Method, system, device and medium for evaluating trafficability of urban bridge network after earthquake

Abstract

The application relates to the technical field of public safety monitoring, and discloses a method, system, device and medium for evaluating the post-earthquake traffic capacity of a city bridge group network, which comprises the following steps: constructing a three-dimensional underground structure speed model of a key area of a target city according to bridge basic information data, generating three-way ground acceleration time histories of each bridge based on the three-dimensional underground structure speed model, calculating structure response data corresponding to the three-way ground acceleration time histories under different seismic scenarios, and calculating residual traffic capacity scores; training a graph neural network by using the structure response data and the residual traffic capacity scores to obtain a target evaluation network; calculating the predicted residual traffic capacity scores of each target bridge node in seismic real-time monitoring data by using the target evaluation network, and solving an optimal traffic path according to the predicted residual traffic capacity scores. The application can quickly and quantitatively evaluate the post-earthquake traffic capacity of the city bridge group network under real scenario ground motion and give an emergency traffic path.

Description

Technical Field

[0001] This invention relates to the field of public safety monitoring technology, specifically to a method, system, equipment, and medium for assessing the post-earthquake accessibility of urban bridge networks. Background Technology

[0002] As cities expand and transportation networks become increasingly complex, bridges, as vital structures spanning rivers, canyons, and existing roads, often occupy key positions within urban road networks. Their post-earthquake traffic capacity directly impacts the smooth operation of emergency rescue, medical evacuation, and material transport. Under strong earthquakes, bridge structures are prone to varying degrees of damage, including pier base cracking, bearing slippage, and beam end displacement. Some critical bridges may even experience severe failures such as beam collapse and pier damage, leading to a significant reduction in the connectivity of urban transportation networks, either locally or overall. Existing earthquake damage events demonstrate that damage to a few critical bridges can cause long-term disruptions to important passageways, severely weakening the seismic resilience of urban infrastructure systems.

[0003] To this end, scholars both domestically and internationally have conducted extensive research on the seismic performance of bridges and the reliability of bridge networks. For the rapid prediction of regional bridge earthquake damage, existing research has proposed using machine learning to establish a mapping relationship between bridge parameters and damage states, enabling rapid assessment of the individual seismic damage status of bridges within a region. Taking the regional bridge macroscopic seismic damage prediction and analysis method described in patent CN118036142A as an example, this method essentially still focuses on the prediction of individual bridges within a region, without considering the cascading effects of the post-earthquake bridge network, making it difficult to support post-earthquake emergency route decision-making. Furthermore, the seismic motion samples are mostly based on data from a typical earthquake event, making it difficult to fully characterize the geological heterogeneity of cities where earthquakes did not occur and the differences in bridge sites. Moreover, the above methods typically rely on offline database construction for one-time modeling, failing to use real-time responses collected by bridge monitoring systems as input for online model updates, thus making it difficult to dynamically correct the assessment results as actual earthquake events occur.

[0004] At the bridge network level, most studies use Bayesian networks or Markov chain methods to assess the connectivity, functional loss, and resilience of bridge networks under different earthquake scenarios. For example, the large-scale bridge network assessment method described in published patent CN109918819B is representative. These methods help to characterize the relationship between bridge damage and traffic network performance holistically, but they generally employ broad bridge function classifications or empirical recovery curves, simplifying bridge nodes to only a few discrete states, making it difficult to finely distinguish the remaining traffic capacity under different levels of earthquake damage. Furthermore, the mutual influence between bridge nodes is mostly described only through topological connectivity, lacking in-depth exploration of the intrinsic connections between node functions and network operational logic.

[0005] Graph Neural Networks (GNNs), as an emerging network architecture in recent years, have been widely applied in fields such as structural health monitoring surrogate models, bridge deck fire prediction, concrete cracking analysis, and fluid mechanics due to their ability to aggregate multi-level information using adjacent nodes on a graph structure. For seismic reliability assessment of transportation infrastructure, some works have attempted to use GNNs as surrogate models for bridge network reliability or connectivity analysis. For example, the method proposed by Tong Liu et al., "Graph Neural Network Surrogate for Seismic Reliability Analysis of Highway Bridge Systems," uses GNNs to quickly predict node connectivity probabilities under seismic scenarios. While such methods have shown some effectiveness in accelerating large-scale network reliability analysis, they often rely on empirical seismic motion prediction equations or simplified site models to generate seismic motion data during training sample construction. This approach fails to adequately consider the complex influences of three-dimensional underground structures and different source mechanisms, making it difficult to reflect the spatial non-uniformity of seismic motion at the urban scale. Furthermore, existing models are typically trained once based on offline-generated seismic damage samples, failing to incorporate real-time response data acquired by bridge safety monitoring systems into node features. This prevents dynamic correction of model predictions based on the physical state during actual earthquakes, resulting in deficiencies in both accuracy and real-time performance in accessibility assessment. Summary of the Invention

[0006] The technical problem to be solved by this invention is how to solve the problem of disaster assessment and emergency rescue in the analysis of bridge network under real earthquake scenarios.

[0007] The present invention solves the above-mentioned technical problems through the following technical means;

[0008] Acquire basic bridge information data for key areas of the target city, and construct a network diagram of the city's bridge group based on the basic bridge information data;

[0009] Based on the array observation data in the bridge basic information data, the three-dimensional underground structure velocity model is inverted, and the three-dimensional ground acceleration time history at the bridge site coordinates of each bridge is generated based on the three-dimensional underground structure velocity model.

[0010] The structural response data of the bridge corresponding to the three-dimensional ground acceleration time history are calculated using the finite element model of the bridge under different seismic scenarios, and the remaining traffic capacity score corresponding to the structural response data is calculated.

[0011] The structural response data and the remaining traffic capacity score are used as the training set to train a pre-built graph neural network to obtain the target evaluation network.

[0012] Real-time earthquake monitoring data of the target area is collected, and the predicted remaining traffic capacity score of each target bridge node in the real-time earthquake monitoring data is calculated using the target evaluation network. The optimal traffic path is then solved based on the predicted remaining traffic capacity score.

[0013] Optionally, the step of inverting the three-dimensional underground structure velocity model based on the array observation data in the bridge foundation information data includes:

[0014] Cross-correlation calculations were performed on the continuous noise records between different station pairs of the array observation data to obtain the dispersion curves of multi-period surface waves;

[0015] The initial shear wave velocity model is obtained by performing a deep inversion on the dispersion curve;

[0016] The velocity structure of the initial shear wave velocity model is constrained and corrected, and the shear wave velocity is adjusted and smoothed to obtain a three-dimensional underground structure velocity model.

[0017] Optionally, generating the three-dimensional ground acceleration time history at the bridge site coordinates of each bridge based on the three-dimensional underground structure velocity model includes:

[0018] Multiple earthquake scenarios and corresponding source parameters were constructed on the three-dimensional underground structure velocity model.

[0019] Based on the source parameters, the three-dimensional finite difference method was used to simulate the source rupture and seismic wave propagation process on the three-dimensional underground structure velocity model to obtain simulation data.

[0020] Interpolation simulations were performed on the simulation data based on the bridge site coordinates of each bridge to obtain the three-dimensional ground acceleration time history.

[0021] Optionally, calculating the remaining traffic capacity score corresponding to the structural response data includes:

[0022] Calculate the expected value and standard deviation of the logarithmic regression of the structural response data under the different earthquake scenarios;

[0023] Calculate the system-level exceedance probability of structural response data at a preset damage level under different earthquake scenarios based on the expected value and the standard deviation.

[0024] The system-level exceedance probability is calculated using the following formula:

[0025]

[0026]

[0027] in, Indicating an earthquake scenario The system-level exceedance probability is below. Earthquake scenario Next Each structural indicator in damage level The probability of exceeding the limit at that time, , These are the expected value and standard deviation of the logarithmic regression, respectively. It is the standard normal distribution function;

[0028] Calculate the graded probabilities of the system-level exceedance probability under different damage levels:

[0029]

[0030] in, For graded probabilities, In the context of an earthquake scenario The system-level exceedance probability is below. Preset damage level The earthquake damage index below;

[0031] Shape parameters of the Beta distribution , As shown in the following formula:

[0032]

[0033]

[0034] The expected value of the Beta distribution is calculated based on the shape parameters and used as the system damage index. :

[0035]

[0036] Calculate the earthquake scenario using the following formula Remaining traffic capacity rating:

[0037]

[0038] in, In the context of an earthquake scenario The remaining traffic capacity rating below, In the context of an earthquake scenario The system-level exceedance probability is below. Indicates earthquake scenario The standard deviation of the lower structure response data.

[0039] Optionally, training a pre-built graph neural network using the structural response data and the remaining capacity score as a training set to obtain the target evaluation network includes:

[0040] The structural response data and the urban bridge network structure are message-passed and neighborhood-aggregated using the graph neural network to obtain the training remaining traffic capacity score for each bridge node.

[0041] Calculate the loss value between the training remaining capacity score and the remaining capacity score;

[0042] The parameters of the graph neural network are updated by backpropagation based on the loss value until the loss value is less than a preset loss value threshold, thus obtaining the target evaluation network.

[0043] Optionally, the step of calculating the predicted remaining traffic capacity score for each target bridge node in the real-time earthquake monitoring data using the target evaluation network includes:

[0044] Construct a real-time urban bridge network structure corresponding to each target bridge node in the real-time earthquake monitoring data;

[0045] Extract the real-time dynamic features and real-time static features of each target bridge node within the set time window;

[0046] The real-time dynamic features and the real-time static features are combined and then used as the target bridge node features of the real-time urban bridge network structure. These features are then input into the target evaluation network to obtain the predicted remaining traffic capacity score.

[0047] Optionally, the step of solving for the optimal travel path based on the predicted remaining capacity score includes:

[0048] The two end nodes of the bridge path are determined based on the target bridge node, and the minimum value of the predicted remaining capacity score corresponding to the two end nodes is used as the control score of the bridge path.

[0049] The edge weights of the bridge path are updated based on the control score to obtain updated edge weight values.

[0050] The updated edge weight values ​​are obtained by using the following formula:

[0051]

[0052]

[0053] in, To update the edge weight values, The penalty magnitude parameter; The preset curvature parameters, The preset normal threshold, The preset closing threshold, Bridge path The corresponding control score, Bridge path Preset original edge weights, To control the penalty factor corresponding to the score;

[0054] The optimal same-path for the target region is calculated using the deterministic weighted shortest path algorithm based on the updated edge weight values.

[0055] To address the aforementioned problems, this invention also proposes a post-earthquake accessibility assessment system for urban bridge networks, the system comprising:

[0056] The bridge network diagram construction module is used to acquire basic bridge information data of key areas in the target city and construct the urban bridge network diagram structure based on the basic bridge information data.

[0057] The model inversion module is used to invert the three-dimensional underground structure velocity model based on the array observation data in the bridge basic information data, and generate the three-dimensional ground acceleration time history of the bridge site coordinates of each bridge based on the three-dimensional underground structure velocity model.

[0058] The remaining traffic capacity score calculation module is used to calculate the structural response data under different seismic scenarios using the finite element model of the bridge corresponding to the three-dimensional ground acceleration time history, and to calculate the remaining traffic capacity score corresponding to the structural response data.

[0059] The graph neural network training module is used to train a pre-built graph neural network using the structural response data and the remaining traffic capacity score as a training set to obtain a target evaluation network.

[0060] The optimal travel path solution module is used to collect real-time earthquake monitoring data of the target area, use the target evaluation network to calculate the predicted remaining traffic capacity score of each target bridge node in the real-time earthquake monitoring data, and solve the optimal travel path based on the predicted remaining traffic capacity score.

[0061] The present invention also provides a processing device, characterized in that it includes at least one processor and at least one memory communicatively connected to the processor, wherein: the memory stores program instructions executable by the processor, and the processor can execute the above-described method for assessing the post-earthquake accessibility of urban bridge networks by calling the program instructions.

[0062] The present invention also provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores computer instructions, the computer instructions causing the computer to execute the above-described method for post-earthquake accessibility assessment of urban bridge network.

[0063] The advantages of this invention are:

[0064] This invention introduces a three-dimensional underground structure velocity model, which can fully consider the influence of factors such as the location, depth, magnitude, and fault strike of the epicenter under different earthquake scenarios on the generation and propagation of seismic waves, as well as the spatial differences caused by different distributions of bridge networks. It effectively solves the limitation of traditional seismic wave selection that relies solely on experience or local standard spectra. By using these high-precision simulation results as training samples for graph neural networks, the model prediction results are closer to reality, providing a reliable basis for assessing the navigability of bridge networks under real earthquake scenarios.

[0065] By utilizing the message passing mechanism of graph neural networks, the seismic damage index of a multi-index cascaded system is closely integrated with the network operation logic. Compared with traditional black-box machine learning models or Markov chain bridge network assessment methods, this invention significantly improves the interpretability of the post-earthquake physical process of a single bridge, while enhancing the ability to express information radiated from a single bridge to the network, which helps to achieve more scientific post-earthquake assessment and decision support in engineering practice.

[0066] Finally, by introducing real-time bridge safety monitoring data as input features in the node feature stage, the shortcomings of traditional models that rely solely on numerical simulation in terms of accuracy are effectively compensated. The network evaluation results can be dynamically updated based on the structural response of the bridge under actual earthquakes, thereby achieving a more intelligent and reliable post-earthquake accessibility assessment. Attached Figure Description

[0067] Figure 1 This is a flowchart illustrating a method for assessing the post-earthquake accessibility of urban bridge networks according to an embodiment of the present invention.

[0068] Figure 2 This is a schematic diagram of a key area of ​​a target city in one embodiment of the present invention;

[0069] Figure 3 This is a schematic diagram of the connection relationship between urban bridge groups and roads and the network structure of urban bridge groups in one embodiment of the present invention;

[0070] Figure 4 This is a time history curve of the vertical acceleration of the site soil simulating a local earthquake in one embodiment of the present invention;

[0071] Figure 5 This is a schematic diagram of the finite element model structure of a bridge in one embodiment of the present invention;

[0072] Figure 6 This is a time history curve of the structural response to the beam end displacement index in one embodiment of the present invention;

[0073] Figure 7 This is a system damage index diagram under different earthquake scenarios in one embodiment of the present invention;

[0074] Figure 8A score map of remaining traffic capacity for bridge nodes under different earthquake scenarios;

[0075] Figure 9 A schematic diagram illustrating the real-time monitoring data update for bridges;

[0076] Figure 10 A visual diagram illustrating the post-earthquake traffic status of the bridge network and the optimal traffic routes;

[0077] Figure 11 This is a functional module diagram of a post-earthquake accessibility assessment system for urban bridge networks provided in one embodiment of the present invention. Detailed Implementation

[0078] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Reference Figure 1 The diagram shown is a flowchart illustrating a method for assessing the post-earthquake accessibility of an urban bridge network according to an embodiment of the present invention. In this embodiment, the method for assessing the post-earthquake accessibility of an urban bridge network includes:

[0080] S1. Obtain basic bridge information data for key areas of the target city, and construct a network diagram of the city's bridge group based on the basic bridge information data.

[0081] In this embodiment of the invention, the bridge basic information data includes short-period dense array observation data of key areas of the target city, finite element analysis models of each bridge, real-time monitoring data of the bridge structure, and the static connection relationship between the bridge group and the road. For example, taking the main urban area of ​​a city as the research area, such as... Figure 2 As shown, a dense array of short-period instruments was deployed in this area, with each station equipped with a three-component short-period instrument (such as QS-5A) with a 5s natural frequency and a sampling rate of 100Hz. The acquired continuous noise records were preprocessed, including detrending, bandpass filtering, and baseline correction, to form the array observation data required for inverting the underground structure.

[0082] Subsequently, finite element models were established for each bridge in the region. The superstructure and substructure were described using a combination of beam and shell elements, and plastic hinges were placed at key locations to accommodate subsequent seismic elastoplastic analysis, ensuring that the deviation between the calculated dominant frequency and the actual bridge dominant frequency was controlled within 5%.

[0083] In terms of monitoring, structural health monitoring sensors are deployed on each bridge: strain gauges are installed at the bottom of the piers, and wire displacement gauges are installed at the supports and expansion joints to record dynamic responses such as pier stress, beam end displacement, and support displacement in real time. The sampling frequency is preferably 10Hz to ensure that the response throughout the entire earthquake process can be continuously extracted.

[0084] Specifically, each bridge in a key area of ​​the target city is used as a node. V The reachable paths connecting the bridge nodes are used as edges of the graph structure. E Constructing a network structure of urban bridge groups Bridge nodes Define static and dynamic features, edge Define static features; the specific structure of the urban bridge network diagram is as follows: Figure 3 As shown, the static characteristics of a node include bridge type, span, construction year, technical condition level, and site category; the dynamic characteristics consist of extreme values, root mean square values, and spectral characteristics extracted from structural monitoring data and finite element responses. The static characteristics of the roadside include road length, road grade, number of lanes, and speed limit, reflecting basic traffic capacity.

[0085] As an example, bridge nodes v 1 and adjacent edges e 1,2 The static and dynamic characteristics of the nodes are shown in Tables 1 and 2 below. Table 1 shows the node characteristics. v The static characteristic parameters of bridge nodes are shown in Table 2. v 1 and v The edge between 2 e 1,2 Static characteristic parameters.

[0086] Table 1

[0087]

[0088] Table 2

[0089]

[0090] S2. Based on the array observation data in the bridge basic information data, invert the three-dimensional underground structure velocity model, and generate the three-dimensional ground acceleration time history of the bridge site coordinates of each bridge based on the three-dimensional underground structure velocity model.

[0091] In this embodiment of the invention, in order to more realistically simulate the ground motion field under different earthquake scenarios, a three-dimensional underground structure velocity model of the key area of ​​the target city is inverted based on the short-period dense array observation data in the bridge basic information data.

[0092] Specifically, the step of inverting the three-dimensional underground structure velocity model based on the array observation data in the bridge foundation information data includes:

[0093] Cross-correlation calculations were performed on the continuous noise records between different station pairs of the array observation data to obtain the dispersion curves of multi-period surface waves;

[0094] The initial shear wave velocity model is obtained by performing a deep inversion on the dispersion curve;

[0095] The velocity structure of the initial shear wave velocity model is constrained and corrected, and the shear wave velocity is adjusted and smoothed to obtain a three-dimensional underground structure velocity model.

[0096] In detail, obtain continuous waveform records from each station in the array observation data, and record any two stations... and The preprocessed waveform is , Then in the first Calculate the cross-correlation function within each time window. for:

[0097]

[0098] In the formula, For the station Hetai Station Between the second The mutual delay within each time window is The cross-correlation function, For time delay; For the length of a single window, For the station In the Preprocessed waveform within a time window For the station In the Within a time window The preprocessed waveform at time 10:00. Let be the time infinitesimal element. Furthermore, by superimposing and averaging the cross-correlation results of multiple time windows, a stable cross-correlation is obtained:

[0099]

[0100] in, For stable cross-correlation, For the station Hetai Station Between the first The latency within each time window is The cross-correlation function, This represents the total number of time windows.

[0101] The stable cross-correlation is used as a representation of the empirical Green's function, and its symmetric component is taken:

[0102]

[0103] in, For stable cross-correlation symmetric components, For delay Stable cross-correlation.

[0104] Furthermore, regarding Perform a Fourier transform and take its phase as:

[0105]

[0106] In the formula, This is the frequency domain spectrum obtained after the Fourier transform. Angular frequency, For the argument operation of complex numbers, For the phase of the stable cross-correlation symmetrical component, then within a given period... Below, the distance between stations is equivalent phase velocity Then it satisfies:

[0107]

[0108]

[0109] In the formula, For a certain period The path dispersion observations below, It is an integer. To stabilize the phase of the cross-correlation symmetric components, For the station Hetai Station The distance between the stations, For a given period Below, the distance between stations is The equivalent phase velocity is used for phase unwrapping and mode consistency constraints.

[0110] Therefore, different stations and In multiple cycles Phase velocity below With period The set of correspondences is represented as the dispersion curve of a multi-period surface wave.

[0111] Further, surface wave travel-time tomography was performed on the dispersion curves to obtain the two-dimensional phase velocity distribution, and the initial shear wave velocity model was obtained by depth inversion of the dispersion curves.

[0112] For each target period The path dispersion observations obtained above By converting to traveltime observations and performing surface wave traveltime tomography inversion, the two-dimensional phase velocity distribution corresponding to that period can be obtained. And the relationship is as follows:

[0113]

[0114] In the formula, For the first p The travel time of the path, To correspond to the propagation path, For path The infinitesimal element, For benchmark period Two-dimensional phase velocity distribution, , These represent the x-axis and y-axis, respectively.

[0115] Discretize the key areas of the target city into a grid, denoted as the... k The slowness of each grid is Then the above equation can be discretized as:

[0116]

[0117] In the formula, The path travel time matrix, For geometric matrices, This is the slowness matrix.

[0118] The slow-degree field can be solved using least squares with smoothing regularization:

[0119]

[0120] In the formula, For observation weights, For spatial smoothing operators, The regularization coefficient is . The path travel time matrix, For geometric matrices, This is the slowness matrix.

[0121] Perform the above steps for each period to obtain a set of two-dimensional phase velocity fields sliced ​​by period. Then, at any grid point in the key area of ​​the target city ( x , y At point ), phase velocity values ​​from multiple cycles can be extracted, and a local dispersion curve for that point can then be constructed. Furthermore, using local dispersion curves as observations, a surface wave dispersion forward modeling operator under layered medium conditions is established:

[0122]

[0123]

[0124] In the formula, For the numerical solution operator of the Love wave dispersion equation, , , These represent shear wave velocity, longitudinal wave velocity, and density as a function of depth. The distribution For layering ratio, This is a local scattering curve.

[0125] Therefore, for each grid point (x, y), by further inverting the difference between the observed dispersion and the forward dispersion, a one-dimensional shear wave velocity profile at that point can be obtained. Merging the initial shear wave velocity profiles of all grid points yields the initial three-dimensional shear wave velocity model. .

[0126] In this embodiment of the invention, geological exploration data is used to constrain and correct the velocity structure of the initial shear wave velocity model from the near-surface to the upper crust:

[0127] Based on geological exploration data (such as borehole data, shallow seismic data, seismic engineering surveys, stratigraphic information, etc.), the reference velocity given by the exploration is directly adopted within a specified depth range. For the initial model By setting upper and lower bounds on the near-surface to upper crust depth range, constraint correction is completed, resulting in a shallow controlled shear wave velocity model. .

[0128] Furthermore, by combining the results of volume wave tomography, a three-dimensional velocity model of the subsurface structure from the surface to the top of the upper mantle is formed by stitching together the overlapping depth zones:

[0129] Let the overlap depth band of the surface wave model and the volume wave model be . The weight function is defined as follows:

[0130]

[0131] Obtaining the velocity model of volume wave tomography The model is then continuously fused according to the weight function:

[0132]

[0133] In the formula, For shear wave velocity, This is a shallow, controlled shear wave velocity model. The overlapping depth band is The weight function at time, This is a velocity model for volume wave tomography.

[0134] Further analysis of the fused 3D Spatial smoothing was performed to match the resolution scales of surface waves and volume waves and to suppress artifacts in the splicing bands, ultimately resulting in a three-dimensional velocity model of the subsurface structure from the Earth's surface to the top of the upper mantle.

[0135] In this embodiment of the invention, by constraining and correcting the velocity structure from the near-surface to the upper crust, the inversion results are made consistent with borehole and seismic sounding data. Simultaneously, by combining body wave tomography results, shear wave velocities are adjusted and smoothed within different depth ranges. Finally, the body wave velocity model and the surface wave velocity model are stitched together in the overlapping depth zone to form a three-dimensional subsurface structure velocity model from the surface to the top of the upper mantle.

[0136] In this embodiment of the invention, the three-dimensional ground acceleration time history is simulated in the three-dimensional underground structure velocity model using the three-dimensional finite difference method to simulate the source rupture and seismic wave field propagation process. The three-dimensional ground acceleration time history is extracted at the free surface according to the coordinates of each bridge site. The results of the extracted three-dimensional ground acceleration time history are as follows: Figure 4 As shown.

[0137] Specifically, the generation of the three-dimensional ground acceleration time history at the bridge site coordinates of each bridge based on the three-dimensional underground structure velocity model includes:

[0138] Multiple earthquake scenarios and corresponding source parameters were constructed on the three-dimensional underground structure velocity model.

[0139] Based on the source parameters, the three-dimensional finite difference method was used to simulate the source rupture and seismic wave propagation process on the three-dimensional underground structure velocity model to obtain simulation data.

[0140] Interpolation simulations were performed on the simulation data based on the bridge site coordinates of each bridge to obtain the three-dimensional ground acceleration time history.

[0141] In this embodiment of the invention, after obtaining the three-dimensional underground structure velocity model, multiple seismic scenarios are set on the model. Each seismic scenario is determined by randomly sampling at least one of the following: source location, source depth, magnitude, fault strike and dip angle, slip component, rupture velocity, and rupture duration, thereby forming a set of source parameters with uncertainty.

[0142] Furthermore, for each set of source parameters, the three-dimensional finite difference method was used in the three-dimensional underground structure velocity model to simulate the source rupture and seismic wave propagation process. The three-dimensional ground acceleration time histories were extracted at the free surface according to the coordinates of each bridge site. The results are as follows: Figure 4 As shown.

[0143] In the case where the bridge site coordinates in the bridge basic information data cannot be aligned to the grid nodes of the three-dimensional finite difference method, the wave field snapshot (grid point values ​​at each time) of the entire computational domain is obtained through the three-dimensional finite difference method, which is the simulation data. For each output time, spatial interpolation is performed at the bridge site coordinate position to form the three-dimensional ground acceleration time history of the bridge site coordinate.

[0144] This invention, through a three-dimensional underground structure velocity model, can fully consider the influence of factors such as the location, depth, magnitude, and fault strike of the seismic source under different earthquake scenarios on the generation and propagation of seismic waves, as well as the spatial differences caused by different distributions of bridge networks. It effectively solves the limitation of traditional seismic wave selection that relies solely on experience or local standard spectra. By using these high-precision simulation results as a training set, the model's prediction results are closer to reality, providing a reliable basis for assessing the navigability of bridge networks under real earthquake scenarios.

[0145] S3. Calculate the structural response data under different seismic scenarios using the finite element model of the bridge corresponding to the three-dimensional ground acceleration time history, and calculate the remaining traffic capacity score corresponding to the structural response data.

[0146] In this embodiment of the invention, the structural response data is the time history curves of the bridge under different seismic scenarios, including at least beam end displacement, support displacement, and pier bottom stress.

[0147] Specifically, the calculation of structural response data using the finite element model of the bridge corresponding to the three-dimensional ground acceleration time history includes:

[0148] Obtain the pre-constructed finite element model of the bridge corresponding to the time history of the three-dimensional ground acceleration;

[0149] The three-dimensional ground acceleration time history is input into the finite element model for nonlinear time history analysis to obtain the structural response data of the bridge under different seismic scenarios.

[0150] Specifically, with nodes v Taking 1 as an example, the three-dimensional ground acceleration time histories obtained from step S2 for each bridge site are input into the corresponding bridge finite element model, such as... Figure 5 As shown, nonlinear time history analysis was performed to obtain structural response data of the bridge under different seismic scenarios. The structural response includes at least time history curves of beam end displacement, support displacement, and pier bottom stress, as shown in the figure. Figure 6 As shown.

[0151] Furthermore, the remaining capacity score corresponding to the structural response data is obtained through Beta distribution mapping. RPI .

[0152] Specifically, calculating the remaining traffic capacity score corresponding to the structural response data includes:

[0153] Calculate the expected value and standard deviation of the logarithmic regression of the structural response data under the different earthquake scenarios;

[0154] Calculate the system-level exceedance probability of structural response data at a preset damage level under different earthquake scenarios based on the expected value and the standard deviation.

[0155] The system-level exceedance probability is mapped using a Beta distribution to obtain the remaining capacity score.

[0156] In detail, under different earthquake scenarios Below, for each structural index in the structural response data: beam end displacement, support displacement, and pier bottom stress. k ∈{1,2,3} according to damage level n Calculate the transcendence probability for ∈{1,2,…,5}.

[0157] The exceedance probability is calculated using the following formula:

[0158]

[0159] in, Earthquake scenario Next Each structural indicator in damage level The probability of exceeding the limit at that time, , These are the expected value and standard deviation of the logarithmic regression, respectively. It is the standard normal distribution function.

[0160] Specifically, the exceedance probability of the beam end displacement structural index under different earthquake scenarios ( k The results of =1) are shown in Table 3. The exceedance probability of the pier bottom stress structural index under different earthquake scenarios ( k The results for =2) are shown in Table 4:

[0161] Table 3

[0162]

[0163] Table 4

[0164]

[0165] The system-level exceedance probability is obtained by using a multi-index cascaded system model. :

[0166]

[0167] Specifically, the system-level exceedance probabilities of the tandem system under different seismic scenarios are shown in Table 5:

[0168] Table 5

[0169]

[0170] In this embodiment of the invention, the step of mapping the system-level exceedance probability to a Beta distribution to obtain the remaining capacity score includes:

[0171] Based on graded probability Perform a Beta distribution mapping on the interval [0,1]:

[0172]

[0173] In the formula, For a preset damage level The earthquake damage index below;

[0174] Then the shape parameter of the Beta distribution , for:

[0175]

[0176]

[0177] Then the expected value of the Beta distribution is taken as the system damage index, and the result is as follows: Figure 7 As shown:

[0178]

[0179] Let the bridge nodes be in this earthquake scenario Below RPI for:

[0180]

[0181] Therefore, it was obtained RPI The mapping relationship between the Beta earthquake damage index and the Beta index.

[0182] For details, please refer to [link / reference]. Figure 8 As shown, the remaining traffic capacity scores of the overall bridge system in key areas of the target city under different earthquake scenarios are obtained. The above operation is repeated, using the bridge structural response data under each earthquake scenario as input samples, to obtain the structural response data corresponding to each bridge node under different earthquake scenarios. RPI value.

[0183] S4. Use the structural response data and the remaining traffic capacity score as the training set to train a pre-built graph neural network to obtain the target evaluation network.

[0184] In this embodiment of the invention, structural response data and corresponding remaining capacity scores under different earthquake scenarios are used as training samples to train the graph neural network. This enables the graph neural network to learn the spatial correlation and structural damage patterns of bridge network capacity under different earthquake scenarios, thereby achieving prediction of remaining capacity scores for unknown earthquake samples or actual monitoring data.

[0185] Specifically, the step of training a pre-constructed graph neural network using the structural response data and the remaining traffic capacity score as a training set to obtain the target evaluation network includes:

[0186] The structural response data and the urban bridge network structure are message-passed and neighborhood-aggregated using the graph neural network to obtain the training remaining traffic capacity score for each bridge node.

[0187] Calculate the loss value between the training remaining capacity score and the remaining capacity score;

[0188] The parameters of the graph neural network are updated by backpropagation based on the loss value until the loss value is less than a preset loss value threshold, thus obtaining the target evaluation network.

[0189] In this embodiment of the invention, the graph neural network is used to perform message passing and neighborhood aggregation on the structural response data and the urban bridge group network graph structure. The message passing includes concatenating the neighbor node representation and edge static features after linear transformation, performing weighted aggregation based on attention weights of node and edge features, and then obtaining the node update representation through nonlinear transformation.

[0190] Among them, the layer to the first The layer is updated as follows:

[0191]

[0192] Attention weights in the formula It is given by the following formula:

[0193]

[0194]

[0195] In the formula, , Bridge nodes and adjacent nodes In the The layer's representation vector, It is a non-linear activation function. , , , , For the mapping matrix, Bridge nodes The set of adjacent nodes, For connection and roadside, For bias vectors, This is a vector concatenation operation. For process parameters, For a linear rectified function with leakage, The attention scoring vector.

[0196] In detail, during the network training phase, for each earthquake scenario, the structural response data of each bridge under that scenario is first input as nodes according to the bridge node number, and the calculated data of each bridge is then used as the network training data. The nodes are labeled for supervision; then, the node inputs and the bridge network graph structure (bridge network topology) are fed into the graph neural network. After message passing and neighborhood aggregation, the network outputs the training predictions for each bridge node. Value; for each bridge node, the "predicted" value will be assigned separately. "and "truth" The "supervised label" is used to calculate the error, and this error is aggregated across all nodes to form the training loss. Finally, backpropagation is performed on the graph neural network parameters based on this loss, so that the network parameters continuously decrease the predicted values ​​of each node in repeated iterations. With reality The deviations between them allow the graph neural network to learn the node structural response, the influence of adjacent nodes, and the combined effects of network topology under different earthquake scenarios. The mapping relationship improves the accuracy of subsequent predictions of remaining capacity scores.

[0197] This invention utilizes a graph neural network message passing mechanism to tightly integrate the seismic damage index of a multi-index cascaded system with the network operation logic. Compared to traditional black-box machine learning models or Markov chain bridge network assessment methods, this invention significantly improves the interpretability of the post-earthquake physical processes of a single bridge and enhances the ability to express information radiated from a single bridge to the network, thus contributing to more scientific post-earthquake assessment and decision support in engineering practice.

[0198] S5. Collect real-time earthquake monitoring data of the target area, use the target evaluation network to calculate the predicted remaining traffic capacity score of each target bridge node in the real-time earthquake monitoring data, and solve for the optimal traffic path based on the predicted remaining traffic capacity score.

[0199] In this embodiment of the invention, when an earthquake occurs, a short-period dense array of monitoring stations is deployed at bridge nodes within the earthquake-affected area to collect real-time monitoring data. The bridge nodes at this time are then used as target bridge nodes to obtain real-time earthquake monitoring data.

[0200] Specifically, the step of calculating the predicted remaining traffic capacity score for each target bridge node in the real-time earthquake monitoring data using the target evaluation network includes:

[0201] Construct a real-time urban bridge network structure corresponding to each target bridge node in the real-time earthquake monitoring data;

[0202] Extract the real-time dynamic features and real-time static features of each target bridge node within the set time window;

[0203] The real-time dynamic features and the real-time static features are combined and then used as the target bridge node features of the real-time urban bridge network structure. These features are then input into the target evaluation network to obtain the predicted remaining traffic capacity score.

[0204] In detail, during an actual earthquake, this invention utilizes real-time monitoring data collected by the bridge safety monitoring system to update the dynamic characteristics of each target bridge node, such as... Figure 9 As shown. Specifically, within a set time window, the maximum value, root mean square value, and spectral characteristics of real-time response indicators such as beam end displacement, support displacement, and pier bottom stress are extracted as real-time dynamic features. Real-time static features include road length, road grade, number of lanes, and speed limit. The real-time dynamic features and real-time static features are concatenated and input into a trained graph neural network model to obtain the predicted remaining traffic capacity score for each target bridge node.

[0205] Further, the step of solving for the optimal travel path based on the predicted remaining capacity score includes:

[0206] The two end nodes of the bridge path are determined based on the target bridge node, and the minimum value of the predicted remaining capacity score corresponding to the two end nodes is used as the control score of the bridge path.

[0207] The edge weights of the bridge path are updated based on the control score to obtain updated edge weight values.

[0208] The optimal same-path for the target region is calculated using the deterministic weighted shortest path algorithm based on the updated edge weight values.

[0209] In detail, the edges in the real-time urban bridge network graph structure are taken as bridge paths, and the target bridge nodes at both ends of the bridge path are taken as the two endpoints of the bridge path. That is, for any side of the bridge path... The scores of the two nodes are respectively and Controlling scores At the closing threshold 0.20, normal threshold When it is 0.80;

[0210] when When the passage is deemed normal, the edge retains its original weight.

[0211] when If the edge is deemed invalid and closed, it will not be included in the path solution.

[0212] when When this condition is met, it is considered a restriction, and a penalty factor is applied to that edge to increase its equivalent cost.

[0213] The formulas for the edge weights and penalty factors are as follows:

[0214] For any bridge path Let the original edge weight be . The updated edge weights are obtained by using the following formula:

[0215]

[0216]

[0217] in, To update the edge weight values, The penalty magnitude parameter; The preset curvature parameters, The preset normal threshold, The preset closing threshold, Bridge path The corresponding control score, Bridge path Preset original edge weights, To control the penalty factor corresponding to the score.

[0218] Further, the step of calculating the optimal same-path path in the target region using the deterministic weighted shortest path algorithm based on the updated edge weight values ​​includes:

[0219] Set the starting target bridge node and the destination node, and establish a candidate set and a confirmed set based on the starting target bridge node and the destination node;

[0220] Calculate the cumulative cost evaluation value for each target bridge node in the candidate set;

[0221] The node with the smallest cumulative cost evaluation value is selected sequentially from the candidate set, expanded, and added to the confirmed set.

[0222] When the destination node is selected, the expansion is terminated and the target bridge node is backtracked along the confirmed set to obtain the optimal travel path.

[0223] Specifically, the minimum cumulative cost evaluation value is calculated using the following formula:

[0224]

[0225] in, The minimum cumulative cost evaluation value, Bridge path The length of the road segment Bridge path The estimated speed, Bridge path Update the edge weight values.

[0226] Furthermore, after obtaining the updated equivalent edge weights, this embodiment adopts A. The deterministic weighted shortest path algorithm solves for the optimal travel path between a given starting point and a destination.

[0227] Finally, the visualization interface shows the bridge nodes. RPI The values ​​are displayed using color mapping, with roadside areas marked by colors according to three states: normal, restricted, and closed. Figure 10 As shown. And the starting target bridge node is overlaid on the map background. v 1 to the destination node v 23 The optimal travel route is determined, providing intuitive route decision support for the dispatch of emergency rescue vehicles and supplies.

[0228] In this embodiment of the invention, the candidate set is the target bridge node other than the starting target bridge node and the destination node, and the confirmed set is the set of nodes with the smallest cumulative cost evaluation value. The confirmed set is initially an empty set. The node with the smallest evaluation value is repeatedly selected from the candidate set for expansion until the destination node is selected. Then, the optimal travel path is obtained by backtracking the confirmed target bridge nodes from the confirmed set. That is, when the destination is selected, the search is terminated and the optimal travel path is obtained by backtracking along the predecessor information.

[0229] In detail, this invention effectively compensates for the lack of accuracy of traditional models that rely solely on numerical simulation by introducing real-time data from bridge safety monitoring as input features in the node feature stage. It can dynamically update the group network evaluation results based on the structural response of the bridge under actual earthquakes, thereby achieving a more intelligent and reliable post-earthquake accessibility assessment.

[0230] like Figure 11The diagram shown is a functional block diagram of a post-earthquake accessibility assessment system for urban bridge networks provided in an embodiment of the present invention.

[0231] The post-earthquake accessibility assessment system 100 for urban bridge networks described in this invention can be installed in a processing device. Depending on the functions implemented, the post-earthquake accessibility assessment system 100 may include a bridge network diagram structure construction module 101, a model inversion module 102, a remaining accessibility scoring calculation module 103, a graph neural network training module 104, and an optimal access path solution module 105. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by an electronic device processor and perform a fixed function, stored in the memory of the electronic device.

[0232] In this embodiment, the functions of each module / unit are as follows:

[0233] The bridge network diagram construction module 101 is used to acquire basic bridge information data of key areas in the target city and construct the city bridge network diagram structure based on the basic bridge information data.

[0234] The model inversion module 102 is used to invert the three-dimensional underground structure velocity model based on the array observation data in the bridge basic information data, and generate the three-dimensional ground acceleration time history of the bridge site coordinates of each bridge based on the three-dimensional underground structure velocity model.

[0235] The remaining traffic capacity score calculation module 103 is used to calculate the structural response data under different seismic scenarios using the finite element model of the bridge corresponding to the three-dimensional ground acceleration time history, and to calculate the remaining traffic capacity score corresponding to the structural response data.

[0236] The graph neural network training module 104 is used to train a pre-built graph neural network using the structural response data and the remaining traffic capacity score as a training set to obtain a target evaluation network.

[0237] The optimal travel path solution module 105 is used to collect real-time earthquake monitoring data of the target area, use the target evaluation network to calculate the predicted remaining traffic capacity score of each target bridge node in the real-time earthquake monitoring data, and solve the optimal travel path based on the predicted remaining traffic capacity score.

[0238] The specific methods for executing the steps in each of the above modules are the same as the corresponding execution steps in the above-mentioned method for assessing the post-earthquake accessibility of urban bridge networks.

[0239] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example.

[0240] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0241] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing the post-earthquake accessibility of urban bridge networks, characterized in that, include: Acquire basic bridge information data for key areas of the target city, and construct a network diagram of the city's bridge group based on the basic bridge information data; Based on the array observation data in the bridge basic information data, the three-dimensional underground structure velocity model is inverted, and the three-dimensional ground acceleration time history at the bridge site coordinates of each bridge is generated based on the three-dimensional underground structure velocity model. The structural response data of the bridge corresponding to the three-dimensional ground acceleration time history are calculated using the finite element model of the bridge under different seismic scenarios, and the remaining traffic capacity score corresponding to the structural response data is calculated. The structural response data and the remaining traffic capacity score are used as the training set to train a pre-built graph neural network to obtain the target evaluation network. Real-time earthquake monitoring data of the target area is collected, and the predicted remaining traffic capacity score of each target bridge node in the real-time earthquake monitoring data is calculated using the target evaluation network. The optimal traffic path is then solved based on the predicted remaining traffic capacity score.

2. The method for assessing the post-earthquake accessibility of urban bridge networks as described in claim 1, characterized in that, The process of inverting the three-dimensional underground structure velocity model based on the array observation data in the bridge foundation information data includes: Cross-correlation calculations were performed on the continuous noise records between different station pairs of the array observation data to obtain the dispersion curves of multi-period surface waves; The initial shear wave velocity model is obtained by performing a deep inversion on the dispersion curve; The velocity structure of the initial shear wave velocity model is constrained and corrected, and the shear wave velocity is adjusted and smoothed to obtain a three-dimensional underground structure velocity model.

3. The method for assessing the post-earthquake accessibility of urban bridge networks as described in claim 1, characterized in that, The generation of the three-dimensional ground acceleration time history at the bridge site coordinates of each bridge based on the three-dimensional underground structure velocity model includes: Multiple earthquake scenarios and corresponding source parameters were constructed on the three-dimensional underground structure velocity model. Based on the source parameters, the three-dimensional finite difference method was used to simulate the source rupture and seismic wave propagation process on the three-dimensional underground structure velocity model to obtain simulation data. Interpolation simulations were performed on the simulation data based on the bridge site coordinates of each bridge to obtain the three-dimensional ground acceleration time history.

4. The method for assessing the post-earthquake accessibility of urban bridge networks as described in claim 1, characterized in that, The calculation of the remaining traffic capacity score corresponding to the structural response data includes: Calculate the expected value and standard deviation of the logarithmic regression of the structural response data under the different earthquake scenarios; Calculate the system-level exceedance probability of structural response data at a preset damage level under different earthquake scenarios based on the expected value and the standard deviation. The system-level exceedance probability is calculated using the following formula: in, In the context of an earthquake scenario The system-level exceedance probability is below. Earthquake scenario Next Each structural indicator in damage level The probability of exceeding the limit at that time, , These are the expected value and standard deviation of the logarithmic regression, respectively. It is the standard normal distribution function; Calculate the graded probabilities of the system-level exceedance probability under different damage levels: in, For graded probabilities, Indicating an earthquake scenario The system-level exceedance probability is below. Preset damage level The earthquake damage index below; Shape parameters of the Beta distribution , As shown in the following formula: The expected value of the Beta distribution is calculated based on the shape parameters and used as the system damage index. : Calculate the earthquake scenario using the following formula Remaining traffic capacity rating: in, In the context of an earthquake scenario The remaining traffic capacity rating below, In the context of an earthquake scenario The system-level exceedance probability is below. Indicates earthquake scenario The standard deviation of the lower structure response data.

5. The method for assessing the post-earthquake accessibility of urban bridge networks as described in claim 1, characterized in that, The step of training a pre-constructed graph neural network using the structural response data and the remaining traffic capacity score as a training set to obtain a target evaluation network includes: The structural response data and the urban bridge network structure are message-passed and neighborhood-aggregated using the graph neural network to obtain the training remaining traffic capacity score for each bridge node. Calculate the loss value between the training remaining capacity score and the remaining capacity score; The parameters of the graph neural network are updated by backpropagation based on the loss value until the loss value is less than a preset loss value threshold, thus obtaining the target evaluation network.

6. The method for assessing the post-earthquake accessibility of urban bridge networks as described in claim 1, characterized in that, The calculation of the predicted remaining traffic capacity score for each target bridge node in the real-time earthquake monitoring data using the target evaluation network includes: Construct a real-time urban bridge network structure corresponding to each target bridge node in the real-time earthquake monitoring data; Extract the real-time dynamic features and real-time static features of each target bridge node within the set time window; The real-time dynamic features and the real-time static features are combined and then used as the target bridge node features of the real-time urban bridge network structure. These features are then input into the target evaluation network to obtain the predicted remaining traffic capacity score.

7. The method for assessing the post-earthquake accessibility of urban bridge networks as described in claim 1, characterized in that, The step of solving for the optimal travel path based on the predicted remaining capacity score includes: The two end nodes of the bridge path are determined based on the target bridge node, and the minimum value of the predicted remaining capacity score corresponding to the two end nodes is used as the control score of the bridge path. The edge weights of the bridge path are updated based on the control score to obtain updated edge weight values. The updated edge weight values ​​are obtained by using the following formula: in, To update the edge weight values, The penalty magnitude parameter; The preset curvature parameters, The preset normal threshold, The preset closing threshold, Bridge path The corresponding control score, Bridge path Preset original edge weights, To control the penalty factor corresponding to the score; The optimal same-path for the target region is calculated using the deterministic weighted shortest path algorithm based on the updated edge weight values.

8. A post-earthquake accessibility assessment system for urban bridge networks, characterized in that, include: The bridge network diagram construction module is used to acquire basic bridge information data of key areas in the target city and construct the urban bridge network diagram structure based on the basic bridge information data. The model inversion module is used to invert the three-dimensional underground structure velocity model based on the array observation data in the bridge basic information data, and generate the three-dimensional ground acceleration time history of the bridge site coordinates of each bridge based on the three-dimensional underground structure velocity model. The remaining traffic capacity score calculation module is used to calculate the structural response data under different seismic scenarios using the finite element model of the bridge corresponding to the three-dimensional ground acceleration time history, and to calculate the remaining traffic capacity score corresponding to the structural response data. The graph neural network training module is used to train a pre-built graph neural network using the structural response data and the remaining traffic capacity score as a training set to obtain a target evaluation network. The optimal travel path solution module is used to collect real-time earthquake monitoring data of the target area, use the target evaluation network to calculate the predicted remaining traffic capacity score of each target bridge node in the real-time earthquake monitoring data, and solve the optimal travel path based on the predicted remaining traffic capacity score.

9. A processing device, characterized in that, It includes at least one processor and at least one memory communicatively connected to the processor, wherein: the memory stores program instructions executable by the processor, and the processor can execute the method as described in any one of claims 1-7 by invoking the program instructions.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause the computer to perform the method as described in any one of claims 1-7.

Images