Method for establishing bridge network sample library and simulating traffic demand based on GAN
The bridge network sample library was established through the GAN model, which solved the problem of unreasonable traffic planning of the bridge network after earthquake, and achieved the accuracy of traffic flow prediction and improved the urban seismic resilience.
Patent Information
- Application Number
- CN202510186981.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-20
- Publication Date
- 2025-07-18
AI Technical Summary
It is difficult for the existing technology to effectively establish a sample library of bridge networks and simulate traffic demands, resulting in unreasonable traffic planning after earthquakes, affecting rescue efficiency and post-disaster reconstruction speed.
A bridge network sample library is established using the GAN model, a virtual sample library is generated through subtopology network database and traffic data, and pre-quake traffic needs are simulated, and the network structure is optimized by combining Louvain algorithm and data enhancement technology.
The established bridge network sample library is consistent with the actual network characteristics, can accurately predict traffic flow changes, help formulate reasonable traffic diversion plans, and improve urban seismic resilience.
Smart Images
Figure CN120336561A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of bridge engineering, and specifically provides a method for establishing a bridge network sample library and simulating traffic demand based on GAN. Background Art
[0002] The bridge network is an important part of the urban lifeline system. Multiple earthquake disasters have shown that the bridge network often becomes the most vulnerable part during earthquakes, and traffic interruptions and the inability to transport resources in a timely manner often occur due to structural damage. However, the bridge network plays a crucial role in post-earthquake rescue. It is not only the main channel for transporting rescue supplies but also an important hub for evacuating affected people. Its operating status directly affects the rescue efficiency and the speed and effect of post-disaster reconstruction. In recent years, with the proposal of the concept of "resilient city", the seismic resilience of the bridge network has received increasing attention. The seismic resilience of the bridge network refers to the ability to maintain its function and quickly resume operation after an earthquake, and it can quickly recover and support the normal operation of the city under the impact of the earthquake disaster. The improvement of seismic resilience is not only a new requirement for bridge design and construction but also a key link for the city to cope with major disasters and ensure the restoration of social and economic order. Therefore, establishing a bridge network sample library and simulating traffic demand have important practical significance for building a resilient city and enhancing the city's disaster prevention and mitigation capabilities. This method will provide a scientific basis for post-disaster emergency planning, resource allocation, and infrastructure reconstruction, and ultimately help improve the comprehensive seismic capacity of the city. Summary of the Invention
[0003] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide a method for establishing a bridge network sample library and simulating traffic demand based on GAN, so as to provide sufficient training data for the seismic resilience assessment of the bridge network, and the simulated traffic demand can help the traffic planning department formulate a reasonable traffic diversion plan before an earthquake.
[0004] To achieve the above purpose, the following technical solutions are provided:
[0005] A method for establishing a bridge network sample library and simulating traffic demand based on GAN, characterized by including the following steps:
[0006] S1. Establish a sub-topology network database using open-source urban road network data;
[0007] S2. Establish a sub-topology network virtual sample library through the established sub-topology network database and GAN network;
[0008] S3. Establish a bridge network sample library based on the sub-topology network sample library;
[0009] S4. Obtain the traffic demand of the sub-topology network before an earthquake using open-source road network traffic data;
[0010] S5. Generate the traffic demand of the virtual sub-topological network through the GAN model.
[0011] Preferably, in S1, the steps of establishing the sub-topological network database are as follows:
[0012] S11. Utilize the open-source urban road network data, convert the road network OD matrix into an adjacency matrix, perform topological processing on the road network data of the research area, and establish the urban large-topological network;
[0013] The specific steps are as follows: The large-topological network takes urban overpasses and large intersections as nodes and roads as edges. When establishing the topological network, the weights of the edges are not considered, and only the connection relationships between nodes are shown in the topological network;
[0014] S12. Use the Louvain algorithm to extract the areas with higher density in the urban road network and form sub-topological networks based on the urban large-topological network;
[0015] The description of this step is as follows: Due to the differences in the connection strength and frequency of nodes in the actual traffic network, as well as the multi-center structure or sparse connections of some road networks, the Louvain algorithm usually gives priority to classifying nodes with high-intensity connections into one category while ignoring nodes with low-intensity connections when processing these networks; Since this method studies the overall seismic resilience of the bridge network rather than individual nodes, the existence of isolated nodes will affect the calculation of network resilience indicators. Therefore, it is necessary to optimize this type of topological network. The specific measure is to remove nodes with a degree value less than or equal to 1;
[0016] S13. Establish a sub-topological network database based on the formed sub-topological network.
[0017] Preferably, in S2, the steps of establishing the virtual database of the sub-topological network are as follows:
[0018] S21. Form more sub-topological samples through data augmentation; the specific steps are as follows: Based on the existing sub-topological network, perform data augmentation by randomly adding or deleting edges to form more sub-topological samples and increase the training samples;
[0019] S22. Construct a GAN network model; the specific steps are as follows: The GAN network consists of a generator and a discriminator. The generator is responsible for generating simulated samples, while the discriminator is responsible for judging whether the input sample is a real sample or a simulated sample. During the training process, the generator continuously optimizes and tries to deceive the discriminator; at the same time, the discriminator also continuously optimizes and tries to identify all simulated samples; during the training iteration process, the two networks confront each other and optimize themselves until a balanced state is reached to obtain the optimal model;
[0020] S23. Generate a virtual sub-topology network using the trained GAN network model. The specific steps are as follows: By introducing gradient penalty and Wasserstein loss to enhance the training stability, use the trained WGAN model to quickly generate a virtual sub-topology network.
[0021] S24. Establish a virtual sample library of the sub-topology network. The specific steps are as follows: Combine the sub-topology network database obtained in S13, the sub-topology samples obtained by data augmentation, and the sub-topology network generated by the GAN model to establish a virtual sample library of the sub-topology network.
[0022] Preferably, in S3, the steps to establish a bridge network sample library are as follows: According to the number of nodes in each network in the sub-topology network sample library, match the same number of parameterized bridge samples to form a bridge topology network, and establish a bridge network sample library.
[0023] Preferably, in S4, the steps to obtain the traffic demand of the pre-earthquake sub-topology network are as follows: The traffic demand of the pre-earthquake sub-topology network is calculated according to the open-source road network traffic data through Equation (1). For the sub-topology network obtained by data augmentation, if an edge is deleted between nodes, the traffic demand between the nodes will decrease by the traffic volume of this edge; if an edge is added, the traffic demand between the nodes will increase by the average traffic volume of all edges between the two nodes in the original topology, so as to obtain the traffic demand matrix after data augmentation.
[0024] D = {q ij}, T = ∑q ij (1)
[0025] In the formula: D is the bridge network traffic demand matrix, q ij represents the traffic demand between nodes i and j.
[0026] Preferably, in S5, the specific steps to generate the traffic demand of the virtual sub-topology network through the GAN model are as follows:
[0027] S51. The GAN model is trained by reading the traffic demand matrix of the existing sub-topology.
[0028] S52. Simulate its traffic characteristics and assign values to the virtual sub-topology network.
[0029] S53. Generate the corresponding virtual sub-topology traffic demand matrix.
[0030] The beneficial effects of the present invention are as follows:
[0031] 1. The present invention uses the GAN model to establish a bridge network sample library based on the real urban road network. The network structure characteristics and traffic characteristics of this sample library are consistent with the actual network, and the distribution range of characteristic parameters is wide, which can ensure the generalization ability of the resilience evaluation model.
[0032] 2. The bridge network sample library established by the present invention can provide richer data for the research of seismic resilience and the health monitoring of bridge structures. Combining the bridge network sample library and the simulated traffic demand can more accurately predict the changing trend of traffic flow on the bridge and formulate traffic diversion plans in advance. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, but do not constitute a limitation to the present invention.
[0034] Figure 1 is a flowchart of the present invention;
[0035] Figure 2 is the Anaheim road network and sub-topology in the present invention;
[0036] Figure 3 is the original topology and the topology after data augmentation processing in the second embodiment of the present invention;
[0037] Figure 4 is the topology sample generated by the GAN network and the bridge topology network sample in the second embodiment of the present invention;
[0038] Figure 5 is the curve graph of the total traffic demand of the actual sub-topology network varying with the edge connectivity and the curve graph of the total traffic demand of the virtual sub-topology network varying with the edge connectivity in the second embodiment of the present invention;
[0039] Figure 6 is the original topology and the topology after data augmentation processing in the third embodiment of the present invention;
[0040] Figure 7 is the topology sample generated by the GAN network and the bridge topology network sample in the third embodiment of the present invention
[0041] Figure 8 is the curve graph of the total traffic demand of the actual sub-topology network varying with the edge connectivity and the curve graph of the total traffic demand of the virtual sub-topology network varying with the edge connectivity in the third embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0042] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
[0043] Embodiment 1
[0044] A method for establishing a bridge network sample library based on GAN and simulating traffic demand, as Figure 1 shown, includes the following steps:
[0045] S1. Use open-source urban road network data to establish a sub-topological network database;
[0046] In S1, the steps for establishing the sub-topological network database are as follows:
[0047] S11. Use open-source urban road network data to convert the road network OD matrix into an adjacency matrix, perform topological processing on the road network data in the research area, and establish a large urban topological network;
[0048] The specific steps are as follows: The large topological network takes urban overpasses and large intersections as nodes and roads as edges. When establishing the topological network, the weights of the edges are not considered, and only the connection relationship between nodes is shown in the topological network;
[0049] S12. Use the Louvain algorithm to extract the areas with higher density in the urban road network and form a sub-topological network based on the large urban topological network;
[0050] The description of this step is as follows: Due to the differences in the connection strength and frequency of nodes in the actual traffic network, as well as the multi-center structure or sparse connection of some road networks, the Louvain algorithm usually gives priority to classifying nodes with high-intensity connections into one category and ignores nodes with low-intensity connections when processing these networks; in view of the fact that this method studies the overall seismic resilience of the bridge network rather than a single node, the existence of isolated nodes will affect the calculation of the network resilience index. Therefore, it is necessary to optimize this type of topological network. The specific measure is to remove nodes with a degree value less than or equal to 1;
[0051] S13. Establish a sub-topological network database based on the formed sub-topological network.
[0052] S2. Establish a sub-topological network virtual sample library through the established sub-topological network database and the GAN network;
[0053] In S2, the steps for establishing the sub-topological network virtual database are as follows:
[0054] S21. Form more sub-topological samples through data augmentation; the specific steps are as follows: On the basis of the existing sub-topological network, perform data augmentation by randomly adding or deleting edges to form more sub-topological samples and increase the training samples;
[0055] S22. Construct a GAN network model. The specific steps are as follows: The GAN network consists of a generator and a discriminator. The generator is responsible for generating simulated samples, while the discriminator is responsible for determining whether the input sample is a real sample or a simulated sample. During the training process, the generator is continuously optimized to try to deceive the discriminator. At the same time, the discriminator is also continuously optimized to strive to identify all simulated samples. During the training iteration process, the two networks compete with each other and optimize themselves until a balanced state is reached to obtain the optimal model.
[0056] S23. Use the trained GAN network model to generate a virtual sub-topology network. The specific steps are as follows: By introducing gradient penalty and Wasserstein loss to enhance training stability, use the trained WGAN model to quickly generate a virtual sub-topology network.
[0057] S24. Establish a virtual sample library for the sub-topology network. The specific steps are as follows: Combine the sub-topology network database obtained in S13, the sub-topology samples obtained by data augmentation, and the sub-topology network generated by the GAN model to establish a virtual sample library for the sub-topology network.
[0058] S3. Establish a bridge network sample library based on the sub-topology network sample library. The steps to establish the bridge network sample library are as follows: According to the number of nodes in each network in the sub-topology network sample library, match the same number of parameterized bridge samples to form a bridge topology network, and establish a bridge network sample library.
[0059] S4. Obtain the traffic demand of the pre-earthquake sub-topology network using the open-source road network traffic data. The steps to obtain the traffic demand of the pre-earthquake sub-topology network are as follows: The traffic demand of the pre-earthquake sub-topology network is calculated according to the open-source road network traffic data through Equation (1). For the sub-topology network obtained by data augmentation, if an edge is deleted between nodes, the traffic demand between the nodes will decrease by the traffic volume of this edge; if an edge is added, the traffic demand between the nodes will increase by the average traffic volume of all edges between the two nodes in the original topology, so as to obtain the traffic demand matrix after data augmentation.
[0060] D = {q ij}, T = ∑q ij (1)
[0061] In the formula: D is the traffic demand matrix of the bridge network, q ij represents the traffic demand between nodes i and j.
[0062] S5. Generate the traffic demand of the virtual sub-topology network through the GAN model. The specific steps to generate the traffic demand of the virtual sub-topology network through the GAN model are as follows:
[0063] S51. The GAN model is trained by reading the traffic demand matrix of the existing sub-topology.
[0064] S52. Simulate its traffic characteristics and assign values to the virtual sub-topology network;
[0065] S53. Generate the corresponding virtual sub-topology traffic demand matrix.
[0066] Embodiment 2
[0067] S1. Taking a certain area in Anaheim as an example, establish a sub-topology network database; use the Louvain algorithm to extract the sub-topology networks of 14 cities. The sub-topology network extracted and optimized from the Anaheim road network is as Figure 2 .
[0068] S2. Establish a virtual database of the sub-topology network.
[0069] First, perform data augmentation. The comparison between the topology after data augmentation and the original topology is as Figure 3 ; then construct a GAN network model; generate virtual sub-topologies based on the constructed GAN network, as Figure 4 (left) shows the randomly selected generated sub-topology samples, which have a high similarity with the real sub-topology structure characteristics, indicating that the performance of the trained WGAN model has reached the expectation. Finally, combine the obtained sub-topology network database, the sub-topology samples obtained by data augmentation, and the sub-topology networks generated by the GAN model to establish a virtual sample library of the sub-topology network.
[0070] S3. Establish a bridge network sample library based on the sub-topology network sample library. According to the number of nodes of each network in the sub-topology network sample library, match the same number of parameterized bridge samples to form a bridge topology network, and establish a bridge network sample library. As Figure 4 (right) The example of the bridge topology network shows that this topology can more realistically reflect the bridge network structure characteristics in the actual road network.
[0071] S4. Obtain the traffic demand of the pre-earthquake sub-topology network using the open-source road network traffic data. As Figure 5 (left) established a curve of the total traffic demand of the actual sub-topology network (calculated according to Equation (1)) changing with the edge connectivity.
[0072] S5. Generate the traffic demand of the virtual sub-topology network through the GAN model. As Figure 5 (right) shows the surface of the total traffic demand of the virtual sub-topology network changing with the edge connectivity. By comparing the two curves, it can be seen that the change trend and value of the virtual sub-topology traffic demand with the connectivity are less different from the actual sub-topology, verifying the rationality of this traffic information simulation method.
[0073] Embodiment 3
[0074] S1. Taking a certain area in Anaheim as an example, a sub-topology network database is established; the sub-topology networks of 14 cities are extracted using the Louvain algorithm. The sub-topology network extracted and optimized from the Anaheim road network is as Figure 2 .
[0075] S2. Establish a virtual database of the sub-topology network.
[0076] First, data augmentation is performed. The comparison between the topology after data augmentation and the original topology is as Figure 6 ; then a GAN network model is constructed; virtual sub-topologies are generated based on the constructed GAN network, such as Figure 7 (left) shows randomly selected generated sub-topology samples, which have a high similarity with the real sub-topology structure characteristics, indicating that the performance of the trained WGAN model has reached the expectation. Finally, a virtual sample library of the sub-topology network is established by combining the obtained sub-topology network database, the sub-topology samples obtained by data augmentation, and the sub-topology networks generated by the GAN model.
[0077] S3. Establish a bridge network sample library based on the sub-topology network sample library. According to the number of nodes in each network in the sub-topology network sample library, parameterized bridge samples with the same number are matched to form a bridge topology network, and a bridge network sample library is established. As Figure 7 (right) The example of the bridge topology network shows that this topology can more realistically reflect the bridge network structure characteristics in the actual road network.
[0078] S4. Obtain the traffic demand of the pre-earthquake sub-topology network using the open-source road network traffic data. As Figure 8 (left) A curve of the total traffic demand of the actual sub-topology network (calculated according to Equation (1)) varying with the edge connectivity is established.
[0079] S5. Generate the traffic demand of the virtual sub-topology network through the GAN model. As Figure 8 (right) shows the surface of the total traffic demand of the virtual sub-topology network varying with the edge connectivity. By comparing the two curves, it can be seen that the change trend and value of the virtual sub-topology traffic demand with the connectivity are less different from those of the actual sub-topology, so it is verified that this traffic information simulation method is reasonable.
[0080] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be encompassed by the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.
Claims
1. A method for establishing a bridge network sample library and simulating traffic demand based on GAN, characterized in that It includes the following steps: S1. Establish a sub-topological network database using open-source urban road network data; S2. Establish a sub-topological network virtual sample library through the established sub-topological network database and GAN network; S3. Establish a bridge network sample library based on the sub-topological network sample library; S4. Obtain the traffic demand of the pre-earthquake sub-topological network using open-source road network traffic data; S5. Generate the traffic demand of the virtual sub-topological network through the GAN model.
2. The method for establishing a bridge network sample library and simulating traffic demand based on GAN according to claim 1, characterized in that, In S1, the steps to establish the sub-topological network database are as follows: S11. Use open-source urban road network data to convert the road network OD matrix into an adjacency matrix, perform topological processing on the road network data of the research area, and establish a large urban topological network; The specific steps are as follows: The large topological network takes urban flyovers and large intersections as nodes and roads as edges. When establishing the topological network, the weights of the edges are not considered, and only the connection relationships between nodes are shown in the topological network; S12. Use the Louvain algorithm to extract the areas with higher density in the urban road network and form a sub-topological network based on the large urban topological network; S13. Establish a sub-topological network database based on the formed sub-topological network.
3. A method for establishing a bridge network sample library and simulating traffic demand based on GAN according to claim 2, characterized in that, In S2, the steps to establish the sub-topological network virtual database are as follows: S21. Form more sub-topological samples through data augmentation. The specific steps are as follows: On the basis of the existing sub-topological network, perform data augmentation by randomly adding or deleting edges to form more sub-topological samples and increase the training samples; S22. Construct a GAN network model. The specific steps are as follows: The GAN network consists of a generator and a discriminator. The generator is responsible for generating simulated samples, while the discriminator is responsible for judging whether the input sample is a real sample or a simulated sample. During the training process, the generator is continuously optimized and tries to deceive the discriminator; at the same time, the discriminator is also continuously optimized and tries to identify all simulated samples; during the training iteration process, the two networks confront each other and optimize themselves until a balanced state is reached to obtain the optimal model; S23. Generate a virtual sub-topological network using the trained GAN network model. The specific steps are as follows: Enhance the training stability by introducing gradient penalty and Wasserstein loss, and use the trained WGAN model to quickly generate a virtual sub-topological network; S24. Establish a sub-topological network virtual sample library. The specific steps are as follows: Combine the sub-topological network database obtained in S13, the sub-topological samples obtained through data augmentation, and the sub-topological network generated by the GAN model to establish a sub-topological network virtual sample library.
4. A method for establishing a bridge network sample library and simulating traffic demand based on GAN according to claim 3, characterized in that In S3, the steps to establish the bridge network sample library are as follows: According to the number of nodes in each network in the sub-topological network sample library, match the same number of parametric bridge samples to form a bridge topological network and establish a bridge network sample library.
5. A method for establishing a bridge network sample library and simulating traffic demand based on GAN according to claim 4, characterized in that, In S4, the steps to obtain the traffic demand of the pre-earthquake sub-topological network are as follows: The traffic demand of the pre-earthquake sub-topological network is calculated by Equation (1) based on the open-source road network traffic data. For the sub-topological network obtained by data augmentation, if an edge is deleted between nodes, the traffic demand between the nodes will decrease by the traffic volume of this edge; if an edge is added, the traffic demand between the nodes will increase by the average traffic volume of all edges between the two nodes in the original topology, so as to obtain the traffic demand matrix after data augmentation. D = {q ij}, T = Σq ij (1) Where: D is the bridge network traffic demand matrix, and q ij represents the traffic demand volume between nodes i and j.
6. A method for establishing a bridge network sample library and simulating traffic demand based on GAN according to claim 5, characterized in that, In S5, the specific steps to generate the traffic demand of the virtual sub-topological network through the GAN model are as follows: S51. The GAN model is trained by reading the traffic demand matrix of the existing sub-topology. S52. Simulate its traffic characteristics and assign values to the virtual sub-topological network. S53. Generate the corresponding virtual sub-topological traffic demand matrix.