Subway station safety risk assessment method combining analytic hierarchy process and data dimension reduction

Abstract

The invention provides a subway station safety risk assessment method combining analytic hierarchy process and data dimension reduction, and relates to the technical field of underground space safety risk assessment. Risk operation and maintenance data of a subway station are called, analytic hierarchy process based on risk indexes is carried out, an index directed graph is mined, and a reverse assessment mode based on the index directed graph is set. Monitoring and training a risk assessment model; monitoring and determining distributed sensing data of the subway station; carrying out data division and slicing dimension reduction based on common risk points; converting and determining a data cognitive map and transmitting the data cognitive map to the risk assessment model; the method is used for solving the technical problem of how to process a complex risk mechanism and carry out high-accuracy analysis in the prior art so as to realize scientific and efficient risk identification and assessment in a subway station with large passenger flow and a complex structure, and the accuracy and reliability of risk assessment can be effectively improved.

Description

[0001] A subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction Technical Field

[0002] The present invention relates to the technical field of underground space safety risk assessment, and in particular to a subway station safety risk assessment method combining hierarchical analysis with data dimensionality reduction. Background Art

[0003] With the rapid development of subway stations in integrated transportation hubs, they have become increasingly important in urban transportation, resulting in an increasing passenger flow and increased management difficulty. Integrated transportation hubs have huge passenger flow, special passenger flow composition, and complex surrounding environment. Subway stations are often built underground, with complex structures and limited space, which increases the difficulty of evacuation and rescue in the event of emergencies in subway stations in integrated transportation hubs. Therefore, subway stations in integrated transportation hubs have greater potential safety hazards.

[0004] Traditional risk assessment methods mainly rely on qualitative analysis and historical data, lacking comprehensive and accurate assessment capabilities, especially the inability to effectively quantify multiple risk factors in complex environments. Existing technologies mostly focus on conventional subway stations, where passenger flow and spatial structure complexity are relatively small compared to integrated transportation hub subway stations. Existing technologies are relatively weak in analyzing the synergy of multiple risk sources and complex risk transfer processes, and are not suitable for safety risk assessment of integrated transportation hub subway stations.

[0005] Therefore, how to deal with complex risk mechanisms and conduct high-accuracy analysis to achieve scientific and efficient risk identification and assessment in subway stations with large passenger flows and complex structures is a problem that needs to be urgently solved by existing technologies. Summary of the invention

[0006] The present application provides a subway station safety risk assessment method that combines hierarchical analysis with data dimensionality reduction, which is used to solve the technical problem of how to deal with complex risk mechanisms and conduct high-accuracy analysis in the prior art, so as to achieve scientific and efficient risk identification and assessment in subway stations with large passenger flows and complex structures.

[0007] In view of the above problems, the present application provides a subway station safety risk assessment method that combines hierarchical analysis with data dimensionality reduction.

[0008] The present application provides a subway station safety risk assessment method that combines hierarchical analysis with data dimensionality reduction, the method comprising: calling the risk operation and maintenance data of the subway station, performing hierarchical analysis based on risk indicators, mining an indicator directed graph, wherein the indicator directed graph is a mapping structure with multiple layers of indicators refined and marked with indicator weights; setting a reverse assessment method based on the indicator directed graph, and obtaining a resilience evaluation matrix of the subway station, supervising the training of a risk assessment model, wherein the reverse assessment method is a hierarchical probability recursion; monitoring and determining the distributed sensor data of the subway station, performing data partitioning and slice dimensionality reduction based on common risk points, and converting and determining a data cognitive map; transmitting the data cognitive map to the risk assessment model, performing probability recursive calculation and extreme value analysis, and determining a risk assessment result, wherein the risk assessment result is represented as a triangular fuzzy number of minimum value-calculated value-maximum value.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages: The embodiment of the present application provides a subway station safety risk assessment method that combines hierarchical analysis and data dimensionality reduction. The method calls the risk operation and maintenance data of the subway station, performs hierarchical analysis based on risk indicators, mines the indicator directed graph, sets an inverse assessment method based on the indicator directed graph, obtains the resilience evaluation matrix of the subway station, supervises the training of the risk assessment model, and the inverse assessment method is hierarchical probability recursion; monitors and determines the distributed sensor data of the subway station, performs data partitioning and slice dimensionality reduction based on common risk points, and converts to determine the data cognitive map; transmits the data cognitive map to the risk assessment model, performs probability recursive calculation and extreme value analysis, and determines the risk assessment result. The risk assessment result is characterized as a triangular fuzzy number of minimum value-calculated value-maximum value, which is used to solve the technical problem of how to deal with complex risk mechanisms and perform high-accuracy analysis in the prior art, so as to realize scientific and efficient risk identification and assessment in subway stations with large passenger flow and complex structure, and can effectively improve the accuracy and reliability of risk assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A schematic flow chart of a subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction is provided for this application; Figure 2 A schematic diagram of the data cognitive map acquisition process in a subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction is provided for this application; Figure 3 A flowchart of probability recursive calculation and extreme value analysis in a subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction is provided for this application. DETAILED DESCRIPTION

[0011] The present application provides a subway station safety risk assessment method that combines hierarchical analysis with data dimensionality reduction, calls the risk operation and maintenance data of the subway station, performs hierarchical analysis based on risk indicators, mines the indicator directed graph, sets a reverse assessment method based on the indicator directed graph, obtains the resilience evaluation matrix of the subway station, supervises the training of the risk assessment model, monitors and determines the distributed sensor data of the subway station, performs data partitioning and slicing dimensionality reduction based on common risk points, converts and determines the data cognitive graph and transmits it to the risk assessment model, performs probabilistic recursive calculation and extreme value analysis, and determines the risk assessment result. The application is used to solve the technical problem of how to deal with complex risk mechanisms and perform high-accuracy analysis in the prior art, so as to realize scientific and efficient risk identification and assessment in subway stations with large passenger flow and complex structure.

[0012] Example: Figure 1 As shown, the present application provides a subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction, the method comprising: S1: Call the risk operation and maintenance data of the subway station, perform hierarchical analysis based on risk indicators, and mine the indicator directed graph, wherein the indicator directed graph is a mapping structure of multi-layer indicator refinement, marked with indicator weights; In the embodiment of the present application, the risk operation and maintenance data of the subway station includes but is not limited to the equipment operation data, personnel management information, emergency records, historical accident data and environmental monitoring data of the subway station, which can be obtained through literature retrieval, on-site investigation and questionnaire survey results. By comprehensively collecting and analyzing them, data support can be provided for the safety risk assessment of the subway station.

[0013] Furthermore, for the risk operation and maintenance data, hierarchical analysis is performed to mine risk indicators. The following is a specific embodiment: taking the subway station as the research object, the safety risk events of the comprehensive transportation hub subway station are summarized into five types of safety risk events: crowded stampede events, fire events, flood events, public health events and large-scale passenger detention events. Then, the various types of safety risk events are divided layer by layer to complete the establishment of the entire evaluation index system of the comprehensive transportation hub subway station. Specifically: the safety of the comprehensive transportation hub subway station is used as the first-level indicator of safety risk evaluation; fire safety, crowded stampede accident safety, public health event safety, large-scale passenger detention event safety, and flood safety are used as second-level indicators of safety risk evaluation; then, based on the risk factor identification results of each second-level indicator from the five aspects of building, personnel, equipment, management and environment, the construction of the third and fourth level indicators is completed according to their own characteristics; the hierarchical establishment of the entire evaluation index system for the comprehensive transportation hub subway station is more in line with the actual risk events of the comprehensive transportation hub subway station.

[0014] Based on this, the indicator directed graph is constructed. The indicator directed graph is a graph structure that connects various risk indicators in a directed manner according to a certain hierarchical structure to form a multi-level mapping relationship from the root node to the leaf node. Among them, the nodes of each layer represent risk indicators, and the direction of the edge represents the dependency and impact path between risk indicators. For example, equipment failure may affect personnel safety management, thereby causing instability in system operation. The relationship between each risk indicator and other indicators is shown in detail in the graph, and the position of each indicator in the graph and its weight reflect the relative importance of the indicator in risk assessment.

[0015] The indicator weight refers to the contribution of each indicator to the risk assessment in the multi-layer indicator refinement mapping structure. For example, equipment failure may be an important factor leading to subway station safety incidents, so the weight of this indicator in the overall assessment may be higher; while risk indicators such as personnel management are equally important, but their weight may be slightly lower. These weight values ​​help to accurately quantify the impact of different risk factors in subway safety assessments and provide a basis for subsequent risk assessments and emergency responses.

[0016] By conducting a hierarchical analysis based on risk indicators on the risk operation and maintenance data of subway stations in comprehensive transportation hubs, an indicator directed graph with a multi-layer refined mapping structure is formed. This process provides a clear and systematic analysis framework for the safety risk assessment of subway stations, and provides a scientific basis for further risk prediction and management measures by effectively identifying and quantifying the weights of various risk indicators.

[0017] Among them, the mining indicator directed graph, step S1 of this application includes: Identify the risk operation and maintenance data, perform event clustering to determine the event clustering cluster, and construct a root event layer; set indicator layer elements, traverse the event clustering cluster to perform multi-layer mining mapping, and construct multiple leaf indicator layers; perform inter-layer mapping between the root event layer and the multiple leaf indicator layers, perform weight distribution based on indicator membership, and generate the indicator directed graph.

[0018] In the embodiment of the present application, the risk operation and maintenance data of the subway station is first identified and extracted, including equipment failure records, personnel operation conditions, environmental change monitoring, accident and event reports, etc. Preferably, data preprocessing and cleaning are performed in advance, such as screening the data based on a preset frequency, removing accidental event data, etc., to ensure the accuracy and reliability of the data.

[0019] Next, event clustering analysis is performed to determine event clusters. Specifically, event clustering refers to grouping similar events according to certain similarity metrics to form multiple clusters. In the risk assessment of subway stations, event clusters can help identify different types of safety risk events, such as stampede incidents, fire incidents, flood incidents, public health incidents, and large-scale passenger detention incidents. Among them, clustering processing usually adopts a distance or similarity-based measurement method to automatically classify similar risk events into one category. Potential safety risk event categories can be clearly divided, providing the necessary basis for building subsequent risk models.

[0020] After completing the event clustering, based on these event clusters, the determined risk event types are used as the root event layer to be constructed. Subsequently, the indicator layer elements are set, and each event cluster is traversed for multi-layer mining mapping. The indicator layer elements refer to the specific indicators used to quantify and describe each risk factor in the risk assessment process. These indicators can be various factors such as equipment operation status, personnel management status, external environment changes, etc.

[0021] For example, the safety of subway stations is divided into fire safety, safety of crowded stampede accidents, safety of public health events, safety of large-scale passenger stranding events, and safety of floods as safety risk assessments, and the indicators are further subdivided from aspects such as buildings, personnel, equipment, management, and environment, and then the levels are refined in turn. By traversing the clusters and combining the actual subway operation data, each cluster is mapped at multiple levels, that is, the complex risk factors are disassembled into multiple sub-level indicators, and they are analyzed layer by layer to construct the multiple leaf indicator layers. This process can help further clarify the specific manifestations of each risk factor and its specific role in risk assessment.

[0022] Through multi-level mining, the internal structure of risk events is refined, thereby improving the granularity and accuracy of risk assessment.

[0023] Next, the root event layer is mapped to multiple leaf indicator layers, and weights are assigned through indicator membership. Interlayer mapping refers to associating indicators at different levels to form a layer-by-layer transmission relationship from the root event to the leaf indicator. The indicator membership reflects the importance or influence of each specific indicator in its level. By calculating the membership, the contribution of each risk indicator to the overall risk assessment can be quantified. The membership is positively correlated with the weight value. The setting of weight distribution ensures that each level and each indicator occupies a reasonable position and role in risk assessment, thereby achieving the rationality and scientificity of multi-level risk analysis.

[0024] In summary, the indicator directed graph is a multi-level graph structure. Each node in each layer of the graph represents a risk indicator, and the directed edges between nodes represent the impact and dependency relationship between the indicators. The graph structure clearly shows the hierarchical relationship from root events to leaf indicators, and quantifies the importance of each indicator through weight allocation, providing a scientific basis for subsequent risk assessment and decision support.

[0025] S2: setting a reverse evaluation method based on the index directed graph, obtaining a resilience evaluation matrix of the subway station, and supervising the training of a risk assessment model, wherein the reverse evaluation method is a hierarchical probability recursion; In the embodiment of the present application, in order to further improve the accuracy and operability of the safety risk assessment of subway stations, a reverse assessment method based on the index directed graph is first set. The so-called reverse assessment method refers to the reverse recursive method, starting from the minimum index unit to conduct risk probability analysis, and gradually conducting upper-level recursive analysis until the analysis of risk events is completed. At the same time, it can effectively discover relatively high-risk indicators in the recursive chain, so as to accurately identify possible risk sources.

[0026] Specifically, the hierarchical probability recursion method is adopted. By analyzing and deducing the probability of risk indicators at different levels, starting from the leaf indicator layer of the smallest indicator unit, based on the mapping relationship between the indicators between layers, the risk probability is transferred upward layer by layer, and finally the various potential risk probabilities of the subway station in the overall operation process are obtained. In this process, the impact of each risk indicator is quantified as a probability value, and through the probability transfer mechanism between levels, it is gradually reflected in the risk assessment at a higher level.

[0027] Next, the resilience evaluation matrix of the subway station is obtained. The resilience evaluation matrix is ​​used to quantify the response and recovery capabilities of the subway station in the comprehensive transportation hub when facing various risk disturbances. Resilience refers to the ability of the subway station to maintain or resume normal operation after being disturbed or impacted. In practical applications, the resilience evaluation matrix evaluates the performance of different links of the subway station and constructs a multidimensional matrix considering the impact of different risk disturbances on system performance. The matrix not only covers the relationship between various risk indicators, but also quantitatively describes the performance of each indicator under specific disturbances and the recovery capacity of the subway station.

[0028] After obtaining the resilience evaluation matrix, the risk assessment model is further supervised and trained. Specifically, the recursive calculation of risk events based on the index directed graph is used as the first training goal, and the impact response based on the resilience evaluation matrix is ​​used as the second training goal. By comparing known historical risk events with the evaluation results in a sample training manner, the parameters of the evaluation model are optimized until the convergence conditions are met, such as the preset analysis accuracy, so that the risk assessment model can learn the probability distribution of different risk events and the evolution law of potential risks from a large amount of historical data, thereby improving the accuracy and robustness of the model. In this process, the model will continuously adjust its parameters to minimize the prediction error and gradually improve its ability to predict the risks of complex subway stations.

[0029] In summary, by setting up a reverse assessment method based on an indicator directed graph, that is, using hierarchical probability recursion to recursively calculate the risk, and combining the resilience evaluation matrix of the subway station, the actual risk conditions caused by the system are analyzed, which effectively improves the accuracy and operability of the subway station safety risk assessment. Through this method, potential risk sources can be accurately identified, system resilience can be quantified, and a scientific basis can be provided for subsequent safety management and emergency response.

[0030] After obtaining the resilience evaluation matrix of the subway station, step S2 of the present application includes: For subway stations, the resilience evaluation matrix is ​​constructed, in which indicators are determined based on risk disturbance-system performance indicators; based on the subway risk event mechanism, a mapping between the indicator directed graph and the resilience evaluation matrix is ​​established, and risk disturbance degradation is used as a training target.

[0031] Specifically, for the subway stations of the comprehensive transportation hub subway stations, the resilience evaluation matrix is ​​constructed according to the system functional characteristics. The resilience evaluation matrix is ​​a tool for quantitatively evaluating the ability of the subway station system to maintain normal operation or rapid recovery when facing various potential risks or disturbances. Specifically, the construction of the resilience evaluation matrix is ​​based on risk disturbances and system performance indicators. Risk disturbances usually refer to external or internal events that may interfere with subway stations, such as equipment failures, natural disasters, accidents, etc.; system performance indicators are parameters that measure the operating status of the subway station system, such as the normal operating time of the equipment, service frequency, passenger safety, etc. Based on these risk disturbances and performance indicators, through reasonable analysis and evaluation, a multidimensional matrix reflecting the overall resilience performance of the subway station is constructed.

[0032] On this basis, based on the subway risk event mechanism, the mapping relationship between the indicator directed graph and the resilience evaluation matrix is ​​further established. The subway risk event mechanism refers to the causal relationship and mutual influence between different risk events and between risk events and the human evaluation matrix in the subway station. For example, equipment failure may cause the train to stop running, thereby affecting passenger flow and normal operation of the system; at the same time, when this risk event exists, the system resilience evaluation index responds to it. The clear analysis of these risk events and their mechanisms helps to determine the dependency between various risk indicators and between risk indicators and resilience indicators. Based on this, the indicator directed graph and the resilience evaluation matrix are mapped and associated. In this process, the risk event mechanism is the core basis for constructing the indicator directed graph, which can reflect the interaction and influence path of various risk factors in the actual operation of subway stations.

[0033] Specifically, in this mapping process, each node (i.e., risk indicator) in the index directed graph is connected to the corresponding part in the resilience evaluation matrix, ensuring that the performance of the subway station under different disturbance scenarios can be reflected by the values ​​in the matrix. This mapping relationship can provide a more detailed and quantitative assessment for the overall resilience analysis of the subway station, and further reveal the specific impact of each risk factor on the system's resilience.

[0034] In addition, the risk disturbance downgrade, that is, the mapped resilience index, is used as the training target for the risk reduction degree of the corresponding risk index to further optimize the risk assessment model. Risk disturbance downgrade refers to classifying the risk events that occur and grading them according to their impact on the subway station. For example, under the influence of the resilience index, the degree of risk events can be reduced to a certain extent. In order to better learn how to adjust the assessment strategy according to different levels of risk disturbances, so as to improve the robustness and flexibility of the risk assessment model, so that the model can respond accurately to different risk scenarios.

[0035] In summary, for subway stations, by constructing a resilience evaluation matrix and determining indicators based on risk disturbances and system performance indicators, and combining the subway risk event mechanism to establish a mapping between the indicator directed graph and the resilience evaluation matrix, and taking risk disturbance degradation as a training target, the accuracy and reliability of subway station risk assessment are effectively enhanced. This comprehensive evaluation method provides scientific and quantitative decision-making support for subway safety management and provides an important basis for the formulation of emergency response measures.

[0036] S3: Monitor and determine the distributed sensor data of subway stations, divide and slice the data based on common risk points, and convert and determine the data cognitive map; In the implementation process of the present invention, the distributed sensor data of the subway station is first monitored and determined. Distributed sensor data refers to real-time data collected by sensors installed at various key locations in the subway station. These sensors may include temperature sensors, humidity sensors, equipment operation monitoring sensors, personnel flow monitoring sensors, etc. The data of these sensors provide detailed information about various systems and links of the subway station, which can reflect the real-time operation status of the subway station and potential safety hazards. Through real-time monitoring of these sensor data, abnormal information that may affect the safe operation of the subway can be obtained in a timely manner, thereby providing data support for subsequent risk assessment and emergency management.

[0037] Next, the data is divided and sliced ​​based on common risk points for dimensionality reduction. Common risk points refer to points in subway stations where different risk indicators intersect or interact with each other. These points are usually key factors where multiple risk factors jointly affect system safety. For example, equipment failure may occur at the same time as human operating errors and environmental changes, leading to serious problems in the system. By identifying and determining these common risk points, complex sensor data can be effectively divided into multiple data groups with strong correlation. The key to this process is to identify the correlation in the risk data, divide the data reasonably, eliminate irrelevant noise data, and ensure that only data that has a practical impact on risk assessment is focused on in the analysis.

[0038] Data slicing and dimensionality reduction refers to the dimensionality reduction processing of the divided data to remove redundant, high-dimensional, and unnecessary variables, thereby simplifying the data structure and improving computational efficiency. Exemplarily, dimensionality reduction technology compresses data through mathematical methods (such as principal component analysis (PCA), linear discriminant analysis (LDA), etc.), converting the original high-dimensional data into a low-dimensional representation, thereby reducing computational complexity while retaining the most useful information for risk assessment. For example, there may be multiple sensor data dimensions in a subway station, but not all data dimensions have a significant impact on risk assessment. Through dimensionality reduction processing, the main influencing factors can be retained, making subsequent analysis more efficient and accurate.

[0039] The distributed sensor data after data partitioning and slicing dimension reduction is used as the data cognitive graph. The data cognitive graph is a structured graph representation after the effective data in the distributed sensor data is classified, which is composed of data nodes and line edge data relationships. For example, equipment failure and personnel management data, environmental monitoring data, etc. may be connected by edges to reflect how they synergistically affect safety risks in subway stations.

[0040] In summary, by monitoring the distributed sensor data of the subway stations in the integrated transportation hub, the data is divided and sliced ​​based on common risk points, and finally converted into a data cognitive map. This process effectively extracts key information related to the safety risks of the subway station, simplifies the data structure through dimensionality reduction, and provides a clear and structured basis for subsequent risk assessment and decision-making.

[0041] Among them, Figure 2 As shown, data partitioning and slicing dimensionality reduction are performed based on common risk points, and the conversion determines the data cognitive map. Step S3 of this application includes: The distributed sensor data is obtained, and common risk points are defined based on top risk, intermediate risk and root risk. Data is divided to determine N data groups, wherein the root risk is an uncontrollable sudden cause, the top risk is an equipment operation type, and the intermediate risk is a personnel management type. The N data groups are traversed to perform data slicing and dimensionality reduction to determine N data cognitive maps.

[0042] Specifically, the distributed sensor data comes from the real-time monitoring of various sensors in the subway station, including information on equipment operation status, environmental factors, personnel flow, etc. The distributed sensor network is widely distributed in key locations of subway stations, collecting various physical quantities in real time, such as temperature and humidity, vibration, noise, power consumption, personnel location and other data. These data provide basic information for subsequent safety risk assessment. Through the centralized management and analysis of these sensor data, it is possible to provide all-round data support for the risk assessment model.

[0043] Next, based on top risks, intermediate risks and root risks, common risk points are defined and data is divided, that is, the data parts corresponding to the top risks, intermediate risks and root risks that are correlated are divided into a group, that is, the divided data of common risk points related to the same risk. In the present invention, top risks, intermediate risks and root risks represent risk factors at different levels respectively. Top risks are usually related to the operating status of equipment, such as equipment failure, aging of facilities and other problems, which are risks that directly affect the operation of subway stations; intermediate risks are related to personnel management, such as operational errors, management negligence and other factors, which may aggravate the occurrence of risks in the context of equipment failure; root risks are uncontrollable sudden inducements, usually external factors or extreme events, such as natural disasters, sudden accidents, etc. These risk factors may cause more serious consequences under the influence of other risk levels.

[0044] The distributed sensor data is divided based on common risk points and reasonably classified into multiple categories, each data group corresponds to the risk characteristics of the same source. For example, a group of data may focus on reflecting the top risks related to equipment failure and the intermediate risks related to personnel management, which are all factors that cause a risk event. Through this division, targeted analysis can be performed on different types of risks.

[0045] After N data groups are determined, data slicing and dimensionality reduction are performed next. Data slicing and dimensionality reduction refers to processing each data group to remove redundant dimensions and retain the most useful key information. The purpose of data dimensionality reduction is to simplify the data structure and reduce computational complexity while ensuring that the core information relied on in risk assessment is retained. The dimensionality reduction process processes each data group through mathematical algorithms (such as principal component analysis PCA or t-SNE techniques) to convert high-dimensional data into low-dimensional representations, thereby removing noise and unnecessary complexity. For example, in some data groups, sensors may record too much environmental change data, which has little impact on risk assessment. After dimensionality reduction, only the data features most relevant to equipment failure and personnel management are retained.

[0046] After slicing and dimensionality reduction, N data cognitive maps were determined. Each data cognitive map reflects the relevant collected data of a specific type of risk. For example, a data cognitive map may include sensor data such as equipment failure and personnel management that caused the risk event. For example, the cognitive map of equipment failure may show the data and relationship between equipment status and operation management.

[0047] In summary, the present invention obtains distributed sensor data from subway stations, and based on the hierarchical structure of top risk, intermediate risk and root risk, defines common risk points to divide the data, and then determines N data groups. Data slicing and dimensionality reduction are performed on each data group, and N data cognitive maps are finally determined. This process can effectively extract potential risk factors and their interrelationships in subway stations, simplify the data structure, and ensure the efficiency and accuracy of risk assessment. Through the construction of data cognitive maps, visual and clear support is provided for subsequent risk analysis, assessment and decision-making.

[0048] S4: Transmitting the data cognitive map to the risk assessment model, performing probability recursive calculation and extreme value analysis, and determining the risk assessment result, wherein the risk assessment result is characterized as a triangular fuzzy number of minimum value-calculated value-maximum value.

[0049] In the implementation process of the present invention, the data cognitive map is firstly transmitted to the risk assessment model. The data cognitive map provides complete input data for the risk assessment model by structuredly representing the sensor data and their interrelationships.

[0050] Next, a probability recursive calculation is performed in the risk assessment model. Probability recursion is a calculation method that gradually derives the probability of an event through a recursive formula, and is usually used in multi-level, multi-dimensional complex systems. In the present invention, the risk assessment model is based on the information in the indicator directed graph and the data cognitive graph, and performs recursive calculations according to certain probability rules. Starting from the leaf indicator layer, the risk indicator probability of the leaf indicator layer of the minimum unit indicator is first determined, and based on the inter-layer mapping relationship, the probability of each risk indicator is passed upward layer by layer. Specifically, based on the inter-layer mapping indicator, a weighted calculation is performed with the risk indicator probability to determine the probability of the corresponding indicator in the previous leaf indicator layer. Through recursive calculation, the probability of risk occurrence at each level is calculated, and the level of overall risk of the system is comprehensively evaluated.

[0051] On the basis of probability recursive calculation, the probability of occurrence of risk events is determined, and extreme value analysis is further performed. Extreme value analysis mainly analyzes the maximum and minimum values ​​of the calculated probability of risk events, specifically including the calculation of the minimum value based on the resilience evaluation matrix and the calculation of the maximum value based on the preset probability scale, i.e., the tolerance interval, to determine the triangular fuzzy number characterized by the minimum value-calculated value-maximum value.

[0052] Through the characterization and analysis of triangular fuzzy numbers, the probability value is converted into a probability interval. The risk assessment results not only provide a quantitative analysis of risk events, but also reflect the uncertainty in the assessment process, making the risk assessment results more comprehensive and flexible. For example, in the assessment of a specific risk event, the minimum value may represent the lowest probability of the event, the calculated value represents the most likely probability of occurrence, and the maximum value represents the highest probability of the event.

[0053] In summary, by combining probability recursion and extreme value analysis, the risk assessment results are represented in the form of triangular fuzzy numbers, thus providing a scientific and accurate assessment basis for the risk management of subway stations. This method can comprehensively and dynamically reflect the risk level of subway stations under different conditions and provide strong support for safety management decisions.

[0054] Among them, Figure 3 As shown, the probability recursive calculation and extreme value analysis are performed, and step S4 of the present application includes: The N data cognitive graphs are transmitted to the risk assessment model, and parallel calculations are performed in combination with the indicator directed graph to determine N event probabilities; for the N event probabilities, risk resistance compensation is performed in combination with the resilience evaluation matrix to determine N probability minima; for the N event probabilities, a preset probability scale is expanded to determine N probability maxima, wherein the preset probability scale represents the uncertain elements of the event process; the N event probabilities, N probability minima and N probability maxima are combined to determine N event evaluation results.

[0055] In the embodiment of the present application, the N data cognitive graphs are first transmitted to the risk assessment model. Each data cognitive graph displays the subway sensor monitoring data in a structured manner, and is combined with the indicator directed graph for parallel calculation to determine the probability of N events. That is, based on the indicator directed graph, risk assessment is performed on each data cognitive graph.

[0056] The indicator directed graph reveals the hierarchical relationship between risk factors through the hierarchical representation of different risk indicators. In the risk assessment model, a parallel computing method is used to associate the node data in the data cognitive graph with the evaluation indicator nodes in the indicator directed graph, and recursive calculation is performed from the minimum indicator probability to obtain the probability of occurrence of each risk event. Parallel computing can greatly improve computing efficiency because the calculation of the probability of each event can be performed simultaneously in different processing units. Using the indicators in the indicator directed graph as the judgment criteria, the data in the data cognitive graph is subjected to risk indicator evaluation and hierarchical recursive calculation, and the probability of occurrence of N risk events is accurately calculated.

[0057] For the N determined event probabilities, the resilience evaluation matrix is ​​then combined to perform risk compensation and determine N probability minima. The resilience evaluation matrix is ​​a tool for evaluating the recovery and risk resistance of subway stations when encountering risk disturbances. By combining the resilience evaluation matrix, the risk assessment model can analyze the system's recovery capabilities after various risk events occur, and then perform risk compensation. The compensation process adjusts the probability value of each event so that it reflects the resilience level of the subway station when facing different risk scenarios. Finally, the N probability minima obtained after compensation represent the probability of risk events that may occur based on the system's risk resistance processing.

[0058] Subsequently, for the determined N event probabilities, the preset probability scale is expanded to determine N probability maxima. The preset probability scale is a parameter used to quantify the uncertainty factors in the event process, which reflects the possible range of event occurrence and the uncertainty of system response. In this process, by expanding the N event probabilities, that is, superimposing the event probabilities on the preset probability scale, N probability maxima can be obtained. These maxima represent the maximum possibility of risk events occurring in the most extreme scenarios. Through this expansion process, the risk assessment model can fully consider various uncertainties in the event process and provide an estimated value of the maximum risk probability for each event.

[0059] Finally, the obtained N event probabilities, N probability minimum values ​​and N probability maximum values ​​are combined to determine the evaluation results of the N events. This reflects the risk level of the event in different scenarios, and combined with the range of minimum and maximum values, it can reflect the most likely scenario of the event and its potential risk fluctuations. It provides a comprehensive, accurate and uncertainty-considered analysis result for the risk assessment of subway stations, which can effectively support safety management decisions and guide the formulation of emergency response measures.

[0060] Among them, in combination with the index directed graph, parallel calculation is performed on N event probabilities. Step S4 of this application includes: Determine a first data cognitive graph, wherein the first data cognitive graph is any one of the N data cognitive graphs; based on the first leaf indicator layer of the indicator directed graph, match the first data cognitive graph with the risk indicators within the layer, determine the risk matching indicators, and determine the prior probability of each risk matching indicator, wherein the first leaf indicator layer is the minimum indicator unit layer; based on the mapping relationship between the second leaf indicator layer and the first leaf indicator layer, perform weighted calculation on the risk matching indicators to determine the second conditional probability; traverse the indicator directed graph until the recursive calculation of the root event layer is completed to determine the first event probability, wherein the indicator probability tree is synchronously output.

[0061] During the implementation of the present invention, the first data cognitive map is first determined. The first data cognitive map is any one of the N data cognitive maps. Then, based on the first leaf indicator layer of the indicator directed graph, the first data cognitive map is matched with the risk indicators within the layer. The first leaf indicator layer is the minimum indicator unit layer, which usually contains basic indicators directly related to the occurrence of risk events. These basic indicators measure the most specific risk elements in subway stations, such as the status of equipment, operating behavior of personnel, etc. By matching the nodes in the first data cognitive map with the risk indicators in the first leaf indicator layer, it can be determined which nodes in the data cognitive map have a direct relationship with these basic risk indicators. The goal of this step is to accurately match the information in the data cognitive map with the risk indicators to ensure that the probability of occurrence of risk events can be accurately calculated based on the underlying indicators.

[0062] The probability of occurrence of the risk indicator matched in the first data cognitive map is used as the prior probability, which represents the probability of occurrence of each risk matching indicator without considering other factors. These prior probabilities can be obtained through historical data analysis or expert judgment, and are intended to provide a preliminary risk measurement for the risk assessment model. By determining the prior probability of each risk matching indicator, a basis is provided for subsequent probability recursive calculation.

[0063] Subsequently, based on the mapping relationship between the second leaf indicator layer and the first leaf indicator layer, the risk matching indicator is weightedly calculated to determine the second conditional probability. The second leaf indicator layer contains indicators at a higher level than the first leaf indicator layer, which usually represent more complex system behaviors or the results of events. For example, the second leaf indicator layer may be related to factors such as the overall safety performance and recovery capability of the system. Based on the mapping relationship between the first leaf indicator layer and the second leaf indicator layer, that is, an indicator in the second leaf indicator layer may correspond to at least one matching indicator in the first leaf indicator layer, the risk matching indicator is weightedly calculated, and the second conditional probability can be further derived in the hierarchical structure. The weighted calculation takes into account the dependency between indicators at different levels, so that the final conditional probability can comprehensively consider the comprehensive impact of risk factors at each level, ensuring the accuracy and systematicness of the evaluation results.

[0064] Next, the indicator directed graph is traversed until the recursive calculation of the root event layer is completed to determine the probability of the first event. The root event layer usually represents the most macro risk event, which is the final result of all risk indicators. By traversing the indicator directed graph, the probability of nodes at each level is gradually calculated, and finally the probability of occurrence of the root event layer is calculated. During the recursive calculation process, the probability of events at each layer will be affected by the probability of events at the previous layer. Therefore, through layer-by-layer recursion, the risk level of the entire system can be finally obtained.

[0065] While completing the recursive calculation, the indicator probability tree is output synchronously. The indicator probability tree is a tree structure that represents the relationship between each risk indicator and its corresponding probability. At the root of the tree is the risk event probability of the entire system, and the leaf nodes of the tree represent the relevant risk indicators. By constructing the indicator probability tree, the probability distribution of indicators at each level and their mutual relationship can be intuitively displayed, so that the risk assessment results not only have numerical assessment results, but also have a visual structured form, which is easy to understand and analyze.

[0066] Similarly, the N data cognitive maps are analyzed separately to determine N event probabilities and N indicator probability trees.

[0067] After determining the N event evaluation results, step S4 of the present application includes: The N event evaluation results are traversed, a threshold is determined for the maximum probability value, and M event evaluation results whose maximum values ​​are less than the threshold are screened; and the M event evaluation results are added to the risk assessment result.

[0068] In the implementation process of the present invention, the N event evaluation results are traversed, that is, each event evaluation result is represented by a set of triangular fuzzy numbers including a minimum probability value, a calculated value and a maximum probability value. These evaluation results reflect the probability of occurrence of each risk event in different scenarios and its uncertainty.

[0069] Next, the maximum probability value in the triangular fuzzy number of each event is thresholded. The maximum probability value is the maximum probability value of an event under the most extreme circumstances. In order to further screen events, a threshold needs to be set according to actual needs or preset safety standards. The threshold can be set based on historical data, expert experience or safety management standards. By comparing the maximum probability value of each event with the set threshold, it is possible to determine which events have risk levels beyond the acceptable range and which events have risk levels within the controllable range.

[0070] In the threshold determination process, if the maximum probability value of an event is less than the preset threshold, the event is considered a low-risk event and can be excluded from the subsequent risk assessment. On the contrary, if the maximum probability value of an event is greater than the threshold, the event is considered a high-risk event and requires further processing. After screening out all events that meet the conditions, the evaluation results of M events with maximum values ​​less than the threshold are obtained.

[0071] Subsequently, the M selected event evaluation results are added to the risk assessment results. The risk assessment result is a comprehensive result that includes all evaluated events and their risk levels, including information such as minimum values, calculated values, and maximum values. It can effectively identify and exclude low-risk events and focus on high-risk events, thereby providing accurate risk assessment results for the safety management of subway stations. This step ensures the scientificity and practicality of the evaluation results, and provides a basis for formulating effective risk response strategies.

[0072] After determining the risk assessment results, the application steps also include: Identify the risk assessment results, sort the risk assessment results according to the risk probability level and diffusion speed, and determine the risk event sequence; generate an emergency plan for the risk event sequence; perform diffusion prediction based on event evolution rules and determine event evolution data; and perform safety risk management for the subway station based on the emergency plan and the event evolution data.

[0073] Specifically, firstly, the risk assessment results are identified and sorted according to the risk probability level and diffusion speed. That is, the priority of the event is determined according to the risk probability level and diffusion speed of each event. The risk probability level reflects the possibility of an event, while the diffusion speed reflects the speed at which the scope of its impact expands after the event occurs. For example, some equipment failures may have a low probability of occurrence, but once they occur, they may quickly trigger a chain reaction and expand the scope of impact. By comprehensively considering these two factors, all risk events can be reasonably sorted to form a risk event sequence, which can reflect the potential threat level of different events to the safety of subway stations.

[0074] Next, generate an emergency plan for the identified risk event sequence. The emergency plan is a set of response measures and operating procedures for each ranked risk event. For high-risk events, the emergency plan will describe in detail the response measures, resource allocation, response time, and responsibility allocation. For example, for a high-risk equipment failure event, the emergency plan may include detailed steps for equipment inspection, troubleshooting, and spare parts replacement. The purpose of the emergency plan is to prepare in advance to ensure that risk events can be handled quickly and effectively when they occur, minimizing the impact of risks on subway station operations.

[0075] After the emergency plan is formulated, diffusion prediction is carried out based on the event evolution rules to determine the event evolution data. Event evolution rules refer to the rules for how various factors in the system interact, expand and affect other parts after the risk event occurs. For example, certain risk events (such as equipment failure) may cause chain reactions in other related systems, thereby triggering a series of subsequent events. By modeling and analyzing the event evolution rules, it is possible to predict the possible development trend of the event and its evolution path in different scenarios. The diffusion prediction process simulates the evolution process of risk events and outputs the data and results of the possible evolution of the event. This process provides a scientific basis for subsequent emergency response and resource allocation, helping decision makers to foresee the risk development trend and take appropriate measures.

[0076] Finally, based on the emergency plan and the event evolution data, the subway station is managed for safety risks. That is, based on the risk assessment and emergency plan, corresponding management measures are taken to control and reduce safety hazards. In this process, the management personnel will combine the measures proposed in the emergency plan, and according to the prediction of the event evolution data, promptly initiate the relevant emergency response mechanism, and carry out resource allocation and personnel command. Safety risk management is not only about responding to emergencies, but also includes daily monitoring and preventive measures. By dynamically updating the risk assessment results and emergency plans, it ensures that the subway station can respond quickly and effectively when various risks occur, and ensure the safety of passengers and staff.

[0077] The present application provides a subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction, which has the following technical effects: 1. By calling the risk operation and maintenance data of subway stations, a multi-level indicator directed graph is constructed based on the hierarchical analysis method, and the weight of each indicator is marked. This indicator directed graph can clearly show the hierarchical structure and dependency relationship of each risk indicator in the subway station. A comprehensive and systematic analysis of the risk factors of subway stations is achieved, so that risk assessment can reflect the safety status of subway stations based on indicators and weight relationships at different levels.

[0078] 2. Set up a reverse assessment method, based on the hierarchical probability recursive algorithm, and combine the resilience evaluation matrix to conduct risk assessment. Train the risk assessment model by mapping risk disturbances to system performance. The combination of the reverse assessment method and the resilience evaluation matrix makes the assessment process more consistent with the actual risk evolution law, and helps to dynamically reflect the recovery capacity and stability of subway stations when encountering different risk disturbances.

[0079] 3. Based on distributed sensor data, common risk points are divided by combining top risks, intermediate risks and root risks, and data slicing and dimensionality reduction are performed to generate multiple data cognitive maps. Through dimensionality reduction and data slicing processing, key risk information can be extracted from a large amount of complex sensor data, improving the efficiency and accuracy of data analysis and ensuring that the assessment model is more accurate and reliable.

[0080] 4. By combining probability recursion with extreme value analysis, the data cognitive map is calculated and represented by triangular fuzzy numbers, covering minimum values, calculated values ​​and maximum values, which can better quantify the uncertainty of risks. This method based on probability recursion and extreme value analysis can comprehensively and accurately reflect the probability of risk events, and take into account extreme changes in different scenarios, thus improving the accuracy and reliability of risk assessment.

[0081] In summary, the comprehensiveness and accuracy of risk assessment are ensured through a multi-level indicator system and reverse assessment method, which can accurately capture various potential risks and predict their impact; through dimensionality reduction, data slicing and risk sorting technology, the data processing efficiency is improved, making emergency response more efficient and operational; combining the resilience evaluation matrix with event evolution rules, the resilience and emergency response capabilities of subway stations in the face of variable risks are improved; through the combination of probability recursion and extreme value analysis, intelligent risk assessment is realized, and real-time updates ensure the dynamic and adaptability of safety management decisions.

[0082] Through the above-mentioned detailed description of a subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction in this specification, those skilled in the art can clearly know a subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction in this embodiment. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0083] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction, characterized in that: The method comprises: Calling the risk operation and maintenance data of the subway station, performing hierarchical analysis based on risk indicators, and mining the indicator directed graph, wherein the indicator directed graph is a mapping structure of multi-layer indicator refinement, marked with indicator weights; Setting a reverse evaluation method based on the index directed graph, obtaining a resilience evaluation matrix of the subway station, and supervising the training of a risk assessment model, wherein the reverse evaluation method is a hierarchical probability recursion; Monitor and determine the distributed sensor data of subway stations, divide and slice the data based on common risk points, and convert and determine the data cognitive map; The data cognitive map is transmitted to the risk assessment model, and probability recursive calculation and extreme value analysis are performed to determine the risk assessment result, wherein the risk assessment result is characterized as a triangular fuzzy number of minimum value-calculated value-maximum value.

2. A subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction as claimed in claim 1, characterized in that: After obtaining the resilience evaluation matrix of the subway station, including: For subway stations, the resilience evaluation matrix is ​​constructed, wherein the indicators are determined based on risk disturbance-system performance indicators; Based on the mechanism of subway risk events, a mapping between the indicator directed graph and the resilience evaluation matrix is ​​established, and risk disturbance degradation is used as a training target.

3. A subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction as claimed in claim 1, characterized in that: Data partitioning and slicing dimensionality reduction are performed based on common risk points, and the transformation determines the data cognitive map, including: Obtain the distributed sensor data, define common risk points based on top risk, intermediate risk and root risk, divide the data, and determine N data groups, wherein the root risk is an uncontrollable sudden cause, the top risk is an equipment operation type, and the intermediate risk is a personnel management type; The N data groups are traversed, data slicing and dimension reduction are performed, and N data cognitive graphs are determined.

4. A subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction as claimed in claim 3, characterized in that: The probability recursive calculation and extreme value analysis include: The N data cognitive graphs are transmitted to the risk assessment model, and parallel calculations are performed in combination with the indicator directed graph to determine N event probabilities; For the N event probabilities, anti-risk compensation is performed in combination with the resilience evaluation matrix to determine N probability minimum values; Expanding the N event probabilities by a preset probability scale to determine N maximum probability values, wherein the preset probability scale represents the uncertainty factor of the event process; The N event probabilities, the N probability minima and the N probability maxima are combined to determine N event evaluation results.

5. A subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction as claimed in claim 4, characterized in that: Combined with the index directed graph, parallel calculation is performed to determine the probabilities of N events, including: Determine a first data cognitive map, wherein the first data cognitive map is any one of the N data cognitive maps; Based on the first leaf indicator layer of the indicator directed graph, the first data cognitive graph is matched with the risk indicators in the layer to determine the risk matching indicators and the prior probability of each risk matching indicator, wherein the first leaf indicator layer is the minimum indicator unit layer; Based on the mapping relationship between the second leaf indicator layer and the first leaf indicator layer, weighted calculation is performed on the risk matching indicator to determine a second conditional probability; The indicator directed graph is traversed until the recursive calculation of the root event layer is completed to determine the first event probability, wherein the indicator probability tree is output synchronously.

6. A subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction as claimed in claim 4, characterized in that: After determining the evaluation results of N events, including: Traversing the N event evaluation results, performing threshold determination on the probability maximum value, and screening M event evaluation results whose maximum values ​​are less than the threshold; The M event evaluation results are added to the risk assessment result.

7. The subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction as claimed in claim 1, characterized in that: The mining index directed graph includes: Identify the risk operation and maintenance data, perform event clustering to determine event clusters, and construct a root event layer; Setting indicator layer elements, traversing the event clusters to perform multi-layer mining mapping, and constructing multiple leaf indicator layers; Inter-layer mapping is performed on the root event layer and the plurality of leaf indicator layers, and weight distribution is performed based on indicator membership to generate the indicator directed graph.

8. The subway station safety risk assessment method combining hierarchical analysis and data dimensionality reduction as claimed in claim 1, characterized in that: After determining the risk assessment results, including: Identify the risk assessment results, sort the risk assessment results according to the risk probability level and diffusion speed, and determine the risk event sequence; Generate an emergency plan for the risk event sequence; Based on the event evolution rules, diffusion prediction is carried out to determine the event evolution data; Based on the emergency plan and the event evolution data, safety risk management is performed on the subway station.

Images