Urban rail transit vehicle system risk point grading method and system

Abstract

This invention provides a method and system for risk point classification in urban rail transit vehicle systems, belonging to the field of urban rail transit operation and maintenance technology. It analyzes and obtains the component composition and connection relationships within the urban rail vehicle system; determines the structural importance index, functional importance index, and critical index based on accident data for each component; uses the K-Means++ algorithm based on a genetic algorithm to determine the number of classifications and clustering results; applies the Gini importance quantification of random forests to quantify the weights of the three types of indicators, calculating the correspondence between the clustering results and the classification levels; and uses set theory as a guide to form the classification results of urban rail vehicle risk point sets. This invention, from the perspective of networked operation and comprehensively considering multiple factors, achieves more accurate risk point classification in urban rail vehicle systems, which has significant theoretical and practical implications for improving rail transit operation safety risk management and ensuring safe and efficient operation.

Description

Technical Field

[0001] This invention relates to the field of urban rail transit operation and maintenance technology, specifically to a method and system for classifying risk points in urban rail transit vehicle systems. Background Technology

[0002] As one of the core equipment systems for ensuring safe and efficient transportation in urban rail transit systems, the vehicle system is characterized by a complex operating environment, frequent starts, and a high concentration of components. Component failures can affect the safe operation of the vehicle. Existing solutions to urban rail vehicle operation safety issues focus primarily on improving component reliability. However, urban rail vehicles do not exist in isolation; during their operation, human factors, the machine itself, and the environment all influence their condition.

[0003] An analysis of domestic and international literature on urban rail transit operation risk assessment reveals that experts and scholars have achieved certain research results. However, current research on urban rail transit operation risks also has some shortcomings, mainly manifested in the following aspects: First, the assessment methods and standards are not uniform, with most methods employing qualitative, quantitative, or a combination of both, which may not be easily promoted in practical applications; second, when determining the weights of risk indicators, expert scoring is often used, which is highly subjective and cannot accurately reflect the actual situation; third, there is relatively little research on risk assessment of urban rail transit vehicle systems, with a greater emphasis on the overall study of the urban rail transit system. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for risk point classification of urban rail transit vehicle systems, so as to solve at least one of the technical problems existing in the background art.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] On one hand, the present invention provides a method for risk point classification of urban rail transit vehicle systems, including:

[0007] The analysis reveals the component composition and connection relationships of the urban rail vehicle system;

[0008] Determine the structural importance index, functional importance index, and critical index based on accident data for the components;

[0009] The K-Means++ algorithm based on genetic algorithms is used to determine the number of levels and the clustering results.

[0010] The Gini importance quantification of the three categories of indicators is applied to random forests to calculate and determine the correspondence between clustering results and hierarchical levels.

[0011] Guided by set theory, a risk point set classification result for urban rail vehicles was generated.

[0012] Preferably, the main components and working principles of the subsystems of the urban rail vehicle system are summarized, and the urban rail vehicle system is mainly divided into six subsystems: car body and interior, auxiliary power supply, train control, traction, braking and bogie. When analyzing the component composition of each subsystem, the smallest maintenance unit is defined as a component. The connection relationship between components is analyzed and summarized, and the connection relationship between components is divided into three forms: mechanical connection, electrical connection and information connection.

[0013] Preferably, the determination of structural importance indicators, functional importance indicators, and key indicators based on accident data for components includes:

[0014] Based on the LeaderRank algorithm, a component function importance evaluation index is introduced that comprehensively considers component function dependency and component reliability.

[0015] The quasi-Laplacian centrality is introduced to measure the structural importance of a component. The quasi-Laplacian centrality method comprehensively considers the importance of the node itself and the importance of its neighbors. Based on the change in quasi-Laplacian energy caused by removing a node, the structural importance of the component is obtained.

[0016] By treating the components of the urban rail vehicle system as keywords, and combining Markov state transition models with word graphs, and integrating TF-IDF and TextRank algorithms, the TF-IDF values ​​obtained are used as the initial values ​​of nodes in the word graph. The frequency of simultaneous occurrence of words in a sentence is used as the node transition probability and participates in the iteration of the TextRank algorithm to obtain key indicators of urban rail vehicle components based on accident data.

[0017] Preferably, the reciprocal of the objective function of K-Means++ is used as the fitness function, and a K-Means++ algorithm based on genetic algorithm is proposed to obtain the number of levels and clustering results.

[0018] Preferably, the optimal clustering result obtained based on the genetic algorithm is used as the training set input for the random forest, and the Gini importance is used to measure the weight of the indicators. The original values ​​of the three types of indicators and the optimal clustering result obtained by the genetic algorithm are used as the input of the random forest to obtain the Gini importance corresponding to each indicator, that is, the weight of each indicator. By inputting the category values ​​K of different sample sets, the optimal K value and clustering result are obtained. The weight values ​​of the three types of indicators are combined with the clustering results to obtain the average importance of each cluster category, and finally obtain the classification result. By comparing the final classification result values, the one with the largest value is the first level of risk point, and so on, with the one with the smallest value being the K level of risk point. The clustering result is correlated with the classification result to obtain the risk point classification result of the urban rail vehicle system.

[0019] Preferably, the urban rail vehicle system includes six subsystems: car body and interior, auxiliary power supply, train control, traction, braking and bogie. Each subsystem has its own set of risk points. Based on the fact that each subsystem has its own set of risk points, and using the risk point classification results as a basis, the risk point set of the urban rail vehicle system is constructed on the basis of set theory.

[0020] Secondly, the present invention provides a risk point classification system for urban rail transit vehicle systems, comprising:

[0021] The acquisition module is used to analyze and acquire the component composition and connection relationships of the components in the urban rail vehicle system.

[0022] The first calculation module is used to determine the structural importance index, functional importance index, and critical index based on accident data of the component.

[0023] The clustering module is used for the K-Means++ algorithm based on genetic algorithms to determine the number of levels and the clustering results;

[0024] The second calculation module is used to apply the Gini importance quantification of random forest to quantify the weights of the three types of indicators and calculate and determine the correspondence between the clustering results and the hierarchical level.

[0025] The module is designed to generate a risk point set classification result for urban rail vehicles, guided by set theory.

[0026] Thirdly, the present invention provides a non-transitory computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the risk point classification method for urban rail transit vehicle systems as described above.

[0027] Fourthly, the present invention provides a computer program product, including a computer program that, when run on one or more processors, is used to implement the risk point classification method for urban rail transit vehicle systems as described above.

[0028] Fifthly, the present invention provides an electronic device, comprising: a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the risk point classification method for urban rail transit vehicle systems as described above.

[0029] The beneficial effects of this invention are as follows: From the perspective of networked operation as a whole, and taking into account multiple factors, it achieves a more accurate classification of risk points in the urban rail vehicle system, which has important theoretical and practical significance for improving the management of rail transit operation safety risks and ensuring safe and efficient operation.

[0030] The advantages of additional aspects of the invention will be set forth more clearly in the following description or will be learned by practice of the invention. Attached Figure Description

[0031] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a schematic diagram of the optimal clustering results (K=2) of the traction subsystem and bogie subsystem based on the genetic algorithm described in this embodiment of the invention.

[0033] Figure 2 This is a schematic diagram of the optimal clustering results (K=3) of the traction subsystem and bogie subsystem based on the genetic algorithm described in this embodiment of the invention.

[0034] Figure 3 This is a schematic diagram of the optimal clustering results (K=4) of the traction subsystem and bogie subsystem based on genetic algorithm according to an embodiment of the present invention.

[0035] Figure 4 This is a schematic diagram of the optimal clustering results (K=5) of the traction subsystem and bogie subsystem based on the genetic algorithm described in this embodiment of the invention.

[0036] Figure 5 This is a schematic diagram of the optimal clustering results (K=6) of the traction subsystem and bogie subsystem based on the genetic algorithm described in this embodiment of the invention.

[0037] Figure 6 This is a schematic diagram of the optimal clustering results (K=7) of the traction subsystem and bogie subsystem based on the genetic algorithm described in this embodiment of the invention.

[0038] Figure 7 This is a graph showing the weighting results of the three types of indicators described in this embodiment of the invention.

[0039] Figure 8 This is a diagram showing the weight representation of the three types of indicators described in this embodiment of the invention. Detailed Implementation

[0040] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0041] To facilitate understanding of the present invention, the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. However, the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0042] Those skilled in the art should understand that the accompanying drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily essential for implementing the present invention.

[0043] Example 1

[0044] This embodiment 1 provides a risk point classification system for urban rail transit vehicle systems, including:

[0045] The acquisition module is used to analyze and acquire the component composition and connection relationships of the components in the urban rail vehicle system.

[0046] The first calculation module is used to determine the structural importance index, functional importance index, and critical index based on accident data of the component.

[0047] The clustering module is used for the K-Means++ algorithm based on genetic algorithms to determine the number of levels and the clustering results;

[0048] The second calculation module is used to apply the Gini importance quantification of random forest to quantify the weights of the three types of indicators and calculate and determine the correspondence between the clustering results and the hierarchical level.

[0049] The module is designed to generate a risk point set classification result for urban rail vehicles, guided by set theory.

[0050] In this embodiment 1, the risk point classification method for urban rail transit vehicle systems is implemented using the above-described system, including:

[0051] The analysis reveals the component composition and connection relationships of the urban rail vehicle system;

[0052] Determine the structural importance index, functional importance index, and critical index based on accident data for the components;

[0053] The K-Means++ algorithm based on genetic algorithms is used to determine the number of levels and the clustering results.

[0054] The Gini importance quantification of the three categories of indicators is applied to random forests to calculate and determine the correspondence between clustering results and hierarchical levels.

[0055] Guided by set theory, a risk point set classification result for urban rail vehicles was generated.

[0056] The study summarizes the main components and working principles of the urban rail vehicle system's subsystems, dividing the system into six main subsystems: car body and interior, auxiliary power supply, train control, traction, braking, and bogie. When analyzing the components of each subsystem, the smallest maintenance unit is defined as a component. The study also analyzes and summarizes the connections between components, classifying them into three types: mechanical connections, electrical connections, and information connections.

[0057] Determine the structural importance index, functional importance index, and critical index based on accident data for components, including:

[0058] Based on the LeaderRank algorithm, a component function importance evaluation index is introduced that comprehensively considers component function dependency and component reliability.

[0059] The quasi-Laplacian centrality is introduced to measure the structural importance of a component. The quasi-Laplacian centrality method comprehensively considers the importance of the node itself and the importance of its neighbors. Based on the change in quasi-Laplacian energy caused by removing a node, the structural importance of the component is obtained.

[0060] By treating the components of the urban rail vehicle system as keywords, and combining Markov state transition models with word graphs, and integrating TF-IDF and TextRank algorithms, the TF-IDF values ​​obtained are used as the initial values ​​of nodes in the word graph. The frequency of simultaneous occurrence of words in a sentence is used as the node transition probability and participates in the iteration of the TextRank algorithm to obtain key indicators of urban rail vehicle components based on accident data.

[0061] By using the reciprocal of the objective function of K-Means++ as the fitness function, a K-Means++ algorithm based on genetic algorithm is proposed to obtain the number of levels and clustering results.

[0062] The optimal clustering result obtained from the genetic algorithm is used as the training set input for the random forest, and the Gini importance is used to measure the weight of the indicators. The original values ​​of the three types of indicators and the optimal clustering result obtained by the genetic algorithm are used as the input for the random forest to obtain the Gini importance corresponding to each indicator, that is, the weight of each indicator. By inputting the category values ​​K of different sample sets, the optimal K value and clustering result are obtained. The weight values ​​of the three types of indicators are combined with the clustering results to obtain the average importance of each cluster category, and finally obtain the classification result. By comparing the final classification result values, the one with the largest value is the first level of risk point, and so on, with the one with the smallest value being the K level of risk point. The clustering result is mapped to the classification result to obtain the risk point classification result of the urban rail vehicle system.

[0063] The urban rail vehicle system comprises six subsystems: car body and interior, auxiliary power supply, train control, traction, braking, and bogie. Each subsystem has its own set of risk points. Based on the risk point classification results of each subsystem, the risk point set of the urban rail vehicle system is constructed on the basis of set theory.

[0064] Example 2

[0065] In this embodiment 2, a risk point classification method for urban rail transit vehicle systems is provided. This method determines the risk point classification method for urban rail transit vehicle systems by determining the calculation methods for the structural importance index of components, the functional importance index of components, and the critical index calculation method based on accident data. It takes the components that affect the safe operation of vehicles as risk points, constructs a risk point set for urban rail transit vehicle systems, and opens up new ideas for risk assessment of urban rail transit vehicle systems.

[0066] In this embodiment 2, the risk point classification method for urban rail transit vehicle systems includes the following steps:

[0067] Step 1: Summarize the main components and working principles of the urban rail vehicle system's subsystems, dividing the urban rail vehicle system into six main subsystems: car body and interior, auxiliary power supply, train control, traction, braking, and bogie. When analyzing the component composition of each subsystem, the smallest maintenance unit is defined as a component.

[0068] (1) Vehicle body and interior subsystem

[0069] The vehicle body provides seating and safety for passengers, as well as space for vehicles with a driver's cab. It also houses numerous components and equipment on the underframe and roof, serving as the foundation for connecting these other parts and equipment. A typical vehicle body includes the roof, end walls, underframe, and side walls. The interior provides safety and comfort for passengers and typically includes seats, lighting, windows, pillars, side wall panels, ceiling, handrails, and other auxiliary facilities.

[0070] (2) Auxiliary power supply system

[0071] The auxiliary power supply system assists the urban rail power supply system in providing power to the vehicles. It typically obtains DC voltage from the power grid, inputs low-voltage electricity via an auxiliary inverter, and powers the auxiliary equipment on board, enabling the power supply of various electrical devices within the vehicle. The auxiliary inverter and the low-voltage power supply are two components of the auxiliary power supply system. The auxiliary inverter converts the voltage into three-phase AC power to power onboard electrical appliances such as lighting, air conditioning, and ventilation systems. The other component is the low-voltage power supply, where batteries store DC power to power the train control system and emergency systems. Therefore, the auxiliary power supply system mainly consists of batteries, auxiliary inverters, and electrical equipment.

[0072] (3) Train control subsystem

[0073] The train control subsystem controls the vehicle's speed and monitors its safe operating status based on information provided by onboard sensors and ground equipment. It is the core subsystem of the vehicle system and mainly consists of equipment such as the Automatic Train Control (ATC) system and the central control unit.

[0074] (4) Traction Subsystem

[0075] The traction subsystem is an important electrical system in the vehicle system, providing the train with the required power and braking force. It is usually powered by electricity and provides power to other electrical equipment on the train. It mainly consists of current collectors, traction motors, traction inverters, high-speed circuit breakers, etc.

[0076] (5) Braking subsystem

[0077] The braking subsystem is one of the key systems for ensuring the safe and timely completion of train transportation tasks. It is a device that controls the train's speed or stops it, and the braking performance of the braking subsystem directly determines the braking effect of the train. The braking subsystem generally includes a dynamic braking section, an air braking section, and a braking control section. The dynamic braking section is connected to the traction subsystem, the air braking section uses compressed air as the driving force for braking, and the braking control section is the core of the braking subsystem.

[0078] (6) Bogie system

[0079] The bogie subsystem provides cushioning for the vehicle through its suspension and damping system, which absorbs unevenness in the track and maintains a comfortable and stable ride. The bogie also has a self-guiding function, allowing the train to run along the track. Simultaneously, the bogie's drive unit provides traction for the urban rail vehicle. The bogie's basic braking system enables smooth braking of the train. Furthermore, the bogie bears the weight of the entire vehicle, effectively distributing its weight to fulfill its load-bearing function. The bogie subsystem generally includes a frame, wheelsets, primary suspension, and secondary suspension.

[0080] Analyzing and summarizing the connection relationships between components, we can categorize them into three types: mechanical connections, which are divided into detachable and non-detachable connections, and are methods of connecting components using fasteners; electrical connections, which are methods of transmitting electrical energy from one component to another using wires, cables, etc., and are achieved by fixing different conductor components together through appropriate mechanical force; and information connections, which are connections in the form of reading feedback signals, are connections that transmit information from one component to another via a channel through wired or wireless means, and are received by the other party.

[0081] Step 2: Determine the calculation methods for the structural importance index of components, the functional importance index of components, and the criticality index calculation methods based on accident data.

[0082] First, we determine the calculation method for the functional importance index of components. Based on the LeaderRank algorithm, we introduce a comprehensive evaluation index that considers both the functional dependence and reliability of components to describe the functional importance of components.

[0083] (1)

[0084] in, For components The set of first-order neighbors, For components Functional dependency Indicates components exist The LR value at time 1.

[0085] Secondly, quasi-Laplacian centrality is introduced to measure the structural importance of components. The quasi-Laplacian centrality method comprehensively considers both the importance of the node itself and the importance of its neighbors. The quasi-Laplacian centrality of a node is measured by the change in quasi-Laplacian energy caused by removing it. The quasi-Laplacian centrality of a node is related not only to its own degree but also to the degrees of other connected nodes. Based on the change in quasi-Laplacian energy caused by removing a node, the structural importance of the component is obtained.

[0086] (2)

[0087] in, For nodes The neighboring nodes, For nodes The degree.

[0088] Finally, the components of the urban rail vehicle system are treated as keywords, which are the nodes in the constructed word graph model of urban rail vehicle system accident text. A Markov state transition model is combined with the word graph, and each sentence is treated as a window. Only when two words exist in the same sentence are the edges connecting the corresponding nodes in the word graph meaningful; the meaning of the edge is the co-occurrence frequency of the words. Based on the fusion of the TF-IDF algorithm and the TextRank algorithm, the solved TF-IDF value is used as the initial value of the nodes in the word graph, and the frequency of simultaneous occurrence of words in a sentence is used as the node transition probability, participating in the iteration of the TextRank algorithm. This yields a method for calculating key indicators of urban rail vehicle components based on accident data.

[0089] (3)

[0090] (4)

[0091] (5)

[0092] in, The total number of texts. For words containing To calculate the number of text words, and to prevent the denominator from being zero due to the presence of words not present in the document, one is incremented in the denominator. For words word frequency, This refers to the node To the node The weight of the edge. For urban rail vehicle components based on accident data Key indicator values.

[0093] Step 3: Using the reciprocal of the objective function of K-Means++ as the fitness function, a K-Means++ algorithm based on genetic algorithm is proposed to obtain the number of levels and clustering results.

[0094] Step 3 specifically includes:

[0095] Step 1: Encoding, using each cluster center as a code, such as... When the dataset is 2-dimensional and 3-dimensional, each cluster center is (1, 4, 9) or (5, 3, 8), so the initial encoding is (1, 4, 9, 5, 3, 8).

[0096] Step 2: Randomly generate the initial population, population size The crossover probability is typically 20-100. The mutation probability is typically 0.25 to 0.75. The value is generally 0.01~0.2; the number of generations is generally 100~500; the number of clusters K is generally [2, ]( (This refers to the size of the dataset).

[0097] Step 3: Classify the data, specifically for the dataset. , Given the number of data points, the K-Means++ algorithm is used to select the initial cluster centers, update the cluster centers, and calculate the mean value of all data points in each class as the cluster center of that class.

[0098] Step 4: Define the fitness function. The fitness value is crucial to the quality of the next generation of the population and is key to the survival of the fittest. The fitness function is the reciprocal of the sum of the Euclidean distances between data points and the cluster centers. The data size for each cluster center is... , For cluster centers, the fitness function is:

[0099] (6)

[0100] Step 5: Selection. Use the roulette wheel selection operator to select individuals with higher fitness. Let the population size be [value missing]. The fitness function of an individual takes the value of Then the probability of it being selected is:

[0101] (7)

[0102] Step 6: Crossover. Single-point crossover is used to divide the population into pairs. For each pair of individuals, a crossover point is set using a random function. Based on the crossover probability, some genes are exchanged at the crossover point to form new individuals and the fitness function value is calculated. The fitness of the new individuals is compared with that of the corresponding parent, and the two chromosomes with higher fitness are selected and saved.

[0103] Step 7: Mutation. The uniform mutation operator is used to perform the operation. Mutation is performed with the cluster center as the basic unit. The cluster center is set as the mutation point. A new sample is randomly selected to replace the original sample as the cluster center based on the mutation probability.

[0104] Step 8: Adjusting the probabilities of crossover and mutation. To reduce the impact of fixed probabilities on the clustering results, an adaptive operator is used for dynamic adjustment. During crossover and mutation, the probability of higher fitness is lower, while the probability of lower fitness is higher. This makes it easier for higher fitness patterns to be preserved and inherited, and for lower fitness patterns to evolve more easily, thus producing superior patterns. , , , Crossover probability and mutation probability for:

[0105] (8)

[0106] (9)

[0107] in, The average fitness of all individuals in each generation of the population. The maximum fitness of an individual within the group. The higher fitness among the two individuals at the intersection It refers to the individual fitness that needs to be mutated.

[0108] Step 9: Terminate the loop and set the number of iterations. and threshold When the number of iterations reaches Or less than the threshold The algorithm ends when the optimal clustering result is obtained.

[0109] The goal of clustering is to minimize intra-cluster distances and maximize inter-cluster distances, by inputting different... Value and for each Each value was tested 100 times, and each was selected. The optimal clustering result corresponds to the best fitness curve under the given value. Calculate each The intra-cluster distance and inter-cluster distance at different values ​​were calculated by averaging the intra-cluster distances and inter-cluster distances from 100 experiments. The results were then compared with different values. The average value is used to determine the minimum average intra-cluster distance and the maximum average inter-cluster distance. The value is the optimal number of levels.

[0110] Step 4: The optimal clustering result obtained from the genetic algorithm is used as the training set input for the random forest, and the weights of the indicators are measured using Gini importance. Gini importance is calculated based on the Gini index, denoted as... .

[0111] (10)

[0112] (11)

[0113] (12)

[0114] (13)

[0115] Where K is the number of categories in the sample set, For nodes The sample belongs to the first The probability value of the class sample, For variables At the node Changes in Gini value before and after splitting For nodes The Gini index value at that location; For nodes The Gini index value of the child nodes after splitting. For variables In the Appearing in the trees The importance of this; This represents the number of classification trees.

[0116] By using the original values ​​of the three categories of indicators and the optimal clustering results obtained through a genetic algorithm as input to the random forest, the Gini importance corresponding to each indicator, i.e., the weight of each indicator, can be obtained. By inputting different The optimal value can be obtained. The values ​​and clustering results are used to determine the ranking results. However, the clustering results are not the same as the ranking results. The clustering results need to be processed to combine the weight values ​​of the three indicators with the clustering results to obtain the average importance of each cluster category, and finally obtain the ranking results.

[0117] (14)

[0118] in, The weights of the three types of indicators, , The first in the cluster level The risk points contained in the level are the first The value of each indicator; The first in the cluster level The total number of risk points contained in the level.

[0119] contrast The highest value indicates a level 1 risk point, and so on, with the lowest value indicating a level 2 risk point. Level, through By correlating the clustering results with the classification results, we can obtain the classification results of risk points in the urban rail vehicle system.

[0120] Step 5: The urban rail vehicle system also consists of six subsystems: car body and interior, auxiliary power supply, train control, traction, braking, and bogie. Each subsystem has its own set of risk points, which can be represented as follows:

[0121] (15)

[0122] (16)

[0123] (17)

[0124] (18)

[0125] (19)

[0126] (20)

[0127] (twenty one)

[0128] in, These are sets of risk points for each subsystem of the urban rail vehicle. For urban rail vehicles The first subsystem One risk point, Represented as the first The number of risk points in each subsystem.

[0129] In a vehicle system, not all components are risk points. The failure of some components, such as the air conditioning unit and side panels, will not affect the safety status of the vehicle system. Therefore, when studying the risks of a vehicle system, it is necessary to identify the safety-related components and term them as risk points in the vehicle system. Thus, each risk point has its own set representation. Based on the above description of the basic concepts and properties of set theory, and using the risk point classification results as a basis, a risk point set for urban rail vehicle systems is constructed using set theory.

[0130] (twenty two)

[0131] in, For the first Subsystem The risk level classification; This represents the number of risk points within the subsystem.

[0132] In summary, the risk point classification method provided in this embodiment, starting from the overall perspective of networked operation, proposes a risk point classification method for urban rail transit vehicle systems, which has important theoretical and practical significance for improving the management of rail transit operation safety risks and ensuring safe and efficient operation.

[0133] Example 3

[0134] In this embodiment 3, taking a certain urban rail vehicle as an example, the risk point classification method of urban rail transit vehicle system is specifically illustrated. The method includes the following steps:

[0135] Step 1: Summarize the main components and working principles of the urban rail vehicle system's subsystems, dividing the urban rail vehicle system into six main subsystems: car body and interior, auxiliary power supply, train control, traction, braking, and bogies. When analyzing the component composition of each subsystem, the smallest maintenance unit is defined as the component. Through discussions with manufacturing design engineers, vehicle safety managers from operating units, and maintenance engineers, and in conjunction with the vehicle system risk lists of each operating unit, the main components affecting the safe operation of the vehicle system are identified, using the smallest maintenance unit of each subsystem as the component, as shown in Table 1. These represent the urban rail vehicle's body and interior subsystems, auxiliary power supply system, train control subsystem, traction subsystem, braking subsystem, and bogie system, respectively. The study examines the connections between vehicle components. Through research on vehicle design and manufacturing companies, the connections between components are analyzed and summarized, categorized into three types: mechanical connection, electrical connection, and information connection.

[0136] Table 1. Key components for safe operation of urban rail vehicle systems

[0137]

[0138]

[0139]

[0140]

[0141] Step 2: Determine the calculation methods for the structural importance index of components, the functional importance index of components, and the criticality index calculation methods based on accident data.

[0142] First, we determine the calculation method for the functional importance index of components. Based on the LeaderRank algorithm, we introduce a comprehensive evaluation index that considers both the functional dependence and reliability of components to describe the functional importance of components.

[0143] (twenty three)

[0144] in, For components The set of first-order neighbors, For components Functional dependency Indicates components exist The LR value at time 1.

[0145] Secondly, quasi-Laplacian centrality is introduced to measure the structural importance of components. The quasi-Laplacian centrality method comprehensively considers both the importance of the node itself and the importance of its neighbors. The quasi-Laplacian centrality of a node is measured by the change in quasi-Laplacian energy caused by removing it. The quasi-Laplacian centrality of a node is related not only to its own degree but also to the degrees of other connected nodes. Based on the change in quasi-Laplacian energy caused by removing a node, the structural importance of the component is obtained.

[0146] (twenty four)

[0147] in, For nodes The neighboring nodes, For nodes The degree.

[0148] Finally, the components of the urban rail vehicle system are treated as keywords, which are the nodes in the constructed word graph model of urban rail vehicle system accident text. A Markov state transition model is combined with the word graph, and each sentence is treated as a window. Only when two words exist in the same sentence are the edges connecting the corresponding nodes in the word graph meaningful; the meaning of the edge is the co-occurrence frequency of the words. Based on the fusion of the TF-IDF algorithm and the TextRank algorithm, the solved TF-IDF value is used as the initial value of the nodes in the word graph, and the frequency of simultaneous occurrence of words in a sentence is used as the node transition probability, participating in the iteration of the TextRank algorithm. This yields a method for calculating key indicators of urban rail vehicle components based on accident data.

[0149] (25)

[0150] (26)

[0151] (27)

[0152] in, The total number of texts. For words containing To calculate the number of text words, and to prevent the denominator from being zero due to the presence of words not present in the document, one is incremented in the denominator. For words word frequency, This refers to the node To the node The weight of the edge. For urban rail vehicle components based on accident data Key indicator values.

[0153] Based on Table 1, the components in the traction subsystem and bogie system are summarized and extracted. For ease of description, the components are numbered. to For nodes of the traction subsystem, to The nodes of the bogie system are shown in Table 2.

[0154] Table 2 List of nodes in the topology network of the traction subsystem and bogie system

[0155]

[0156] The normalized values ​​of the three categories of indicators—component structural importance, functional importance, and criticality based on accident data—obtained through calculation of the traction subsystem and bogie subsystem are shown in Table 3.

[0157] Table 3 Normalized values ​​of the three types of indicators for each component

[0158]

[0159]

[0160] Step 3: Using the reciprocal of the objective function of K-Means++ as the fitness function, a K-Means++ algorithm based on genetic algorithm is proposed to obtain the number of levels and clustering results.

[0161] Preferably, step 3 includes:

[0162] Step 1: Encoding, using each cluster center as a code, such as... When the dataset is 2-dimensional and 3-dimensional, each cluster center is (1, 4, 9) or (5, 3, 8), so the initial encoding is (1, 4, 9, 5, 3, 8).

[0163] Step 2: Randomly generate the initial population, population size The crossover probability is typically 20-100. The mutation probability is typically 0.25 to 0.75. The value is generally 0.01~0.2; the number of generations is generally 100~500; the number of clusters K is generally [2, ]( (This refers to the size of the dataset).

[0164] Step 3: Classify the data, specifically for the dataset. , Given the number of data points, the K-Means++ algorithm is used to select the initial cluster centers, update the cluster centers, and calculate the mean value of all data points in each class as the cluster center of that class.

[0165] Step 4: Define the fitness function. The fitness value is crucial to the quality of the next generation of the population and is key to the survival of the fittest. The fitness function is the reciprocal of the sum of the Euclidean distances between data points and the cluster centers. The data size for each cluster center is... The fitness function is:

[0166] (28)

[0167] Step 5: Selection. Use the roulette wheel selection operator to select individuals with higher fitness. Let the population size be [value missing]. The fitness function of an individual takes the value of Then the probability of it being selected is:

[0168] (29)

[0169] Step 6: Crossover. Single-point crossover is used to divide the population into pairs. For each pair of individuals, a crossover point is set using a random function. Based on the crossover probability, some genes are exchanged at the crossover point to form new individuals and the fitness function value is calculated. The fitness of the new individuals is compared with that of the corresponding parent, and the two chromosomes with higher fitness are selected and saved.

[0170] Step 7: Mutation. The uniform mutation operator is used to perform the operation. Mutation is performed with the cluster center as the basic unit. The cluster center is set as the mutation point. A new sample is randomly selected to replace the original sample as the cluster center based on the mutation probability.

[0171] Step 8: Adjusting the probabilities of crossover and mutation. To reduce the impact of fixed probabilities on the clustering results, an adaptive operator is used for dynamic adjustment. During crossover and mutation, the probability of higher fitness is lower, while the probability of lower fitness is higher. This makes it easier for higher fitness patterns to be preserved and inherited, and for lower fitness patterns to evolve more easily, thus producing superior patterns. , , , Crossover probability and mutation probability for:

[0172] (30)

[0173] (31)

[0174] in, The average fitness of all individuals in each generation of the population. The maximum fitness of an individual within the group. The higher fitness among the two individuals at the intersection It refers to the individual fitness that needs to be mutated.

[0175] Step 9: Terminate the loop and set the number of iterations. and threshold When the number of iterations reaches Or less than the threshold The algorithm ends when the optimal clustering result is obtained.

[0176] The goal of clustering is to minimize intra-cluster distances and maximize inter-cluster distances, by inputting different... Value and for each Each value was tested 100 times, and each was selected. The optimal clustering result corresponds to the best fitness curve under the given value. Calculate each The intra-cluster distance and inter-cluster distance at different values ​​were calculated by averaging the intra-cluster distances and inter-cluster distances from 100 experiments. The results were then compared with different values. The average value is used to determine the minimum average intra-cluster distance and the maximum average inter-cluster distance. The value is the optimal number of levels.

[0177] Safety-related components in the traction subsystem and bogie system are risk points. These are the nodes in the topology network of the traction subsystem and bogie subsystem. A clustering algorithm based on a genetic algorithm is used to determine the number of levels and the clustering results. The population size is then set. Number of iterations Cluster number The value is [2, For the same metric dataset, for each Each value was subjected to 100 experiments, and the clustering results under the optimal fitness curve were compared to obtain the optimal clustering result. The optimal clustering result corresponding to the value, such as Figures 1 to 6 As shown in the figure, indicator 1 corresponds to the node functional importance indicator, indicator 2 corresponds to the node structural importance indicator, and indicator 3 corresponds to the node criticality indicator based on accident data (the circles in the figure correspond to the cluster centers, the asterisks correspond to the nodes, and the asterisks of different colors indicate that the data points belong to different cluster categories).

[0178] Each Each value was tested 100 times, and each was compared... The cluster target mean value is shown in Table 4. It was found that when... At this time, the intra-cluster distance is the smallest and the inter-cluster distance is the largest; therefore, choose... The optimal number of gradations is determined by the optimal fitness function obtained after 100 experiments, which represents the optimal clustering result.

[0179] Table 4 shows the average clustering target values ​​for each level.

[0180]

[0181] Step 4: The optimal clustering result obtained from the genetic algorithm is used as the training set input for the random forest, and the weights of the indicators are measured using Gini importance. Gini importance is calculated based on the Gini index, denoted as... .

[0182] (32)

[0183] (33)

[0184] (33)

[0185] (34)

[0186] Where K is the number of categories in the sample set, For nodes The sample belongs to the first The probability value of the class sample, For variables At the node Changes in Gini value before and after splitting For nodes The Gini index value at that location; For nodes The Gini index value of the child nodes after splitting. For variables In the Appearing in the trees The importance of this; This represents the number of classification trees.

[0187] By using the original values ​​of the three categories of indicators and the optimal clustering results obtained through a genetic algorithm as input to the random forest, the Gini importance corresponding to each indicator, i.e., the weight of each indicator, can be obtained. By inputting different The optimal value can be obtained. The values ​​and clustering results are used to determine the ranking results. However, the clustering results are not the same as the ranking results. The clustering results need to be processed to combine the weight values ​​of the three indicators with the clustering results to obtain the average importance of each cluster category, and finally obtain the ranking results.

[0188] (35)

[0189] in, The weights of the three types of indicators, , The first in the cluster level The risk points contained in the level are the first The value of each indicator; The first in the cluster level The total number of risk points contained in the level.

[0190] contrast The highest value indicates a level 1 risk point, and so on, with the lowest value indicating a level 2 risk point. Level, through By correlating the clustering results with the classification results, we can obtain the classification results of risk points in the urban rail vehicle system.

[0191] To accurately characterize the importance of risk points in the traction subsystem and bogie subsystem, the clustering results were used as input to a random forest to obtain the weights of various indicators. The results are shown below. Figure 7 and Figure 8 As shown, the weights of the three types of indicators are not significantly different. The functional importance indicator has the greatest impact on the clustering categories, followed by the structural importance indicator.

[0192] Based on the weights of the three types of indicators obtained from the Gini importance of random forest, they are combined with the clustering results obtained from the genetic algorithm to obtain the average of the sum of the weights and indicator values ​​of all risk points in different cluster categories, that is, the average importance of all nodes in each cluster category, as shown in Table 5.

[0193] Table 5. Average Cluster Level

[0194]

[0195] from The highest value represents risk point level one, followed by risk point level two, and so on, ultimately yielding the risk point classification for the traction subsystem. Table 5 shows the clustering levels within the traction subsystem and bogie subsystem. Corresponding to risk level 1, cluster level Corresponding risk level 2, cluster level Corresponding risk level three, cluster level The corresponding risk level is level four.

[0196] Step 5: The urban rail vehicle system also consists of six subsystems: car body and interior, auxiliary power supply, train control, traction, braking, and bogie. Each subsystem has its own set of risk points, which can be represented as follows:

[0197] (36)

[0198] (37)

[0199] (38)

[0200] (39)

[0201] (40)

[0202] (41)

[0203] (42)

[0204] in, These are sets of risk points for each subsystem of the urban rail vehicle. For urban rail vehicles The first subsystem One risk point.

[0205] In a vehicle system, not all components are risk points. The failure of some components, such as the air conditioning unit and side panels, will not affect the safety status of the vehicle system. Therefore, when studying the risks of a vehicle system, it is necessary to identify the safety-related components and term them as risk points in the vehicle system. Thus, each risk point has its own set representation. Based on the above description of the basic concepts and properties of set theory, and using the risk point classification results as a basis, a risk point set for urban rail vehicle systems is constructed using set theory.

[0206] (43)

[0207] in, For the first Subsystem The risk level classification; This represents the number of risk points within the subsystem. Based on the risk point classification results in the traction subsystem and bogie subsystem, the final risk point set is... See Table 6.

[0208] Table 6. Risk Points for Traction Subsystem and Bogie System

[0209]

[0210]

[0211] The lower the risk level, the more important the risk point. Based on the risk point sets of the traction subsystem and bogie subsystem, it can be seen that in risk point classification I, the main circuit breaker, oil tank, driver controller and temperature sensor are the most important risk points, followed by the traction control unit, traction motor, gearbox boom, frame, connecting sleeper beam and secondary lateral damper as secondary risk points. The number of risk points in risk point classification III and IV is more than the first two levels.

[0212] Example 4

[0213] Embodiment 4 of the present invention provides a non-transitory computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, a risk point classification method for urban rail transit vehicle systems is implemented.

[0214] Example 5

[0215] Embodiment 5 of the present invention provides a computer program (product), including a computer program that, when run on one or more processors, is used to implement a risk point classification method for urban rail transit vehicle systems.

[0216] Example 6

[0217] Embodiment 6 of the present invention provides an electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to execute instructions for implementing a risk point classification method for urban rail transit vehicle systems.

[0218] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solutions disclosed in the present invention, various modifications or variations that can be made by those skilled in the art without creative effort should be included within the scope of protection of the present invention.

Claims

1. A method for classifying risk points in an urban rail transit vehicle system, characterized in that, include: The analysis reveals the component composition and connection relationships of the urban rail vehicle system; The study identifies structural importance indices, functional importance indices, and criticality indices based on accident data for components. These include: introducing a component functional importance assessment index that comprehensively considers component functional dependency and reliability, building upon the LeaderRank algorithm; introducing quasi-Laplacian centrality to measure component structural importance, which comprehensively considers the importance of a node itself and the importance of its neighbors, obtaining the component's structural importance based on the change in quasi-Laplacian energy caused by removing a node; and treating components of the urban rail vehicle system as keywords, combining a Markov state transition model with a word graph, integrating the TF-IDF and TextRank algorithms, using the solved TF-IDF values ​​as initial values ​​for nodes in the word graph, and using the frequency of simultaneous occurrence of words in a sentence as node transition probabilities to participate in the iteration of the TextRank algorithm, thus obtaining criticality indices for urban rail vehicle components based on accident data. The K-Means++ algorithm based on genetic algorithms is used to determine the number of levels and the clustering results. The Gini importance quantification of the three categories of indicators is applied to random forests to calculate and determine the correspondence between clustering results and hierarchical levels. Guided by set theory, a risk point set classification result for urban rail vehicles was generated.

2. The risk point classification method for urban rail transit vehicle systems according to claim 1, characterized in that, The method also includes: summarizing the main components and working principles of the subsystems of the urban rail vehicle system, dividing the urban rail vehicle system into six subsystems: car body and interior, auxiliary power supply, train control, traction, braking, and bogie; when analyzing the component composition of each subsystem, defining the smallest maintenance unit as a component; analyzing and summarizing the connection relationships between components, classifying the connection relationships between components into three forms: mechanical connection, electrical connection, and information connection.

3. The risk point classification method for urban rail transit vehicle systems according to claim 1, characterized in that, The K-Means++ algorithm based on genetic algorithm is proposed to determine the number of levels and clustering results. This includes using the reciprocal of the objective function of K-Means++ as the fitness function to obtain the number of levels and clustering results.

4. The risk point classification method for urban rail transit vehicle systems according to claim 1, characterized in that, The Gini importance quantification of the three categories of indicators is applied to random forests to calculate and determine the correspondence between clustering results and hierarchical levels. Guided by set theory, a risk point classification result for urban rail transit vehicles is generated, including: using the optimal clustering result obtained by a genetic algorithm as the training set input for a random forest, and using Gini importance to measure the weight of the indicators; using the original values ​​of the three types of indicators and the optimal clustering result obtained by the genetic algorithm as the input for the random forest to obtain the Gini importance corresponding to each indicator, i.e., the weight of each indicator; obtaining the optimal K value and clustering result by inputting the category values ​​K of different sample sets; combining the weight values ​​of the three types of indicators with the clustering results to obtain the average importance of each cluster category, and finally obtaining the classification result; comparing the final classification result values, the one with the largest value is the first-level risk point, and so on, with the one with the smallest value being the K-level risk point, and mapping the clustering results with the classification results to obtain the risk point classification result of the urban rail transit vehicle system.

5. The risk point classification method for urban rail transit vehicle systems according to claim 1, characterized in that, The urban rail vehicle system comprises six subsystems: car body and interior, auxiliary power supply, train control, traction, braking and bogie. Each subsystem has its own set of risk points. Based on the fact that each subsystem has its own set of risk points, and using the risk point classification results as a basis, a set of risk points for the urban rail vehicle system is constructed on the basis of set theory.

6. A risk point classification system for urban rail transit vehicle systems, characterized in that, include: The acquisition module is used to analyze and acquire the component composition and connection relationships of the components in the urban rail vehicle system. The first calculation module is used to determine the structural importance index, functional importance index, and critical index based on accident data of components. This includes: introducing a component functional importance evaluation index that comprehensively considers component functional dependency and component reliability, based on the LeaderRank algorithm; introducing quasi-Laplacian centrality to measure the structural importance of components. The quasi-Laplacian centrality method comprehensively considers the importance of the node itself and the importance of its neighbors, and obtains the structural importance of the component based on the change in quasi-Laplacian energy caused by removing a node; treating components of the urban rail vehicle system as keywords, combining a Markov state transition model with a word graph, and integrating the TF-IDF algorithm and the TextRank algorithm. The solved TF-IDF value is used as the initial value of the node in the word graph, and the frequency of simultaneous occurrence of words in a sentence is used as the node transition probability to participate in the iteration of the TextRank algorithm, thus obtaining the critical index of urban rail vehicle components based on accident data. The clustering module is used for the K-Means++ algorithm based on genetic algorithms to determine the number of levels and the clustering results; The second calculation module is used to apply the Gini importance quantification of random forest to quantify the weights of the three types of indicators and calculate and determine the correspondence between the clustering results and the hierarchical level. The module is designed to generate a risk point set classification result for urban rail vehicles, guided by set theory.

7. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement the risk point classification method for urban rail transit vehicle systems as described in any one of claims 1-5.

8. A computer program product, characterized in that, Includes a computer program, which, when run on one or more processors, is used to implement the risk point classification method for urban rail transit vehicle systems as described in any one of claims 1-5.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the risk point classification method for urban rail transit vehicle systems as described in any one of claims 1-5.

Images