Graph network-based bridge structure service performance spatio-temporal evolution prediction method and device

Abstract

The invention provides a bridge structure service performance spatio-temporal evolution prediction method and device based on a graph network, and relates to the technical field of bridge structure service performance prediction. The method comprises the following steps: constructing a bridge graph model; constructing a basic bridge structure service performance evolution prediction model based on the bridge graph model and the space-time attention graph convolutional neural network; embedding a parameterized graph learning module for adaptively updating the space-time dependency relationship of the bridge structural members, obtaining a structural member service performance evolution time sequence sample data set of the to-be-predicted bridge, and training the embedded bridge structure service performance evolution prediction model to obtain a trained bridge structure service performance evolution prediction model; and according to the model, obtaining a bridge structure service performance spatio-temporal evolution prediction result. The method aims at effectively capturing the space-time dependency relationship between structural members of the same type and between structural members of different types in a bridge, and the precision of a bridge structure service performance space-time evolution prediction model is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent operation and maintenance of bridges and prediction of the service performance of bridge structures, and in particular to a method and device for predicting the spatiotemporal evolution of the service performance of bridge structures based on a graph network. Background Art

[0002] As a key structure that crosses natural obstacles, bridges are an important part of the transportation network and play a significant role in promoting national economic development and improving people's quality of life. The service performance of bridge structures directly affects the service level and safety of transportation networks. During the long-term service of bridge structures, the service performance will degrade with the increase of service time due to the corrosion of the service environment, vehicle loads and their coupling effects. At the same time, the evolution of the service performance of bridge structures shows obvious spatiotemporal dependence. Specifically, the service environment, material properties and mechanical properties of the same type of structural components in the bridge, as well as between different types of structural components, are interrelated and jointly affect the service performance of the bridge. Therefore, predicting the spatiotemporal evolution of the service performance of bridge structures is crucial for making intelligent maintenance decisions throughout the life cycle of bridges.

[0003] Traditional prediction methods for the evolution of service performance of bridge structures mainly include regression analysis and probability statistics. These methods can predict the evolution of service performance of bridge structures to a certain extent, but they often require a lot of assumptions and simplifications, making it difficult to deal with the influence of multi-factor coupling, and ignoring the spatiotemporal dependence between structural components of the same type and between different types of structural components in the bridge. Regression analysis processes the time series data of bridge structures through techniques such as polynomial fitting, characterizes the nonlinear relationship between the service performance of bridge structures and time, and constructs a model of their evolution over time. Existing studies have determined the influence of multiple variables such as load, environment, and design characteristics on the service performance of bridge structures through multivariate regression analysis, and established and verified the service performance evolution model of prestressed concrete bridge superstructures and substructures (Srikanth I, Arockiasamy M. Remaining service life prediction of agingconcrete bridges based on multiple relevant explanatory variables[J]. Practice Periodical on Structural Design and Construction, 2021, 26(4):04021036.). However, such methods do not consider the randomness and uncertainty in the evolution of service performance, and ignore the spatiotemporal dependencies between structural components of the same type and between different types of structural components in the bridge. The prediction model based on probabilistic statistical methods captures the randomness in the evolution of bridge structures by simulating the changes in random variables such as the service performance of bridge structures or the duration of the structure in a specific service performance. It can be mainly divided into two categories: one is a state-based model, which is based on the probability of transitioning from one state to another in a discrete time interval, such as the Markov model, which captures the randomness in its evolution process by treating the evolution of the service performance of the bridge structure as a random process, and realizes the prediction of bridge performance. Based on the historical inspection data of bridges, the existing research established a semi-Markov continuous time process bridge service performance evolution prediction model and verified its performance (Furtado F, Ribeiro D. Railway Bridge Management System Based on Visual Inspections with Semi-Markov Continuous Time Process[J]. KSCE Journal of Civil Engineering, 2023, 27(1): 233-250.).The other type is a time-based model, which treats the duration of a bridge structure in a certain state as a random variable and uses a probability distribution function to describe the degradation process. The disadvantage of this type of method is that it relies heavily on the orderliness and standardization of the data and ignores the spatiotemporal dependencies between structural components of the same type and between structural components of different types in the bridge. Since the above method cannot handle the influence of multi-factor coupling during the evolution of service performance and the spatiotemporal dependencies between structural components, there are certain limitations in its application.

[0004] In recent years, with the development of big data and artificial intelligence, especially the breakthroughs of deep learning technology in image recognition and complex pattern analysis, the prediction methods of bridge service performance evolution based on machine learning have gradually attracted attention. Machine learning methods can process high-dimensional, nonlinear and noisy data and capture the potential patterns and trends of complex data. The current prediction models based on machine learning can be roughly divided into artificial neural network models, fuzzy set theory models and deep learning models, etc. The prediction performance of these models has been verified by a large number of data sets. The prediction model of bridge service performance evolution based on traditional machine learning, represented by artificial neural network models, can accurately predict the overall service performance of bridges. Deep learning methods are more capable than traditional machine learning methods in processing large-scale data and obtaining the evolution laws under corrosion, vehicle loads and their coupling. The convolutional neural network-based bridge service performance evolution prediction model learns tens of thousands of historical data from different bridges to predict the evolution of bridge structure service performance under the coupling of multiple factors (Liu H, Zhang Y. Bridge condition rating data modeling using deep learning algorithm[J]. Structure and Infrastructure Engineering, 2020, 16(10): 14471460.; Liu H, Nehme J, Lu P. An application of convolutional neural network fordeterioration modeling of highway bridge components in the United States[J]. Structure and Infrastructure Engineering, 2023, 19(6): 731-744.). The bridge service performance evolution prediction model based on recurrent neural network learns the evolution law from several years of integrated data, and better completes the prediction of bridge structure service performance evolution (Liu K, El-Gohary N. Deep learning-based analyticsof multisource heterogeneous bridge data for enhanced data-driven bridge deterioration prediction[J]. Journal of Computing in Civil Engineering, 2022,36(5): 04022023.).However, the above model does not take into account the temporal and spatial dependencies between structural components of the same type and between structural components of different types in the bridge, making it difficult to accurately predict the service performance of the bridge structure and having certain limitations.

[0005] The evolution of service performance of bridge structures is affected by corrosion in the service environment, vehicle loads and their coupling effects, and shows obvious spatiotemporal dependence. Specifically, the service environment, material properties and mechanical properties of the same type of structural components and different types of structural components in the bridge are interrelated, and jointly affect the service performance of the bridge structure. The prediction model of service performance evolution of bridge structures based on regression analysis can characterize the nonlinear relationship between the variables affecting the evolution of service performance and service performance, but it cannot predict the evolution of service performance under the coupling of multiple factors, and ignores the spatiotemporal dependence between the same type of structural components and different types of structural components in the bridge. The prediction model of service performance evolution of bridge structures based on probability and statistics methods, such as the Markov chain model, has a high dependence on the orderliness and standardization of data, and the prediction results obtained are very sensitive to the model parameters. At the same time, this method ignores the spatiotemporal dependence between the same type of structural components and different types of structural components in the bridge. The prediction model of service performance evolution of bridge structures based on machine learning can process high-dimensional, nonlinear and noisy data, capture the potential patterns and trends of complex data, and the prediction effect of the model is better than the above-mentioned traditional prediction models. Among them, the deep learning-based prediction model for the evolution of service performance of bridge structures can process large-scale bridge structure data, and its performance is better than other traditional machine learning models. However, the existing deep learning-based prediction models for the evolution of service performance of bridge structures are all based on time series data for training. The spatiotemporal dependencies between structural components of the same type and between structural components of different types in the bridge are not considered in the model, making it difficult to accurately predict the service performance of the bridge structure. Therefore, the spatiotemporal evolution prediction model for the service performance of bridge structures based on deep learning needs to be further developed.

[0006] Existing machine learning methods are still insufficient in predicting the evolution of bridge structure service performance. Research mainly focuses on the evolution prediction in the time dimension, ignoring the relationship between the same type of structural components and different types of structural components in the bridge in the spatial dimension. It is necessary to develop a bridge structure service performance prediction model that can consider the temporal and spatial dependencies of different structural components of the bridge, so as to improve the accuracy and efficiency of service performance prediction, thereby better supporting intelligent maintenance decisions for bridges. Summary of the invention

[0007] In order to solve the shortcomings of existing machine learning methods in predicting the evolution of bridge structure service performance. The research mainly focuses on the evolution prediction of the time dimension, ignoring the relationship between the same type of structural components and different types of structural components in the bridge in the spatial dimension. It is necessary to develop a bridge structure service performance prediction model that can consider the spatiotemporal dependencies of different structural components of the bridge to improve the accuracy and efficiency of service performance prediction, so as to better support the technical problem of intelligent maintenance decision-making of bridges. The embodiment of the present invention provides a method and device for predicting the spatiotemporal evolution of bridge structure service performance based on a graph network. The technical solution is as follows:

[0008] On the one hand, a method for predicting the spatiotemporal evolution of service performance of a bridge structure based on a graph network is provided. The method is implemented by a device for predicting the spatiotemporal evolution of service performance of a bridge structure. The method comprises:

[0009] S1. Construct a bridge graph model.

[0010] S2. Based on the bridge graph model and spatiotemporal attention graph convolutional neural network, a basic bridge structure service performance evolution prediction model is constructed.

[0011] S3. For the basic bridge structure service performance evolution prediction model, a parameterized graph learning module that adaptively updates the spatiotemporal dependencies of bridge structure components is embedded to obtain the embedded bridge structure service performance evolution prediction model.

[0012] S4. Obtain a sample data set of the service performance evolution time series of the structural components of the bridge to be predicted, train the embedded bridge structure service performance evolution prediction model, obtain the optimal bridge graph model and the network parameters of the spatiotemporal attention graph convolutional neural network, and obtain a trained bridge structure service performance evolution prediction model.

[0013] S5. Obtain the service performance evolution time series data of the structural components of the bridge to be predicted, input it into the trained bridge structure service performance evolution prediction model, and obtain the spatiotemporal evolution prediction result of the bridge structure service performance.

[0014] Optionally, building a bridge graph model in S1 includes:

[0015] S11. Based on the existing theories, simulate the evolution of service performance of important structural components of bridges, construct a finite element model of bridge structure, conduct static analysis on the bridge structure with service performance evolution, obtain the spatial dependency relationship between structural components in the bridge, map the spatial dependency relationship into the topological relationship of the graph neural network, and construct the bridge graph structure according to the topological relationship.

[0016] The bridge graph structure is represented as , node set Nodes in represents the structural components of the bridge. Indicates the number of nodes; edge set The edges in represent the connection relationship between the bridge structure components; Represents the adjacency matrix.

[0017] S12. Parameterize the material properties of the bridge structure components in the bridge diagram structure to obtain a bridge diagram model.

[0018] Optionally, the basic bridge structure service performance evolution prediction model is constructed based on the bridge graph model and the spatiotemporal attention graph convolutional neural network in S2, including:

[0019] S21. Obtain a sample data set of time series of service performance evolution of bridge structure components.

[0020] S22. Construct a spatiotemporal attention graph convolutional neural network; wherein the spatiotemporal attention graph convolutional neural network includes: a spatiotemporal attention module and a spatiotemporal graph convolution module.

[0021] The spatiotemporal attention module includes: a temporal attention layer and a spatial attention layer. The temporal attention layer is used to extract the temporal attention matrix in the data, and the spatial attention layer is used to extract the spatial attention matrix in the data.

[0022] The spatiotemporal graph convolution module includes: temporal convolution layer and spatial convolution layer. The temporal convolution layer is used to identify the temporal dependency of adjacent time points, and the spatial convolution layer is used to capture the spatial dependency of the neighborhood.

[0023] S23. Based on the time series sample dataset of bridge structure component service performance evolution, the network parameters of the spatiotemporal attention graph convolutional neural network are trained to obtain a basic bridge structure service performance evolution prediction model.

[0024] Optionally, the process of acquiring the temporal attention matrix in S22 includes:

[0025] According to the following formula (1), the time attention matrix is ​​obtained :

[0026] (1)

[0027] The temporal attention matrix Normalized by the softmax function, the temporal attention matrix output by the temporal attention layer is obtained , as shown in the following formula (2):

[0028] (2)

[0029] In the formula, , , , , represents the learnable parameters, represents the activation function, Represents the temporal data information of the r-th spatiotemporal block of the model input at time step h, represents the transpose of a matrix, Represents the temporal attention matrix The elements in The value of represents the temporal attention matrix Medium time and The degree of dependence between Represents the length of the time dimension of the r-th layer of space-time block.

[0030] Optionally, the process of acquiring the spatial attention matrix in S22 includes:

[0031] The spatial attention matrix is ​​obtained according to the following formula (3): :

[0032] (3)

[0033] The spatial attention matrix The spatial attention matrix output by the spatial attention layer is obtained by normalizing with the softmax function. , as shown in the following formula (4):

[0034] (4)

[0035] In the formula, , , , , represents the learnable parameters, represents the transpose of a matrix, represents the activation function, Represents the temporal data information of the r-th spatiotemporal block of the model input at time step h, Represents the spatial attention matrix The elements in The value of represents the spatial attention matrix Middle position and The degree of dependence between them indicates that Represents the size of the spatial dimension of the r-th layer of spatiotemporal blocks.

[0036] Optionally, a parameterized graph learning module in S3 for adaptively updating spatiotemporal dependencies of bridge structural components is used to extract bidirectional dependencies between nodes of structural components in the bridge.

[0037] Extract bidirectional dependencies between nodes of structural components in bridges, including:

[0038] The affinity matrix is ​​generated by the following formula (5): :

[0039] (5)

[0040] In the formula, , represents the learnable parameters, represents the transpose of a matrix, Weights used to generate diagonal positions.

[0041] Obtaining the old affinity matrix through the parameterized graph learning module , the old affinity matrix is ​​transformed into and the resulting affinity matrix To combine:

[0042] (6)

[0043] In the formula, represents the adaptive score matrix obtained by the adaptive aggregation module, which is used to determine the combination weight between the newly learned new and old affinity matrices. represents a nonlinear activation function, Represents one or more 1×1 convolutional layers.

[0044] The old affinity matrix and affinity matrix Fusion to generate a new affinity matrix :

[0045] (7)

[0046] In the formula, Represents element-wise multiplication.

[0047] According to the new affinity matrix , the sparse affinity matrix is ​​obtained by the following formula (8): :

[0048] (8)

[0049] In the formula, the matrix Elements , represents the rows of the matrix, represents the columns of the matrix, Indicates the threshold value.

[0050] The sparse affinity matrix Normalize and get the affinity matrix :

[0051] (9)

[0052] In the formula, the matrix Elements .

[0053] Optionally, obtaining a sample data set of service performance evolution time series of structural components of a bridge to be predicted in S4, training an embedded bridge structure service performance evolution prediction model, obtaining an optimal bridge graph model and network parameters of a spatiotemporal attention graph convolutional neural network, and obtaining a trained bridge structure service performance evolution prediction model includes:

[0054] S41. Obtain a sample data set of the time series of the service performance evolution of bridge structure components, and perform preprocessing to obtain a preprocessed sample data set. According to the preprocessed sample data set, train and optimize the final bridge structure service performance evolution prediction model to obtain a trained final bridge structure service performance evolution prediction model.

[0055] S42, based on step S44, the optimal affinity matrix is ​​obtained, and the final bridge structure service performance evolution prediction model is trained based on the optimal affinity matrix to obtain the network parameters of the optimized bridge structure service performance evolution prediction model. .

[0056] S43, fixed optimized network parameters of the bridge structure service performance evolution prediction model , train the parameterized graph learning module to learn and update the spatial dependencies between nodes.

[0057] S44, updating the optimal affinity matrix according to the output result of the trained parameterized graph learning module;

[0058] S45, judging whether the preset number of iterations has been reached, if so, outputting the optimal model configuration of the final bridge structure service performance evolution prediction model; if not, going to step S41.

[0059] On the other hand, a device for predicting the spatiotemporal evolution of service performance of a bridge structure based on a graph network is provided. The device is applied to a method for predicting the spatiotemporal evolution of service performance of a bridge structure based on a graph network. The device comprises:

[0060] The bridge graph model building module is used to build the bridge graph model.

[0061] The prediction model building module is used to build a basic bridge structure service performance evolution prediction model based on the bridge graph model and spatiotemporal attention graph convolutional neural network.

[0062] The embedding module is used to embed the basic bridge structure service performance evolution prediction model into a parameterized graph learning module that adaptively updates the spatiotemporal dependencies of bridge structure components to obtain the embedded bridge structure service performance evolution prediction model.

[0063] The training module is used to obtain a sample data set of the service performance evolution time series of the structural components of the bridge to be predicted, train the embedded bridge structure service performance evolution prediction model, obtain the optimal bridge graph model and the network parameters of the spatiotemporal attention graph convolutional neural network, and obtain a trained bridge structure service performance evolution prediction model.

[0064] The output module obtains the time series data of the service performance evolution of the structural components of the bridge to be predicted, and inputs it into the trained bridge structure service performance evolution prediction model to obtain the spatiotemporal evolution prediction results of the bridge structure service performance.

[0065] On the other hand, a device for predicting the spatiotemporal evolution of the service performance of a bridge structure is provided, and the device for predicting the spatiotemporal evolution of the service performance of a bridge structure comprises: a processor; and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, any one of the above-mentioned graph network-based methods for predicting the spatiotemporal evolution of the service performance of a bridge structure is implemented.

[0066] On the other hand, a computer-readable storage medium is provided, in which at least one instruction is stored. The at least one instruction is loaded and executed by a processor to implement any one of the above-mentioned graph network-based methods for predicting the spatiotemporal evolution of service performance of bridge structures.

[0067] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0068] In this invention, a graph network-based bridge structure service performance evolution prediction method is proposed, which aims to effectively capture the spatiotemporal dependencies between structural components of the same type and between structural components of different types in the bridge, and improve the accuracy of the spatiotemporal evolution prediction model of the bridge structure service performance. The technical effects are as follows:

[0069] 1. Improve prediction accuracy: Different types of structural components (such as main beams, piers, and cables) in bridges are interrelated in space and load, and the same types of structural components are interrelated in space and mechanical properties. Traditional prediction models mainly focus on the analysis of the time dimension, ignoring the spatial dependency between bridge structural components. The present invention constructs a bridge graph model to accurately characterize the spatial dependency between bridge structural components. First, the spatial dependency between structural components in the bridge is mapped to the topological relationship of the graph neural network, and the connection relationship between components is characterized by the edges in the graph structure. Secondly, the material properties of the bridge structural components are parameterized to characterize the spatial dependency between structural components of the same type. The graph model can simultaneously reflect the spatial dependency between different types of structural components and between structural components of the same type in the bridge. Based on the bridge graph model and the spatiotemporal attention graph convolutional neural network, a bridge structure service performance evolution prediction model is constructed. The model takes the time series data of the service performance of the bridge structure as input, and uses the spatiotemporal attention module and the graph convolution module to capture the spatiotemporal dependency between the structural components implicit in the data. Therefore, the prediction results obtained by the model are more in line with the actual state.

[0070] 2. Adaptability and model optimization: As the service time increases, the spatiotemporal dependencies between bridge structure components will change. Based on the service performance evolution prediction model, the present invention embeds a parameterized graph learning algorithm that adaptively updates the spatiotemporal dependencies between bridge structure components. The algorithm dynamically captures and updates the spatiotemporal dependencies between structural components during the service process, improves the bridge graph model in the graph convolutional neural network, and adapts to the dynamically changing bridge structure characteristics during the service process. This dynamic adjustment mechanism enables the model to adapt to the changes in the spatiotemporal dependencies of structural components caused by factors such as the environment and load during the long-term service of the bridge. Combining the spatiotemporal attention graph convolutional neural network and the parameterized graph learning algorithm, a prediction model for the evolution of the service performance of the bridge structure is constructed. The prediction model training, parameterized graph learning and affinity matrix update in the model together constitute a single optimization loop. By setting multiple iterative loops to continuously adjust and improve the model parameters, the optimal model configuration for predicting the service performance of the bridge structure can be obtained, further improving the model performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0072] Figure 1 It is a flow chart of a method for predicting the spatiotemporal evolution of service performance of a bridge structure based on a graph network provided by an embodiment of the present invention;

[0073] Figure 2 It is an architecture diagram of a method for predicting the spatiotemporal evolution of service performance of a bridge structure based on a graph network provided in an embodiment of the present invention;

[0074] Figure 3 is a schematic diagram of a bridge graph model provided by an embodiment of the present invention;

[0075] Figure 4 It is an architecture diagram of a bridge structure service performance prediction model based on a spatiotemporal attention graph convolutional neural network provided by an embodiment of the present invention;

[0076] Figure 5 It is an architecture diagram of a prediction model embedded with a parameterized graph learning algorithm provided by an embodiment of the present invention;

[0077] Figure 6 It is a diagram of the model training optimization process of the graph network-based bridge structure service performance spatiotemporal evolution prediction method provided in an embodiment of the present invention;

[0078] Figure 7 It is a block diagram of a device for predicting the spatiotemporal evolution of service performance of a bridge structure based on a graph network provided by an embodiment of the present invention;

[0079] Figure 8 It is a structural schematic diagram of a device for predicting the spatiotemporal evolution of service performance of a bridge structure provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0080] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0081] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0082] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0083] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0084] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0085] The embodiment of the present invention provides a method for predicting the spatiotemporal evolution of service performance of a bridge structure based on a graph network. The method can be implemented by a device for predicting the spatiotemporal evolution of service performance of a bridge structure. The device can be a terminal or a server. Figure 1 , 2 The flowchart of the method for predicting the spatiotemporal evolution of service performance of bridge structures based on graph networks is shown. The processing flow of the method may include the following steps:

[0086] S1. Construct a bridge graph model.

[0087] Optionally, the above step S1 may include the following steps S11-S12:

[0088] S11. Based on the existing theories, simulate the evolution of service performance of important structural components of bridges, construct a finite element model of bridge structure, conduct static analysis on the bridge structure with service performance evolution, obtain the spatial dependency relationship between structural components in the bridge, map the spatial dependency relationship into the topological relationship of the graph neural network, and construct the bridge graph structure according to the topological relationship.

[0089] The bridge graph structure is represented as , node set Nodes in represents the structural member of the bridge. Indicates the number of nodes; edge set The edges in represent the connection relationship between the bridge structure components; Represents the adjacency matrix.

[0090] S12. Parameterize the material properties of the bridge structure components in the bridge graph structure by means of one-hot encoding, etc., assign eigenvalues ​​to nodes, and then obtain a bridge graph model.

[0091] In a feasible implementation, different types of structural components (such as main beams, piers, and cables) in the bridge are interrelated in space and load, and the same type of structural components are interrelated in space and mechanical properties.

[0092] Specifically, a bridge graph model is constructed to characterize the spatial dependency relationship between bridge structural components. First, the spatial dependency relationship between structural components in the bridge is mapped to the topological relationship of the graph neural network, and the connection relationship between components is represented by the edges in the graph structure. and edge sets Composition, expressed as Among them, nodes represent bridge structure components, and edges represent the connection relationship between components. Representation diagram The adjacency matrix of and If there is an edge between ,otherwise Secondly, the material properties of bridge structural components are parameterized to characterize the spatial dependence between structural components of the same type. and The material properties of ,otherwise The schematic diagram of the bridge model is shown in the attached Figure 3 As shown in the figure, if the node and If there is a service performance evolution space dependency relationship between them, the node and Furthermore, dotted lines are used to represent the intra-class dependencies between components of the same type, and solid lines are used to represent the inter-class dependencies between components of different types.

[0093] S2. Based on the bridge graph model and spatiotemporal attention graph convolutional neural network, a basic bridge structure service performance evolution prediction model is constructed.

[0094] In a feasible implementation, based on the bridge graph model and the spatiotemporal attention graph convolutional neural network, the spatiotemporal attention module and the graph convolution module are used to capture the spatiotemporal dependencies between the structural components implicit in the data, and a bridge structure service performance evolution prediction model is constructed. The overall architecture of the model is shown in the attached figure. Figure 4 As shown in the figure. The model takes the time series data of bridge structure service performance as input, uses the spatiotemporal attention module and the graph convolution module to capture the spatiotemporal dependencies between structural components implicit in the data, and outputs the prediction results of service performance evolution in the future. The spatiotemporal attention module consists of a temporal attention layer and a spatial attention layer, which can extract the temporal attention matrix and the spatial attention matrix in the data. The information processed by the attention module is input into the spatiotemporal graph convolution module, which combines the graph convolution in the spatial dimension and the standard convolution in the temporal dimension. The graph convolution in the spatial dimension is used to capture the spatial dependencies of the neighborhood, and the standard convolution in the temporal dimension is used to identify the temporal dependencies of adjacent time points.

[0095] Optionally, the above step S2 may include the following steps S21-S23:

[0096] S21. Obtain a sample data set of time series of service performance evolution of bridge structure components.

[0097] S22. Construct a spatiotemporal attention graph convolutional neural network; wherein the spatiotemporal attention graph convolutional neural network includes: a spatiotemporal attention module and a spatiotemporal graph convolution module.

[0098] The spatiotemporal attention module includes: a temporal attention layer and a spatial attention layer. The temporal attention layer is used to extract the temporal attention matrix in the data, and the spatial attention layer is used to extract the spatial attention matrix in the data.

[0099] The spatiotemporal graph convolution module includes: temporal convolution layer and spatial convolution layer. The temporal convolution layer is used to identify the temporal dependency of adjacent time points, and the spatial convolution layer is used to capture the spatial dependency of the neighborhood.

[0100] Optionally, the process of acquiring the temporal attention matrix in S22 includes:

[0101] According to the following formula (1), the time attention matrix is ​​obtained :

[0102] (1)

[0103] The temporal attention matrix Normalized by the softmax function, the temporal attention matrix output by the temporal attention layer is obtained , as shown in the following formula (2):

[0104] (2)

[0105] In the formula, , , , , represents the learnable parameters, represents the activation function, Represents the temporal data information of the r-th spatiotemporal block of the model input at time step h, represents the transpose of a matrix, Represents the temporal attention matrix The elements in The value of represents the temporal attention matrix Medium time and The degree of dependence between Represents the length of the time dimension of the r-th layer of space-time block.

[0106] Optionally, the process of acquiring the spatial attention matrix in S22 includes:

[0107] The spatial attention matrix is ​​obtained according to the following formula (3): :

[0108] (3)

[0109] The spatial attention matrix The spatial attention matrix output by the spatial attention layer is obtained by normalizing with the softmax function. , as shown in the following formula (4):

[0110] (4)

[0111] In the formula, , , , , represents the learnable parameters, represents the transpose of a matrix, represents the activation function, Represents the temporal data information of the r-th spatiotemporal block of the model input at time step h, Represents the spatial attention matrix The elements in The value of represents the spatial attention matrix Middle position and The degree of dependence between them indicates that Represents the size of the spatial dimension of the r-th layer of spatiotemporal blocks.

[0112] Furthermore, the information processed by the attention module is input into the spatiotemporal graph convolution module, which combines graph convolution in the spatial dimension and standard convolution in the temporal dimension. The graph convolution in the spatial dimension is used to capture the spatial dependencies of the neighborhood, and the standard convolution in the temporal dimension is used to identify the temporal dependencies of adjacent time points.

[0113] S23. Based on the time series sample dataset of bridge structure component service performance evolution, the network parameters of the spatiotemporal attention graph convolutional neural network are trained to obtain a prediction model for the evolution of bridge structure service performance.

[0114] S3. For the basic bridge structure service performance evolution prediction model, a parameterized graph learning module that adaptively updates the spatiotemporal dependencies of bridge structure components is embedded to obtain the embedded bridge structure service performance evolution prediction model.

[0115] In a feasible implementation, based on the service performance evolution prediction model, a parameterized graph learning algorithm for adaptively updating the spatiotemporal dependencies of bridge structural components is embedded. The overall architecture of the model is shown in the attached figure. Figure 5 As shown in the figure. After the parameterized graph learning module is embedded, the entire model is divided into a prediction module and a graph learning module. Through input data training, the optimal graph model output by the graph learning module and the optimal network weight parameters of the prediction module are obtained. The algorithm dynamically captures and updates the spatiotemporal dependencies between structural components during service, improves the bridge graph model in the graph convolutional neural network, adapts to the dynamically changing characteristics of bridge structures during service, and further improves the accuracy and robustness of the spatiotemporal evolution prediction model of bridge structure service performance.

[0116] Optionally, a parameterized graph learning module in S3 for adaptively updating spatiotemporal dependencies of bridge structural components is used to extract bidirectional dependencies between nodes of structural components in the bridge.

[0117] Extract bidirectional dependencies between nodes of structural components in bridges, which can include:

[0118] The rough affinity matrix is ​​generated by the following formula (5): :

[0119] (5)

[0120] In the formula, , represents the learnable parameters, represents the transpose of a matrix, Weights used to generate diagonal positions.

[0121] Use the adaptive aggregation module to combine the old spatial dependencies with the new spatial dependencies:

[0122] (6)

[0123] In the formula, represents the adaptive score matrix obtained by the adaptive aggregation module, which is used to determine the combination weight between the newly learned new and old affinity matrices. represents a nonlinear activation function, Represents one or more 1×1 convolutional layers.

[0124] Use the adaptive aggregation module to merge the old and new affinity matrices to generate a new affinity matrix :

[0125] (7)

[0126] In the formula, Represents element-wise multiplication.

[0127] In order to enhance the sparsity of the generated matrix, the sparse affinity matrix is ​​obtained by the following formula (8): :

[0128] (8)

[0129] In the formula, the matrix Elements , represents the rows of the matrix, represents the columns of the matrix, Represents the threshold, used to filter out weak relationship.

[0130] The sparse affinity matrix Normalize and get the affinity matrix :

[0131] (9)

[0132] In the formula, the matrix Elements .

[0133] The algorithm can dynamically capture and update the spatiotemporal dependencies of bridge structure components during service, thereby improving the initial bridge graph model defined in the graph convolutional neural network and adapting to the dynamically changing structural characteristics of the bridge during service.

[0134] Furthermore, a prediction model for the evolution of the service performance of bridge structures is constructed by combining the spatiotemporal attention graph convolutional neural network and the parameterized graph learning algorithm. The training and optimization process of the model is divided into three steps. First, the prediction model is trained and optimized based on the preprocessed data set to ensure that the model can effectively capture the spatiotemporal dependencies between structural components implicit in the data and output accurate prediction results. Second, the network parameters of the current prediction model are fixed, and the parameterized graph learning module is trained to accurately learn and update the spatial dependencies between nodes through an adaptive algorithm. Third, the current optimal affinity matrix is ​​updated based on the results output by the parameterized graph learning module to ensure that the model adapts to the dynamically changing characteristics of the bridge structure during service. The above three steps together constitute a single optimization loop of the model. In order to further improve the performance of the model, multiple iterative cycles are performed to continuously adjust and improve the model parameters until the optimal model configuration for predicting the service performance of the bridge structure is found. The training and optimization process of the model is shown in the attached figure. Figure 6 As shown, the following steps S41-S45 may be specifically included:

[0135] S41. Obtain a sample data set of the time series of the service performance evolution of bridge structure components, and perform preprocessing to obtain a preprocessed sample data set. According to the preprocessed sample data set, train and optimize the final bridge structure service performance evolution prediction model to obtain a trained final bridge structure service performance evolution prediction model.

[0136] In a feasible implementation, the prediction model is trained and optimized based on the preprocessed data set to ensure that the model can effectively capture the temporal and spatial dependencies between structural components implicit in the data and output accurate prediction results.

[0137] S42, obtaining the optimal affinity matrix through the parameterized graph learning module, optimizing the trained final bridge structure service performance evolution prediction model according to the optimal affinity matrix, and obtaining the network parameters of the optimized bridge structure service performance evolution prediction model .

[0138] In a feasible implementation, the optimal affinity matrix is ​​obtained from the affinity matrix set. , which is used to optimize the prediction model and obtain the network parameters As shown in the following formula (10), select the predicted value With the true value Minimizing the L1 loss between is used as the training objective of the prediction model:

[0139] (10)

[0140] Where: is the true value; is the predicted value.

[0141] S43, fixed optimized network parameters of the bridge structure service performance evolution prediction model , train the parameterized graph learning module to learn and update the spatial dependencies between nodes.

[0142] In one feasible implementation, the network parameters of the current prediction model are fixed as The estimated value of is used to train the parameterized graph learning module, and the spatial dependencies between nodes are accurately learned and updated through an adaptive algorithm. To ensure the sparsity of the generated dependencies, the loss function is given by To optimize:

[0143] (11)

[0144] In the formula, is a hyperparameter used to control The sparsity rate of exist and Set the value at position 1 and set the value at other positions to 0. Represents the dimensions of the affinity matrix.

[0145] S44. Update the optimal affinity matrix according to the output results of the trained parameterized graph learning module.

[0146] In one feasible implementation, the current optimal affinity matrix is ​​updated based on the output of the parameterized graph learning module. , to ensure that the model adapts to the dynamic changes in bridge structure characteristics during service. First, the generated adjacency matrix The income affinity matrix set A. Then, the graphs in the set and the validation set are input into the prediction model respectively, and the loss function , calculate the loss of all sub-graphs according to formula (12):

[0147] (12)

[0148] In the formula, is the prediction function, represents the kth matrix in the affinity matrix set, is a learnable parameter in the prediction model.

[0149] make is the vector of all prediction losses, is the maximum loss value, the weight vector As shown in the following formula (13):

[0150] (13)

[0151] represents the weight vector.

[0152] The current affinity matrix can be obtained by weighting all subgraphs:

[0153] (14)

[0154] S45, determine whether the preset number of iterations has been reached. If so, output the optimal model configuration of the final bridge structure service performance evolution prediction model; if not, go to step S41. By setting multiple iterations to continuously adjust and improve the model parameters, the optimal model configuration for predicting the service performance of the bridge structure is obtained to improve the model performance. The training process of the entire model is shown in the attached figure. Figure 6 shown.

[0155] S5. Obtain the service performance evolution time series data of the structural components of the bridge to be predicted, input it into the trained bridge structure service performance evolution prediction model, and obtain the spatiotemporal evolution prediction result of the bridge structure service performance.

[0156] The present invention proposes a bridge structure service performance evolution prediction method based on graph network, which includes the following two parts: a bridge structure service performance evolution prediction model based on spatiotemporal attention graph neural network, and a parameterized graph learning algorithm that adaptively updates the spatiotemporal dependency of bridge structure components. The technical problems to be solved include: describing the topological relationship between bridge structure components through graph structure, constructing a bridge graph model, and accurately characterizing the spatial dependency between components; constructing a bridge structure service performance evolution prediction model based on the bridge graph model and spatiotemporal attention graph convolutional neural network; and embedding a parameterized graph learning algorithm that adaptively updates the spatiotemporal dependency of bridge structure components on the basis of the service performance evolution prediction model. The algorithm dynamically captures and updates the spatiotemporal dependency of bridge structure components during service, improves the initial bridge graph model defined in the graph convolutional neural network, adapts to the dynamically changing bridge structure characteristics during service, and further improves the prediction performance of the model.

[0157] In the embodiment of the present invention, a graph network-based bridge structure service performance evolution prediction method is proposed, which aims to effectively capture the spatiotemporal dependencies between structural components of the same type and between structural components of different types in the bridge, and improve the accuracy of the spatiotemporal evolution prediction model of the bridge structure service performance. The technical effects are as follows:

[0158] 1. Improve prediction accuracy: Different types of structural components (such as main beams, piers, and cables) in bridges are interrelated in space and load, and the same types of structural components are interrelated in space and mechanical properties. Traditional prediction models mainly focus on the analysis of the time dimension, ignoring the spatial dependency between bridge structural components. The present invention constructs a bridge graph model to accurately characterize the spatial dependency between bridge structural components. First, the spatial dependency between structural components in the bridge is mapped to the topological relationship of the graph neural network, and the connection relationship between components is characterized by the edges in the graph structure. Secondly, the material properties of the bridge structural components are parameterized to characterize the spatial dependency between structural components of the same type. The graph model can simultaneously reflect the spatial dependency between different types of structural components and between structural components of the same type in the bridge. Based on the bridge graph model and the spatiotemporal attention graph convolutional neural network, a bridge structure service performance evolution prediction model is constructed. The model takes the time series data of the service performance of the bridge structure as input, and uses the spatiotemporal attention module and the graph convolution module to capture the spatiotemporal dependency between the structural components implicit in the data. Therefore, the prediction results obtained by the model are more in line with the actual state.

[0159] 2. Adaptability and model optimization: As the service time increases, the spatiotemporal dependencies between bridge structure components will change. Based on the service performance evolution prediction model, the present invention embeds a parameterized graph learning algorithm that adaptively updates the spatiotemporal dependencies between bridge structure components. The algorithm dynamically captures and updates the spatiotemporal dependencies between structural components during the service process, improves the bridge graph model in the graph convolutional neural network, and adapts to the dynamically changing bridge structure characteristics during the service process. This dynamic adjustment mechanism enables the model to adapt to the changes in the spatiotemporal dependencies of structural components caused by factors such as the environment and load during the long-term service of the bridge. Combining the spatiotemporal attention graph convolutional neural network and the parameterized graph learning algorithm, a prediction model for the evolution of the service performance of the bridge structure is constructed. The prediction model training, parameterized graph learning and affinity matrix update in the model together constitute a single optimization loop. By setting multiple iterative loops to continuously adjust and improve the model parameters, the optimal model configuration for predicting the service performance of the bridge structure can be obtained, further improving the model performance.

[0160] Figure 7 The present invention is a block diagram of a device for predicting the spatiotemporal evolution of service performance of a bridge structure based on a graph network according to an exemplary embodiment. The device is used in a method for predicting the spatiotemporal evolution of service performance of a bridge structure based on a graph network. Figure 7 The device includes a bridge graph model building module 310, a prediction model building module 320, an embedding module 330, a training module 340 and an output module 350. Among them:

[0161] The bridge graph model building module 310 is used to build a bridge graph model.

[0162] The prediction model building module 320 is used to build a basic bridge structure service performance evolution prediction model based on the bridge graph model and the spatiotemporal attention graph convolutional neural network.

[0163] The embedding module 330 is used to embed a parameterized graph learning module that adaptively updates the spatiotemporal dependency of bridge structure components into the basic bridge structure service performance evolution prediction model to obtain the embedded bridge structure service performance evolution prediction model.

[0164] The training module 340 is used to obtain a sample data set of the service performance evolution time series of the structural components of the bridge to be predicted, train the embedded bridge structure service performance evolution prediction model, obtain the optimal bridge graph model and network parameters of the spatiotemporal attention graph convolutional neural network, and obtain a trained bridge structure service performance evolution prediction model.

[0165] The output module 350 obtains the service performance evolution time series data of the structural components of the bridge to be predicted, and inputs it into the trained bridge structure service performance evolution prediction model to obtain the spatiotemporal evolution prediction result of the bridge structure service performance.

[0166] In the embodiment of the present invention, a graph network-based bridge structure service performance evolution prediction method is proposed, which aims to effectively capture the spatiotemporal dependencies between structural components of the same type and between structural components of different types in the bridge, and improve the accuracy of the spatiotemporal evolution prediction model of the bridge structure service performance. The technical effects are as follows:

[0167] 1. Improve prediction accuracy: Different types of structural components (such as main beams, piers, and cables) in bridges are interrelated in space and load, and the same types of structural components are interrelated in space and mechanical properties. Traditional prediction models mainly focus on the analysis of the time dimension, ignoring the spatial dependency between bridge structural components. The present invention constructs a bridge graph model to accurately characterize the spatial dependency between bridge structural components. First, the spatial dependency between structural components in the bridge is mapped to the topological relationship of the graph neural network, and the connection relationship between components is characterized by the edges in the graph structure. Secondly, the material properties of the bridge structural components are parameterized to characterize the spatial dependency between structural components of the same type. The graph model can simultaneously reflect the spatial dependency between different types of structural components and between structural components of the same type in the bridge. Based on the bridge graph model and the spatiotemporal attention graph convolutional neural network, a bridge structure service performance evolution prediction model is constructed. The model takes the time series data of the service performance of the bridge structure as input, and uses the spatiotemporal attention module and the graph convolution module to capture the spatiotemporal dependency between the structural components implicit in the data. Therefore, the prediction results obtained by the model are more in line with the actual state.

[0168] 2. Adaptability and model optimization: As the service time increases, the spatiotemporal dependencies between bridge structure components will change. Based on the service performance evolution prediction model, the present invention embeds a parameterized graph learning algorithm that adaptively updates the spatiotemporal dependencies between bridge structure components. The algorithm dynamically captures and updates the spatiotemporal dependencies between structural components during the service process, improves the bridge graph model in the graph convolutional neural network, and adapts to the dynamically changing bridge structure characteristics during the service process. This dynamic adjustment mechanism enables the model to adapt to the changes in the spatiotemporal dependencies of structural components caused by factors such as the environment and load during the long-term service of the bridge. Combining the spatiotemporal attention graph convolutional neural network and the parameterized graph learning algorithm, a prediction model for the evolution of the service performance of the bridge structure is constructed. The prediction model training, parameterized graph learning and affinity matrix update in the model together constitute a single optimization loop. By setting multiple iterative loops to continuously adjust and improve the model parameters, the optimal model configuration for predicting the service performance of the bridge structure can be obtained, further improving the model performance.

[0169] Figure 8 is a schematic diagram of a structure of a device for predicting the spatiotemporal evolution of service performance of a bridge structure provided by an embodiment of the present invention, such as Figure 8As shown, the bridge structure service performance spatiotemporal evolution prediction device may include the above Figure 7 The bridge structure service performance spatiotemporal evolution prediction device based on graph network is shown. Optionally, the bridge structure service performance spatiotemporal evolution prediction device 410 may include a first processor 2001.

[0170] Optionally, the bridge structure service performance spatiotemporal evolution prediction device 410 may further include a memory 2002 and a transceiver 2003 .

[0171] The first processor 2001, the memory 2002 and the transceiver 2003 may be connected via a communication bus.

[0172] Combine the following Figure 8 The components of the bridge structure service performance spatiotemporal evolution prediction device 410 are specifically introduced:

[0173] The first processor 2001 is the control center of the bridge structure service performance spatiotemporal evolution prediction device 410, which can be a processor or a general term for multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPUs), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present invention, such as one or more microprocessors (digital signal processors, DSPs), or one or more field programmable gate arrays (field programmable gate arrays, FPGAs).

[0174] Optionally, the first processor 2001 can perform various functions of the bridge structure service performance spatiotemporal evolution prediction device 410 by running or executing a software program stored in the memory 2002 and calling data stored in the memory 2002 .

[0175] In a specific implementation, as an embodiment, the first processor 2001 may include one or more CPUs, such as Figure 8 CPU0 and CPU1 are shown in FIG.

[0176] In a specific implementation, as an embodiment, the bridge structure service performance spatiotemporal evolution prediction device 410 may also include multiple processors, such as Figure 8The first processor 2001 and the second processor 2004 are shown in FIG. Each of these processors may be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0177] The memory 2002 is used to store the software program for executing the solution of the present invention, and is controlled to be executed by the first processor 2001. The specific implementation method can refer to the above method embodiment, which will not be repeated here.

[0178] Optionally, the memory 2002 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 may be integrated with the first processor 2001, or may exist independently, and access the first processor 2001 through the interface circuit ( Figure 8 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0179] The transceiver 2003 is used to communicate with a network device or a terminal device.

[0180] Optionally, the transceiver 2003 may include a receiver and a transmitter ( Figure 8 The receiver is used to implement a receiving function, and the transmitter is used to implement a sending function.

[0181] Optionally, the transceiver 2003 may be integrated with the first processor 2001, or may exist independently, and may be connected to the first processor 2001 through the interface circuit ( Figure 8 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0182] It should be noted that Figure 8 The structure of the bridge structure service performance spatiotemporal evolution prediction device 410 shown in the figure does not constitute a limitation on the router, and the actual knowledge structure identification device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0183] In addition, the technical effects of the bridge structure service performance spatiotemporal evolution prediction device 410 can refer to the technical effects of the bridge structure service performance spatiotemporal evolution prediction method based on graph network described in the above method embodiment, and will not be repeated here.

[0184] It should be understood that the first processor 2001 in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0185] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0186] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0187] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0188] In the present invention, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0189] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0190] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0191] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0192] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0193] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0194] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0195] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0196] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for predicting the spatiotemporal evolution of service performance of bridge structures based on graph networks, characterized in that: The method comprises: S1, construct a bridge graph model; S2. Based on the bridge graph model and the spatiotemporal attention graph convolutional neural network, a basic bridge structure service performance evolution prediction model is constructed; S3, embedding a parameterized graph learning module that adaptively updates the spatiotemporal dependency relationship of bridge structure components into the basic bridge structure service performance evolution prediction model to obtain an embedded bridge structure service performance evolution prediction model; S4, obtaining a sample data set of service performance evolution time series of structural components of the bridge to be predicted, training the embedded bridge structure service performance evolution prediction model, obtaining the optimal bridge graph model and network parameters of the spatiotemporal attention graph convolutional neural network, and obtaining a trained bridge structure service performance evolution prediction model; S5. Obtaining the service performance evolution time series data of the structural components of the bridge to be predicted, inputting the data into the trained bridge structure service performance evolution prediction model, and obtaining the spatiotemporal evolution prediction result of the bridge structure service performance.

2. The method for predicting the spatiotemporal evolution of service performance of bridge structures based on graph networks according to claim 1 is characterized in that: The construction of the bridge graph model in S1 includes: S11. Based on the existing theory, simulate the service performance evolution of important structural components of the bridge, construct a finite element model of the bridge structure, perform static analysis on the bridge structure with service performance evolution, obtain the spatial dependency relationship between the structural components in the bridge, map the spatial dependency relationship into the topological relationship of the graph neural network, and construct the bridge graph structure according to the topological relationship; The bridge graph structure is represented as , node set Nodes in represents the structural member of the bridge. Indicates the number of nodes; edge set The edges in represent the connection relationship between the bridge structure components; represents the adjacency matrix; S12, parameterizing the material properties of the bridge structure components in the bridge diagram structure, thereby obtaining a bridge diagram model.

3. The method for predicting the spatiotemporal evolution of service performance of bridge structures based on graph networks according to claim 1 is characterized in that: The S2 constructs a basic bridge structure service performance evolution prediction model based on the bridge graph model and the spatiotemporal attention graph convolutional neural network, including: S21. Obtain a sample data set of service performance evolution time series of bridge structural components; S22, constructing a spatiotemporal attention graph convolutional neural network; wherein the spatiotemporal attention graph convolutional neural network includes: a spatiotemporal attention module and a spatiotemporal graph convolution module; The spatiotemporal attention module includes: a temporal attention layer and a spatial attention layer, wherein the temporal attention layer is used to extract a temporal attention matrix from the data, and the spatial attention layer is used to extract a spatial attention matrix from the data; The spatiotemporal graph convolution module includes: a temporal convolution layer and a spatial convolution layer, wherein the temporal convolution layer is used to identify the temporal dependency of adjacent time points, and the spatial convolution layer is used to capture the spatial dependency of the neighborhood; S23. Based on the time series sample dataset of bridge structure component service performance evolution, the network parameters of the spatiotemporal attention graph convolutional neural network are trained to obtain a basic bridge structure service performance evolution prediction model.

4. The method for predicting the spatiotemporal evolution of service performance of bridge structures based on graph networks according to claim 3 is characterized in that: The process of acquiring the temporal attention matrix in S22 includes: According to the following formula (1), the time attention matrix is ​​obtained : （1） The temporal attention matrix Normalized by the softmax function, the temporal attention matrix output by the temporal attention layer is obtained , as shown in the following formula (2): （2） In the formula, , , , , represents the learnable parameters, represents the activation function, Represents the temporal data information of the r-th spatiotemporal block of the model input at time step h, represents the transpose of a matrix, Represents the temporal attention matrix The elements in The value of represents the temporal attention matrix Medium time and The degree of dependence between Represents the length of the time dimension of the r-th layer of space-time block.

5. The method for predicting the spatiotemporal evolution of service performance of bridge structures based on graph networks according to claim 3 is characterized in that: The process of acquiring the spatial attention matrix in S22 includes: The spatial attention matrix is ​​obtained according to the following formula (3): : （3） The spatial attention matrix The spatial attention matrix output by the spatial attention layer is obtained by normalizing with the softmax function. , as shown in the following formula (4): （4） In the formula, , , , , represents the learnable parameters, represents the transpose of a matrix, represents the activation function, Represents the temporal data information of the r-th spatiotemporal block of the model input at time step h, Represents the spatial attention matrix The elements in The value of represents the spatial attention matrix Middle position and The degree of dependence between them indicates that Represents the size of the spatial dimension of the r-th layer of spatiotemporal blocks.

6. The method for predicting the spatiotemporal evolution of service performance of bridge structures based on graph networks according to claim 1 is characterized in that: The parameterized graph learning module for adaptively updating the spatiotemporal dependency of bridge structural components in S3 is used to extract the bidirectional dependency between the nodes of the structural components in the bridge; The step of extracting the bidirectional dependency relationship between nodes of the structural components in the bridge includes: The affinity matrix is ​​generated by the following formula (5): : （5） In the formula, , represents the learnable parameters, represents the transpose of a matrix, The weights used to generate the diagonal positions; Obtaining the old affinity matrix through the parameterized graph learning module , the old affinity matrix is ​​transformed into and the resulting affinity matrix To combine: （6） In the formula, represents the adaptive score matrix obtained by the adaptive aggregation module, which is used to determine the combination weight between the newly learned new and old affinity matrices. represents a nonlinear activation function, Represents one or more 1×1 convolutional layers; The old affinity matrix and affinity matrix Fusion to generate a new affinity matrix : （7） In the formula, represents element-wise multiplication; According to the new affinity matrix , the sparse affinity matrix is ​​obtained by the following formula (8): : （8） In the formula, the matrix Elements , represents the rows of the matrix, represents the columns of the matrix, Indicates the threshold value; The sparse affinity matrix Normalize and get the affinity matrix : （9） In the formula, the matrix Elements .

7. The method for predicting the spatiotemporal evolution of service performance of bridge structures based on graph networks according to claim 1 is characterized in that: The step S4 of obtaining a sample data set of service performance evolution time series of structural components of the bridge to be predicted, training the embedded bridge structure service performance evolution prediction model, obtaining the optimal bridge graph model and network parameters of the spatiotemporal attention graph convolutional neural network, and obtaining a trained bridge structure service performance evolution prediction model includes: S41, obtaining a sample data set of the time series of service performance evolution of bridge structure components, and preprocessing it to obtain a preprocessed sample data set, and training and optimizing a final bridge structure service performance evolution prediction model according to the preprocessed sample data set to obtain a trained final bridge structure service performance evolution prediction model; S42, based on step S44, the optimal affinity matrix is ​​obtained, and the final bridge structure service performance evolution prediction model is trained based on the optimal affinity matrix to obtain the network parameters of the optimized bridge structure service performance evolution prediction model. ; S43, fixed optimized network parameters of the bridge structure service performance evolution prediction model , train the parameterized graph learning module to learn and update the spatial dependencies between nodes; S44, updating the optimal affinity matrix according to the output result of the trained parameterized graph learning module; S45, judging whether the preset number of iterations has been reached, if so, outputting the optimal model configuration of the final bridge structure service performance evolution prediction model; if not, going to step S41.

8. A graph network-based device for predicting the spatiotemporal evolution of service performance of a bridge structure, the graph network-based device for predicting the spatiotemporal evolution of service performance of a bridge structure is used to implement the graph network-based method for predicting the spatiotemporal evolution of service performance of a bridge structure as claimed in any one of claims 1 to 7, characterized in that: The device comprises: A bridge graph model building module, used to build a bridge graph model; A prediction model building module, used to build a basic bridge structure service performance evolution prediction model based on the bridge graph model and the spatiotemporal attention graph convolutional neural network; An embedding module is used to embed a parameterized graph learning module for adaptively updating the spatiotemporal dependency relationship of bridge structure components into the basic bridge structure service performance evolution prediction model to obtain an embedded bridge structure service performance evolution prediction model; A training module is used to obtain a sample data set of service performance evolution time series of structural components of a bridge to be predicted, train the embedded bridge structure service performance evolution prediction model, obtain the optimal bridge graph model and network parameters of the spatiotemporal attention graph convolutional neural network, and obtain a trained bridge structure service performance evolution prediction model; The output module obtains the service performance evolution time series data of the structural components of the bridge to be predicted, and inputs it into the trained bridge structure service performance evolution prediction model to obtain the spatiotemporal evolution prediction result of the bridge structure service performance.

9. A device for predicting the spatiotemporal evolution of service performance of a bridge structure, characterized in that: The bridge structure service performance spatiotemporal evolution prediction device comprises: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program codes, which can be called by a processor to execute the method according to any one of claims 1 to 7.

Images