CN-121997009-A - Large-scale urban bridge tunnel network importance rapid ordering method based on mixed graph attention network

CN121997009ACN 121997009 ACN121997009 ACN 121997009ACN-121997009-A

Abstract

The invention discloses a method for rapidly sequencing the importance of a large-scale urban bridge tunnel network based on a mixed graph attention network, which comprises the following steps of firstly, defining monomer importance indexes considering the structural failure probability and economic losses generated in safety, society, function and environmental dimensions after failure; the method comprises the steps of calculating total travel quantity and total arrival quantity of each node of a bridge and tunnel network and travel requirements among nodes, realizing initial traffic flow distribution based on a Frank-Wolfe algorithm, generating a reassigned traffic flow manufacturing data set by destroying edges which possibly fail in the bridge and tunnel network one by one, designing a mixed graph attention network, training the mixed graph attention network designed in the step five, and rapidly sequencing the importance of the large-scale urban bridge and tunnel network by using the trained mixed graph attention network. The method solves the defect of low efficiency caused by repeated solving of the high-dimensional optimization problem in the calculation of the importance index of the bridge-tunnel network monomer.

Inventors

LI SHUNLONG
WANG ANDONG
GUO YAPENG
CUI HONGTAO

Assignees

哈尔滨工业大学

Dates

Publication Date: 20260508
Application Date: 20260116

Claims (9)

1. A method for rapidly sequencing importance of a large-scale urban bridge and tunnel network based on a mixed graph attention network is characterized by comprising the following steps: step one, defining monomer importance indexes considering the structural failure probability and economic losses generated in safety, society, function and environmental dimensions after failure; Step two, calculating the total travel amount, the total arrival amount and the travel demand among nodes of each node of the bridge tunnel network; Step three, realizing initial traffic flow distribution based on Frank-Wolfe algorithm; Step four, generating a reassigned traffic flow making data set by destroying the sides which possibly fail in the bridge-tunnel network one by one; step five, designing a mixed graph attention network, and modeling a topological association mode between a bridge-tunnel network side capacity reduction vector and a flow redistribution vector; Training the mixed graph attention network designed in the step five, and rapidly sequencing the importance of the large-scale urban bridge tunneling network by using the trained mixed graph attention network.
2. The method for rapidly sequencing importance of a large-scale urban bridge-tunnel network based on a mixed graph attention network according to claim 1, wherein the monomer structure importance index is calculated by the following formula: in the formula, Is of a single structure Is used for the importance index of (a), Is of a single structure Is used for the failure probability of the (c) in the (c), Is of a single structure The economic loss caused by the failure of the device, In order to be a safety dimension economic loss, As an economic loss in the social dimension, In order to be a functional dimension economic loss, Economic losses are environmental dimensions.
3. The method for rapidly sequencing importance of large-scale urban bridge tunneling network based on mixed graph attention network according to claim 2, characterized in that said security dimension economic loss Calculated from the following formula: in the formula, Is the construction cost of a unit area structure, And The width and the length of the plane of the single structure are respectively; Economic loss of said social dimension Calculated from the following formula: in the formula, Is of a single structure under normal operation The daily average traffic volume of the road on which the traffic is located, Is the flow rate of Time monomer structure The speed of the traffic on the road where the road is located, And Is the average number of passengers of a unit passenger car and a truck, Is of a single structure The trucks on the road account for the total traffic ratio, For the personnel death reimbursement amount standard, For the proportion of the population with the age group of [ i, i+1), Compensating annual standard for population of the age group [ i, i+1), wherein i is age years; economic loss of the functional dimension Calculated from the following formula: in the formula, In order to be able to bypass the time costs additionally, In order to be a cost of the distance, Is of a single structure The number of detouring days due to failure, And The average unit time value of the passenger car and the freight car respectively, And The average unit distance value of the passenger car and the freight car respectively, And Respectively of a single structure Front and rear edge of failure Is used for the daily average traffic volume of (1), Is a side The flow rate of (2) is The time of passage of the time-dependent traffic, Is a side Is provided for the length of (a), Is a bridge-tunnel network edge set; economic loss of the environmental dimension Calculated from the following formula: in the formula, For the structural carbon-emission costs, In order to additionally bypass the carbon-carbon emissions costs, And Respectively market carbon unit price and unit area structural carbon emission, And Average carbon emissions per unit distance for passenger and freight cars, respectively.
4. The method for rapidly sequencing importance of a large-scale urban bridge and tunnel network based on a mixed graph attention network according to claim 1, wherein in the second step, the total travel amount of each node of the bridge and tunnel network is calculated according to the resident population of the administrative district (county) where the node is located, the total travel amount of each node is equal to the total arrival amount, the resident population of the administrative district (county) where the node is located is divided by the number of nodes falling on the administrative district (county), and the travel demand among the nodes is calculated according to an attraction model, and the specific formula is as follows: in the formula, Is a node Sum node The amount of attraction between the two, And Respectively nodes Total departure and node of (a) Is used to determine the total arrival amount of the (c), Is a node 、 The distance between the two plates is set to be equal, Is a super-parameter of the gravitational model, And To meet the constraint factors of travel and arrival constraints.
5. The method for rapidly sequencing importance of large-scale urban bridge and tunnel networks based on mixed graph attention network according to claim 3, wherein in the third step, the optimization objective and constraint condition formula of initial traffic flow distribution are as follows: in the formula, Is a vector consisting of the flow rates of all sides, Is a path Above the node 、 For the traffic to go out and to the place, As a set of paths, For 0-1 coefficient, if the path Through the edge The value is 1 and vice versa is 0.
6. The method for rapidly sequencing importance of large-scale urban bridge tunneling network based on mixed graph attention network according to claim 1, wherein in said step three, the main implementation procedure of Frank-Wolfe algorithm is as follows: The first step, initializing, to make all the travel time of all sides be the travel time under the free stream state, namely On the basis, all-in-all AON allocation is carried out to obtain the road section flow Order-making ; Step two, updating travel time, namely substituting a BPR formula to update travel time of each side on the basis of initializing acquired flow, namely ; Third step, determining the descending direction, namely, at the time after updating On the basis of (a) performing AON allocation to obtain auxiliary traffic Subtracting the road section flow from the auxiliary flow to obtain a descending direction; Fourth step, determining iteration step length, solving Obtaining the optimal step length ; Fifth step, moving, making ; Sixth, checking convergence if meeting the convergence condition , If the accuracy index is the preset accuracy index, the algorithm is terminated, and at the moment I.e. traffic flow, otherwise And returning to the second step of calculation again until the convergence condition is met.
7. The method for rapidly sequencing importance of large-scale urban bridge and tunnel networks based on mixed graph attention network according to claim 1, wherein in the fourth step, the data set is made by reducing the capacities of the sides which may fail in the bridge and tunnel networks one by using random capacity reduction coefficients, and the specific formula is as follows: in the formula, And Respectively, the structure of the rear axle tunnel monomer under normal operation and failure The capacity of the side where the water is located, Is a reduction coefficient.
8. The method for rapidly sequencing importance of large-scale urban bridge and tunnel networks based on the mixed graph attention network according to claim 1, wherein in the fourth step, the mixed graph attention network takes bridge and tunnel network side capacity reduction coefficient vectors as input and traffic redistribution normalization vectors at each side as output, and the structure comprises a node-side characteristic transformation layer, a full connection layer and a mixed graph attention block, specifically: L0 layer, node-edge feature transformation layer, input data scale size is The two-dimensional data represent batches, edges, respectively, the layers operating in an edge dimension, and an associated matrix Is to convert the side information into node information, and output data scale is , The total number of nodes in the bridge tunneling network; l1 layer, namely full connection layer, upper connection layer L0 layer, wherein the layer increases dimension of data to three dimensions, increases feature dimension after batch and node dimension, then operates in feature dimension, inputs feature number as 1, and outputs feature number as The output data has a size of ; L2-1 layer, static attention layer, upper L1 layer, which operates in space and feature dimension, and has input feature number of Output feature number is The output data has a size of ; L2-2 layer, dynamic attention layer, upper L1 layer, which operates in space and feature dimension, and input feature number is Output feature number is The output data has a size of ; L3-1 layer, static attention layer, upper L1 layer, which operates in space and feature dimension, and has input feature number of Output feature number is The output data has a size of ; L3-2 layer, dynamic attention layer, upper L1 layer, which operates in space and feature dimension, and input feature number is Output feature number is The output data has a size of ; L4 layer is a full connection layer, and is connected with L2-1 layer, L2-2 layer, L3-1 layer and L3-2 layer, said layer firstly operates in characteristic dimension, and the input characteristic number is Outputting the characteristic number of 1, then reducing the data dimension to two dimensions, and outputting the data dimension of 1 ; L5 layer, node-edge feature transformation layer, upper connection L4 layer, the layer is operated in node dimension and connected with the incidence matrix Multiplying, converting node information into side information, and outputting data with size of scale of ; Wherein, the L0 layer and the L5 layer have no activation function, and the other layer activation functions are all ReLU.
9. The method for rapidly ordering importance of large-scale urban bridge tunneling network based on mixed graph attention network according to claim 8, wherein the mixed graph attention block formula is: in the formula, 、 The attention block input and output for the first hybrid graph respectively, For the operation of the multi-head attention, In order to perform the convolution operation of the drawing, And Respectively an adjacency matrix and a learnable adjacency matrix of the bridge-tunnel network, And Is a learnable embedding matrix.

Description

Large-scale urban bridge tunnel network importance rapid ordering method based on mixed graph attention network Technical Field The invention relates to a bridge-tunnel network service performance evaluation method, in particular to a large-scale urban bridge-tunnel network importance rapid ordering method based on a mixed graph attention network. Background The bridge tunnel structure is used as a life line engineering of a traffic network, has the outstanding function of connecting different areas, and plays an irreplaceable role in meeting the convenient and efficient travel demands of people. As a weak link in a traffic network, in the long-term use process, the bridge-tunnel structure will inevitably deteriorate due to load, structural defects, environmental conditions and the like, resulting in structural damage, casualties, network function degradation and environmental pollution, and producing significant negative effects in various aspects. Under limited maintenance fund constraints, scientific planning of fund distribution is of paramount importance, and the premise of achieving the task is to quantify and rank the importance of a single bridge-tunnel structure at the network level. The importance metrics are used to quantify the importance of different objects in a given system, which can identify bottlenecks in the system and present a scientific optimization solution for system improvement or maintenance activities. At present, a common importance ranking method is a risk-based importance ranking method, and importance ranking is achieved by calculating failure probability of a structure and multiplying economic loss caused by failure. However, the additional detour distance and detour time penalty resulting from the reduced road capacity due to failure must be included in the economic losses calculated in the traffic network. They are directly related to traffic flow redistribution solutions, whose computational dimension grows exponentially with increasing network size, with computation time even in days in large-scale networks. The essence of traffic flow redistribution calculation is that under the condition that given network topology characteristics are taken as input, specific output is solved, the characteristics of data driving performance exist, and quick calculation can be realized through a deep learning agent model. The graph neural network is used as a deep learning model of a special modeling topological structure, and has the feasibility of completing the task. Therefore, the method for rapidly sequencing the importance of the large-scale urban bridge and tunnel network based on the graph neural network can achieve great efficiency improvement compared with the traditional explicit algorithm, and has remarkable supporting significance for intelligent operation and maintenance of the bridge and tunnel network. Disclosure of Invention The invention provides a large-scale urban bridge tunnel network importance quick sequencing method based on a mixed graph attention network, which aims to solve the problem that the current urban bridge tunnel network monomer importance sequencing calculation is time-consuming. The method solves the defect of low efficiency caused by repeated solving of the high-dimensional optimization problem in the calculation of the importance index of the bridge-tunnel network monomer, realizes the accurate and rapid calculation of the importance index of the large-scale urban bridge-tunnel network, and is suitable for the efficient management and maintenance of the large-scale urban bridge-tunnel network. The invention aims at realizing the following technical scheme: a method for rapidly sequencing importance of a large-scale urban bridge tunnel network based on a mixed graph attention network comprises the following steps: step one, defining monomer importance indexes considering the structural failure probability and economic losses generated in safety, society, function and environmental dimensions after failure; Step two, calculating the total travel amount, the total arrival amount and the travel demand among nodes of each node of the bridge tunnel network; Step three, realizing initial traffic flow distribution based on Frank-Wolfe algorithm; Step four, generating a reassigned traffic flow making data set by destroying the sides which possibly fail in the bridge-tunnel network one by one; step five, designing a mixed graph attention network, and modeling a topological association mode between a bridge-tunnel network side capacity reduction vector and a flow redistribution vector; Training the mixed graph attention network designed in the step five, and rapidly sequencing the importance of the large-scale urban bridge tunneling network by using the trained mixed graph attention network. Compared with the prior art, the invention has the following advantages: 1. Aiming at the problem of time consumption of single importance sorting calculation of the urba