Risk level determination method for planning and construction period of non-stop energy supplement zero-carbon transportation system

Abstract

The present application relates to the technical field of risk assessment, in particular to a risk grade determination method for planning and construction period of a no-parking energy supplement zero-carbon transportation system, comprising the following steps: obtaining facility vibration frequency time series data, performing difference processing and identifying abnormal points by using isolated forest, combining corresponding stress peak value, fatigue grade and slope parameter weighted analysis to output risk correction value, normalizing and weighting to generate risk score and labeling, extracting path node stress and risk value, using K-means clustering analysis to output risk node, screening and spatial clustering the risk node elastic recovery ratio and stress peak value, and outputting the risk grade determination result. In the present application, the vibration frequency data is processed by subsection difference, the mutation nodes are accurately positioned by combining abnormal detection, the multi-dimensional stress, fatigue and terrain parameters are weighted and matched, the risk is normalized and graded and labeled, the structure and environmental characteristics are integrated by sliding window clustering analysis, the risk monitoring granularity and response speed are refined, and the dynamic monitoring precision and linkage ability are improved.

Description

Technical Field

[0001] This invention relates to the field of risk assessment technology, and in particular to a method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system. Background Technology

[0002] The field of risk assessment technology involves identifying, analyzing, and classifying uncertainties faced by projects or systems during the planning, construction, and operation phases. Its core aspects include risk identification, risk level classification, risk source determination, risk quantification methods, and the construction of an assessment indicator system. This technology is widely used in infrastructure construction, energy management, transportation, and other fields. Particularly in complex systems engineering projects with multiple intertwined factors, long cycles, and large investment scales, establishing systematic and standardized risk assessment methods helps improve the scientific rigor and foresight of risk management.

[0003] The traditional risk level assessment method for the planning and construction phase of non-stop refueling zero-carbon transportation systems refers to classifying key risks during the planning and construction phases based on risk factors affecting the system's accessibility, reliability, safety, policy compatibility, and environmental compatibility. This is achieved through methods such as comparing original project data, expert qualitative scoring, weighting engineering experience, and summarizing on-site survey results. A risk level assessment system is formed based on subcategories such as route accessibility assessment standards, refueling facility construction difficulty assessment tables, policy support document consistency review records, and construction environment adaptability survey data.

[0004] Traditional risk assessment models, which rely primarily on subjective experience and data comparison, are limited in their ability to identify risks when faced with multi-source environmental changes and sudden structural changes. Periodic manual intervention and a lack of dynamic data support make it difficult to reflect abnormal changes in risk nodes in a timely manner. Existing methods are limited to static classification and experience-based weighting, resulting in information lag, omission of high-risk nodes, and insufficient timely early warning responses. This restricts accurate risk management and dynamic decision-making, and affects the effectiveness of closed-loop risk control during major engineering construction. Summary of the Invention

[0005] To address the limitations of traditional risk assessment methods, which rely primarily on subjective experience and data comparison, in the face of multi-source environmental changes and abrupt structural changes, including limited identification capabilities, periodic manual intervention, and a lack of dynamic data support, leading to difficulties in timely reflection of abnormal changes in risk nodes, and because existing methods are limited to static grading and experience-weighted approaches with lagging information updates, resulting in the omission of high-risk nodes and insufficient timely early warning responses, thus hindering accurate risk management and dynamic decision-making and affecting the effectiveness of closed-loop risk control during major engineering construction, this invention provides a method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system. The technical solution is as follows: On the one hand, a method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system is provided. This method includes: S1: Obtain the structural vibration frequency time series data of overhead contact network, roadside poles, and power transmission and transformation line facilities under wind load, rainfall, and temperature disturbance during the planning and construction period, perform segmented differential processing, use the isolated forest algorithm to identify abnormal mutation points of monitoring nodes, and output frequency anomaly point data; S2: For the frequency anomaly point data, obtain the peak value of structural stress response, material fatigue level and terrain slope parameters for the corresponding time period, and use interval matching and weighted calculation to analyze the node risk correction value of the anomaly point and output the node risk correction dataset. S3: Based on the node risk correction dataset, the risk score calculation function is used to normalize and weight the risk correction values, construct a risk trigger score sequence, and label the facility nodes with risk scores in combination with the grading standard, and output the node risk labeling results; S4: Call the node risk labeling results, extract the original stress loading data of the facility monitoring nodes on the operation path, group the path nodes, use a sliding window to combine the risk score and stress loading peak value, perform cluster analysis on the combined data based on K-means clustering, and output the construction period risk node list.

[0006] As a further embodiment of the present invention, the frequency anomaly data includes anomaly frequency values, monitoring node numbers, and anomaly occurrence times; the node risk correction dataset includes structural safety factors, stress response peak values, and fatigue levels; the node risk labeling results include risk classification identifiers, risk scores, and node identifiers; and the construction period risk node list includes a list of high-risk nodes, node spatial locations, and cluster numbers.

[0007] As a further aspect of the present invention, the specific steps of S1 include: S101: Obtain the structural vibration frequency time series data of overhead contact network, roadside poles, and power transmission and transformation line facilities under wind load, rainfall, and temperature disturbance during the planning and construction period. Divide the structural vibration frequency time series data into time intervals and perform differential calculation on the continuous frequency values ​​within the intervals to obtain segmented frequency difference values. S102: Call the segmented frequency difference value, organize the segmented difference data of the monitoring nodes, take the difference sequence after node segmentation as the input variable, use the isolated forest algorithm, output the anomaly coefficient based on the difference characteristics of each monitoring node segmentation interval, and obtain the sub-node anomaly coefficient sequence. S103: Based on the sub-node anomaly coefficient sequence, compare the anomaly coefficients of the monitoring node segment intervals with the isolation forest algorithm discrimination threshold, filter the frequency intervals and node numbers where the anomaly coefficients exceed the set discrimination threshold, and generate frequency anomaly point data; The discrimination threshold is a boundary value used to determine whether the anomaly coefficient of the segmented interval of the monitoring node reaches the anomaly standard. The optimal threshold is determined by statistical analysis of the original monitoring data and is set to 1.5 times the average value of the anomaly coefficient of the isolated forest.

[0008] As a further aspect of the present invention, the specific steps of S2 include: S201: Obtain the frequency anomaly point data, filter the structural stress response peak value, material fatigue level and terrain slope parameters of the corresponding time period, pair the parameters according to the time period of the anomaly point, classify them into the corresponding data set, count the interval range of each parameter set, and generate the interval parameter set. S202: Call the set of interval parameters, use interval matching to weight the peak value of structural stress response, material fatigue level and terrain slope parameters in the same interval according to the influence ratio of the parameters in the node risk, calculate the risk value of the outlier point and generate the node risk weighted value. S203: Call the node risk weighting value, collect the risk value under each node according to the weighting result corresponding to all anomalies, and summarize to obtain the node risk correction dataset.

[0009] As a further aspect of the present invention, the specific steps of S3 include: S301: Based on the node risk correction dataset, apply the risk score calculation function to extract the risk correction value of each node in sequence, convert the node risk correction value using the calculation formula, and normalize the value range of the node risk score to generate a normalized risk score sequence. The normalization process employs the Min-Max normalization method; S302: Call the normalized risk score sequence, adjust the distribution of each normalized risk score using the risk weight parameter, set the influence of the weight parameter according to the node attribute, correct the weight of the risk score, and generate a weighted risk trigger score sequence. S303: Based on the weighted risk trigger score sequence, the score of each node and the grading standard threshold are partitioned and judged, the risk score range of the node is labeled and assigned, and the node risk labeling result is generated.

[0010] As a further aspect of the present invention, the normalized risk score refers to the normalization of the risk correction value obtained through the risk score calculation function to obtain a standardized risk score for unified comparison. The risk score calculation function is a mathematical function that outputs the original risk score of a node based on its risk correction value and attributes. The weighted risk trigger score refers to the weighted score obtained by combining the normalized risk score with the node weight parameter, which reflects the real-time risk priority of the node. The grading standard threshold refers to the preset score limit used to distinguish weighted risk trigger scores into differentiated risk levels.

[0011] As a further aspect of the present invention, the specific steps of S4 include: S401: Call the node risk labeling result, obtain the original stress loading data of the facility monitoring nodes on the operation path, classify the path nodes according to the physical location and number of the nodes, integrate the grouped nodes and the original stress loading data, and generate path node stress grouping data. S402: Call the stress grouping data of the path nodes, use a fixed-length sliding window, extract the risk score and stress loading peak value in each window for the data of each group of nodes, and use a traversal method to count the risk score and stress peak value data pairs under the same window to obtain the joint feature value interval of the window. S403: Based on the window joint feature value interval, the K-means clustering algorithm is used to classify the node window combination data, summarize the risk scores and stress peak data of multi-category node groups, and filter the node data with large risk scores and stress peak averages from the clustering results to obtain the construction period risk node list.

[0012] As a further aspect of the present invention, the window joint feature value refers to the data combination of risk score and stress loading peak value extracted simultaneously within a fixed-length sliding window, reflecting the risk and stress characteristics of the node within a time period. The K-means clustering algorithm groups the joint feature value data of the node windows, where the K value is determined based on the Elbow method and the silhouette coefficient method combined with the distribution characteristics of the training dataset.

[0013] As a further aspect of the present invention, the method includes step S5: S5: The elastic recovery ratio and stress loading peak value of the nodes in the construction period risk node list are judged, and risk nodes with a ratio greater than the risk elasticity threshold are screened. The spatial clustering method is used to analyze the spatial distribution characteristics of the nodes, and the risk level judgment result of the non-stop energy replenishment zero-carbon transportation system planning and construction period is output. The risk level determination results include the overall risk classification results, spatial distribution characteristics, and resilience ratio.

[0014] As a further aspect of the present invention, the specific steps of S5 include: S501: Obtain the elastic recovery ratio and stress loading peak value of nodes in the construction period risk node list, calculate the node ratio and compare it with the risk elasticity threshold, filter nodes with ratio values ​​greater than the threshold, and generate an elasticity threshold filter node ratio range. The risk resilience threshold is the minimum ratio standard for measuring whether the resilience recovery capability of risk nodes during the construction period meets the standard. It is set to 0.75. The risk level is divided into three segments based on the comparison between the weighted risk trigger score and the threshold. S502: Call the elastic threshold to filter the node ratio range, use spatial clustering method to classify the spatial distribution information of the nodes, combine the node coordinates and spatial clustering results, quantify the concentration degree and distribution characteristics of the spatial distribution of nodes under clustering, and obtain the node distribution characteristic coefficient; The spatial clustering method uses the DBSCAN density clustering algorithm to perform spatial partitioning based on node geographic coordinates and risk indicators. S503: Based on the node distribution characteristic coefficients, the risk level of the node distribution characteristics under spatial clustering is classified and judged. Combined with the real-time risk attributes and distribution patterns of the nodes during the construction period, the risk level judgment result of the planning and construction period is generated.

[0015] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By segmenting and differentially processing the vibration frequency time-series data of structural facilities under differentiated environmental disturbances, and combining it with anomaly detection algorithms to accurately locate abrupt changes in monitoring nodes, and by using interval matching and weighting of parameters such as multidimensional stress, fatigue, and topography, the granularity of risk node identification and response speed are significantly enhanced. The normalized risk correction results are graded and labeled in a serialized manner, enabling dynamic display of facility risk status. The sliding window combined with cluster analysis of risk scores and stress peak values ​​effectively integrates structural characteristics and environmental dynamics, sorts out high-risk distribution, and demonstrates the ability to accurately capture sudden risks and subdivided node control, thereby improving the precision and real-time linkage level of risk monitoring in complex environments. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the workflow of the present invention; Figure 2 This is a detailed flowchart of S1 of the present invention; Figure 3 This is a detailed flowchart of the S2 process of the present invention; Figure 4 This is a detailed flowchart of the S3 process of the present invention; Figure 5 This is a detailed flowchart of the S4 process of the present invention; Figure 6 This is a detailed flowchart of S5 of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] Please see Figure 1 This invention provides a method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system. The processing flow of this method may include the following steps: S1: Obtain the structural vibration frequency time series data of overhead contact network, roadside poles, and power transmission and transformation line facilities under wind load, rainfall, and temperature disturbance during the planning and construction period, perform segmented differential processing, use the isolated forest algorithm to identify abnormal mutation points of monitoring nodes, and output frequency anomaly point data; S2: For frequency anomaly data, obtain the peak value of structural stress response, material fatigue level and terrain slope parameters for the corresponding time period, use interval matching and weighted calculation to analyze the node risk correction value of the anomaly point, and output the node risk correction dataset. S3: Based on the node risk correction dataset, the risk score calculation function is used to normalize and weight the risk correction values, construct a risk trigger score sequence, and label the facility nodes with risk scores in combination with the classification standard, and output the node risk labeling results; S4: Call the node risk labeling results, extract the original stress loading data of facility monitoring nodes on the work path, group the path nodes, use a sliding window to combine the risk score and stress loading peak value, perform cluster analysis on the combined data based on K-means clustering, and output the construction period risk node list; S5: Determine the elastic recovery ratio and stress loading peak execution ratio of the nodes in the risk node list during the construction period, screen out risk nodes with ratios greater than the risk elasticity threshold, analyze the spatial distribution characteristics of the nodes using spatial clustering methods, and output the risk level determination results for the planning and construction period of the non-stop energy replenishment zero-carbon transportation system. The frequency anomaly data includes the abnormal frequency value, monitoring node number, and anomaly occurrence time. The node risk correction dataset includes the structural safety factor, stress response peak value, and fatigue level. The node risk labeling results include risk classification identifier, risk score, and node identifier. The construction period risk node list includes a list of high-risk nodes, node spatial location, and cluster number. The risk level determination results include the overall risk classification results, spatial distribution characteristics, and elastic recovery ratio.

[0023] Specifically, such as Figure 2 As shown, the specific steps of S1 are as follows: S101: Obtain the structural vibration frequency time series data of overhead contact network, roadside poles, and power transmission and transformation line facilities under wind load, rainfall, and temperature disturbance during the planning and construction period. Divide the structural vibration frequency time series data into time intervals and perform differential calculation on the continuous frequency values ​​within the intervals to obtain segmented frequency difference values. When acquiring time-series data on the structural vibration frequencies of overhead contact lines, roadside poles, and power transmission and transformation lines under wind load, rainfall, and temperature disturbances during the planning and construction period, monitoring points are first deployed along major transportation corridors and key nodes within the project area. Each monitoring point is equipped with a frequency sensor, collecting structural vibration frequency data once per hour. For example, from June 1st to June 7th, 2025, 24 sets of frequency data were recorded daily for monitoring points A1, A2, and A3. The frequency data is then segmented according to time, with each segment divided into 6-hour intervals. Each interval consists of frequency values ​​at four times. For example, the frequency data for point A1 in the first 4 hours is [8.6, 8.7, 8.9, 9.1] Hz. Next, within each segment, the frequency change is calculated for adjacent times, specifically by subtracting the frequency value of the previous time from the frequency value of the next time interval. The internal difference result [0.1, 0.2, 0.2] Hz is obtained, and the frequency difference sequence of the monitoring point interval is organized into a difference array. In this process, if the frequencies recorded at monitoring point A1 at 00:00, 06:00, 12:00, and 18:00 on June 1 are 8.5, 8.7, 9.0, and 9.3 Hz respectively, then their segmented differences are 0.2, 0.3, and 0.3 Hz respectively. The segmented difference values ​​are recorded in the form of "monitoring point number - interval number - difference value", as shown in Table 1. The monitoring point numbers in Table 1 correspond to the specific locations of the overhead contact network, roadside poles, and transmission and transformation lines in real time. For example, A1 is at kilometer 960 of the contact network, A2 is pole No. 3 in area A, and A3 is section S1 of the transmission line. After the sorting is completed, a segmented frequency difference dataset including monitoring points and time intervals is formed. This step lays the data foundation for subsequent risk level determination.

[0024] Table 1: Segmented Frequency Difference Values ​​for Monitoring Points

[0025] As shown in Table 1, the frequency data of the Q1 interval at monitoring point A1 are 8.6Hz, 8.7Hz, 8.9Hz, and 9.1Hz respectively. In the differential step, the first difference is 8.7-8.6=0.1Hz, the second difference is 8.9-8.7=0.2Hz, and the third difference is 9.1-8.9=0.2Hz. If a negative difference occurs, it indicates a frequency decrease. All differential data are recorded. This method meticulously decomposes the data acquisition, segmentation, differential, and recording process for each monitoring point and each interval, ensuring that the parameters are closely connected and that there are no omissions in the data. Through the above process, the segmented frequency differential values ​​of the monitoring points are sorted out, providing detailed data support for the subsequent determination of abnormal structural response intervals.

[0026] S102: Call the segmented frequency difference value, organize the segmented difference data of the monitoring nodes, take the difference sequence after node segmentation as the input variable, use the isolated forest algorithm, output the anomaly coefficient based on the difference characteristics of each monitoring node segmentation interval, and obtain the sub-node anomaly coefficient sequence. After retrieving the differential sequences obtained by the monitoring nodes within time segments, the segmented differential data of each monitoring node are first categorized. Differential sequences of the same node within differentiated time partitions are placed into the corresponding node's dataset. Then, for each node's interval differential sequence, the segmented differential values ​​are read one by one according to the node number, and the differential data is input in array form. To identify abnormal responses of the structure under the influence of factors such as wind load, rainfall, or temperature changes during the construction period, each differential value in each differential sequence needs to be compared with the original normal differential value of the same period. Assuming the differential sequence of interval Q1 for node A1 is [0.1, 0.2, 0.2], it is compared with the original mean and fluctuation range of the same period for node A1. If the original mean is 0.15Hz and the standard deviation is 0.03Hz, then the absolute difference of each item in the current interval's differential sequence is compared with the mean, and the absolute value of its deviation from the mean is calculated as |0.1-0.15|=0.05H. z, |0.2-0.15|=0.05Hz, |0.2-0.15|=0.05Hz, then the absolute deviation values ​​of each node and each segment interval are added together to obtain the total deviation of the interval. For example, the total deviation of the interval Q1 of node A1 is 0.15Hz. Then the total deviation of the interval is sorted by node, and the interval with higher deviation is selected. When distinguishing the deviation interval, the low deviation interval is set to 0~0.06Hz, the medium deviation interval is 0.06~0.12Hz, and the high deviation interval is above 0.12Hz. If the total deviation of the interval Q1 of node A1 is 0.15Hz, it is judged as a high deviation. Then, each node and each segment interval is assigned an anomaly score. The higher the value, the greater the degree of anomaly. The node segment interval and its anomaly score are organized into a node anomaly coefficient sequence. Finally, a set of anomaly coefficient sequences with node number and segment interval as index and anomaly score as content are obtained, which facilitates subsequent interval anomaly screening and risk assessment.

[0027] S103: Based on the sub-node anomaly coefficient sequence, the anomaly coefficients of the segmented intervals of the monitoring nodes are compared with the discrimination threshold of the isolated forest algorithm. Frequency intervals and node numbers with anomaly coefficients exceeding the set discrimination threshold are selected to generate frequency anomaly point data. The discrimination threshold is a boundary value used to determine whether the anomaly coefficient of the segmented interval of the monitoring node reaches the anomaly standard. The optimal threshold is determined by statistical analysis of the original monitoring data and is set to 1.5 times the average value of the anomaly coefficient of the isolated forest. Based on the obtained sub-node anomaly coefficient sequence, the anomaly score for each segment interval of each monitoring node is first directly compared with a preset discrimination threshold. The discrimination threshold is determined based on the statistical average of the anomaly scores of all nodes and all intervals, and is set to 1.5 times the average anomaly score. For example, if the statistical average of the anomaly scores of all intervals of a node is 0.08Hz, then the discrimination threshold is set to 0.08 × 1.5 = 0.12Hz. Subsequently, the anomaly scores of the monitoring node intervals are traversed sequentially, and the node numbers and interval numbers with anomaly scores greater than or equal to 0.12Hz are filtered out to form a frequency anomaly point data list, such as A. If the abnormal score of monitoring point Q1 is 0.15Hz, which is higher than the threshold of 0.12Hz, then A1-Q1 is marked as abnormal and its abnormal score of 0.15Hz is recorded. If the abnormal score of monitoring point A2 in the Q1 interval is 0.05Hz, which is lower than the threshold, then it is not marked as abnormal. Throughout the screening process, the node number, interval number, and abnormal score of the abnormal data points are organized and archived one by one to form a dataset of abnormal structural vibration frequencies. At the same time, the abnormal point data can be used for subsequent risk level classification and response measure formulation. This process ensures that risk points higher than the discrimination threshold are accurately screened out, realizing dynamic monitoring and data support for structural risks.

[0028] Specifically, such as Figure 3 As shown, the specific steps of S2 are as follows: S201: Obtain frequency anomaly data, filter the structural stress response peak, material fatigue level and terrain slope parameters for the corresponding time period, pair the parameters according to the time period of the anomaly points, classify them into the corresponding data sets, calculate the interval range of each parameter set, and generate interval parameter sets. After acquiring the frequency anomaly data, the first step is to sequentially search for the peak structural stress response, material fatigue level, and terrain slope parameter within the corresponding time period based on the node number and interval identifier of each anomaly point. For example, if node A1 in interval Q1 is determined to be an anomaly, then the peak stress response data of point A1 within interval Q1 needs to be searched. Assuming the daily frequency anomaly occurred between 12:00 and 18:00 on June 2, 2025, with a peak stress response of 165 MPa, a material fatigue level converted to 0.79 using the period counting method, and a terrain slope parameter obtained from topographic mapping data of 2.8%, these three parameters are paired with the time period of the anomaly point. This search and pairing process is repeated to organize the parameters of the frequency anomaly points, forming a structural stress response indexed by the time period of the anomaly point. Peak value arrays, material fatigue level arrays, and terrain slope arrays are categorized into datasets and archived separately according to the time periods of the differential anomalies. For example, for anomalies in interval Q2 of node A2, the peak stress is paired with a value of 142 MPa, a fatigue level of 0.68, and a terrain slope of 3.9%. Subsequently, interval statistics are performed on each dataset to calculate the minimum, maximum, and average values ​​of the structural stress response peak within the anomaly interval. The material fatigue level and terrain slope parameters are also calculated separately for interval ranges. For example, the structural stress range is from 120 MPa to 172 MPa, the material fatigue level range is from 0.55 to 0.85, and the terrain slope range is from 2.1% to 4.8%. The parameter statistics results are all paired and categorized into their respective datasets, ultimately generating an interval parameter set based on the time period of the anomaly.

[0029] S202: Call the interval parameter set, use the interval matching method to weight the peak value of structural stress response, material fatigue level and terrain slope parameters in the same interval according to the influence ratio of the parameters in the node risk, calculate the risk value of the outlier point and generate the node risk weighted value. After calling the interval parameter set, for each anomaly time period, the corresponding structural stress response peak value, material fatigue grade, and terrain slope parameters are extracted. The influence ratio of these three parameters on node risk is determined according to the correlation standard, with the stress response peak value weighted at 0.5, the material fatigue grade weighted at 0.3, and the terrain slope weighted at 0.2. These weights are statistically allocated based on the original failure cases. Subsequently, for anomalies A1-Q1, the stress response peak value (165 MPa), fatigue grade (0.79), and slope (2.8%) are read. All three parameters are standardized, and the stress response peak value within the interval 120 MPa to 172 MPa is linearly normalized to obtain standardized values. The material fatigue level is standardized as follows: The terrain slope is standardized as Then, multiply by the weights respectively to calculate the node risk value of the A1-Q1 interval. Repeat the above weighted calculation to calculate the node risk weight value for the anomaly interval, and finally generate a set of node risk weight values ​​indexed by node number and interval identifier.

[0030] S203: Call the node risk weighting value, collect the risk value under each node based on the weighting results of all anomalies, and summarize to obtain the node risk correction dataset; After calling the node risk weighting value, the weighted results for all intervals corresponding to outliers are aggregated node by node. First, based on the node number, the risk weighting values ​​within the interval are summarized by node. For each node, the set of risk weighting values ​​for all its outlier intervals is calculated, and the average risk weighting value, maximum risk weighting value, and sum are calculated. For example, the risk weighting value for the outlier interval of node A1 is... Then the average risk weighted value of node A1 is ( The maximum value is 0.7950, and the sum is 2.2014. The risk weights of the nodes are all collected in accordance with the above method to form the risk value dataset under each node. Finally, the node risk correction dataset is obtained, which is used for subsequent risk level determination and management.

[0031] Specifically, such as Figure 4 As shown, the specific steps of S3 are as follows: S301: Based on the node risk correction dataset, apply the risk score calculation function to extract the risk correction value of each node in sequence, convert the node risk correction value using the calculation formula, and normalize the value range of the node risk score to generate a normalized risk score sequence. The normalization process uses the Min-Max normalization method; Based on the collected node risk correction dataset, the original risk correction data for each node is retrieved one by one. Assuming the risk correction values ​​for nodes A1, A2, and A3 are 0.7338, 0.6425, and 0.8012 respectively, the risk score calculation function is called sequentially for each node. The extracted risk correction value is input into the function for score conversion. Taking a linear function as an example, the risk correction value is directly assigned to the risk score. Node A1 inputs 0.7338 and outputs a risk score of 0.7338; node A2 inputs 0.6425 and outputs 0.6425; node A3 inputs 0.8012 and outputs 0.8012. Subsequently, a score set [0.7338, 0.6425, 0.8012] is formed for the node risk scores. To ensure consistency in score scale, normalization processing is required. The maximum value of the set is 0.8012, and the minimum value is 0.6425. Using the Min-Max normalization method, the normalized score for node A1 is: ; Node A2 is: ; Node A3 is: ; After normalization, the node values ​​are 0.5755, 0, and 1. The normalized node scores are numbered and recorded to form a normalized risk score sequence [0.5755, 0, 1]. This sequence is the basis for subsequent risk weight correction and classification analysis.

[0032] S302: Call the normalized risk score sequence, adjust the distribution of each normalized risk score using the risk weight parameter, set the influence of the weight parameter according to the node attribute, correct the weight of the risk score, and generate a weighted risk trigger score sequence. The normalized risk score sequence is invoked, and the risk weight is adjusted for the normalized risk score of each node. First, a risk weight parameter, a node attribute correction item, a risk influence coefficient between nodes, and a weight parameter under the attribute category are assigned to each node. Specifically, for node A1, the risk weight parameter is set to 0.7, the attribute correction item is set to 0.02, the influence coefficient between A1 and A2 is 0.11, and between A1 and A3 is 0.14. A1 contains two types of attributes, which are assigned weight parameters of 0.35 and 0.21 respectively. The weighted risk trigger score is calculated using the formula: ; in, The weighted risk trigger score represents the node with the ID 'a'. The normalized risk score represents the node with the node number 'a'. The risk weight parameter represents the node with node number 'a'. This represents the attribute modification item for the node with the ID 'a'. represents the risk impact coefficient between node a and node b, and N represents the total number of nodes. M represents the attribute weight parameter of node number a under attribute category c, and M represents the total number of attribute categories involved in node number a.

[0033] The calculation of node A1 will be explained in detail below: First calculate ; The influence terms between nodes are calculated separately: ; Squaring and then adding: The square root yields 0.14; Calculate the denominator: 1 + 0.35 + 0.21 = 1.56; final ; Nodes A2 and A3 are calculated sequentially using their respective parameters. Let's assume node A2: ; A3 node: The specific parameters and calculations are shown in the table below.

[0034] Table 2: Node Parameters and Weighted Risk Scores

[0035] As shown in Table 2, the weighted risk trigger scores for nodes A1, A2, and A3 are 0.3607, 0.1550, and 0.4988, respectively.

[0036] S303: Based on the weighted risk trigger score sequence, the score of each node is partitioned and judged according to the grading standard threshold. The risk score range of the node is labeled and assigned to generate node risk labeling results. Based on the obtained weighted risk trigger score sequence, the score of each node is compared with the threshold of the classification standard for interval determination. Assuming the risk level partitioning standard is: low risk interval is score less than 0.2, medium risk interval is 0.2-0.4, and high risk interval is greater than 0.4, the weighted score of node A1 (0.3607) is compared with the threshold. 0.2 < 0.3607 ≤ 0.4, so A1 is determined to be medium risk. The score of node A2 (0.1550) is less than 0.2, so A2 is determined to be low risk. The score of node A3 (0.4988) is greater than 0.4, so A3 is determined to be high risk. The interval comparison and risk level assignment are performed for each node one by one. Finally, node A1 is marked as "medium risk", A2 as "low risk", and A3 as "high risk", forming a node risk labeling result sequence. The partitioning results are output according to the node number and the corresponding score, and finally a risk level list of nodes is generated to ensure that the risk control partitioning information is complete and traceable.

[0037] Specifically, such as Figure 5 As shown, the specific steps of S4 are as follows: S401: Call the node risk labeling results, obtain the original stress loading data of the facility monitoring nodes on the operation path, classify the path nodes according to the physical location and number of the nodes, integrate the grouped nodes and the original stress loading data, and generate path node stress grouping data. The node risk labeling results are retrieved to obtain the original stress loading data of the facility monitoring nodes on the work path. First, the risk labeling results are matched one-to-one with the facility monitoring node number list. For each node on the path, its physical location coordinates are read and combined with the number. Nodes A1, A2, and A3 are selected as examples, with A1 numbered 001, A2 numbered 002, and A3 numbered 003. Their physical locations are set to (10, 5), (15, 8), and (20, 12) respectively. Node A1 is then retrieved from the original monitoring database in sequence. The stress loading data for nodes A1, A2, and A3 are assumed to be as follows: A1 node stress loading data are 25.5 MPa, 26.0 MPa, and 27.2 MPa; A2 node stress loading data are 21.8 MPa, 22.4 MPa, and 22.9 MPa; and A3 node stress loading data are 28.7 MPa, 29.5 MPa, and 30.1 MPa. These data are categorized according to the physical location of the nodes. Nodes located close to each other (e.g., less than 10 meters apart) are grouped together. The distance between A1 and A2 is... If the distance is less than 10 meters, then A1 and A2 are grouped together, and A3 is grouped separately because it is located far away. For each group of nodes, their numbers and original stress loading data are integrated to form a grouped data table, as shown in Table 3. The number, location, and original stress loading data of the path nodes are accurately recorded in the table to ensure that the node information in each group is complete. The original stress loading data of each group of nodes are arranged in time series to facilitate subsequent sliding window analysis.

[0038] Table 3: Stress Grouping Data for Path Nodes

[0039] As shown in Table 3, nodes A1 and A2 are assigned to group 1, and node A3 is assigned to group 2. The node information and stress loading data in each group have been clearly archived, and the path node stress group data are finally obtained.

[0040] S402: Call the stress grouping data of the path nodes, use a fixed-length sliding window, extract the risk score and stress loading peak value in each window for the data of each group of nodes, and use the traversal method to count the risk score and stress peak value data pairs under the same window to obtain the joint feature value range of the window. The stress grouping data of the path nodes is retrieved using a fixed-length sliding window. The data for each group of nodes is traversed within this window. Assuming the sliding window length is set to 2, data from two consecutive time points are selected each time. First, the window is divided for group 1 (A1, A2). The stress loading data for node A1 at time points 1 and 2 are 25.5 MPa and 26.0 MPa, respectively, and for node A2, they are 21.8 MPa and 22.4 MPa. The stress data combination for nodes A1 and A2 corresponding to window 1 is (25.5, 21.8) and (26.0, 22.4). Simultaneously, for each node, the risk score is retrieved within the window. Assuming the risk score for A1 is 0.5755, for A2 it is 0, and for A3 it is 1, the data for group 1 in window 1 has a risk score of 0 for A1. 5755, A2 risk score 0, corresponding to stress peak values ​​of 25.5 and 21.8 MPa respectively. For node in group 2 (A3), stress loading data of 28.7 and 29.5 MPa are taken at time points 1 and 2, with a risk score of 1. The two data points are (1, 28.7) and (1, 29.5) respectively. This method is used to traverse the sliding window. The risk score of each node in each window is statistically paired with the stress loading peak value to form the joint characteristic value interval of the window. Specifically, taking the first window of group 1 as an example, the total number of nodes in the window is 2. The A1 risk score is paired with the stress peak values ​​(0.5755, 25.5) and A2 (0, 21.8) respectively. The stress peak values ​​are 25.5 MPa (A1) and 21.8 MPa (A2). The joint characteristic value interval of the window is calculated using the formula: ; in, Representative node group and sliding window The joint eigenvalue interval, Representing the The first node in the group The risk score of each node reflects the degree of hidden danger or potential problem of that node in a specific environment. Representing the The first node in the group The absolute value of the risk score for each node is used to remove the positive or negative influence of the risk score and ensure the stability of the calculation results. Representing the The first sliding window The peak stress loading at each node represents the maximum load or pressure that the node withstands within a specific time window. This represents the total number of nodes within the sliding window, used for normalization calculations and adjusting results to reduce the impact of bias in the number of nodes.

[0041] Substitute the above parameters into the calculation: Multiply by the cumulative term: ; ; Total of numerators: ; Denominator: ; ; The results show that the joint eigenvalue of the first window in group 1 is 0.309, representing the joint intensity of risk stress in the group nodes under this window. Subsequent steps will perform similar calculations on the windows and groups to obtain the complete range of joint eigenvalues ​​for the windows. The advantage of the formula is that by introducing the term-by-term product of the absolute value of the risk score and the peak stress, the risk superposition effect is comprehensively reflected in the numerator. At the same time, by correcting the peak stress and the number of nodes in the denominator, a balanced integration of risk and stress is achieved, improving the discriminative ability of the comprehensive eigenvalue.

[0042] S403: Based on the joint feature value interval of the window, the K-means clustering algorithm is used to classify the node window combination data, summarize the risk scores and stress peak data of the multi-category node groups, and select the node data with large risk scores and stress peak averages from the clustering results to obtain the construction period risk node list; By combining the joint eigenvalue intervals of the windows, the data from multiple node window combinations are grouped and classified. First, the joint eigenvalue results of the aforementioned groups and windows are organized into a one-dimensional array or a two-dimensional matrix. Each group of data is evaluated sequentially, and the node window group with higher joint eigenvalues ​​is selected. Assuming the joint eigenvalue intervals are 0.309, 0.322, and 0.341 for the three windows in group 1, and 0.468, 0.495, and 0.510 for the three windows in group 2, the eigenvalues ​​are compared and sorted according to their numerical values. Values ​​with joint eigenvalues ​​higher than 0.45 are classified as high-risk. The risk score is divided into two zones: 0.3-0.45 is classified as medium risk zone, and less than 0.3 is classified as low risk zone. For high risk zones (such as window 2 and window 3 in group 2, with joint feature values ​​of 0.495 and 0.510 respectively), the risk scores and stress peak values ​​of the corresponding nodes are statistically analyzed. The node number, window number and corresponding original data are extracted to form a risk node list. The list includes information such as node number, group number, window number, risk score, stress peak value, and joint feature value. Node window groups with a value higher than 0.45 are all included in the risk node list as the screening result for risk nodes during the construction period.

[0043] Specifically, such as Figure 6 As shown, the specific steps of S5 are as follows: S501: Obtain the elastic recovery ratio and stress loading peak value of nodes in the risk node list during the construction period, calculate the node ratio and compare it with the risk elasticity threshold, filter nodes with ratio values ​​greater than the threshold, and generate an elasticity threshold filter node ratio range. The risk resilience threshold is the minimum ratio standard for measuring whether the resilience recovery capability of risk nodes during the construction period meets the standard. It is set at 0.75. The risk level is divided into three stages based on the comparison between the weighted risk trigger score and the threshold. To obtain the elastic recovery ratio and peak stress loading of nodes in the risk node list during the construction period, first retrieve the node number and its corresponding elastic recovery ratio from the risk node list. and peak stress loading Taking nodes A1, A2, and A3 as examples, the elastic recovery ratio of A1 is... The peak stress loading is The elastic recovery ratio of A2 is The peak stress loading is The elastic recovery ratio of A3 is The peak stress loading is The node list is traversed sequentially, and for each node, the ratio of elastic recovery ratio to peak stress loading is calculated, specifically as follows: ; Taking A1 as an example, A2 is A3 is ; The ratios of the nodes are then organized into arrays or tables for interval recording. Next, the risk resilience threshold is called to set the threshold. Set as The ratio for each node With threshold Compare and determine the node ratio Is it greater than the threshold? Nodes with a ratio greater than a threshold are considered qualified nodes for elasticity; otherwise, they are considered unqualified nodes. For example, if the ratios of A1, A2, and A3 are all less than a threshold... If the ratio is greater than a certain threshold, then all nodes are considered unqualified in terms of elasticity. Nodes are then filtered according to the elasticity threshold, and those with a ratio greater than a certain threshold are selected. The node number, elastic recovery ratio, stress loading peak value, and ratio are recorded in the "Elastic Threshold Filtering Node Ratio Range". Nodes that do not meet the conditions are not included in this range. The node data is given in the table below, as shown in Table 4.

[0044] Table 4: Screening Table for Nodal Elastic Recovery Ratio and Peak Stress Loading

[0045] As shown in Table 4, the node ratios are all below the elasticity threshold. None of them entered the range of nodes filtered by the elasticity threshold. The elasticity threshold is the minimum standard for measuring the resilience of risk nodes during the construction period, and its set value is... Combining industry-recommended standards with real-time elastic distribution settings for nodes, risk levels are classified based on a comparison of the weighted risk score with the range of this threshold; scores below this threshold are considered risk-averse. The range that does not meet the standard is higher than the standard. To achieve the target range, the final range of elastic threshold filtering node ratios is formed.

[0046] S502: Call the elastic threshold to filter the node ratio range, use the spatial clustering method to classify the spatial distribution information of the nodes, and combine the node coordinates and spatial clustering results to quantify the concentration and distribution characteristics of the spatial distribution of nodes under the clustering, and obtain the node distribution characteristic coefficient. The spatial clustering method uses the DBSCAN density clustering algorithm to perform spatial partitioning based on node geographic coordinates and risk indicators; To filter the node ratio range using the elastic threshold, first collect the location information of the nodes that meet the criteria, and obtain the node coordinate data. Assuming node numbers A4 and A5 are nodes that meet the elastic threshold, the coordinates of A4... A5 coordinates The geographic coordinates of these two nodes are organized into vectors. The node set is then traversed sequentially, and the Euclidean distance between the coordinate points is calculated as follows: ; The distance between A4 and A5 is The distance matrix of nodes is organized, and spatial grouping is performed based on the geographical distribution of nodes. The node set is traversed, and the spatial distance is compared with the preset density parameter for each pair. Comparison, Set as If the distance between two nodes is less than If they are in the same spatial cluster group, then A4 and A5 are spatially distant. If nodes are grouped into the same cluster, the number of nodes in each cluster is counted, and the spatial distribution concentration of nodes in each group is calculated. The ratio of the maximum distance to the average distance of nodes in each group is taken as the node distribution characteristic coefficient. Taking A4 and A5 as examples, the maximum distance The average distance is also , The node feature coefficients under clustering are archived and organized into a distribution feature coefficient array. The spatial distribution information and its feature coefficients are recorded. The larger the node distribution feature coefficient, the more dispersed the spatial distribution of the nodes, and vice versa. After the node distribution feature coefficient values ​​are summarized, a node distribution feature coefficient set is formed.

[0047] S503: Based on the node distribution characteristic coefficient, the risk level classification of node distribution characteristics under spatial clustering is determined, and combined with the real-time risk attributes and distribution patterns of nodes during the construction period, the risk level determination result for the planning and construction period is generated. Based on the node distribution characteristic coefficients, the risk level of node distribution characteristics under spatial clustering is classified and determined. First, the node distribution characteristic coefficients obtained in the previous stage are arranged sequentially, and the classification intervals are set. This is a low-risk distribution area. This area is classified as a medium-risk area. For high-risk distribution areas, the distribution characteristic coefficients of each cluster are individually determined to determine their respective intervals. Assuming the A4 / A5 cluster... If a set of nodes is classified as a low-risk area, and its distribution characteristic coefficient is... If the distribution characteristic coefficient of another group of nodes is classified as medium risk, then... The risk level of each cluster is statistically analyzed, and the risk level and node information of each cluster are compiled. Combined with the real-time risk attributes of the nodes during the construction period, the current risk level of each node is determined, and finally, the risk level determination result of the planning and construction period, including cluster number, node number, distribution characteristic coefficient, and risk level, is generated.

[0048] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system, characterized in that... Includes the following steps: S1: Obtain the structural vibration frequency time series data of overhead contact network, roadside poles, and power transmission and transformation line facilities under wind load, rainfall, and temperature disturbance during the planning and construction period, perform segmented differential processing, use the isolated forest algorithm to identify abnormal mutation points of monitoring nodes, and output frequency anomaly point data; S2: For the frequency anomaly point data, obtain the peak value of structural stress response, material fatigue level and terrain slope parameters for the corresponding time period, and use interval matching and weighted calculation to analyze the node risk correction value of the anomaly point and output the node risk correction dataset. S3: Based on the node risk correction dataset, the risk score calculation function is used to normalize and weight the risk correction values, construct a risk trigger score sequence, and label the facility nodes with risk scores in combination with the grading standard, and output the node risk labeling results; S4: Call the node risk labeling results, extract the original stress loading data of the facility monitoring nodes on the operation path, group the path nodes, use a sliding window to combine the risk score and stress loading peak value, perform cluster analysis on the combined data based on K-means clustering, and output the construction period risk node list; The specific steps of S4 include: S401: Call the node risk labeling result, obtain the original stress loading data of the facility monitoring nodes on the operation path, classify the path nodes according to the physical location and number of the nodes, integrate the grouped nodes and the original stress loading data, and generate path node stress grouping data. S402: Call the stress grouping data of the path nodes, use a fixed-length sliding window, extract the risk score and stress loading peak value in each window for the data of each group of nodes, and use a traversal method to count the risk score and stress peak value data pairs under the same window to obtain the joint feature value interval of the window. S403: Based on the window joint feature value interval, the K-means clustering algorithm is used to classify the node window combination data, summarize the risk scores and stress peak data of multi-category node groups, and filter the node data with large risk scores and stress peak averages from the clustering results to obtain the construction period risk node list; S5: The elastic recovery ratio and stress loading peak value of the nodes in the construction period risk node list are judged, and risk nodes with a ratio greater than the risk elasticity threshold are screened. The spatial clustering method is used to analyze the spatial distribution characteristics of the nodes, and the risk level judgment result of the non-stop energy replenishment zero-carbon transportation system planning and construction period is output. The risk level determination results include the overall risk classification results, spatial distribution characteristics, and elastic recovery ratio.

2. The method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system according to claim 1, characterized in that, The frequency anomaly data includes anomaly frequency values, monitoring node numbers, and anomaly occurrence times. The node risk correction dataset includes structural safety factors, stress response peak values, and fatigue levels. The node risk labeling results include risk classification identifiers, risk scores, and node identifiers. The construction period risk node list includes a list of high-risk nodes, node spatial locations, and cluster numbers.

3. The method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system according to claim 1, characterized in that, The specific steps of S1 include: S101: Obtain the structural vibration frequency time series data of overhead contact network, roadside poles, and power transmission and transformation line facilities under wind load, rainfall, and temperature disturbance during the planning and construction period. Divide the structural vibration frequency time series data into time intervals and perform differential calculation on the continuous frequency values ​​within the intervals to obtain segmented frequency difference values. S102: Call the segmented frequency difference value, organize the segmented difference data of the monitoring nodes, take the difference sequence after node segmentation as the input variable, use the isolated forest algorithm, output the anomaly coefficient based on the difference characteristics of each monitoring node segmentation interval, and obtain the sub-node anomaly coefficient sequence. S103: Based on the sub-node anomaly coefficient sequence, the anomaly coefficients of the monitoring node segment intervals are compared with the discrimination threshold of the isolated forest algorithm, and frequency intervals and node numbers whose anomaly coefficients exceed the set discrimination threshold are filtered to generate frequency anomaly point data.

4. The method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system according to claim 3, is characterized in that... The specific steps of S2 include: S201: Obtain the frequency anomaly point data, filter the structural stress response peak value, material fatigue level and terrain slope parameters of the corresponding time period, pair the parameters according to the time period of the anomaly point, classify them into the corresponding data set, count the interval range of each parameter set, and generate the interval parameter set. S202: Call the set of interval parameters, use interval matching to weight the peak value of structural stress response, material fatigue level and terrain slope parameters in the same interval according to the influence ratio of the parameters in the node risk, calculate the risk value of the outlier point and generate the node risk weighted value. S203: Call the node risk weighting value, collect the risk value under each node according to the weighting result corresponding to all anomalies, and summarize to obtain the node risk correction dataset.

5. The method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system according to claim 4, characterized in that, The specific steps of S3 include: S301: Based on the node risk correction dataset, apply the risk score calculation function to extract the risk correction value of each node in sequence, convert the node risk correction value using the calculation formula, and normalize the value range of the node risk score to generate a normalized risk score sequence. S302: Call the normalized risk score sequence, adjust the distribution of each normalized risk score using the risk weight parameter, set the influence of the weight parameter according to the node attribute, correct the weight of the risk score, and generate a weighted risk trigger score sequence. S303: Based on the weighted risk trigger score sequence, the score of each node and the grading standard threshold are partitioned and judged, the risk score range of the node is labeled and assigned, and the node risk labeling result is generated.

6. The method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system according to claim 5, characterized in that, The normalized risk score refers to the normalization of the risk correction value obtained through the risk score calculation function to obtain a standardized risk score for unified comparison. The risk score calculation function is a mathematical function that outputs the original risk score of a node based on its risk correction value and attributes. The weighted risk trigger score refers to the weighted score obtained by combining the normalized risk score with the node weight parameter, which reflects the real-time risk priority of the node. The grading standard threshold refers to the preset score limit used to distinguish weighted risk trigger scores into differentiated risk levels.

7. The method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system according to claim 1, characterized in that, The window joint feature value refers to the data combination of risk score and stress loading peak value extracted simultaneously within a fixed-length sliding window, reflecting the risk and stress characteristics of the node within a time period. The K-means clustering algorithm groups the joint feature value data of the node windows, where the K value is determined based on the Elbow method and the silhouette coefficient method combined with the distribution characteristics of the training dataset.

8. The method for determining the risk level during the planning and construction phase of a non-stop refueling zero-carbon transportation system according to claim 1, characterized in that, The specific steps of S5 include: S501: Obtain the elastic recovery ratio and stress loading peak value of nodes in the construction period risk node list, calculate the node ratio and compare it with the risk elasticity threshold, filter nodes with ratio values ​​greater than the threshold, and generate an elasticity threshold filter node ratio range. S502: Call the elastic threshold to filter the node ratio range, use spatial clustering method to classify the spatial distribution information of the nodes, combine the node coordinates and spatial clustering results, quantify the concentration degree and distribution characteristics of the spatial distribution of nodes under clustering, and obtain the node distribution characteristic coefficient; S503: Based on the node distribution characteristic coefficients, the risk level of the node distribution characteristics under spatial clustering is classified and judged. Combined with the real-time risk attributes and distribution patterns of the nodes during the construction period, the risk level judgment result of the planning and construction period is generated.

Images