identifying damage degree of a stay cable on a target cable-stayed bridge comprises the following steps:
acquiring position information of each stay cable in a target cable-stayed bridge and condition data of the target cable-stayed bridge; the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data;
normalizing the acceleration data to generate a corresponding feature matrix;
establishing an adjacency matrix according to the position information of each cable-stayed bridge;
and constructing a neural network model combining a graph neural network and a gate control cycle unit according to the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix.
Preferably, the step of normalizing the acceleration data to generate a corresponding feature matrix includes:
screening out the acceleration with the largest numerical value and the acceleration with the smallest numerical value in the acceleration data;
generating a normalized acceleration corresponding to each stay cable according to the acceleration with the maximum numerical value, the acceleration with the minimum numerical value and the acceleration data of each stay cable;
and constructing the corresponding characteristic matrix according to the normalized acceleration of each stay cable.
Preferably, the step of establishing an adjacency matrix according to the position information of each cable-stayed bridge includes:
numbering each stay cable according to the position information of each stay cable, and establishing the association strength between the stay cables according to the numbering;
and establishing the adjacency matrix according to the correlation strength between the stay cables.
Preferably, the step of establishing the strength of association between the stay cables according to the number includes:
setting the stay cables with the serial number difference value within a preset threshold value as an incidence relation;
setting the stay cables of which the serial number difference values are not within a preset threshold value as an unassociated relation;
and generating the association strength between the stay cables according to the association relationship and the non-association relationship.
Preferably, the step of constructing a neural network model combining a neural network and a gated cyclic unit according to the feature matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data, and the adjacency matrix includes:
and setting a model parameter learning rate and the maximum iteration times, and constructing a neural network model combining the neural network of the graph and the gate control circulation unit by using the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix.
In order to realize the method, the method for identifying the damage degree of the stay cable is used for identifying the damage degree of the stay cable on the target cable-stayed bridge, and comprises the following steps:
acquiring actual position information of each stay cable in a target cable-stayed bridge and actual condition data of the target cable-stayed bridge; the actual condition data comprises actual acceleration data of each stay cable in the target cable-stayed bridge;
normalizing the actual acceleration data of the stay cable to generate an actual characteristic matrix;
establishing an actual adjacency matrix according to the actual position information of each stay cable;
and inputting the actual characteristic matrix and the actual adjacent matrix into a neural network model to generate the loss degree of each stay cable of the target cable-stayed bridge.
Preferably, the neural network includes three first neural networks connected in series, and the step of inputting the actual feature matrix and the actual adjacency matrix into the neural network model to generate the damage to each stay cable includes:
inputting the actual feature matrix and the actual adjacency matrix into the first neural network model to generate a first feature matrix;
inputting the first feature matrix and the actual adjacency matrix into the first neural network model to generate a second feature matrix;
inputting the second feature matrix and the actual adjacency matrix into the first neural network model to generate a third feature matrix;
and generating the loss degree of each stay cable of the target cable-stayed bridge according to the third characteristic matrix.
Preferably, the first neural network model comprises a graph neural network and a gated round robin unit, and the step of inputting the actual feature matrix and the actual adjacency matrix into the first neural network model to generate a first feature matrix comprises:
inputting the actual feature matrix and the actual adjacency matrix into the graph neural network for feature extraction to generate hidden layer features;
inputting the hidden layer feature and the actual feature matrix into the gating circulation unit to generate the first feature matrix.
Still include a suspension cable damage degree identification model construction equipment for realizing this application, the device is used for constructing the damage degree identification model of suspension cable on the target cable-stay bridge, include:
the data module is used for acquiring the position information of each stay cable in a target cable-stayed bridge and the condition data of the target cable-stayed bridge; the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data;
the characteristic matrix module is used for normalizing the acceleration data to generate a corresponding characteristic matrix;
the adjacency matrix module is used for establishing an adjacency matrix according to the position information of each cable-stayed bridge;
and the neural network model module is used for constructing a neural network model combining a graph neural network and a gating circulation unit according to the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix.
Still include a suspension cable damage degree recognition device for realizing this application, the device is used for the damage degree discernment of suspension cable on the target cable-stay bridge, include:
the actual data module is used for acquiring the actual position information of each stay cable in the target cable-stayed bridge and the actual condition data of the target cable-stayed bridge; the actual condition data comprises actual acceleration data of each stay cable in the target cable-stayed bridge;
the actual characteristic matrix module is used for normalizing the actual acceleration data of the stay cable to generate an actual characteristic matrix;
the actual adjacency matrix module is used for establishing an actual adjacency matrix according to the actual position information of each stay cable;
and the loss degree module is used for inputting the loss degree of each stayed cable of the target cable-stayed bridge into the neural network model according to the actual characteristic matrix and the actual adjacent matrix to generate the loss degree of each stayed cable of the target cable-stayed bridge.
The application has the following advantages:
in the embodiment of the application, the position information of each stay cable in a target cable-stayed bridge and the condition data of the target cable-stayed bridge are obtained; the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data; normalizing the acceleration data to generate a corresponding feature matrix; establishing an adjacency matrix according to the position information of each cable-stayed bridge; and constructing a neural network model combining a graph neural network and a gated circulation unit according to the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix. By constructing a neural network model combining a graph convolution neural network and a gating circulation unit, the extraction of spatial information is realized, and the problem of excessive smoothness caused by node assimilation is solved; constructing an adjacency matrix to express the relevance between different stay cables; and constructing a stay cable damage identification model based on a graph convolution neural network and a gating circulation unit, and realizing the whole network structure from input to output. The neural network model provided by the application can make full use of the relevance between the stay cables, well retains the information of the nodes, and realizes perfect combination of global and local.
Drawings
In order to more clearly illustrate the technical solutions of the present application, the drawings needed to be used in the description of the present application will be briefly introduced below, and it is apparent that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive labor.
Fig. 1 is a flowchart illustrating steps of a stay cable damage degree identification method according to an embodiment of the present application;
fig. 2 is a flowchart illustrating overall steps of a stay cable damage degree identification method according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of a GCN-GRU module of a stay cable damage degree identification method according to an embodiment of the present application;
fig. 4 is a schematic diagram of a GCN-GRU neural network model framework of a stay cable damage degree identification method according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of a target cable-stayed bridge three-dimensional model of a method for identifying damage degree of a stay cable according to an embodiment of the present application;
fig. 6 is a schematic structural diagram of a target cable-stayed bridge according to a method for constructing a model for identifying damage to a stay cable according to an embodiment of the present application;
fig. 7 is a block diagram illustrating a structure of a model building apparatus for identifying damage degree of a stay cable according to an embodiment of the present application;
fig. 8 is a schematic structural diagram of a target cable-stayed bridge according to a method for identifying damage to a stay cable provided in an embodiment of the present application;
fig. 9 is a block diagram illustrating a structure of a stay cable damage degree identification apparatus according to an embodiment of the present application;
fig. 10 is a schematic structural diagram of a computer device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, features and advantages of the present application more comprehensible, the present application is described in further detail with reference to the accompanying drawings and the detailed description. It should be apparent that the embodiments described are some, but not all embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Referring to fig. 1, a flowchart illustrating steps of a construction method of a model for identifying damage to a stay cable according to an embodiment of the present application is shown, and specifically includes the following steps:
s110, acquiring position information of each stay cable in a target cable-stayed bridge and condition data of the target cable-stayed bridge; the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data;
s120, normalizing the acceleration data to generate a corresponding feature matrix;
s130, establishing an adjacency matrix according to the position information of each cable-stayed bridge;
and S140, constructing a neural network model combining a graph neural network and a gate control cycle unit according to the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix.
Next, a construction method of the stay cable damage degree identification model in the present exemplary embodiment will be further described.
As described in step S110, obtaining position information of each stay cable in the target cable-stayed bridge and status data of the target cable-stayed bridge; and the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data.
In an embodiment of the present invention, the step S110 of "obtaining position information of each cable-stayed cable in the target cable-stayed bridge and status data of the target cable-stayed bridge" may be further described with reference to the following description; the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data; "is used herein.
As an example, acquiring position information of each stay cable in a target cable-stayed bridge and condition data of the target cable-stayed bridge; wherein the condition data comprises at least 90 sets; each group of condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data.
As one example, the operating condition data includes a type of operating condition and a degree of operating condition. The working condition types comprise 83 damaged working conditions and healthy working conditions, and the damage degree reference values of the working condition degrees are 0.01, 0.02, 0.03, 0.04, 0.05, 0.1 and 0.2.
As an example, a data set is prepared, a finite element model of the cable-stayed bridge is established, and vertical acceleration of each stay cable of the cable-stayed bridge under different working conditions is acquired. Specifically, the damage of the stay cable is simulated by adopting a mode of reducing the elastic modulus; and each stay cable is provided with a plurality of acceleration acquisition points, and the corresponding node positions of different stay cables are ensured to be at the same height.
In a specific embodiment, the study object selection and the sensor installation are carried out, the cable-stayed bridge selected in the experiment is of a front-back and left-right symmetrical structure, and a total of 108 stay cables are shown in fig. 5, only 1/4 stay cables are selected for study due to the symmetrical structure, and a total of 27 stay cables are shown in a dashed line frame in fig. 6. Numbering the stay cables, wherein the cable close to the central shaft is No. 1, and the cable far away from the central shaft is No. 27; 12 acceleration sensors are mounted on each cable in a collected mode, the acceleration sensors are 0 # sensors close to the bridge floor and 11 # sensors far away from the bridge floor, and the corresponding sensors are located at the same horizontal position. A total of 324 sensors are mounted to collect vertical acceleration.
In one embodiment, the injury and health conditions are simulated and data collected using finite element software ANSYS (which is a multi-purpose finite element method computer design program). The experiment simulates 83 damaged working conditions and healthy working conditions, applies white noise excitation to the cable-stayed bridge model under each working condition, and collects the vertical acceleration of each node. In the experiment, the elastic modulus is selected as a damage variable, the stay cable damage with different degrees is simulated by reducing the elastic modulus with different sizes, and damage degree reference values of 0.01, 0.02, 0.03, 0.04, 0.05, 0.1 and 0.2 are selected as basic conditions of working conditions. And combining the collected acceleration data of each node, wherein each sample in the combined data set is characterized by the vertical acceleration of 324 nodes of 27 stay cables, and the output variable is the damage degree of each stay cable.
As described in step S120, the acceleration data is normalized to generate a corresponding feature matrix.
In an embodiment of the present invention, the specific process of "normalizing the acceleration data to generate the corresponding feature matrix" in step S120 may be further described with reference to the following description.
Screening acceleration with the largest value and acceleration with the smallest value in the acceleration data according to the following steps; generating a normalized acceleration corresponding to each stay cable according to the acceleration with the maximum numerical value, the acceleration with the minimum numerical value and the acceleration data of each stay cable; and constructing the corresponding characteristic matrix according to the normalized acceleration of each stay cable.
In a specific embodiment, the acceleration data of each node of the stay cable is normalized to obtain preprocessed data, and the preprocessed data is divided into a training set and a test set. Normalization enables each acceleration value to fall within a certain range, the purposes of simplifying calculation and preventing gradient explosion can be achieved, and the formula of normalization is as follows:
wherein x 0 ,x 1 Before and after normalization, respectively, and a and b are the minimum and maximum values of the sample data, respectively.
As an example, the feature matrix is divided into a training feature matrix and a testing feature matrix according to a preset proportion; wherein the preset proportion is 7: 3.
in a specific embodiment, the normalized data set is normalized according to 7: and 3, randomly dividing the ratio into a training set and a test set, and respectively converting the training set and the test set into a form of a feature matrix to form the feature matrix of the training set and the test set. The total 27 stay cables are provided, each stay cable is provided with 12 acceleration acquisition nodes, and the characteristic matrix of one sample is X ═ X 1 ,X 2 ,…,X i ,…,X 27 ],
As mentioned above, in step S130, an adjacency matrix is established according to the position information of each cable-stayed bridge.
In an embodiment of the present invention, the step S130 "establishing an adjacency matrix according to the position information of each cable-stayed bridge may be further described with reference to the following description. "
Numbering each stay cable according to the position information of each stay cable, and establishing the association strength between the stay cables according to the numbering; and establishing the adjacency matrix according to the correlation strength between the stay cables.
In an embodiment of the present invention, the specific process of "establishing the association strength between the stay cables according to the number" may be further described with reference to the following description.
Setting stay cables with the number difference value within a preset threshold value as an incidence relation; setting the stay cables of which the serial number difference values are not within a preset threshold value as an unassociated relation; generating the association strength between the stay cable and the stay cable according to the association relation and the non-association relation; specifically, the preset threshold is three.
In one embodiment, the adjacency matrix a is constructed to reflect the relationship between different stay cables. An excellent adjacent structure can filter out training consumption among nodes with small relevance, and can extract the relation among nodes with strong relevance, so that the training of the model is facilitated. The strength of association between the stay cables is mainly reduced along with the increase of the distance, a threshold value is set by taking one stay cable as a center, and the stay cable within the threshold value is considered to have a dependency relationship with the center stay cable. Because the intersection points of the stay cables and the bridge floor are equidistantly distributed on a straight line and are numbered according to the sequence of geographical positions, the distance between the stay cables can be reflected by the number difference of the stay cables. The graph structure determined by taking the stay cable number difference value as a threshold value is shown as the following formula:
wherein d is i,j Is weaved by stay cables i and jThe difference of the numbers, Dis is the threshold value of the difference of the numbers of the stay cables, a i,j Shows that the two stay cables have a dependency relationship, a i,j The value of (d) depends on the difference d between the numbers of the stay cables i,j Compared with the threshold Dis, a value of 1 indicates that the two stay cables are dependent, and a value of 0 indicates that the two stay cables are not dependent and do not affect each other. Therefore, by filtering the stay cables, each stay cable can integrate the stay cable information related to the stay cable, rather than considering the factors of the stay cable.
In one embodiment, a 27 × 27 dimensional adjacency matrix is constructed based on the connection relationship between the stay cables, using a i,j The stay cable i and the stay cable j are related to each other, and if 1, the stay cable i and the stay cable j are not related to each other. In the experiment, only the relevance is established among the stay cables with the difference not more than 3, if the number difference of the stay cables is not more than 3, the corresponding value in the matrix is 1, otherwise, the number difference is 0, and the established adjacent matrix A is shown as the following
In one embodiment, a target cable-stayed bridge outputs an adjacent matrix of integral stay cables.
And as described in the step S140, a neural network model combining a neural network and a gated loop unit is constructed according to the feature matrix, the operating condition data of the target cable-stayed bridge corresponding to the acceleration data, and the adjacency matrix.
In an embodiment of the present invention, a specific process of "constructing a neural network model combining a neural network and a gated loop unit according to the feature matrix, the operating condition data of the target cable-stayed bridge corresponding to the acceleration data, and the adjacency matrix" in step S140 may be further described with reference to the following description.
And setting a model parameter learning rate and the maximum iteration times, and constructing a neural network model combining a neural network and a gated cyclic unit by using the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix.
In an embodiment of the present invention, a specific process of "setting a model parameter learning rate and a maximum iteration number, and constructing a neural network model combining a neural network of a graph and a gated loop unit from the feature matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data, and the adjacency matrix" may be further described with reference to the following description.
Constructing a feature extraction model according to the combination of the graph convolution neural network and the gating circulation unit; setting three feature extraction models, connecting the three feature extraction models in series, and setting a full connection layer at the last feature extraction model so as to generate a neural network to be trained; setting a model parameter learning rate and the maximum iteration times, and inputting the training characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix into the neural network model to be trained for training to generate a trained neural network; and inputting the test characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix into the trained neural network model to generate the neural network.
As an example, setting a model parameter learning rate and a maximum iteration number, and constructing a first neural network model combining a neural network of a graph and a gating cycle unit by using the feature matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix; and connecting three first neural network model sets in series, and setting a full connection layer at the last first neural network so as to generate the neural network.
In one embodiment, a feature extraction module (GCN-GRU) is constructed that combines a graph convolution neural network and a gated round robin unit. The method comprises the steps of firstly extracting the spatial features of a graph by adopting a graph convolution neural network, enabling each node to contain information of other nodes, and then adding a GRU structure to strengthen the information of the node. The GCN network layer can enable the central node to aggregate information of adjacent nodes, but the problem of excessive smoothness is easy to occur after multiple times of graph convolution, namely the nodes are similar in characteristics and cannot be distinguished. In order to solve the problem, the invention adopts the GRU to strengthen the information of the node itself and prevent the over-smoothing, as shown in the following formula:
X (l+1) =GRU(GCN(X (l) ,A;θ (l) ),X (l) )
for the GCN-GRU network layer of the l layer, firstly, one GCN layer is used for outputting X of the upper layer (l) Performing convolution, if the current layer is the first GCN-GRU network layer, performing convolution on the original characteristic X, taking the result of GCN convolution as a hidden layer state of the GRU input unit, and keeping the input characteristic of the GRU of the current layer to be X (l) Therefore, the information of the current node is ensured, and the information of the adjacent nodes is fused. The structure of a GCN-GRU module is shown in FIG. 3.
As an example, a GCN-GRU neural network model is established, 3 GCN-GRU network layers are connected in series for feature extraction, and the results of each node are output through a full connection layer, and the specific structure is shown in fig. 4. And selecting the mean square error as a loss function, and selecting an Adam algorithm to optimize model parameters.
As an example, a GCN-GRU neural network model is established, 3 GCN-GRU network layers are connected in series for feature extraction, and the results of each node are output through a full connection layer, and the specific structure is shown in fig. 3. And selecting the mean square error as a loss function, and selecting an Adam algorithm to optimize model parameters.
Training a GCN-GRU neural network model, and setting a model parameter learning rate, a maximum iteration number and an iteration termination condition. And continuously updating the model parameters until the training stopping condition is met, and obtaining the trained stay cable damage identification model. The training process is as follows:
TABLE 1 GCN-GRU model training process table
And inputting the test data and the adjacency matrix into the trained model to obtain a result of the model on the test set, and judging the effect of the model.
It should be noted that GCN is a neural network for studying graphs, which essentially acts as CNN, and is a feature extractor, except that its object is graph data. The GCN ingeniously designs a method for extracting features from the graph data, so that the features can be used for node classification and prediction of the graph data, and the embedded representation of the graph can be obtained incidentally. The core idea of the graph convolution neural network is to use the information of other nodes to deduce the information of the node to be predicted, and the graph convolution neural network is a process of feature propagation in each node.
The training process of the graph convolution neural network needs to describe the propagation process of a graph structure by using an arrival degree matrix and an adjacent matrix, and multi-layer aggregation is carried out by using the information of edges existing in the matrixes, so that a final characteristic matrix which tends to be stable is generated. The degree matrix represents the number of nodes connected with other nodes, only the diagonal has numerical values, and other positions are zero. The connection matrix represents the whole graph structure, and it can be seen from the connection matrix which nodes are associated before, and two nodes are connected with a value of 1, and the other is 0. Finally, a Laplace matrix fusing the information of the two matrixes can be obtained by connecting the matrix and the degree matrix, and the calculation process is shown by the following formula:
in the formula, D represents a degree matrix, and A represents an adjacency matrix. F is a laplace matrix, which has the advantage that the laplace matrix is a symmetric matrix and can be subjected to eigen decomposition.
The graph convolution neural network has the advantages that unnecessary training processes among points can be reduced through graph structure information, all relevant nodes are continuously iterated and integrated in a weighting mode, a characteristic matrix which tends to be stable is finally trained, the matrix contains information among all nodes, and finally a predicted value can be obtained through feature extraction of a full connection layer. The propagation calculation process between layers in GCN is shown by the following formula:
in the formula, D represents a degree matrix, A represents an adjacency matrix, H is a characteristic matrix of each layer, if the first layer is the layer, the characteristic matrix is the input X of the graph convolution neural network, W is a parameter matrix of the l-th layer, and ReLU is a nonlinear activation function.
Gru (gate recovery unit) is one of Recurrent Neural Networks (RNN). Like LSTM (Long-Short Term Memory), it is proposed to solve the problems of Long-Term Memory and gradient in back-propagation.
In a specific embodiment, a GCN-GRU stay cable damage identification model is designed, and the GCN-GRU neural network model designed in the step is formed by connecting 3 GCN-GRU modules in series and then connecting the GCN-GRU neural network model with a full connection layer. The specific GCN-GRU module extracts the spatial characteristics of the stay cable by a convolutional neural network, the extracted result of the characteristics is used as a hidden state and is input into a GRU structure, and the input characteristics of the GRU are still the input characteristics of the GCN network.
And inputting the training characteristic matrix and the adjacency matrix into a GCN-GRU stay cable damage identification model for training. And continuously updating the model parameters until the training stopping condition is met, and obtaining the trained traffic flow prediction model. The mean square error MSE is selected as a loss function, and the Adam algorithm is selected as a parameter optimizer.
Inputting the test characteristic matrix and the adjacency matrix into the trained GCN-GRU to predict through a network model, obtaining the loss degree of the stay cable, and selecting three error functions of average absolute error, average absolute percentage error and root mean square error as indexes for evaluating the prediction effect of the model.
In one embodiment, referring to fig. 2, the present application provides a stayed cable damage identification method (GCN-GRU) combining a graph convolutional neural network and a gated cyclic unit. The GCN is adopted to extract the spatial features of the stay cable, the GCN extraction result is used as a hidden state and is input to the GRU, the GRU still has the features of not extracting the GCN, and the global information of the stay cable can be effectively extracted through the GCN-GRU module for a plurality of times, and the local information of the stay cable is also kept.
The method and the device can identify the damage degree of all the stay cables by using the damage working condition data of part of typical stay cables. Because the damage working conditions of the stay cables are very many, the stay cables can be single cables or multiple cables, the combination conditions of different cables are very many, the damage degree of each cable is different, and the difficulty in acquiring all the damage working conditions is very high.
This application can utilize the associativity between the suspension cable, because the cable-stay bridge is a whole, has received the damage when certain cable, and adjacent cable also can receive certain influence.
The method needs to effectively combine global information and local information of the stay cables, each stay cable is related, but the excessive emphasis on the relationship can assimilate different stay cables, so that the characteristics of the stay cables cannot be ignored while the information related to other stay cables is fused.
Referring to fig. 8, a flowchart illustrating steps of a stay cable damage degree identification method according to an embodiment of the present application is shown, and specifically includes the following steps:
s810, acquiring actual position information of each stay cable in a target cable-stayed bridge and actual condition data of the target cable-stayed bridge; the actual condition data comprises actual acceleration data of each stay cable in the target cable-stayed bridge;
s820, normalizing the actual acceleration data of the stay cable to generate an actual characteristic matrix;
s830, establishing an actual adjacency matrix according to the actual position information of each stay cable;
and S840, inputting the actual characteristic matrix and the actual adjacent matrix into a neural network model to generate the loss degree of each stayed cable of the target cable-stayed bridge.
As described in the above step S810, acquiring actual position information of each stay cable in the target cable-stayed bridge and actual condition data of the target cable-stayed bridge; and the actual condition data comprises actual acceleration data of each stay cable in the target cable-stayed bridge.
In an embodiment of the present invention, the step S810 of "obtaining actual position information of each cable-stayed cable in the target cable-stayed bridge and actual status data of the target cable-stayed bridge" may be further described with reference to the following description; and the actual condition data comprises actual acceleration data of each stayed cable in the target cable-stayed bridge. "is used herein.
In a specific embodiment, the study object selection and sensor installation are performed, the cable-stayed bridge selected in the experiment is a front-back and left-right symmetrical structure, and the total number of 108 cable-stayed bridges is shown in fig. 5, and only 1/4 cable is selected for study due to the symmetrical structure, and the total number of 27 cable-stayed bridges is shown in a dotted line frame in fig. 6. Numbering the stay cables, wherein the cable close to the central shaft is No. 1, and the cable far away from the central shaft is No. 27; 12 acceleration sensors are mounted on each cable in a collected mode, the acceleration sensors are 0 # sensors close to the bridge floor and 11 # sensors far away from the bridge floor, and the corresponding sensors are located at the same horizontal position. A total of 324 sensors are mounted to collect vertical acceleration.
As described in step S820, the actual acceleration data of the stay cable is normalized to generate an actual feature matrix.
In an embodiment of the present invention, a specific process of "normalizing the actual acceleration data of the stay cable to generate the actual feature matrix" in step S820 may be further described with reference to the following description.
Screening out the actual acceleration with the largest numerical value and the actual acceleration with the smallest numerical value in the actual acceleration data according to the following steps; generating actual normalized acceleration corresponding to each stay cable according to the actual acceleration with the maximum numerical value, the actual acceleration with the minimum numerical value and the actual acceleration data of each stay cable; and constructing the corresponding actual characteristic matrix according to the actual normalized acceleration of each stay cable.
In a specific embodiment, the acceleration data of each node of the stay cable is normalized to obtain preprocessed data, and the preprocessed data is divided into a training set and a test set. Normalization enables each acceleration value to fall within a certain range, the purposes of simplifying calculation and preventing gradient explosion can be achieved, and the formula of normalization is as follows:
wherein x 0 ,x 1 The values before and after normalization, respectively, and a and b are the minimum and maximum values of the sample data, respectively.
As described in step S830, an actual adjacency matrix is established according to the actual position information of each cable-stayed bridge.
In an embodiment of the present invention, the step S830 "establishing an actual adjacency matrix according to the actual position information of each cable-stayed bridge" can be further described with reference to the following description. "
Numbering each stay cable according to the actual position information of each stay cable, and establishing the actual association strength between the stay cables according to the numbering; and establishing the actual adjacency matrix according to the actual correlation strength between the stay cables.
In an embodiment of the present invention, the specific process of "establishing the actual association strength between the stay cables according to the number" may be further described with reference to the following description.
Setting the stay cables with the number difference value within a preset threshold value as an actual association relation as follows; setting the stay cables of which the serial number difference values are not within a preset threshold value as an actual unassociated relation; and generating the actual association strength between the stay cable and the stay cable according to the actual association relation and the actual non-association relation.
In one embodiment, a database 27 is constructed based on the connection relationship between the stay cables27 dimensional actual adjacency matrix, using a i,j The stay cable i and the stay cable j are related to each other, and if 1, the stay cable i and the stay cable j are not related to each other. In the experiment, only the relations between the stayed cables with the difference not more than 3 are established, if the number difference of the stayed cables is not more than 3, the corresponding value in the matrix is 1, otherwise, the number difference is 0, and the established actual adjacent matrix A is shown as the following
As described in step S840, the loss degree of each stay cable of the target cable-stayed bridge is generated according to the actual feature matrix and the actual adjacency matrix input neural network model.
In an embodiment of the present invention, the specific process of "generating the loss degree of each stay cable of the target cable-stayed bridge according to the actual feature matrix and the actual adjacency matrix input to the neural network model" in step S840 may be further described with reference to the following description.
Inputting the actual feature matrix and the actual adjacency matrix into the first neural network model to generate a first feature matrix; inputting the first feature matrix and the actual adjacency matrix into the first neural network model to generate a second feature matrix; inputting the second feature matrix and the actual adjacency matrix into the first neural network model to generate a third feature matrix; and generating the loss degree of each stay cable of the target cable-stayed bridge according to the third characteristic matrix.
In an embodiment of the present invention, the step of inputting the actual feature matrix and the actual adjacency matrix into the first neural network model to generate the first feature matrix may be further described with reference to the following description.
Inputting the actual feature matrix and the actual adjacent matrix into the graph neural network for feature extraction to generate hidden layer features; inputting the hidden layer feature and the actual feature matrix into the gating circulation unit to generate the first feature matrix.
For the device embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.
Referring to fig. 7, a device for constructing a model for identifying damage degree of a stay cable according to an embodiment of the present application is shown, which specifically includes the following modules,
the data module 710: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring position information of each stay cable in a target cable-stayed bridge and condition data of the target cable-stayed bridge; the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data;
the feature matrix module 720: the characteristic matrix is used for normalizing the acceleration data to generate a corresponding characteristic matrix;
adjacency matrix module 730: the system comprises a plurality of cable-stayed bridges, a plurality of adjacent matrixes and a plurality of control modules, wherein the adjacent matrixes are used for establishing adjacent matrixes according to the position information of each cable-stayed bridge;
the neural network model module 740: and the neural network model is used for constructing a neural network model combining a graph neural network and a gate control cycle unit according to the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix.
In an embodiment of the present invention, the feature matrix module 720 includes:
an acceleration submodule: the acceleration screening device is used for screening the acceleration with the largest numerical value and the acceleration with the smallest numerical value in the acceleration data;
a normalized acceleration submodule: the acceleration data acquisition unit is used for acquiring the acceleration data of each inclined stay cable and the acceleration data of the corresponding inclined stay cable;
a feature matrix submodule: and the characteristic matrix is used for constructing the corresponding characteristic matrix according to the normalized acceleration of each stay cable.
In an embodiment of the present invention, the adjacency matrix module 730 includes:
the correlation strength submodule: the system comprises a plurality of stay cables, a plurality of sensors and a controller, wherein the stay cables are used for being numbered according to the position information of the stay cables and establishing the association strength between the stay cables according to the numbers;
adjacency matrix submodule: and the adjacency matrix is established according to the correlation strength between the stay cables.
In an embodiment of the present invention, the association strength sub-module includes:
a correlation submodule: the stay cables with the serial number difference value within a preset threshold value are set to be in an incidence relation;
unassociated sub-modules: the system comprises a plurality of stay cables, a plurality of sensors and a controller, wherein the stay cables are used for setting the number difference value between the stay cables not within a preset threshold value as an unassociated relation;
and an association strength generation submodule: and the system is used for generating the association strength between the stay cable and the stay cable according to the association relation and the non-association relation.
In an embodiment of the present invention, the neural network model module 740 includes:
a neural network model submodule: and the neural network model is used for setting the learning rate and the maximum iteration times of model parameters, and combining the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix construction graph neural network with the gating circulation unit.
Referring to fig. 9, a stay cable damage degree identification apparatus provided in an embodiment of the present application is shown, which specifically includes the following modules,
the actual data module 910: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring actual position information of each stay cable in a target cable-stayed bridge and actual condition data of the target cable-stayed bridge; the actual condition data comprises actual acceleration data of each stay cable in the target cable-stayed bridge;
actual feature matrix module 920: the actual characteristic matrix is generated by normalizing the actual acceleration data of the stay cable;
actual adjacency matrix module 930: the actual adjacency matrix is established according to the actual position information of each stayed cable;
loss degree module 940: and the loss degree of each stayed cable of the target cable-stayed bridge is generated by inputting the actual characteristic matrix and the actual adjacent matrix into a neural network model.
In an embodiment of the present invention, the loss degree module 940 includes:
a first feature matrix submodule: the actual feature matrix and the actual adjacency matrix are input into the first neural network model to generate a first feature matrix;
a second feature matrix submodule: the actual adjacency matrix is input into the first neural network model to generate a first feature matrix;
a third feature matrix submodule: the actual adjacency matrix is input into the first neural network model to generate a second feature matrix;
loss level submodule: and the loss degree of each stay cable of the target cable-stayed bridge is generated according to the third characteristic matrix.
In an embodiment of the present invention, the first feature matrix sub-module includes:
hidden layer feature submodule: the actual feature matrix and the actual adjacency matrix are input into the graph neural network for feature extraction to generate hidden layer features;
a first feature matrix generation submodule: for inputting the hidden layer feature and the actual feature matrix into the gated round unit to generate the first feature matrix.
It should be noted that for simplicity of description, the method embodiments are shown as a series of combinations of acts, but those skilled in the art will recognize that the embodiments are not limited by the order of acts, as some steps may occur in other orders or concurrently in accordance with the embodiments. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred and that no particular act is required to implement the invention.
The present embodiment and the above embodiments have repeated operation steps, and the present embodiment is only described briefly, and the rest of the schemes may be described with reference to the above embodiments.
For the device embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.
The application further includes an electronic device, which includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, where the computer program is executed by the processor to implement the steps of the stay cable damage degree identification method.
The present application further includes a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the stay cable damage degree identification method as described above.
Referring to fig. 10, a computer device for identifying damage degree of a stay cable according to the present application is shown, which may specifically include the following:
the computer device 12 described above is embodied in the form of a general purpose computing device, and the components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a memory 28, and a bus 18 that couples various system components including the memory 28 and the processing unit 16.
The memory 28 may include computer system readable media in the form of volatile memory, such as random access memory 30 and/or cache memory 32. Computer device 12 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 34 may be used to read from and write to non-removable, nonvolatile magnetic media (commonly referred to as "hard drives"). Although not shown in FIG. 10, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 by one or more data media interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 42, with the program modules 42 configured to carry out the functions of embodiments of the application.
A program/utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules 42, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. Program modules 42 generally perform the functions and/or methodologies of the embodiments described herein.
The processing unit 16 executes various functional applications and data processing by running the program stored in the memory 28, for example, implementing a method for identifying damage level of a stay cable according to the embodiment of the present application.
That is, the processing unit 16 implements, when executing the program,: acquiring position information of each stay cable in a target cable-stayed bridge and condition data of the target cable-stayed bridge; the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data; normalizing the acceleration data to generate a corresponding feature matrix; establishing an adjacency matrix according to the position information of each cable-stayed bridge; and constructing a neural network model combining a graph neural network and a gated circulation unit according to the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix.
In the embodiments of the present application, the present application further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements a method for identifying damage level of a stay cable according to all embodiments of the present application.
That is, the program when executed by the processor implements: acquiring position information of each stay cable in a target cable-stayed bridge and condition data of the target cable-stayed bridge; the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data; normalizing the acceleration data to generate a corresponding feature matrix; establishing an adjacency matrix according to the position information of each cable-stayed bridge; and constructing a neural network model combining a graph neural network and a gate control cycle unit according to the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix.
Any combination of one or more computer-readable media may be employed. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Computer program code for carrying out operations for aspects of the present application may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the operator's computer, partly on the operator's computer, as a stand-alone software package, partly on the operator's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the operator's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider). The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
While preferred embodiments of the present application have been described, additional variations and modifications of these embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including the preferred embodiment and all such alterations and modifications as fall within the true scope of the embodiments of the application.
Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or terminal that comprises the element.
The method for identifying the damage degree of the stay cable and the method for constructing the model are introduced in detail, specific examples are applied to explain the principle and the implementation mode of the method, and the description of the examples is only used for helping to understand the method and the core idea of the method; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Claims (10)
1. A method for building a model for identifying damage degree of a stay cable is used for building a model for identifying damage degree of a stay cable on a target cable-stayed bridge, and is characterized by comprising the following steps:
acquiring position information of each stay cable in a target cable-stayed bridge and condition data of the target cable-stayed bridge; the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data;
normalizing the acceleration data to generate a corresponding characteristic matrix;
establishing an adjacency matrix according to the position information of each cable-stayed bridge;
and constructing a neural network model combining a graph neural network and a gate control cycle unit according to the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix.
2. The method for constructing a model for identifying damage degree of a stay cable according to claim 1, wherein the step of normalizing the acceleration data to generate a corresponding feature matrix comprises the steps of:
screening out the acceleration with the largest numerical value and the acceleration with the smallest numerical value in the acceleration data;
generating a normalized acceleration corresponding to each stay cable according to the acceleration with the maximum numerical value, the acceleration with the minimum numerical value and the acceleration data of each stay cable;
and constructing the corresponding characteristic matrix according to the normalized acceleration of each stay cable.
3. The method for constructing a model for identifying damage to a cable-stayed cable according to claim 1, wherein the step of establishing an adjacency matrix based on the position information of each cable-stayed bridge includes:
numbering each stay cable according to the position information of each stay cable, and establishing the association strength between the stay cables according to the numbering;
and establishing the adjacency matrix according to the correlation strength between the stay cables.
4. A method for constructing a model for identifying damage to a stay cable according to claim 3, wherein the step of establishing the strength of association between the stay cable and the stay cable according to the number includes:
setting the stay cables with the serial number difference value within a preset threshold value as an incidence relation;
setting the stay cables of which the serial number difference values are not within a preset threshold value as an unassociated relation;
and generating the association strength between the stay cables according to the association relationship and the non-association relationship.
5. The method for constructing a model for identifying damage degree of a stay cable according to claim 1, wherein the step of constructing a neural network model combining a neural network and a gated loop unit according to the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data, and the adjacency matrix comprises:
and setting a model parameter learning rate and the maximum iteration times, and constructing a neural network model combining the neural network of the graph and the gate control circulation unit by using the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix.
6. A method for identifying damage degree of a stay cable on a target cable-stayed bridge is characterized by comprising the following steps:
acquiring actual position information of each stay cable in a target cable-stayed bridge and actual condition data of the target cable-stayed bridge; the actual condition data comprises actual acceleration data of each stay cable in the target cable-stayed bridge;
normalizing the actual acceleration data of the stay cable to generate an actual characteristic matrix;
establishing an actual adjacency matrix according to the actual position information of each stay cable;
and inputting the actual characteristic matrix and the actual adjacent matrix into a neural network model to generate the loss degree of each stayed cable of the target cable-stayed bridge.
7. The method for identifying damage degree of a stay cable according to claim 6, wherein the neural network includes three first neural networks connected in series, and the step of generating damage of each stay cable by inputting the actual feature matrix and the actual adjacency matrix into the neural network model includes:
inputting the actual feature matrix and the actual adjacency matrix into the first neural network model to generate a first feature matrix;
inputting the first feature matrix and the actual adjacency matrix into the first neural network model to generate a second feature matrix;
inputting the second feature matrix and the actual adjacency matrix into the first neural network model to generate a third feature matrix;
and generating the loss degree of each stayed cable of the target cable-stayed bridge according to the third characteristic matrix.
8. The stay cable damage degree identification method according to claim 7, wherein the first neural network model includes a graph neural network and a gated cyclic unit, and the step of inputting the actual feature matrix and the actual adjacency matrix into the first neural network model to generate the first feature matrix includes:
inputting the actual feature matrix and the actual adjacent matrix into the graph neural network for feature extraction to generate hidden layer features;
inputting the hidden layer feature and the actual feature matrix into the gating circulation unit to generate the first feature matrix.
9. The utility model provides a stay cable damage degree identification model construction equipment, the device is used for constructing the damage degree identification model of stay cable on the target cable-stay bridge, its characterized in that includes:
the data module is used for acquiring the position information of each stay cable in a target cable-stayed bridge and the condition data of the target cable-stayed bridge; the condition data comprises acceleration data of each stay cable in the target cable-stayed bridge and working condition data of the target cable-stayed bridge corresponding to the acceleration data;
the characteristic matrix module is used for normalizing the acceleration data to generate a corresponding characteristic matrix;
the adjacent matrix module is used for establishing an adjacent matrix according to the position information of each cable-stayed bridge;
and the neural network model module is used for constructing a neural network model combining a graph neural network and a gating circulation unit according to the characteristic matrix, the working condition data of the target cable-stayed bridge corresponding to the acceleration data and the adjacency matrix.
10. A stay cable damage degree recognition device, the device is used for the damage degree recognition of the stay cable on the target cable-stayed bridge, which is characterized by comprising:
the actual data module is used for acquiring actual position information of each stay cable in the target cable-stayed bridge and actual condition data of the target cable-stayed bridge; the actual condition data comprises actual acceleration data of each stay cable in the target cable-stayed bridge;
the actual characteristic matrix module is used for normalizing the actual acceleration data of the stay cable to generate an actual characteristic matrix;
the actual adjacency matrix module is used for establishing an actual adjacency matrix according to the actual position information of each stay cable;
and the loss degree module is used for inputting the loss degree of each stayed cable of the target cable-stayed bridge into the neural network model according to the actual characteristic matrix and the actual adjacent matrix to generate the loss degree of each stayed cable of the target cable-stayed bridge.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210608594.9A CN114969934B (en) | 2022-05-31 | 2022-05-31 | Stay cable damage degree identification method and model construction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210608594.9A CN114969934B (en) | 2022-05-31 | 2022-05-31 | Stay cable damage degree identification method and model construction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114969934A true CN114969934A (en) | 2022-08-30 |
CN114969934B CN114969934B (en) | 2023-09-05 |
Family
ID=82957192
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210608594.9A Active CN114969934B (en) | 2022-05-31 | 2022-05-31 | Stay cable damage degree identification method and model construction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114969934B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108090295A (en) * | 2017-12-27 | 2018-05-29 | 武汉光谷北斗控股集团有限公司 | A kind of long-span cablestayed bridges Damages in Stay Cables recognition methods |
CN111783212A (en) * | 2020-07-09 | 2020-10-16 | 长沙理工大学 | Typical damage identification method for cable-stayed bridge |
CN112733933A (en) * | 2021-01-08 | 2021-04-30 | 北京邮电大学 | Data classification method and device based on unified optimization target frame graph neural network |
CN112890827A (en) * | 2021-01-14 | 2021-06-04 | 重庆兆琨智医科技有限公司 | Electroencephalogram identification method and system based on graph convolution and gate control circulation unit |
CN114169374A (en) * | 2021-12-10 | 2022-03-11 | 湖南工商大学 | Cable-stayed bridge stay cable damage identification method and electronic equipment |
US20220101103A1 (en) * | 2020-09-25 | 2022-03-31 | Royal Bank Of Canada | System and method for structure learning for graph neural networks |
-
2022
- 2022-05-31 CN CN202210608594.9A patent/CN114969934B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108090295A (en) * | 2017-12-27 | 2018-05-29 | 武汉光谷北斗控股集团有限公司 | A kind of long-span cablestayed bridges Damages in Stay Cables recognition methods |
CN111783212A (en) * | 2020-07-09 | 2020-10-16 | 长沙理工大学 | Typical damage identification method for cable-stayed bridge |
US20220101103A1 (en) * | 2020-09-25 | 2022-03-31 | Royal Bank Of Canada | System and method for structure learning for graph neural networks |
CN112733933A (en) * | 2021-01-08 | 2021-04-30 | 北京邮电大学 | Data classification method and device based on unified optimization target frame graph neural network |
CN112890827A (en) * | 2021-01-14 | 2021-06-04 | 重庆兆琨智医科技有限公司 | Electroencephalogram identification method and system based on graph convolution and gate control circulation unit |
CN114169374A (en) * | 2021-12-10 | 2022-03-11 | 湖南工商大学 | Cable-stayed bridge stay cable damage identification method and electronic equipment |
Non-Patent Citations (1)
Title |
---|
杨杰;李爱群;缪长青;: "BP神经网络在大跨斜拉桥的斜拉索损伤识别中的应用", no. 05, pages 72 - 77 * |
Also Published As
Publication number | Publication date |
---|---|
CN114969934B (en) | 2023-09-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Lin et al. | Structural damage detection with automatic feature‐extraction through deep learning | |
Xu et al. | Identification framework for cracks on a steel structure surface by a restricted Boltzmann machines algorithm based on consumer‐grade camera images | |
CN109583501B (en) | Method, device, equipment and medium for generating image classification and classification recognition model | |
CN111723732B (en) | Optical remote sensing image change detection method, storage medium and computing equipment | |
CN113779675B (en) | Physical-data driven intelligent shear wall building structure design method and device | |
Zou et al. | Multicategory damage detection and safety assessment of post‐earthquake reinforced concrete structures using deep learning | |
CN111126202A (en) | Optical remote sensing image target detection method based on void feature pyramid network | |
CN106570522B (en) | Object recognition model establishing method and object recognition method | |
CN111292195A (en) | Risk account identification method and device | |
CN113761250A (en) | Model training method, merchant classification method and device | |
CN110675954A (en) | Information processing method and device, electronic equipment and storage medium | |
Liao et al. | A channel-spatial-temporal attention-based network for vibration-based damage detection | |
CN112819024B (en) | Model processing method, user data processing method and device and computer equipment | |
Nikose et al. | Dynamic wind response of tall buildings using artificial neural network | |
CN113553356A (en) | Drilling parameter prediction method and system | |
CN114418189A (en) | Water quality grade prediction method, system, terminal device and storage medium | |
CN112445957A (en) | Social network abnormal user detection method, system, medium, equipment and terminal | |
CN116628903A (en) | Optimal arrangement method for urban wind field environment monitoring sensors | |
CN111352926B (en) | Method, device, equipment and readable storage medium for data processing | |
Zhu et al. | Loan default prediction based on convolutional neural network and LightGBM | |
CN117688509A (en) | Multi-mode false news detection method based on multi-level fusion and attention mechanism | |
CN111158918B (en) | Supporting point parallel enumeration load balancing method, device, equipment and medium | |
CN112508687A (en) | AI credit evaluation method, system, electronic device and storage medium | |
CN114969934A (en) | Stay cable damage degree identification method and model construction method | |
CN116977271A (en) | Defect detection method, model training method, device and electronic equipment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |