A vulnerability-based method for identifying critical road sections in highway network connectivity

Abstract

The present invention belongs to the field of highway network traffic safety and relates to a vulnerability-based method for identifying key road sections of highway network connectivity, including obtaining multi-source data of a target highway network area; setting road alignment change points as road segmentation points based on road alignment data to perform static coarse-grained segmentation of the road sections; statistically calculating characteristic parameters of the minimum road section unit based on different characteristics of the road sections, constructing a road section unit characteristic vector, and using an ordered clustering algorithm to perform dynamic fine-grained segmentation of the road sections; quantifying the vulnerability of the road sections according to the road section characteristic values; and constructing a highway network Bayesian network structure model to identify key road sections. The present invention integrates the influence of multiple factors on the road sections, effectively overcoming the limitations of relying solely on a single evaluation index for identifying key road sections, ensuring the comprehensiveness and accuracy of the identification results; scientifically evaluating the importance of each road section in the highway network from the perspective of the relationship between the potential interruption risk of the road section and the overall failure of the road network, thereby improving the reliability and practicality of the identification results.

Description

Technical Field

[0001] The present invention belongs to the field of highway network traffic safety, and in particular relates to a vulnerability-based method for identifying key road section connectivity in a highway network. Background Art

[0002] Road network connectivity is crucial to the operation level of the highway network. Good road network connectivity helps improve the efficiency of road network operation and enhance people's travel convenience. In emergency situations, such as natural disasters or sudden accidents, good road network connectivity can ensure that rescue and emergency services can reach the required locations quickly and improve disaster resistance. Therefore, it is of great significance to identify key sections that affect road network connectivity.

[0003] Among the existing road network connectivity analysis methods, Zhou Ruitao et al. proposed a road connectivity analysis method based on urban waterlogging conditions (CN113627817A). This method considers the impact of road section failure on road network connectivity changes based on the impact of different rainfall amounts on road sections. However, it only considers the single influencing factor of rainfall and has certain limitations. Zhong Jilong et al. proposed a method for identifying important nodes in urban road networks based on local structural traffic (CN112598305B). This method identifies node importance based on road network structure and road traffic, analyzes road network connectivity from the perspective of node importance, and does not consider the relationship between changes in road section operating status and road network connectivity. Summary of the Invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a vulnerability-based method for identifying key road sections of highway network connectivity.

[0005] The present invention can be achieved through the following technical solutions:

[0006] A vulnerability-based method for identifying critical road sections in highway network connectivity includes the following steps:

[0007] Step 1: Obtain multi-source data of the target road network area;

[0008] Step 2: Based on the road alignment data, the road alignment change points are set as road segment division points to perform static coarse-grained division of the road segments;

[0009] Step 3: Based on the static coarse-grained division in step 2, the characteristic parameters of the smallest road section unit are statistically calculated based on the different characteristics of the road section, and the road section unit feature vector is constructed. The ordered clustering algorithm is used to perform dynamic fine-grained division of the road section.

[0010] Step 4: Based on the results of dynamic fine-grained segmentation in step 3, the vulnerability of the road section is quantified by the road section characteristic value;

[0011] Step 5: Construct a Bayesian network structure model of the highway network to identify key sections.

[0012] In step 1, the multi-source data obtained include highway network data, road alignment data, road accident data, the proportion of "two passenger and one hazardous materials" vehicles, road surface quality data, and weather data. Highway network data includes road length data and road network pile number data; road alignment data includes intersection locations, curve lengths, ramp lengths, and the number of lanes; road accident data includes the pile number data of the road section corresponding to the accident location and the number of casualties in the accident; the proportion of "two passenger and one hazardous materials" (chartered tourist buses, Class III and above regular passenger buses, and special road vehicles transporting dangerous chemicals, fireworks, and civilian explosives) vehicles includes the total traffic volume data of the road section and the traffic volume data of "two passenger and one hazardous materials"; road surface quality data includes the pile number data of the road section corresponding to the road inspection section and the comprehensive road surface damage rate; and weather data includes rainfall intensity data and road visibility data.

[0013] In step 2, the road line shape changes include intersections, curves, ramps and changes in the number of lanes; for intersections, the intersection position is set as the dividing point; for curves, the curves and smooth curves are divided into the same road section; for ramps, the section from the starting point to the end point of the ramp is divided into the same road section; for changes in the number of lanes, the position where the number of lanes changes is set as the dividing point.

[0014] In step 3, the process of performing dynamic fine-grained segmentation of road sections based on the ordered clustering algorithm is specifically as follows:

[0015] Step 301: Obtaining a granularity based on multi-source data as the minimum road segment statistical unit, and statistically analyzing the characteristic parameters of each source data for each minimum road segment unit to form a characteristic vector for each road segment unit. The characteristic parameters include the number of equivalent accidents, the proportion of "two passenger and one hazardous" vehicles, the road condition index, the rainfall intensity, and the visibility range.

[0016] The calculation formula for the equivalent number of accidents is:

[0017] AN=r1X1+r2X2+r3X3+X4 (1)

[0018] Among them, AN represents the equivalent number of accidents, X1 represents the number of deaths in accidents, X2 represents the number of serious injuries in accidents, X3 represents the number of minor injuries, X4 represents the number of accidents, and r1, r2, and r3 represent the weights of the number of deaths, serious injuries, and minor injuries, respectively.

[0019] The calculation formula for the proportion of "two passenger and one dangerous goods" vehicles is:

[0020]

[0021] Among them, R arepresents the proportion of “two passenger and one dangerous goods” vehicles, V "两客一危” represents the traffic volume of “two passenger and one dangerous goods” vehicles within the statistical unit, V 总 Indicates the total vehicle flow in the statistical unit.

[0022] The calculation formula of the road condition index is:

[0023] PCI=100-15DR (3)

[0024] Among them, PCI is the pavement condition index and DR is the comprehensive pavement damage rate.

[0025] Among them, the statistical unit of rainfall intensity is mm / h, and the statistical unit of visibility range is m.

[0026] Step 302: After normalizing the feature vector, an ordered sequence (x1, x2, ..., x n );

[0027] Step 303: Input the ordered sequence into the ordered clustering algorithm to obtain the number of division segments μ0 and the road segment division result

[0028] In step 303, the process of inputting the ordered sequence into the ordered clustering algorithm to obtain the number of segmentation μ0 is specifically as follows:

[0029] For an ordered sequence consisting of the smallest statistical units (x1, x2, ..., x n ), set the possible partition point λ (2≤λ≤n-1), calculate the sum of squares of the deviations before and after the partition point, and the expression for the sum of squares of the deviations before and after the possible partition point λ is:

[0030]

[0031] in, The mean of the data before the division point, The mean of the data after the division point, x i is the i-th data in the ordered sequence, S λ is the sum of squares of deviations before the division point, S n-λ is the sum of squares of deviations before the division point, and n is the total number of minimum statistical units.

[0032] Calculate the total deviation square sum SS of the partition point at this time λ , SS λ The calculation formula is:

[0033] SS λ =S λ +S n-λ (6)

[0034] When the total deviation square sum reaches the minimum value, the possible division point λ is considered to be the optimal point and is determined as the division point. The total deviation square sum is calculated under different numbers of division segments. As the number of division segments increases, the total deviation square sum decreases, indicating that the characteristic differences within the road segment after division become smaller. When the difference between the total deviation square sum before and after the change is less than σ, the final number of road segment divisions μ0 is determined. The characteristic value of the road segment when the number of segment divisions is μ0 is obtained. Where σ is the total deviation squared sum difference threshold.

[0035] In step 4, the vulnerability of the road section is quantified based on the road section characteristic value. The vulnerability of the road section is an important characteristic that measures the ability of the road section to maintain road operation in the face of various potential threats and uncertainties. The specific quantification method is as follows:

[0036] Step 401: Calculate the road section interruption risk value based on the index value of the road section characteristic parameter. The relationship between the index value and the road section interruption risk value is as follows:

[0037] Index value z max ~z1 corresponds to the first-level risk level, and the corresponding risk value threshold is 100~f1;

[0038] The indicator values ​​z1 to z2 correspond to the second-level risk level, and the corresponding risk value thresholds are f1 to f2;

[0039] Index value z2~z ... Corresponding to the third level of risk, the corresponding risk value threshold is f2~f ... ;

[0040] Index value z ... ~0 corresponds to the nth level of risk, and the corresponding risk value threshold is f...~f n .

[0041] Among them, z i is the characteristic parameter index value classification threshold, f i is the threshold value for the road interruption risk value classification, n is the number of risk level classifications, and the corresponding relationship between the specific index value and the road interruption risk value is expressed by a continuous piecewise function:

[0042]

[0043] Among them, F represents the road interruption risk value corresponding to different road section characteristics.

[0044] Step 402: Quantify the vulnerability of the road section according to the road section characteristic risk value. The vulnerability calculation formula is as follows:

[0045] SI=w1F1+w2F2+w3F3+w4F4+w5F5 (8)

[0046] Among them, SI represents the vulnerability of the road section; F1~F5 correspond to the accident risk value, the "two passenger and one dangerous goods" flow risk value, the road condition risk value, the rainfall impact risk value, and the visibility range risk value, respectively; w1~w5 correspond to the historical accident risk level weight, the "two passenger and one dangerous goods" flow risk level weight, the road condition risk level weight, the rainfall impact risk level weight, and the visibility range risk level weight, respectively.

[0047] In step 5, a Bayesian network model of the highway network is constructed to identify key road segments that affect highway network connectivity. Highway network connectivity refers to the degree of connection between road segments, paths, and OD pairs in the highway network. The specific identification process is as follows:

[0048] Step 501: establishing a road network topology structure based on the road network structure data, wherein nodes in the topology structure represent intersections in the road network, and line segments in the topology structure represent road segments in the road network;

[0049] Step 502: Define road segments as root nodes of the Bayesian network. The set of road segments constitutes the node variable set. The Bayesian network structure model of the highway network is determined by the connection relationship between road segments. The relationship between road segments forming paths, paths forming OD pairs, and OD pairs forming the road network determines the four levels of the Bayesian network structure model: road segments, paths, OD pairs, and road network.

[0050] Step 503: The priori disruption probabilities of each road segment are input into the Bayesian network model to calculate the distribution of disruption probabilities for each path and OD pair, as well as the probability of network failure. A high probability of network failure indicates a high likelihood of network connectivity disruption. The posterior conditional probability of each road segment in the event of network failure is then inferred. The importance of each road segment in the network is ranked from highest to lowest based on their posterior conditional probabilities, identifying key road segments that affect network connectivity.

[0051] The formula for calculating the priori probability of road section interruption is as follows:

[0052]

[0053] Among them, R i is the a priori interruption probability of road section i; l is the length of the road section; L is the total length of the road network; SI is the vulnerability of the road section.

[0054] This paper proposes an innovative vulnerability-based method for identifying critical road sections for highway network connectivity based on multi-source data from the "people-vehicle-road-environment" model for inter-city highway networks. This method accurately identifies critical road sections that affect highway network connectivity. The present invention has the following advantages:

[0055] (1) The present invention integrates the influence of multiple factors such as “people-vehicles-roads-environment” on road sections, effectively overcoming the limitation of relying on a single evaluation index to identify key road sections, and ensuring the comprehensiveness and accuracy of the identification results;

[0056] (2) The present invention scientifically evaluates the importance of each road section in the highway network from the perspective of the relationship between the potential interruption risk of the road section and the overall failure of the road network, thereby improving the reliability and practicality of the identification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a flow chart of the present invention.

[0058] Figure 2 It is a schematic diagram of the static coarse-grained division result of the road section in step 2 of the specific implementation method of the present invention.

[0059] Figure 3 It is a schematic diagram of the result of dynamic fine-grained division of road sections in step 3 of a specific implementation method of the present invention.

[0060] Figure 4 It is a schematic diagram of the Bayesian network structure of the highway network in step 5 of the specific implementation method of the present invention. DETAILED DESCRIPTION

[0061] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and technical solutions.

[0062] The present invention provides a vulnerability-based method for identifying key road sections of highway network connectivity. The method is based on multi-source data of the highway network. First, the road sections are statically coarse-grained according to the road alignment data of the highway network. Based on the static division, the ordered clustering algorithm is used to perform dynamic fine-grained division of the road sections based on other dynamic characteristic parameters. The vulnerability of the road sections is quantified according to the characteristic values ​​of the road sections. Finally, a highway network Bayesian network structure model is constructed to identify key road sections that affect the connectivity of the highway network. Figure 1 As shown, the method includes the following steps:

[0063] Step 1: Obtain multi-source data for the target highway network area. The multi-source data includes: annual number of traffic accidents (occurrences / year), the proportion of "two passenger and one hazardous" vehicles on the road, the comprehensive road surface damage rate, rainfall intensity (mm / day), and road visibility range (m);

[0064] Step 2: Based on the road alignment data, the road alignment change points are set as road segment division points. The road alignment includes intersections, curves, ramps, and the number of lanes, and the road segments are divided into static coarse-grained sections.

[0065] Step 3: Based on the static coarse-grained segmentation in Step 2, the characteristic parameters of the smallest road segment unit are calculated. These parameters include: the number of equivalent accidents, the proportion of "two passenger and one hazardous vehicle" vehicles, the road condition index (PCI), rainfall intensity, and visibility range. The road segment unit feature vector is constructed and input into the ordered clustering algorithm for dynamic fine-grained segmentation.

[0066] Step 4: Based on the results of the dynamic fine-grained segmentation in step 3, the vulnerability of the road section is quantified by the road section characteristic value;

[0067] Step 5: Construct a Bayesian network structure model of the highway network to identify key sections.

[0068] In step 1, the multi-source data obtained includes highway network data, road alignment data, road accident data, the proportion of "two passenger and one hazardous materials" vehicles, road surface quality data, and weather data. Highway network data includes road length data and road network stake number data; road alignment data includes intersection locations, curve lengths, ramp lengths, and the number of lanes; road accident data includes the stake number data of the road section corresponding to the accident location and the number of casualties in the accident; the proportion of "two passenger and one hazardous materials" vehicles includes the total traffic volume data of the road section and the traffic volume data of "two passenger and one hazardous materials"; road surface quality data includes the stake number data of the road section corresponding to the road inspection section and the comprehensive road surface damage rate; and weather data includes rainfall intensity data and road visibility range data.

[0069] In step 2, the points where the road line shape changes are set as division points to perform static coarse-grained segmentation of the road section. For intersections, the intersection location is set as the division point; for curves, the curve and smooth curve are divided into the same road section; for ramps, the section from the starting point to the end point of the ramp is divided into the same road section; for changes in the number of lanes, the location where the number of lanes changes is set as the division point. The results of the static coarse-grained segmentation are as follows: Figure 2 As shown in the figure, the road is statically coarse-grained according to the locations of road curves and road intersections.

[0070] In step 3, the distance between adjacent pile numbers in the road network is used as the minimum road section statistical unit, and the characteristic parameters of each source data are counted according to the minimum road section unit to form the characteristic vector x of each road section unit. i The characteristic parameters include the number of equivalent accidents, the proportion of "two passenger and one dangerous goods" vehicles, the road condition index, rainfall intensity and visibility range. The specific meanings of the characteristic parameters are as follows:

[0071] The calculation formula for the equivalent number of accidents is:

[0072] AN=r1X1+r2X2+r3X3+X4 (1)

[0073] Where AN represents the equivalent number of accidents, X1 represents the number of fatalities, X2 represents the number of serious injuries, X3 represents the number of minor injuries, and X4 represents the number of accidents. r1, r2, and r3 represent the weights of the number of fatalities, serious injuries, and minor injuries, respectively. In this embodiment, r1, r2, and r3 are 2, 1.5, and 1.2, respectively.

[0074] The formula for calculating the proportion of "two passenger and one dangerous goods" vehicles is:

[0075]

[0076] Among them, R a represents the proportion of “two passenger and one dangerous goods” vehicles, V "两客一危” V represents the annual traffic volume of “two passenger and one hazardous vehicle” within the statistical unit. 总 Indicates the total vehicle flow in the statistical unit / year.

[0077] The calculation formula of road condition index is:

[0078] PCI=100-15DR (3)

[0079] Among them, PCI is the pavement condition index and DR is the comprehensive pavement damage rate.

[0080] The statistical unit for rainfall intensity is mm / day, and the statistical unit for visibility range is m.

[0081] After the feature vector is normalized, an ordered sequence of characteristic parameters of the entire road (x1, x2, ..., x n ), input the ordered sequence into the ordered clustering algorithm to obtain the segment number μ0 and the dynamic fine-grained segment segment feature value

[0082] The process of obtaining the segment number μ0 by inputting the sequence into the ordered clustering algorithm is as follows: According to the ordered sequence (x1, x2, ..., x n ), set the possible partition point λ (2≤λ≤n-1), and the expression for the sum of squares of the deviations before and after the possible partition point λ is:

[0083]

[0084]

[0085] in, The mean of the data before the division point, The mean of the data after the division point, x i is the i-th data in the ordered sequence, S λ is the sum of squares of deviations before the division point, S n-λ is the sum of squares of deviations before the division point, and n is the total number of minimum statistical units.

[0086] Calculate the total square sum of deviations SS of the mutation point at this time λ , SS λ The calculation formula is:

[0087] SS λ =S λ +S n-λ (6)

[0088] When the total deviation square sum reaches the minimum value, the possible division point λ is considered to be the optimal point and is determined as the division point. The total deviation square sum is calculated under different numbers of division segments. As the number of division segments increases, the total deviation square sum decreases, indicating that the feature differences within the road segment after division become smaller. When the difference between the total deviation square sum before and after the change is less than σ, the final number of road segment divisions μ0 is determined, and the feature value of the road segment after dynamic fine-grained division is obtained. Wherein, σ is the total deviation square sum difference threshold, which can be set to σ=10.

[0089] Figure 3 The dynamic fine-grained segmentation results are determined by setting the segment points from K1 to Kn based on accident characteristics, "two passenger and one dangerous goods" characteristics, road conditions, rainfall intensity, and visibility range characteristics. K1 and Kn represent the starting and ending points of the segment.

[0090] In step 4, based on the dynamic fine-grained segmentation results, the road interruption risk value corresponding to each characteristic parameter value of the road section is calculated. The corresponding risk index level of each characteristic index is shown in Tables 1 to 5 below:

[0091] The threshold value of the equivalent accident number index is determined by the 95th percentile, 85th percentile, and 80th percentile values.

[0092] Table 1. Relationship between the index value of equivalent accident number and the level of road interruption risk value

[0093]

[0094] The threshold value of the "two passengers and one dangerous goods" proportion index is determined by the 95th percentile, 85th percentile, and 80th percentile values.

[0095] Table 2. Relationship between the proportion of “two passenger and one dangerous goods” and the level of road interruption risk value

[0096]

[0097] The road condition index value thresholds are divided according to the Highway Technical Condition Assessment Standard (JTG5210-2018):

[0098] Table 3. Relationship between road condition index values ​​and road section interruption risk value levels

[0099]

[0100] The threshold value of rainfall intensity index is divided according to the standards of the Interim Measures for the Administration of Weather Forecast Release:

[0101] Table 4. Relationship between rainfall intensity index values ​​and road interruption risk value levels Unit: mm / day

[0102]

[0103] The threshold values ​​of visibility range index values ​​are divided according to the standard of "Grades of Highway Traffic Meteorological Conditions" (QX / T 111-2010):

[0104] Table 5. Relationship between visibility range index values ​​and road interruption risk value levels Unit: m

[0105]

[0106] The relationship between the indicator value and the risk value is represented by a continuous piecewise function, as shown below:

[0107]

[0108] Among them, z i is the characteristic parameter index value classification threshold, f i is the threshold value for the road interruption risk value classification, n is the number of risk level classifications, and F represents the road interruption risk value corresponding to different road section characteristics.

[0109] The vulnerability of a road section is quantified based on the road interruption risk value of each road feature. The vulnerability calculation formula is as follows:

[0110] SI=w1F1+w2F2+w3F3+w4F4+w5F5 (8)

[0111] Among them, SI is the vulnerability of the road section; F1~F5 correspond to the accident risk value, the "two passengers and one dangerous goods" flow risk value, the road condition risk value, the rainfall impact risk value, and the visibility range risk value, respectively; w1~w5 correspond to the accident risk level weight, the "two passengers and one dangerous goods" flow risk level weight, the road condition risk level weight, the rainfall intensity risk level weight, and the visibility range risk level weight, respectively.

[0112] The specific weight reference values ​​are shown in Tables 6 to 10

[0113] Table 6. Accident risk level weight reference

[0114] Accident risk level <![CDATA[w1]]> Danger 1.5 More dangerous 1 Safer 0.8 Safety 0.6

[0115] Table 7. Reference weights for risk levels of “two passenger and one dangerous goods” traffic

[0116] "Two Passenger and One Dangerous Goods" Traffic Risk Level <![CDATA[w2 <!-- 7 -->]]> many 1.2 More 1 generally 0.8 less 0.6

[0117] Table 8. Road risk level weight reference

[0118] Road risk level <![CDATA[w3]]> Difference 1.5 Second-rate 1.2 middle 1 good 0.8 excellent 0.6

[0119] Table 9. Rainfall intensity risk level weight reference

[0120] Rainfall intensity risk level <![CDATA[w4]]> Heavy rain 2 rainstorm 1.5 heavy rain 1.2 moderate rain 1 light rain 0.8

[0121] Table 10. Visibility range risk level weight reference

[0122]

[0123]

[0124] In step 5, based on the road network structure data, the intersections in the road network are defined as nodes, and the road sections in the road network are defined as line segments to establish the road network topology. Then, the road sections in the road network topology are defined as the root nodes of the Bayesian network structure, forming a node variable set. A Bayesian network structure model is established, including four levels: road sections, paths, OD pairs, and road networks. The Bayesian network structure relationship of the road network is as follows: Figure 4 shown.

[0125] The a priori probability of road section interruption is calculated according to the following formula:

[0126]

[0127] Among them, R i is the a priori interruption probability of road section i; l is the length of the road section; L is the total length of the road network; SI is the vulnerability of the road section.

[0128] The priori outage probabilities of each road section were input into the established Bayesian network model of the highway network. The distribution of outage probabilities for each path and OD pair, as well as the probability of network failure, were calculated. A higher network failure probability indicates poorer connectivity within the network. The posterior conditional probability of each road section was calculated when the network failed. The posterior conditional probability values ​​of the road sections were arranged in descending order to identify key sections that affect highway network connectivity.

[0129] In step 503, the calculation principle of the Bayesian network structure model for calculating the road network failure probability is as follows:

[0130]

[0131] Among them, P(M|N) is the posterior probability, which represents the probability of M occurring when N is known to occur; P(N|M) is the likelihood ratio, which represents the probability of N occurring when M is known to occur; P(M) is the prior probability, which represents the probability when M occurs; and P(N) is the unconditional probability, which represents the probability when N occurs.

[0132] Assume that M exists m n variable nodes, we can get the following formula:

[0133]

[0134] The highway network key section identification method proposed in the present invention can be well applied to actual road network management. Compared with the traditional identification method based on single indicators such as road network connectivity, number of road section accidents, and saturation, it starts from multi-source data and adopts a combination of static and dynamic methods to reasonably divide road sections, effectively quantify the vulnerability of road sections, and identify key sections that affect highway network connectivity with reliability and accuracy.

Claims

1. A vulnerability-based method for identifying critical road sections in highway network connectivity, characterized in that: The following steps are involved: Step 1: Obtain multi-source data of the target road network area; Step 2: Based on the road alignment data, the road alignment change points are set as road segment division points to perform static coarse-grained division of the road segments; Step 3: Based on the static coarse-grained division in step 2, the characteristic parameters of the smallest road section unit are statistically calculated based on the different characteristics of the road section, and the road section unit feature vector is constructed. The ordered clustering algorithm is used to perform dynamic fine-grained division of the road section. Step 4: Based on the results of dynamic fine-grained segmentation in step 3, the vulnerability of the road section is quantified by the road section characteristic value; Step 5: Construct a Bayesian network structure model of the highway network to identify key sections; In step 1, the multi-source data obtained include highway network data, road alignment data, road accident data, "two passenger and one hazardous" vehicle ratio data, road surface quality data, and weather data; wherein the highway network data includes road length data and road network pile number data; the road alignment data includes intersection location, curve length, ramp length, and number of lanes; the road accident data includes the pile number data of the road section corresponding to the accident location and the number of casualties in the accident; the "two passenger and one hazardous" vehicle ratio data includes the total traffic flow data of the road section and the "two passenger and one hazardous" vehicle flow data; the road surface quality data includes the pile number data of the road section corresponding to the road inspection section and the comprehensive road surface damage rate; the weather data includes rainfall intensity data and road visibility data; In step 2, the road alignment changes include intersections, curves, ramps, and changes in the number of lanes; for intersections, the intersection location is set as the dividing point; for curves, the curves and smooth curves are divided into the same road section; for ramps, the section from the starting point to the end point of the ramp is divided into the same road section; for changes in the number of lanes, the location where the number of lanes changes is set as the dividing point; In step 3, the process of performing dynamic fine-grained segmentation of road sections based on the ordered clustering algorithm is specifically as follows: Step 301: Based on the multi-source data, a granularity is obtained as the minimum road segment statistical unit. The characteristic parameters of the data from each source are counted for each minimum road segment unit to form a characteristic vector for each road segment unit. The characteristic parameters include the number of equivalent accidents, the proportion of "two passenger and one hazardous materials" vehicles, the road condition index, the rainfall intensity, and the visibility range. The calculation formula for the equivalent number of accidents is: AN=r1X1+r2X2+r3X3+X4 (1) Where AN represents the equivalent number of accidents, X1 represents the number of deaths in accidents, X2 represents the number of serious injuries in accidents, X3 represents the number of minor injuries, X4 represents the number of accidents, r1, r2, and r3 represent the weights of the number of deaths, serious injuries, and minor injuries, respectively; The calculation formula for the proportion of "two passenger and one dangerous goods" vehicles is: Among them, R a represents the proportion of "two passenger and one dangerous goods" vehicles, V "两客一危” represents the traffic volume of "two passenger and one hazardous vehicle" within the statistical unit, V 总 Indicates the total vehicle flow in the statistical unit; The calculation formula of the road condition index is: PCI=100-15DR (3) Among them, PCI is the pavement condition index, DR is the comprehensive pavement damage rate; The statistical unit for rainfall intensity is mm / h, and the statistical unit for visibility range is m; Step 302: After normalizing the feature vector, an ordered sequence (x1, x2, ..., x n ); Step 303: Input the ordered sequence into the ordered clustering algorithm to obtain the number of division segments μ0 and the road segment division result The specific process is: For an ordered sequence consisting of the smallest statistical units (x1, x2, ..., x n ), set the possible partition point λ, 2≤λ≤n-1, calculate the sum of squares of the deviations before and after the partition point, and the expression for the sum of squares of the deviations before and after the possible partition point λ is: in, The mean of the data before the division point, The mean of the data after the division point, x i is the i-th data in the ordered sequence, S λ is the sum of squares of deviations before the division point, S n-λ is the sum of squares of deviations before the division point, and n is the total number of minimum statistical units; Calculate the total square sum of the deviations SS of the partition points at this time λ , SS λ The calculation formula is: SS λ =S λ +S n-λ (6) When the total deviation square sum reaches the minimum value, the possible division point λ is considered to be the optimal point and is determined as the division point; the total deviation square sum under different division numbers is calculated; as the number of divisions increases, the total deviation square sum decreases, indicating that the characteristic difference within the road section becomes smaller after the division; when the difference between the total deviation square sum before and after the change is less than σ, the final number of road section divisions μ0 is determined; the section characteristic value when the number of road section divisions is μ0 is obtained Where σ is the total deviation square sum difference threshold; In step 4, the vulnerability of the road section is quantified according to the road section characteristic value; the specific quantification method is as follows: Step 401: Calculate the road section interruption risk value based on the index value of the road section characteristic parameter; the relationship between the index value and the road section interruption risk value is as follows: Index value z max ~z1 corresponds to the first-level risk level, and the corresponding risk value threshold is 100~f1; The indicator values ​​z1 to z2 correspond to the second-level risk level, and the corresponding risk value thresholds are f1 to f2; Index value z2~z ... Corresponding to the third level of risk, the corresponding risk value threshold is f2~f ... ; Index value z ... ~0 corresponds to the nth level of risk, and the corresponding risk value threshold is f ... ~f n ; Among them, z i is the characteristic parameter index value classification threshold, f i is the threshold value for the road interruption risk value classification, n is the number of risk level classifications, and the corresponding relationship between the specific index value and the road interruption risk value is expressed by a continuous piecewise function: Among them, F represents the road interruption risk value corresponding to different road section characteristics; Step 402: Quantify the vulnerability of the road section according to the road section characteristic risk value. The vulnerability calculation formula is as follows: SI=w1F1+w2F2+w3F3+w4F4+w5F5 (8) Among them, SI represents the vulnerability of the road section; F1 to F5 correspond to the accident risk value, the "two passenger and one dangerous goods" flow risk value, the road condition risk value, the rainfall impact risk value, and the visibility range risk value, respectively; w1 to w5 correspond to the historical accident risk level weight, the "two passenger and one dangerous goods" flow risk level weight, the road condition risk level weight, the rainfall impact risk level weight, and the visibility range risk level weight, respectively. In step 5, a Bayesian network structure model of the highway network is constructed to identify key road sections that affect highway network connectivity. Highway network connectivity refers to the degree of connection between road sections, paths, and OD pairs in the highway network. The specific identification process is as follows: Step 501: establishing a highway network topology structure based on highway network structure data, wherein nodes in the topology structure represent intersections in the highway network, and line segments in the topology structure represent road segments in the highway network; Step 502: Define the road segment as the root node of the Bayesian network, and the set formed by each road segment as the node variable set. The Bayesian network structure model of the highway network is determined by the connection relationship between the road segments. The relationship between the road segments forming paths, the paths forming OD pairs, and the OD pairs forming the road network determines the four levels of the Bayesian network structure model, namely, road segments, paths, OD pairs, and road network. Step 503: Input the priori interruption probability of each road section into the Bayesian network structure model to calculate the distribution of interruption probabilities for each path and OD pair and the road network failure probability. A high road network failure probability indicates a high possibility of road network connectivity interruption. Then, reversely infer the posterior conditional probability of each road section when the road network fails. The road sections are ranked from the highest to the lowest according to their posterior conditional probabilities in the road network, and the key road sections that affect the road network connectivity are identified. The formula for calculating the priori probability of road section interruption is as follows: Among them, R i is the a priori interruption probability of road section i; l is the length of the road section; L is the total length of the road network; SI is the vulnerability of the road section.

Images