Traffic network anti-seismic toughness evaluation method fusing seismic motion prediction and machine learning

Abstract

The invention provides a traffic road network anti-seismic toughness evaluation method fusing seismic motion prediction and machine learning, and the method comprises the steps: determining a probability seismic demand model (PSDM) parameter of each road section in a target road network based on an agent model, so as to determine the vulnerability of each road section; determining a road section post-earthquake function index of each road section based on the vulnerability; determining a path post-earthquake function index of each feasible path according to the post-earthquake function indexes of the plurality of road sections corresponding to each feasible path; and calculating a post-earthquake function index of each group of OD pairs according to the path post-earthquake function index of each feasible path of each group of OD pairs so as to determine a road network post-earthquake function index of the target road network, and determining an anti-seismic toughness index of the target road network based on the road network post-earthquake function index. According to the method, machine learning, PSDM and GMPE are efficiently coupled, and multi-scene, whole-network and segment-path-network three-level rapid integrated evaluation is realized on the premise of not depending on large-scale simulation.

Description

Technical Field

[0001] This invention relates to the field of transportation engineering technology, and in particular to a method for assessing the seismic resilience of transportation networks that integrates seismic motion prediction and machine learning. Background Technology

[0002] Traditional road network seismic assessments typically rely on finite element analysis or high-fidelity numerical simulations to obtain the post-earthquake structural response and traffic capacity of road sections. These methods are computationally expensive, have poor scalability, and are difficult to rapidly assess in large-scale networks with multiple seismic events and multiple paths.

[0003] Existing network layer aggregation methods mostly use shortest path or simple bottleneck rules, which make it difficult to take into account path redundancy, multi-path selection behavior, and the probabilistic and continuous nature of segment functions.

[0004] Therefore, there is still a need for an end-to-end, physically-compliant, and computationally parallelizable fast evaluation method to achieve efficient computation in large-scale scenarios. Summary of the Invention

[0005] This invention provides a method for assessing the seismic toughness of transportation networks that integrates ground motion prediction and machine learning, in order to address the shortcomings of existing technologies that lack efficient and rapid methods for assessing the seismic toughness of transportation networks, and to realize an end-to-end rapid assessment framework for the seismic toughness of transportation networks.

[0006] This invention provides a method for assessing the seismic toughness of transportation networks that integrates seismic motion prediction and machine learning, comprising: Based on the surrogate model, the probabilistic seismic demand model (PSDM) parameters of each road segment in the target road network are determined, and the vulnerability of each road segment is determined based on the PSDM parameters and the preset ground motion. The surrogate model outputs the PSDM parameters of the road segment based on the input road segment parameters. Post-earthquake functional indicators for each road segment are determined based on its vulnerability. The feasible paths for each OD pair in the target road network are determined, and the post-earthquake functional index of each feasible path is determined based on the post-earthquake functional index and penalty factor of multiple road segments corresponding to each feasible path, so that it is dominated by the worst road segment post-earthquake functional index of the corresponding road segment. The post-earthquake functional index of each OD pair is calculated by weighting the post-earthquake functional index of each feasible path and the path weight of each feasible path, so that it is dominated by the optimal post-earthquake functional index of the corresponding feasible path. The average post-earthquake functional index of all OD pairs in the target road network is used as the post-earthquake functional index of the target road network, and the seismic toughness index of the target road network is determined based on the post-earthquake functional index of the road network.

[0007] According to the present invention, a method for assessing the seismic resilience of a transportation network that integrates seismic motion prediction and machine learning, the step of determining the post-earthquake functional indicators of each road segment based on vulnerability specifically includes: Determine the speed retention rate and capacity retention rate of road segments based on their vulnerability. The post-earthquake functional indicators of the road section are determined based on the speed retention rate and the capacity retention rate.

[0008] According to the present invention, a method for assessing the seismic resilience of a transportation network that integrates seismic motion prediction and machine learning, the step of determining the post-seismic function index of each feasible path based on the post-seismic function index and penalty factor of multiple road segments corresponding to each feasible path specifically includes: The worst-case path principle function is used to calculate the first path function for each feasible path. softmin Functions are constructed to make the first path function of each feasible path approximate the post-earthquake function index of the worst road segment corresponding to the road segment; The product of the penalty factor and the function of the first path is used as the post-earthquake function index of each feasible path. The penalty factor is negatively correlated with the overlap between the current feasible path and the candidate feasible path, and the current feasible path and the candidate feasible path together constitute all feasible paths of each OD pair.

[0009] According to the present invention, a method for assessing the seismic resilience of a transportation network integrating seismic motion prediction and machine learning, the step of calculating the post-seismic function index of each OD pair based on the post-seismic function index of each feasible path and the path weight of each feasible path specifically includes: The path weight of each feasible path is calculated based on the optimal path principle function, which is based on... softmax Functions are constructed to give higher weights to feasible paths with higher post-earthquake functional indicators. The post-earthquake functional index of each feasible path in each OD pair is the weighted sum of the path post-earthquake functional index and the corresponding path weight.

[0010] According to the present invention, a method for assessing the seismic resilience of a traffic network that integrates seismic motion prediction and machine learning includes the following steps for determining feasible paths for each OD pair in the target road network: In each OD pair, retain the first type of path whose path length does not exceed a length threshold among all paths corresponding to each OD pair, wherein the length threshold is the product of the shortest path length corresponding to each OD pair and the relaxation factor. The first type of paths are filtered so that the path overlap between any two paths in the retained second type of paths does not exceed the preset overlap threshold. The second type of path is taken as the feasible path.

[0011] The method for assessing the seismic toughness of transportation networks by integrating seismic motion prediction and machine learning, provided by the present invention, further includes: The roadbed parameters and geometric parameters of each road segment in the road network are collected as road segment parameters, and numerical simulation is performed on the road segment parameters to obtain the mechanical response of the road segment parameters under seismic loading. The PSDM parameters for each group of road segment parameters are obtained by fitting the mechanical response of each group of road segment parameters under different seismic actions; Each set of road segment parameters and the corresponding PSDM parameters are used as a data set to construct a dataset, and the proxy model is trained on the dataset.

[0012] According to the present invention, a method for assessing the seismic resilience of a transportation network that integrates seismic motion prediction and machine learning is provided, wherein the preset seismic motion is generated by a seismic motion prediction equation.

[0013] This invention also provides a system for assessing the seismic toughness of transportation networks that integrates seismic motion prediction and machine learning, comprising: The parameter determination module is used to determine the probabilistic earthquake demand model (PSDM) parameters of each road segment in the target road network based on the surrogate model, and to determine the vulnerability of each road segment based on the PSDM parameters and the preset ground motion. The surrogate model outputs the PSDM parameters of the road segment based on the input road segment parameters. The road segment index determination module is used to determine the post-earthquake functional index of each road segment based on its vulnerability. The path index determination module is used to determine the feasible path of each OD pair in the target road network, and to determine the path post-earthquake function index of each feasible path based on the post-earthquake function index and penalty factor of multiple road segments corresponding to each feasible path, so that it is dominated by the worst road segment post-earthquake function index of the corresponding road segment. The path index determination module is also used to calculate the post-earthquake function index of each OD pair based on the post-earthquake function index of each feasible path and the path weight of each feasible path, so that it is dominated by the optimal post-earthquake function index of the corresponding feasible path. The road network resilience assessment module is used to take the average post-earthquake functional index of all OD pairs in the target road network as the post-earthquake functional index of the target road network, and determine the seismic resilience index of the target road network based on the post-earthquake functional index of the road network.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method for assessing the seismic toughness of a transportation network by integrating seismic motion prediction and machine learning.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for assessing the seismic toughness of a transportation network that integrates seismic motion prediction and machine learning as described above.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for assessing the seismic resilience of transportation networks by integrating seismic motion prediction and machine learning.

[0017] The present invention provides a method for assessing the seismic toughness of transportation networks by integrating ground motion prediction and machine learning. By efficiently coupling machine learning, PSDM and GMPE, it can achieve rapid integrated toughness assessment at three levels: multi-scenario, full network, and road segment-path-network, without relying on large-scale simulation. It has low computational overhead, high scalability and stability, and finally outputs a toughness index based on the integral form of function curve as the result of the road network toughness assessment. The result has high interpretability and can effectively support the formulation of repair strategies and resource optimization. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is one of the flowcharts of the method for assessing the seismic toughness of transportation networks that integrates seismic motion prediction and machine learning provided by the present invention. Figure 2 This is a schematic diagram of the resilience curve obtained by the seismic resilience assessment method for transportation networks that integrates seismic motion prediction and machine learning provided by the present invention. Figure 3 This is the second flowchart of the method for assessing the seismic resilience of transportation networks that integrates seismic motion prediction and machine learning provided by this invention. Figure 4 This is a schematic diagram of the structure of the traffic network seismic toughness assessment system that integrates seismic motion prediction and machine learning provided by the present invention. Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] The following is combined Figures 1 to 3 This invention introduces a method for assessing the seismic toughness of transportation networks that integrates seismic motion prediction and machine learning. Figure 1 As shown, it includes: Step 101: Determine the probabilistic earthquake demand model (PSDM) parameters for each road segment in the target road network based on the surrogate model, and determine the vulnerability of each road segment based on the PSDM parameters and the preset ground motion. The surrogate model outputs the PSDM parameters of the road segment based on the input road segment parameters. A surrogate model is pre-trained to predict the probabilistic seismic demand model (PSDM) parameters for each road segment based on the segment parameters. The PSDM parameters include at least the PSDM intercept. a and slope b In this embodiment, the output of the surrogate model also includes the mean squared error. σ and effectiveness r 2 .

[0022] Alternatively, the surrogate model can be trained using one or more machine learning algorithms such as neural networks, random forests, gradient boosting trees, support vector machines, and decision trees.

[0023] By inputting the road segment parameters into the trained surrogate model, a PSDM model with determined parameters for that road segment can be obtained, which is used to describe the seismic intensity parameters. IM The probabilistic relationship between the engineering requirement parameter EDP and the following: ln ( EDP )= ln ( a )+ b * ln ( IM )+ ; In the formula, To conform to a mean of 0 and a standard deviation of σ The probability distribution.

[0024] Based on this, the vulnerability curves of the corresponding road segments are generated using the parameter-determined PSDM model, and the DS values ​​for different damage states are obtained. k The transcendental probability function.

[0025] Specifically, before plotting the vulnerability curve, it is necessary to determine a seismic intensity parameter that covers the range of possible seismic intensity of the target road network. IM Sequence. Regarding the above... IM Each in the sequence IM The intensity value is calculated using a PSDM model with defined parameters and a preset damage state threshold. The conditional probability that the engineering requirement parameter EDP exceeds this threshold is then used to plot the road segment at a specific damage state threshold DS. k The vulnerability curves under different damage states are obtained by analyzing the vulnerability curves.

[0026] In this embodiment, the exceedance probability functions of three damage state thresholds DS1, DS2 and DS3 are plotted, where DS1, DS2 and DS3 are 0.037, 0.413 and 1.14, respectively.

[0027] In a preferred embodiment, in order to perform damage assessment under a specific seismic scenario, the seismic intensity index of each road segment in the target road network under any seismic scenario event is calculated based on the General Seismic Prediction Equation (GMPE) to characterize the preset seismic event.

[0028] Optionally, the ground motion scenario generation can use Chiou and Youngs' ground motion prediction equation, employing a combined model that includes magnitude scaling terms, distance attenuation terms, hanging wall effects, site nonlinearity terms, and Z1.0 depth terms to predict the actual ground motion intensity experienced by each road segment under seismic action.

[0029] The intensity index of the preset seismic motion IM Substituting the generated vulnerability curve set, the exceedance probability of the road segment under the preset seismic motion can be obtained. Corresponding to the three determined damage state thresholds, the corresponding exceedance probability P[DS≥DS] is obtained in this embodiment. k Let P1, P2 and P3 be the three groups, and satisfy 1≥P1≥P2≥P3≥0.

[0030] Step 102: Determine the post-earthquake functional indicators for each road segment based on its vulnerability. The post-earthquake traffic capacity of a road segment can be characterized by the post-earthquake traffic speed and / or flow carrying capacity. Therefore, in a feasible implementation, the post-earthquake functional index of a road segment can be defined by the speed retention rate / capacity retention rate.

[0031] Among them, the speed retention rate indicates the degree to which the traffic speed of a road segment can be retained after an earthquake compared to before the earthquake, and the capacity retention rate indicates the degree to which the traffic volume per unit time of a road segment can be retained after an earthquake compared to before the earthquake.

[0032] It is understandable that the degree of speed retention / capacity retention varies depending on the post-earthquake damage state of the road segment. Therefore, we can first determine the speed retention parameters and capacity retention parameters of the road under each damage state, and then calculate the probability of each damage state from no damage to complete damage based on the exceedance probability of each damage state. Finally, we can calculate the sum of the product of the probability of each damage state and the speed retention parameter / capacity retention parameter for each damage state as the speed retention rate and capacity retention rate of the road segment.

[0033] Based on this, the post-earthquake functional indicators of the road section are determined according to the speed retention rate / capacity retention rate, so that the higher the value of the post-earthquake functional indicator of the road section, the better the traffic function of the road section after the earthquake.

[0034] In other feasible implementations, other known methods may also be used to determine the post-earthquake functional indicators of road sections.

[0035] Step 103: Determine the feasible path for each OD pair in the target road network, and determine the post-earthquake functional index of each feasible path based on the post-earthquake functional index and penalty factor of multiple road segments corresponding to each feasible path, so that it is dominated by the worst road segment post-earthquake functional index of the corresponding road segment. A set of OD pairs (Origin-Destination) represents a set of starting points and ending points in a road network. For the same set of starting points and ending points, there will be multiple different paths in the road network. For each set of OD pairs, first screen and determine the multiple feasible paths corresponding to them, and then calculate the post-earthquake function index of the feasible paths.

[0036] It is understandable that each route consists of multiple road segments, and the post-earthquake traffic capacity of a route is determined by the road segment with the worst traffic capacity. Therefore, in a feasible implementation, the minimum value of the post-earthquake functional index of the road segment can be directly used as the basis for calculating the post-earthquake functional index of the route, and multiplied by the penalty factor to obtain the post-earthquake functional index of the route.

[0037] The penalty factor is determined based on the degree of overlap between the currently calculated path and other paths among the multiple feasible paths corresponding to OD. The higher the degree of overlap between the currently calculated path and other paths, the smaller the post-earthquake functional index of the currently calculated path.

[0038] Step 104: Calculate the post-earthquake functional index of each OD pair based on the post-earthquake functional index of each feasible path and the path weight of each feasible path, so that it is dominated by the optimal post-earthquake functional index of the corresponding feasible path. Furthermore, the post-earthquake functional indicators of each feasible path corresponding to each OD pair can be calculated, and the traffic capacity of the OD pair is determined by the path with the best traffic capacity among the multiple feasible paths.

[0039] Therefore, the post-earthquake functional indicators of each feasible path are weighted by path weight, and the weighted sum is calculated as the post-earthquake functional indicators of the OD pair. The path weight gives higher weight to paths with better post-earthquake functional indicators.

[0040] Step 105: Take the average post-earthquake functional index of all OD pairs in the target road network as the post-earthquake functional index of the target road network, and determine the seismic toughness index of the target road network based on the post-earthquake functional index of the road network.

[0041] Calculate the post-earthquake functional index of all OD pairs in the target road network, and use their average value as the post-earthquake functional index of the target road network. NF : NF = mean ({ F od}).in, F od This represents the post-earthquake functional index of the OD pair.

[0042] Based on this, by normalizing the post-earthquake functional indicators of the road network and combining them with a given repair strategy, a resilience index based on the integral form of the functional curve is used to calculate the seismic resilience of the target road network. The repair strategy determines the repair sequence for each path in the road network.

[0043] Optionally, resilience index R As shown in the following formula: ; In the formula, Indicates the preset time of earthquake occurrence. Indicates the time it takes for the repair to be completed based on the repair strategy; This indicates the post-earthquake functional indicators of the road network after a pre-settled seismic event. Indicates repair based on repair strategy t Post-earthquake functional indicators of the road network at any given time; its visualization results are as follows: Figure 2 As shown. It is understandable that as the timing of repair changes based on the repair strategy, the path to completion of the repair also changes accordingly.

[0044] This resilience index can be used to quickly calculate the seismic resilience index of the target road network.

[0045] This invention achieves rapid, integrated resilience assessment across multiple scenarios, the entire network, and three levels (segment-path-network) without relying on large-scale simulations by efficiently coupling machine learning, PSDM, and GMPE. It features low computational overhead, high scalability and stability, and ultimately outputs a resilience index based on the integral form of the function curve as the road network resilience assessment result. The result is highly interpretable and can effectively support the formulation of repair strategies and resource optimization.

[0046] In the method for assessing the seismic resilience of transportation networks that integrates seismic motion prediction and machine learning, the step of determining the post-earthquake functional indicators of each road segment based on vulnerability specifically includes: Determine the speed retention rate and capacity retention rate of road segments based on their vulnerability. Optionally, speed retention rate and capacity retention rate Calculate according to the following formula: ; ; In the formula, Indicates the first i Damage-like state, Indicates the first i The probability of a damage-like state occurring. For the velocity retention rate parameter, This is the capacity retention rate parameter.

[0047] Specifically, the three damage state thresholds will result in four damage states: ; ; ; ; , , and These correspond to the probabilities of no damage, minor damage, moderate damage, and severe damage, respectively.

[0048] Optionally, corresponding to the above four damage states, the velocity retention parameters are 1.0, 0.75, 0.50 and 0.0, respectively; and the capacity retention parameters are 1.0, 0.70, 0.30 and 0.0, respectively.

[0049] The post-earthquake functional indicators of the road section are determined based on the speed retention rate and the capacity retention rate.

[0050] In this embodiment, the post-earthquake functional indicators of the road section are determined jointly based on the speed retention rate and the capacity retention rate.

[0051] Optionally, the mean, weighted average, or harmonic average of the speed retention rate and capacity retention rate can be calculated to determine the post-earthquake functional indicators of the road section.

[0052] Preferably, the post-earthquake functional indicators of the road section are: ; In the formula, SF For post-earthquake functional indicators of road sections, For speed weighting, For speed retention rate, For capacity weights, The capacity retention rate is defined as follows: the speed retention rate is determined based on the overtaking probability of the road segment under different damage states and the speed retention parameter under the corresponding damage states; the capacity retention rate is determined based on the overtaking probability of the road segment under different damage states and the capacity retention parameter under the corresponding damage states.

[0053] Specifically, the post-earthquake functional indicators of the road section in this implementation method SF Determined based on the harmonic mean of velocity retention rate and capacity retention rate.

[0054] Optionally, velocity weights and capacity weight This is an empirical value, and in this implementation method... and Both are 0.5.

[0055] Using the above method, the speed retention rate and capacity retention rate of the road segment can be calculated from the exceedance probabilities P1, P2, and P3 for each loss state. Then, based on the speed retention rate and capacity retention rate, the post-earthquake functional indicators of the road segment can be determined. SF ,make SF The larger the section, the more completely its post-earthquake functionality is preserved.

[0056] In the method for assessing the seismic resilience of transportation networks that integrates seismic motion prediction and machine learning, the step of determining the post-seismic functional index of each feasible path based on the post-seismic functional index and penalty factor of multiple road segments corresponding to each feasible path specifically includes: The worst-case path principle function is used to calculate the first path function for each feasible path. softmin Functions are constructed to make the first path function of each feasible path approximate the post-earthquake function index of the worst road segment corresponding to the road segment; In this implementation, the principles of connectivity, traffic flow, network traffic, and intelligent simulation are used to analyze the road segments included in each path. SF Aggregation is performed according to the worst-case path principle function to obtain the first path function. PF : ,in These are control parameters.

[0057] In the above formula softmin The worst-case road segment principle function defined for this implementation is shown in the following formula: ; In the formula, m express Find the minimum value and crop the output to [0,1]. This is the set of post-earthquake functional indicators for all road segments corresponding to the path. Take 0.01.

[0058] The smaller the control parameter, PF The closer the path is to being dominated by the segment with the worst post-earthquake function, the larger the control parameter, and the closer the power factor (PF) is to the average of multiple segments. Through the above method, this implementation method ensures the first path function of the path. PF The results are derived from the worst-performing post-earthquake functional index of the corresponding road segment. m The dominant factor is the post-earthquake functional index of the route, considering only the impact of road segments on the route.

[0059] In other feasible implementations, the worst-case path principle function may also be in other forms. softmin A variant function, or any other function of any form, is required that makes the first path function of each feasible path approximate the worst post-earthquake function index of the corresponding road segment.

[0060] The product of the penalty factor and the function of the first path is used as the post-earthquake function index of each feasible path. The penalty factor is negatively correlated with the overlap between the current feasible path and the candidate feasible path, and the current feasible path and the candidate feasible path together constitute all feasible paths of each OD pair.

[0061] Furthermore, each OD pair corresponds to multiple feasible paths. The currently calculated path is taken as the current feasible path, and the other paths in the feasible path are taken as candidate feasible paths.

[0062] Optionally, calculate the Jaccard overlap between the current feasible path and the candidate feasible paths. : ; In the formula, p Indicates the currently feasible path. q Indicates candidate feasible paths, and These represent the sets of road segments for the current feasible path and the sets of candidate feasible paths, respectively.

[0063] For each currently feasible path, the penalty factor is determined by its maximum overlap with candidate feasible paths.

[0064] Based on this, a uniqueness penalty factor is applied to each feasible path: .

[0065] in, To determine the penalty intensity, a value of 10 is used in this embodiment. After weighting the first path function using the penalty factor, the post-earthquake function index of the feasible path is obtained. .

[0066] The post-earthquake functional index obtained by the above method takes into account the impact of path overlap on the post-earthquake function of OD pairs, and can serve as a better basis for calculating the post-earthquake functional index of OD pairs.

[0067] In the method for assessing the seismic resilience of transportation networks that integrates seismic motion prediction and machine learning, the step of calculating the post-seismic function index of each OD pair based on the post-seismic function index of each feasible path and the path weight of each feasible path specifically includes: The path weight of each feasible path is calculated based on the optimal path principle function, which is based on... softmax Functions are constructed to give higher weights to feasible paths with higher post-earthquake functional indicators. In this embodiment, path weight . For a set of feasible paths for a group of OD pairs, the post-earthquake functional indicators are used. This refers to the temperature parameter.

[0068] In the above formula softmax The optimal path principle function defined in this implementation: ; In the formula, Post-earthquake functional indicators representing currently feasible paths. Post-earthquake functional indicators representing feasible paths.

[0069] Optionally, Take 0.05.

[0070] In other feasible implementations, the optimal path principle function may also be in other forms. softmax Variant functions, or other functions of any form, can be used to assign higher weights to feasible paths with higher post-earthquake functional indicators.

[0071] The post-earthquake functional index of each feasible path in each OD pair is the weighted sum of the path post-earthquake functional index and the corresponding path weight.

[0072] Based on this, the post-earthquake functional indicators of each OD pair It can be represented as: ; Using the above method, the post-earthquake functional index of each OD pair can be calculated based on the post-earthquake functional index of all feasible paths for each OD pair.

[0073] In the method for assessing the seismic toughness of traffic networks that integrates seismic motion prediction and machine learning, the step of determining the feasible path for each OD pair in the target road network includes: In each OD pair, retain the first type of path whose path length does not exceed a length threshold among all paths corresponding to each OD pair, wherein the length threshold is the product of the shortest path length corresponding to each OD pair and the relaxation factor. Understandably, considering all paths corresponding to each OD pair as feasible paths would require significant computational resources. Therefore, this implementation uses the shortest path length corresponding to each OD pair. L min Based on this, all paths are filtered.

[0074] Specifically, determine the relaxation factor. α The product of the relaxation factor and the shortest path length is used as a length threshold to filter all paths, and the paths whose length does not exceed the length threshold are retained as the first type of path.

[0075] Optionally, the relaxation factor can be an empirical value, which is 1.3 in this embodiment.

[0076] The first type of paths are filtered so that the path overlap between any two paths in the retained second type of paths does not exceed the preset overlap threshold. The second type of path is taken as the feasible path.

[0077] Further, determine the preset overlap threshold. β The overlap threshold is an empirical value, and in this embodiment it is taken as 0.8.

[0078] The first type of path is cyclically filtered using an overlap threshold until the overlap between any two paths in the remaining paths does not exceed the preset overlap threshold, thus obtaining the second type of path as a feasible path.

[0079] The above method allows for path selection for each OD pair before evaluating the seismic resilience of the target road network. Combined with parallel batch processing, this can significantly reduce computational overhead and improve scalability and stability.

[0080] The present invention, which integrates seismic motion prediction and machine learning for assessing the seismic toughness of transportation networks, further includes: The roadbed parameters and geometric parameters of each road segment in the road network are collected as road segment parameters, and numerical simulation is performed on the road segment parameters to obtain the mechanical response of the road segment parameters under seismic loading. In this embodiment, data is first collected in order to train the proxy model.

[0081] The collected data includes the roadbed parameters and geometric parameters of each road segment in the road network.

[0082] Optionally, the subgrade parameters include the density, cohesion, internal friction angle, maximum shear modulus of the embankment soil, as well as the height, slope angle, and top width of the embankment.

[0083] Optionally, the geometric parameters include roadbed height, slope angle, and top width.

[0084] Based on the road network topology, a corresponding three-dimensional geological model or equivalent parameterized model is constructed using the collected road segment parameters.

[0085] Based on this, a corresponding mechanical library is constructed through extensive finite element simulations or other numerical methods.

[0086] In one specific implementation, the collected road segment parameters are subjected to Latin hypercube sampling to establish 1,000 sets of embankment finite element models, and then 100 suitable ground motion data are selected to form 100,000 sets of road segment parameter-ground motion pairs.

[0087] Finite element analysis was used to simulate each road segment parameter-seismic motion pair to obtain the mechanical response of this paired road segment parameters under the corresponding seismic motion.

[0088] The PSDM parameters for each group of road segment parameters are obtained by fitting the mechanical response of each group of road segment parameters under different seismic actions; Each set of road segment parameters and the corresponding PSDM parameters are used as a data set to construct a dataset, and the proxy model is trained on the dataset.

[0089] Based on this, for each set of road segment parameters, 100 corresponding mechanical responses can be obtained. Log-linear regression can be performed on the data points of these 100 mechanical responses to fit the PSDM parameters corresponding to this set of road segment parameters, resulting in a set of road segment parameter-PSDM parameter pairs, which can be used as a data point in the dataset used to train the surrogate model.

[0090] Optionally, the XGBoost algorithm can be used to train on the obtained dataset, with road segment parameters as input and the corresponding PSDM parameters as labels, to obtain the surrogate model used to evaluate road network resilience.

[0091] In one specific embodiment, a flowchart of the present invention is as follows: Figure 3 As shown, a surrogate model is trained based on the subgrade parameters of the road segment. The PSDM parameters of each road segment in the target road network are determined based on the surrogate model. The vulnerability curve of the road segment is determined based on the PSDM model with determined parameters and the preset ground motion simulated by GMPE. The probability of the road segment under different damage states is obtained, and the post-earthquake functional index of the road segment is determined. On this basis, the post-earthquake functional index of the path and the post-earthquake functional index of the road network are determined in sequence. Finally, the resilience evaluation result of the target road network is obtained based on the repair strategy adopted.

[0092] The following describes the seismic resilience assessment system for transportation networks that integrates ground motion prediction and machine learning provided by this invention. The seismic resilience assessment system for transportation networks that integrates ground motion prediction and machine learning described below can be referred to in correspondence with the seismic resilience assessment method for transportation networks that integrates ground motion prediction and machine learning described above.

[0093] like Figure 4 As shown, the traffic network seismic resilience assessment system integrating seismic motion prediction and machine learning of the present invention includes a parameter determination module 401, a road segment index determination module 402, a route index determination module 403, and a road network resilience assessment module 404.

[0094] The parameter determination module 401 is used to determine the probabilistic earthquake demand model (PSDM) parameters of each road segment in the target road network based on the surrogate model, and to determine the vulnerability of each road segment based on the PSDM parameters and the preset ground motion. The surrogate model outputs the PSDM parameters of the road segment based on the input road segment parameters. A surrogate model is pre-trained to predict the probabilistic seismic demand model (PSDM) parameters for each road segment based on the segment parameters. The PSDM parameters include at least the PSDM intercept. a and slope b In this embodiment, the output of the surrogate model also includes the mean squared error. σ and effectiveness r 2 .

[0095] Alternatively, the surrogate model can be trained using one or more machine learning algorithms such as neural networks, random forests, gradient boosting trees, support vector machines, and decision trees.

[0096] By inputting the road segment parameters into the trained surrogate model, a PSDM model with determined parameters for that road segment can be obtained, which is used to describe the seismic intensity parameters. IM The probabilistic relationship between the engineering requirement parameter EDP and the following: ln ( EDP )= ln ( a )+ b *ln ( IM )+ ; In the formula, To conform to a mean of 0 and a standard deviation of σ The probability distribution.

[0097] Based on this, the vulnerability curves of the corresponding road segments are generated using the parameter-determined PSDM model, and the DS values ​​for different damage states are obtained. k The transcendental probability function.

[0098] Specifically, before plotting the vulnerability curve, it is necessary to determine a seismic intensity parameter that covers the range of possible seismic intensity of the target road network. IM Sequence. Regarding the above... IM Each in the sequence IM The intensity value is calculated using a PSDM model with defined parameters and a preset damage state threshold. The conditional probability that the engineering requirement parameter EDP exceeds this threshold is then used to plot the road segment at a specific damage state threshold DS. k The vulnerability curves under different damage states are obtained by analyzing the vulnerability curves.

[0099] In this embodiment, the exceedance probability functions of three damage state thresholds DS1, DS2 and DS3 are plotted, where DS1, DS2 and DS3 are 0.037, 0.413 and 1.14, respectively.

[0100] Furthermore, in order to conduct loss assessment under specific earthquake scenarios, this embodiment calculates the seismic intensity index of each road segment in the target road network under any seismic scenario event based on the General Ground Motion Prediction Equation (GMPE) to characterize the preset seismic event.

[0101] Among them, the generation of ground motion scenarios can use the ground motion prediction equation of Chiou and Youngs, and adopt a combined model that includes magnitude scaling term, distance attenuation term, hanging wall effect, site nonlinear term and Z1.0 depth term to predict the actual ground motion intensity of each road section under the action of earthquake.

[0102] The intensity index of the preset seismic motion IM Substituting the generated vulnerability curve set, the exceedance probability of the road segment under the preset seismic motion can be obtained. Corresponding to the three determined damage state thresholds, the corresponding exceedance probability P[DS≥DS] is obtained in this embodiment. k Let P1, P2 and P3 be the three groups, and satisfy 1≥P1≥P2≥P3≥0.

[0103] The road segment index determination module 402 is used to determine the post-earthquake functional index of each road segment based on its vulnerability. The post-earthquake traffic capacity of a road segment can be characterized by the post-earthquake traffic speed and / or flow carrying capacity. Therefore, in a feasible implementation, the post-earthquake functional index of a road segment can be defined by the speed retention rate / capacity retention rate.

[0104] Among them, the speed retention rate indicates the degree to which the traffic speed of a road segment can be retained after an earthquake compared to before the earthquake, and the capacity retention rate indicates the degree to which the traffic volume per unit time of a road segment can be retained after an earthquake compared to before the earthquake.

[0105] It is understandable that the degree of speed retention / capacity retention varies depending on the post-earthquake damage state of the road segment. Therefore, we can first determine the speed retention parameters and capacity retention parameters of the road under each damage state, and then calculate the probability of each damage state from no damage to complete damage based on the exceedance probability of each damage state. Finally, we can calculate the sum of the probability of each damage state and the product of the speed retention parameter / capacity retention parameter for each damage state as the speed retention rate and capacity retention rate of the road segment.

[0106] Based on this, the post-earthquake functional indicators of the road section are determined according to the speed retention rate / capacity retention rate, so that the higher the value of the post-earthquake functional indicator of the road section, the better the traffic function of the road section after the earthquake.

[0107] In other feasible implementations, other known methods may also be used to determine the post-earthquake functional indicators of road sections.

[0108] The path index determination module 403 is used to determine the feasible path of each OD pair in the target road network, and to determine the path post-earthquake function index of each feasible path based on the post-earthquake function index and penalty factor of multiple road segments corresponding to each feasible path, so that it is dominated by the worst road segment post-earthquake function index of the corresponding road segment. A set of OD pairs (Origin-Destination) represents a set of starting points and ending points in a road network. For the same set of starting points and ending points, there will be multiple different paths in the road network. For each set of OD pairs, first screen and determine the multiple feasible paths corresponding to them, and then calculate the post-earthquake function index of the feasible paths.

[0109] It is understandable that each route consists of multiple road segments, and the post-earthquake traffic capacity of a route is determined by the road segment with the worst traffic capacity. Therefore, in a feasible implementation, the minimum value of the post-earthquake functional index of the road segment can be directly used as the basis for calculating the post-earthquake functional index of the route, and multiplied by the penalty factor to obtain the post-earthquake functional index of the route.

[0110] The penalty factor is determined based on the degree of overlap between the currently calculated path and other paths among the multiple feasible paths corresponding to OD. The higher the degree of overlap between the currently calculated path and other paths, the smaller the post-earthquake functional index of the currently calculated path.

[0111] The path index determination module 403 is further configured to calculate the post-earthquake function index of each OD pair based on the post-earthquake function index of each feasible path and the path weight of each feasible path, so that it is dominated by the optimal post-earthquake function index of the corresponding feasible path. Furthermore, the post-earthquake functional indicators of each feasible path corresponding to each OD pair can be calculated, and the traffic capacity of the OD pair is determined by the path with the best traffic capacity among the multiple feasible paths.

[0112] Therefore, the post-earthquake functional indicators of each feasible path are weighted by path weight, and the weighted sum is calculated as the post-earthquake functional indicators of the OD pair. The path weight gives higher weight to paths with better post-earthquake functional indicators.

[0113] The road network resilience assessment module 404 is used to take the average post-earthquake functional index of all OD pairs in the target road network as the post-earthquake functional index of the target road network, and determine the seismic resilience index of the target road network based on the post-earthquake functional index of the road network.

[0114] Calculate the post-earthquake functional index of all OD pairs in the target road network, and use their average value as the post-earthquake functional index of the target road network. NF : NF = mean ({ F od}).in, F od This represents the post-earthquake functional index of the OD pair.

[0115] Based on this, by normalizing the post-earthquake functional indicators of the road network and combining them with a given repair strategy, a resilience index based on the integral form of the functional curve is used to calculate the seismic resilience of the target road network. The repair strategy determines the repair sequence for each path in the road network.

[0116] Optionally, resilience index R As shown in the following formula: ; In the formula, Indicates the preset time of earthquake occurrence. Indicates the time it takes for the repair to be completed based on the repair strategy; This indicates the post-earthquake functional indicators of the road network after a pre-settled seismic event. Indicates repair based on repair strategy tPost-earthquake functional indicators of the road network at any given time; its visualization results are as follows: Figure 2 As shown. It is understandable that as the timing of repair changes based on the repair strategy, the path to completion of the repair also changes accordingly.

[0117] This resilience index can be used to quickly calculate the seismic resilience index of the target road network.

[0118] This invention achieves rapid, integrated resilience assessment across multiple scenarios, the entire network, and three levels (segment-path-network) without relying on large-scale simulations by efficiently coupling machine learning, PSDM, and GMPE. It features low computational overhead, high scalability and stability, and ultimately outputs a resilience index based on the integral form of the function curve as the road network resilience assessment result. The result is highly interpretable and can effectively support the formulation of repair strategies and resource optimization.

[0119] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530, and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a traffic network seismic resilience assessment method that integrates ground motion prediction and machine learning. This method includes: determining the probabilistic earthquake demand model (PSDM) parameters of each road segment in the target road network based on a surrogate model, and determining the vulnerability of each road segment based on the PSDM parameters and a preset ground motion, wherein the surrogate model outputs the PSDM parameters of the road segment based on the input road segment parameters; determining the post-earthquake functional index of each road segment based on the vulnerability; determining the feasible path for each OD pair in the target road network, and determining the feasible path based on each feasible path. The post-earthquake functional indicators and penalty factors for multiple road segments are used to determine the post-earthquake functional indicators for each feasible path, so that they are dominated by the worst road segment post-earthquake functional indicators of the corresponding road segment. The post-earthquake functional indicators for each OD pair are calculated by weighting the post-earthquake functional indicators of each feasible path and the path weight of each feasible path, so that they are dominated by the best path post-earthquake functional indicators of the corresponding feasible path. The average post-earthquake functional indicators of all OD pairs in the target road network are used as the road network post-earthquake functional indicators of the target road network, and the seismic toughness index of the target road network is determined based on the road network post-earthquake functional indicators.

[0120] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0121] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the above-described method for assessing the seismic resilience of a traffic network by integrating seismic motion prediction and machine learning. This method includes: determining the probabilistic earthquake demand model (PSDM) parameters of each road segment in the target road network based on a surrogate model, and determining the vulnerability of each road segment based on the PSDM parameters and a preset seismic motion, wherein the surrogate model outputs the PSDM parameters of the road segment based on the input road segment parameters; and determining the post-earthquake power utilization of each road segment based on the vulnerability. The seismic performance indicators are as follows: Feasible paths for each OD pair in the target road network are determined, and the post-earthquake functional index of each feasible path is determined based on the post-earthquake functional index and penalty factor of multiple road segments corresponding to each feasible path, so that it is dominated by the worst road segment post-earthquake functional index of the corresponding road segment; the post-earthquake functional index of each OD pair is calculated by weighting the post-earthquake functional index of each feasible path and the path weight of each feasible path, so that it is dominated by the optimal path post-earthquake functional index of the corresponding feasible path; the average post-earthquake functional index of all OD pairs in the target road network is used as the road network post-earthquake functional index of the target road network, and the seismic toughness index of the target road network is determined based on the road network post-earthquake functional index.

[0122] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for assessing the seismic resilience of a transportation network by integrating ground motion prediction and machine learning. This method includes: determining the probabilistic earthquake demand model (PSDM) parameters for each road segment in the target road network based on a surrogate model, and determining the vulnerability of each road segment based on the PSDM parameters and a preset ground motion, wherein the surrogate model outputs the PSDM parameters for the road segment based on the input road segment parameters; determining the post-earthquake functional index of each road segment based on the vulnerability; and determining the vulnerability index of each road segment in the target road network. Feasible paths for each OD pair are identified, and the post-earthquake functional index of each feasible path is determined based on the post-earthquake functional index and penalty factor of multiple road segments corresponding to each feasible path, so that it is dominated by the worst road segment post-earthquake functional index of the corresponding road segment. The post-earthquake functional index of each OD pair is calculated by weighting the post-earthquake functional index of each feasible path and the path weight of each feasible path, so that it is dominated by the best path post-earthquake functional index of the corresponding feasible path. The average post-earthquake functional index of all OD pairs in the target road network is used as the road network post-earthquake functional index of the target road network, and the seismic toughness index of the target road network is determined based on the road network post-earthquake functional index.

[0123] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0124] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; 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; and these 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 seismic toughness of transportation networks by integrating seismic motion prediction and machine learning, characterized in that, include: Based on the surrogate model, the probabilistic seismic demand model (PSDM) parameters of each road segment in the target road network are determined, and the vulnerability of each road segment is determined based on the PSDM parameters and the preset ground motion. The surrogate model outputs the PSDM parameters of the road segment based on the input road segment parameters. Post-earthquake functional indicators for each road segment are determined based on its vulnerability. The feasible paths for each OD pair in the target road network are determined, and the post-earthquake functional index of each feasible path is determined based on the post-earthquake functional index and penalty factor of multiple road segments corresponding to each feasible path, so that it is dominated by the worst road segment post-earthquake functional index of the corresponding road segment. The post-earthquake functional index of each OD pair is calculated by weighting the post-earthquake functional index of each feasible path and the path weight of each feasible path, so that it is dominated by the optimal post-earthquake functional index of the corresponding feasible path. The average post-earthquake functional index of all OD pairs in the target road network is used as the post-earthquake functional index of the target road network, and the seismic toughness index of the target road network is determined based on the post-earthquake functional index of the road network.

2. The method for assessing the seismic toughness of transportation networks by integrating seismic motion prediction and machine learning as described in claim 1, characterized in that, The steps for determining the post-earthquake functional indicators of each road segment based on vulnerability specifically include: Determine the speed retention rate and capacity retention rate of road segments based on their vulnerability. The post-earthquake functional indicators of the road section are determined based on the speed retention rate and the capacity retention rate.

3. The method for assessing the seismic toughness of transportation networks by integrating seismic motion prediction and machine learning as described in claim 1, characterized in that... The step of determining the post-earthquake functional index of each feasible path based on the post-earthquake functional index and penalty factor of multiple road segments corresponding to each feasible path specifically includes: The worst-case path principle function is used to calculate the first path function for each feasible path. softmin Functions are constructed to make the first path function of each feasible path approximate the post-earthquake function index of the worst road segment corresponding to the road segment; The product of the penalty factor and the function of the first path is used as the post-earthquake function index of each feasible path. The penalty factor is negatively correlated with the overlap between the current feasible path and the candidate feasible path, and the current feasible path and the candidate feasible path together constitute all feasible paths of each OD pair.

4. The method for assessing the seismic toughness of transportation networks by integrating seismic motion prediction and machine learning as described in claim 1, characterized in that, The step of calculating the post-earthquake functional index of each OD pair based on the post-earthquake functional index of each feasible path and the path weight of each feasible path specifically includes: The path weight of each feasible path is calculated based on the optimal path principle function, which is based on... softmax Functions are constructed to give higher weights to feasible paths with higher post-earthquake functional indicators. The post-earthquake functional index of each feasible path in each OD pair is the weighted sum of the path post-earthquake functional index and the corresponding path weight.

5. The method for assessing the seismic toughness of transportation networks by integrating seismic motion prediction and machine learning according to any one of claims 1-4, characterized in that, The steps for determining feasible paths for each OD pair in the target road network include: In each OD pair, retain the first type of path whose path length does not exceed a length threshold among all paths corresponding to each OD pair, wherein the length threshold is the product of the shortest path length corresponding to each OD pair and the relaxation factor. The first type of paths are filtered so that the path overlap between any two paths in the retained second type of paths does not exceed the preset overlap threshold. The second type of path is taken as the feasible path.

6. The method for assessing the seismic toughness of transportation networks by integrating seismic motion prediction and machine learning according to any one of claims 1-4, characterized in that, Also includes: The roadbed parameters and geometric parameters of each road segment in the road network are collected as road segment parameters, and numerical simulation is performed on the road segment parameters to obtain the mechanical response of the road segment parameters under seismic loading. The PSDM parameters for each group of road segment parameters are obtained by fitting the mechanical response of each group of road segment parameters under different seismic actions; Each set of road segment parameters and the corresponding PSDM parameters are used as a data set to construct a dataset, and the proxy model is trained on the dataset.

7. The method for assessing the seismic toughness of transportation networks by integrating seismic motion prediction and machine learning according to any one of claims 1-4, characterized in that, The preset ground motion is generated by the ground motion prediction equation.

8. A system for assessing the seismic toughness of transportation networks that integrates seismic motion prediction and machine learning, characterized in that, include: The parameter determination module is used to determine the probabilistic earthquake demand model (PSDM) parameters of each road segment in the target road network based on the surrogate model, and to determine the vulnerability of each road segment based on the PSDM parameters and the preset ground motion. The surrogate model outputs the PSDM parameters of the road segment based on the input road segment parameters. The road segment index determination module is used to determine the post-earthquake functional index of each road segment based on its vulnerability. The path index determination module is used to determine the feasible path of each OD pair in the target road network, and to determine the path post-earthquake function index of each feasible path based on the post-earthquake function index and penalty factor of multiple road segments corresponding to each feasible path, so that it is dominated by the worst road segment post-earthquake function index of the corresponding road segment. The path index determination module is also used to calculate the post-earthquake function index of each OD pair based on the post-earthquake function index of each feasible path and the path weight of each feasible path, so that it is dominated by the optimal post-earthquake function index of the corresponding feasible path. The road network resilience assessment module is used to take the average post-earthquake functional index of all OD pairs in the target road network as the post-earthquake functional index of the target road network, and determine the seismic resilience index of the target road network based on the post-earthquake functional index of the road network.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method for assessing the seismic resilience of transportation networks that integrates seismic motion prediction and machine learning as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for assessing the seismic resilience of transportation networks that integrates seismic motion prediction and machine learning as described in any one of claims 1 to 6.

Images