Urban rail operation event analysis algorithm based on heinrich and bayes and application

Abstract

The application discloses a kind of city railcar driving event analysis algorithm and application based on heinrich and bayes, in combination with heinrich rule, Kant's cognitive theory and bayes theorem, by building five levels of data acquisition model, design quantitative analysis algorithm, develop data acquisition and analysis system, after algorithm and system test verification, applied to city railcar driving safety management.The algorithm uses probability value and proportion value to quantify abnormal phenomenon evolution correlation, can be combined with computer vision technology, applied to double prevention mechanism landing, safety card control and training plan formulation, realize risk accurate control, reduce accident probability, with good technical practicability, economic benefit and industry promotion value.

Description

Technical Field

[0001] This invention belongs to the field of urban rail transit traffic safety management technology, specifically the urban rail transit traffic event analysis algorithm and application based on Heinrich and Bayes. Background Technology

[0002] In urban rail transit operations, the occurrence and development of traffic incidents are complexly interconnected. Timely identification of the evolutionary patterns of abnormal phenomena in traffic incidents and accurate location of risk nodes are crucial to improving traffic safety management and reducing the probability of accidents. Currently, the closest existing technology in the field of urban rail transit traffic safety management is Heinrich's Law. As a classic industrial accident prevention theory, this law combines mathematical statistical theory with actual production work. Its core logic is that "small accidents can prevent big accidents." However, modern applications place more emphasis on this qualitative logic rather than strictly relying on its original quantitative ratio (1:29:300).

[0003] However, the application of Heinrich's law in the analysis of urban rail transit traffic events has obvious shortcomings, specifically in the following four aspects:

[0004] I. The statistical foundation is weak, and the universality of the proportions is questionable. The original quantitative proportions of Heinrich's Law are derived from statistics of specific industrial scenarios, which are difficult to adapt to the complex scenarios of dynamic changes in multiple variables such as personnel, equipment, environment, and work processes in urban rail operations. Therefore, the proportions are not universally applicable in the statistics of urban rail operation events and cannot be used as a quantitative basis for accurate early warning.

[0005] Second, the one-way probability relationship ignores the proportion of sources. This rule can only reflect the one-way logic that "small anomalies may accumulate and cause big accidents". It cannot quantify the interrelationship between different anomalies, and it ignores the proportion of multiple sources of a certain anomaly, making it difficult to trace the root cause of the anomaly.

[0006] Third, it may lead to erroneous resource allocation and a false sense of security. Due to the lack of precise quantitative analysis, safety management based on this principle is prone to a "spreading of effort" resource allocation pattern, failing to focus on high-risk areas. At the same time, it may create a false sense of security due to the apparent reduction of minor anomalies, ignoring potential major risks.

[0007] Fourth, the statistics are too general and cannot pinpoint specific problems. The statistics on driving incidents in this law are too macroscopic. It can only distinguish between incident categories of different severity levels, but it cannot be refined to specific abnormal phenomena, let alone pinpoint the key nodes in the evolution of abnormal phenomena, making it difficult to achieve targeted safety control.

[0008] Furthermore, current urban rail transit risk classification and control, as well as hazard investigation and mitigation, primarily rely on personnel experience to identify obvious abnormal behaviors and states. They fail to delve into minor anomalies, which are often the root causes of major accidents. Simultaneously, the lack of standardized models for train operation data collection and the reliance on limited data analysis methods prevent the digital and visual representation of the evolutionary patterns of anomalies, hindering the provision of scientific and accurate data support for safety decision-making. Therefore, there is an urgent need for a technological solution that can overcome the shortcomings of existing technologies and achieve precise analysis, risk visualization, and optimized management of urban rail transit incidents. Summary of the Invention

[0009] The purpose of this invention is to overcome the limitations of the application of Heinrich's law in the prior art, and to provide an urban rail transit event analysis algorithm and application based on Heinrich and Bayesian methods. This algorithm enables the digital and quantitative analysis of urban rail transit events, accurately identifies the evolutionary relationships between abnormal phenomena, identifies high-risk nodes, and provides a scientific basis for safety control, training plan formulation, and resource allocation. Ultimately, it improves the efficiency of urban rail transit safety management and reduces the probability of accidents.

[0010] To achieve the above objectives, the present invention employs the following technical means:

[0011] An algorithm for analyzing urban rail transit events based on Heinrich's law and Bayes' theorem is proposed. Combining the core logic of Heinrich's law, Kantian epistemology, and Bayes' theorem, this algorithm achieves precise analysis of the evolution patterns of abnormal phenomena in urban rail transit events by constructing a standardized data collection model for train operation events, designing a quantitative analysis algorithm, and developing a data collection and analysis system. The specific steps are as follows:

[0012] (1) Construction of driving event data collection model: Combining Kantian epistemology and set theory, the definitions of driving events and abnormal phenomena are clarified. Abnormal phenomena correspond to sample points in set theory, and driving events correspond to sample space. That is, a driving event is a collection of multiple abnormal phenomena in a certain scenario. A five-level data collection model including phenomenon layer, handling layer, attribute layer, evaluation layer and relation symbol is constructed.

[0013] (2) Basic data preparation: Based on the established data collection model, create basic data templates and fill in basic information related to urban rail transit operation, including vehicle type, station, section, operation process, etc.; develop event information collection function to support multi-terminal input of train operation event data;

[0014] (3) Bayesian analysis algorithm design: Based on Bayes' theorem and combined with the logic of Heinrich's law that "accumulation of anomalies leads to accidents", a driving event data analysis algorithm is designed. Two calculation methods, probability value and proportion value, are used to quantify the evolutionary relationship between abnormal phenomena.

[0015] (4) Algorithm verification and optimization: Apply the developed algorithm to historical driving event data. If the verification fails, redesign the algorithm. If the verification succeeds, proceed to the next step.

[0016] (5) Data acquisition and analysis system development: Using computer programming technology and combined with third-party open source modules, develop a driving event data acquisition and analysis system to realize event information input, automatic data analysis, visualization display functions, and generate a panoramic view of the evolution of driving events;

[0017] (6) System testing and application: Input train operation event data to test the system. If the test fails, rebuild the data acquisition model. If the test passes, apply it to urban rail train operation safety management.

[0018] Preferably, in the five-level data acquisition model: the phenomenon layer records abnormal phenomena and related information observed during driving; the handling layer records standardized emergency response measures for abnormal phenomena; the attribute layer records the time, location, personnel, and equipment attribute information corresponding to each abnormal phenomenon; the evaluation layer performs qualitative evaluation of the safety attributes of driving events; and the relational symbols adopt parallel, evolution, addition, subtraction, right, and wrong relationships to associate elements within and between each level.

[0019] Preferably, the specific calculation logic of the Bayesian analysis algorithm is as follows: Let Z be the total number of occurrences of anomaly B within a period, Y be the total number of occurrences of anomaly A, and X be the number of occurrences of anomaly A immediately following anomaly B within a period; the probability value of anomaly B converting into anomaly A = X / Z, and the proportion of anomaly A caused by anomaly B = X / Y.

[0020] Preferably, the algorithm can also be combined with computer vision technology. By installing cameras, microcontrollers, power supplies, network cards and other devices in the driver's cab, the algorithm can automatically collect the driver's physiological indicators, behavioral characteristics and on-site environmental data. Combined with Bayesian algorithm analysis of data correlation, it can provide data support for safety management and training.

[0021] Preferably, the visualization display function of the data acquisition and analysis system includes an event evolution panorama and a probability / proportion table. The event evolution panorama allows for a macroscopic view of the overall risk of the line and a microscopic view of the correlation between specific abnormal phenomena. The probability / proportion table can separately present the probability value and proportion value of two related phenomena.

[0022] An application of an urban rail transit event analysis algorithm based on Heinrich and Bayesian methods is presented. This algorithm is applied to urban rail transit safety management. Specific application scenarios include the implementation of a dual prevention mechanism, the setting of safety control measures, and the formulation of training plans. This enables precise control of urban rail transit risks, optimized allocation of training resources, and a reduction in the probability of accidents.

[0023] Preferably, the specific application of the dual prevention mechanism is as follows: combining the calculation formulas for risk value and hidden danger value, using the probability value and proportion value calculated by the algorithm, the abnormal phenomenon is digitally assigned a value, thereby realizing the digitalization and systematization of risk classification control and hidden danger investigation and management; the risk value = probability value × severity of subsequent phenomenon / risk value, and the hidden danger value = proportion value × severity of subsequent phenomenon / risk value × rectification difficulty coefficient.

[0024] Preferably, the specific application of the safety control setting is as follows: by generating a panoramic view of the evolution of driving events through an algorithm, the evolution relationship of abnormal phenomena on the line is macroscopically displayed, focusing on a single abnormal phenomenon, analyzing the proportion of its precursor sources and the probability of its subsequent evolution, locating the risk control position, and setting targeted control measures.

[0025] Preferably, the specific application method of the training plan is as follows: combining the probability value and proportion value of abnormal phenomena, and the human-machine-environment-management driving data associated with the algorithm, a refined training plan is formulated for high-risk abnormal phenomena, taking into account different vehicle types, stations, passenger flow periods, and passenger-carrying status scenarios.

[0026] The present invention has the following beneficial effects:

[0027] I. Beneficial effects from a technical perspective

[0028] 1. The digital and quantitative application of Heinrich's Law has been realized: Breaking through the limitations of the traditional Heinrich's Law, which is only used for safety awareness education and is difficult to quantify for early warning, the evolution probability between abnormal phenomena is accurately calculated through Bayes' formula, realizing the transformation from qualitative logic to quantitative analysis, and enabling Heinrich's Law to be accurately adapted to the complex scenarios of urban rail transit operation.

[0029] 2. An evolution map of abnormal phenomena was constructed to achieve risk visualization: Based on event data, an "evolution panorama of abnormal phenomena" (event map) was generated, which intuitively shows the relationship and development path between abnormal phenomena. It supports macro (overall risk of the line) and micro (related to specific phenomena) dual perspective analysis, making risk nodes clearly identifiable.

[0030] 3. Enhanced the depth of analysis and scientific basis of decision-making for driving incident data: By calculating both probability and proportion values, high-risk nodes in the evolution of abnormal phenomena are accurately identified, providing precise data support for safety management and training resource allocation, avoiding blind resource allocation, and achieving "precise prevention and control".

[0031] 4. Support the digital implementation of the dual prevention mechanism: By combining the calculation formulas for risk value and hazard value, abnormal phenomena are quantitatively assessed, and the dual prevention mechanism is transformed from "experience-driven" to "data-driven", realizing the data-driven and systematic management of risk classification and control and hazard investigation and management.

[0032] II. Beneficial Effects on Safety and Operations Management

[0033] 1. Enhance driving safety early warning capabilities: Through the probability analysis of abnormal phenomenon evolution, high-risk links can be identified in advance, and targeted control measures can be implemented; at the same time, it supports the source analysis of accident events, which can accurately trace the source of abnormal phenomena and fundamentally cut off the accident evolution chain.

[0034] 2. Optimize training plans and resource allocation: Develop training content based on high-risk and high-potential abnormal phenomena, optimize training programs in combination with specific scenarios, improve the relevance and effectiveness of training, avoid the waste of resources in "all-encompassing" training, and achieve the rational allocation of training resources.

[0035] 3. Reduce the probability of safety accidents: Through data-driven safety management and control, high-risk nodes can be precisely managed, reducing safety accidents caused by human error and equipment failure, and improving the stability and reliability of the urban rail transit operation system.

[0036] III. Economic Benefits

[0037] 1. Reduce management costs: Enable paperless reporting and automatic analysis of driving events, saving on paper, printing and other consumable costs; at the same time, reduce the time spent on manual data entry and statistical analysis, improve management efficiency and reduce labor costs.

[0038] 2. Save on system development and consulting costs: Developing the data analysis system in-house avoids the high costs of purchasing external software or consulting services; the system architecture is flexible and scalable, supports multi-line and multi-dimensional data analysis, has low promotion costs, and offers good economic benefits.

[0039] IV. Social and Organizational Benefits

[0040] 1. Improve passenger travel safety and service quality: Reduce operational accidents, improve train punctuality and driving safety, and enhance public trust in the rail transit system through advance warning and process control.

[0041] 2. Reduce employee workload: Simplify the process of reporting driving incidents, reduce non-core work time for drivers and managers, and improve employee job satisfaction and work efficiency.

[0042] 3. It has industry promotion value: The system architecture is flexible and the data model can be adapted to rail transit systems of different lines and cities. It provides the urban rail transit industry with a replicable and scalable smart safety management system solution, promoting the overall improvement of the industry's safety management level. Attached Figure Description

[0043] Figure 1This is a schematic diagram of the driving event data model of the present invention;

[0044] Figure 2 This is a probability diagram illustrating the evolution of abnormal phenomena in driving events according to the present invention.

[0045] Figure 3 This is a diagram illustrating the statistical steady-state control of the present invention;

[0046] Figure 4 This is a schematic diagram of the Bayesian formula of the present invention;

[0047] Figure 5 This is a schematic diagram of the driving event data analysis algorithm of the present invention;

[0048] Figure 6 The algorithm application diagram of this invention is shown (including the event map interface).

[0049] Figure 7 A graph showing the assignment of risk values ​​and potential hazards to abnormal phenomena in this invention;

[0050] Figure 8 This is a partial diagram illustrating the evolution of abnormal phenomena in this invention.

[0051] Figure 9 This is a table showing the probability / percentage of abnormal phenomena in this invention;

[0052] Figure 10 This is a schematic diagram of the security card control system of the present invention;

[0053] Figure 11 This is a training plan template for the present invention;

[0054] Figure 12 This is the multi-terminal data collection interface for driving events of the present invention;

[0055] Figure 13 This is the web application interface of the present invention;

[0056] Figure 14 This is a schematic diagram illustrating a related phenomenon of the present invention. Detailed Implementation

[0057] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] An algorithm for analyzing urban rail transit events based on Heinrich's law and Bayes' theorem is proposed. Combining the core logic of Heinrich's law, Kantian epistemology, and Bayes' theorem, this algorithm achieves precise analysis of the evolution patterns of abnormal phenomena in urban rail transit events by constructing a standardized data collection model for train operation events, designing a quantitative analysis algorithm, and developing a data collection and analysis system. The specific steps are as follows:

[0059] 1. Construction of a Driving Event Data Acquisition Model: Combining Kantian epistemology and set theory, the definitions of driving events and abnormal phenomena are clarified—abnormal phenomena correspond to sample points in set theory, while driving events correspond to sample spaces; that is, a driving event is a set of multiple abnormal phenomena in a certain scenario; refer to... Figure 1 (Diagram of driving event data model) Construct a five-level data acquisition model including phenomenon layer, handling layer, attribute layer, evaluation layer, and relational symbol, where (each level corresponds to...) Figure 1 (Middle Phenomenon Layer, Processing Layer, Attribute Layer, Evaluation Layer, Relationship Symbol)

[0060] - Phenomenon layer: Records abnormal phenomena observed during driving and related information obtained, serving as the basis for event analysis;

[0061] - Response layer: Standardized records of emergency response measures taken in response to abnormal phenomena, forming complete response data;

[0062] - Attribute layer: Records the time, location, personnel, equipment, and other attribute information corresponding to each anomaly to ensure data traceability;

[0063] - Evaluation layer: Qualitatively evaluates the safety attributes of driving events to provide a basis for data analysis and event determination;

[0064] - Relational operators: Using relational operators such as parallel, evolution, addition, subtraction, right, and wrong, elements within and between levels are associated, laying the foundation for data axiomatics.

[0065] 2. Basic Data Preparation: Based on the established data collection model, create a basic data template and fill it with basic information related to urban rail transit operation, including train type, stations, sections, and operational procedures; develop an event information collection function to support multi-terminal (web page, mobile device, etc.) input of train operation event data, ensuring standardized and diversified data collection. The input interface can be referenced. Figure 12 (Multi-terminal data collection interface for driving events) Figure 13 (Web application interface) design.

[0066] 3. Bayesian Analysis Algorithm Design: Based on Bayes' theorem and combined with Heinrich's rule ("accumulated anomalies lead to accidents"), and referring to... Figure 4 (Diagram of Bayes' formula) and Figure 5(Diagram of driving event data analysis algorithm) The driving event data analysis algorithm is designed, using both probability and proportion values ​​to quantify the evolutionary correlation between abnormal phenomena. The specific calculation logic is as follows (the calculation results can be referenced). Figure 9 (Table showing probability / percentage of abnormal phenomena)

[0067] Through the above calculations, high-risk nodes in the evolution of anomalies are accurately identified, providing a quantitative basis for risk management. The probability of anomaly evolution can be referenced. Figure 2 (Probability diagram of the evolution of abnormal phenomena in driving events) is shown.

[0068] - Define parameters: Let Z be the total number of occurrences of anomaly B within the period, Y be the total number of occurrences of anomaly A, and X be the number of occurrences of anomaly A immediately following anomaly B within the period.

[0069] - Probability calculation: The probability of anomaly B transforming into anomaly A = X / Z, representing the likelihood of anomaly A occurring after anomaly B occurs;

[0070] - Proportion Calculation: The proportion of abnormal phenomenon A caused by abnormal phenomenon B = X / Y, which represents the proportion of abnormal phenomenon A caused by abnormal phenomenon B.

[0071] 4. Algorithm Verification and Optimization: Design an algorithm verification process, apply the developed algorithm to historical driving event data, and refer to... Figure 5 (Diagram of driving event data analysis algorithm) The algorithm is validated by calculating the probability and proportion of abnormal phenomena, and then compared with... Figure 9 Data is organized in the format of an anomaly probability / percentage table, compared with the actual accident evolution, and algorithm parameters are adjusted to ensure the accuracy of the calculation results. After successful verification, the next step of application development begins. The algorithm verification process can be combined with... Figure 3 (Statistical steady state control diagram) Monitoring algorithm stability.

[0072] 5. Data Acquisition and Analysis System Development: Utilizing computer programming technology and combining third-party open-source modules (such as the Graphviz module), develop a driving event data acquisition and analysis system to realize functions such as event information entry, automatic data analysis, and visualization; develop event information display functions to generate a panoramic view of the evolution of driving events (event map), supporting both macro (overall route risk) and micro (correlation of specific phenomena) dual-perspective analysis. The event map interface can be found in [reference needed]. Figure 6 (Algorithm application illustration (including event map interface)), the local evolution relationship can be referred to. Figure 8 (Diagram of the evolution of the abnormal phenomenon (partial)).

[0073] 6. System Testing and Application: Input driving event data and pass... Figure 12 (Multi-terminal data acquisition interface for driving events) Complete data entry, test the system, and generate a panoramic view of the event evolution (corresponding to...) Figure 6 (Algorithm application illustrations (including event map interface)) and probability / proportion tables (corresponding to) Figure 9 (Table showing probability / percentage of abnormal phenomena) Test results indicate that the system operates stably and the data calculation is accurate. The system is applied to urban rail transit safety management, enabling practical applications such as safety control and training plan development. The system's operational stability can be referenced... Figure 3 (Statistical steady state control diagram) Monitoring.

[0074] Furthermore, the application of the driving event data acquisition model includes two parts: data acquisition and data analysis. Data acquisition is conducted through a multi-terminal input interface (see...). Figure 12 (Multi-terminal data collection interface for driving events) Figure 13 (Web application interface) enables standardized collection of data such as incident time, location, people / objects involved, incident phenomena, and handling; data analysis focuses on time as the core dimension, combined with Bayesian algorithms (see...). Figure 4 (Diagram of Bayes' formula) Figure 5 (Diagram of driving event data analysis algorithm) This algorithm enables correlation analysis between statistical values ​​and transformed values ​​of abnormal phenomena, and supports multi-dimensional logical reasoning. The correlation relationships between abnormal phenomena can be found by referring to... Figure 14 (Illustrative diagram of related phenomena).

[0075] Furthermore, the event mapping function, through a third-party open-source module, evolves and correlates anomalies over time, calculates the probability and proportion of anomalies on a specific operating route over a period of time, and generates an evolutionary graphical diagram (corresponding to...). Figure 6 (Algorithm application illustrations (including event map interface)) Figure 8 (Evolution diagram of abnormal phenomena (partial)); It also provides a table display function, which can separately present the probability values ​​and percentage values ​​of two related phenomena (corresponding to...) Figure 9 (Probability / Percentage Table of Abnormal Phenomena) makes it easy for data users to quickly locate key information.

[0076] Furthermore, the algorithm can also be combined with computer vision technology to achieve automatic acquisition and analysis of visual data. By installing cameras, microcontrollers, power supplies, network cards, and other devices in the driver's cab, it can automatically capture the driver's physiological indicators (such as eye-closing time and number of eye-closing times), behavioral characteristics (such as lookout intervals and time spent away from the driver's cab), and on-site environmental data, combined with Bayesian algorithms (see...). Figure 4 (Diagram of Bayes' formula) Figure 5 (Diagram of driving event data analysis algorithm) Analyzes data correlations to identify high-risk nodes (corresponding to) caused by factors such as fatigued driving. Figure 8(Evolution diagram of abnormal phenomena (partial)) provides more data support for safety management and training.

[0077] Applications of Algorithms and Systems

[0078] The analysis algorithm and the developed system of this invention are mainly applied in the field of urban rail transit safety management, specifically including the following three aspects:

[0079] 3.1 Digital Application of Dual Prevention Mechanism

[0080] Combining the risk value and hidden danger value calculation formulas in the dual prevention mechanism, the probability value and proportion value (corresponding to) calculated using the algorithm of this invention are... Figure 9 (Table of Probability / Percentage of Abnormal Phenomena) Abnormal phenomena are assigned digital values ​​to achieve digital and systematic risk classification management and hidden danger investigation and control. The specific calculation formula is as follows, and the assignment logic can be found in the table. Figure 7 (Chart assigning values ​​to risk and hazard values ​​for abnormal phenomena):

[0081] Risk value = Probability value × Severity of aftereffect / Risk value

[0082] Hazard value = Percentage value × Severity of subsequent phenomena / Risk value × Difficulty coefficient of rectification

[0083] In this context, the severity and risk value of post-event phenomena are interchangeable. Only abnormal phenomena (terminal phenomena) included in safety incidents and accidents are assigned severity, which can be determined or assumed based on industry standard risk matrix tables. The rectification difficulty coefficient is set by the relevant specialty of each abnormal phenomenon in conjunction with the actual situation. This is achieved by assigning risk and hazard values ​​to all abnormal phenomena (referencing...). Figure 7 (Assignment chart of risk values ​​and hidden danger values ​​for abnormal phenomena) accurately identifies high-risk and high-hidden-risk abnormal phenomena, providing a targeted basis for hidden danger rectification.

[0084] 3.2 Security Card Control Settings

[0085] Using the event map function (corresponding) Figure 6 (Algorithm application illustration (including event map interface))) This macroscopically displays the evolutionary relationship of all anomalies on a specific operating route over a period of time (corresponding to...) Figure 8 (Anomaly Evolution Diagram (Partial)) allows managers to have a comprehensive understanding of the overall risk situation of the line; at the same time, it allows them to focus on a specific anomaly, combined with... Figure 14 (Illustrative diagram of related phenomena) Analyze the relationships between them, and analyze in detail the proportion of preceding phenomena and the probability of subsequent evolutionary phenomena (corresponding to...) Figure 9 (Table of Probability / Percentage of Abnormal Phenomena) , accurately pinpointing risk control points, refer to Figure 10(Safety control diagram) Set up targeted safety control measures, intervene in high-risk links in real time, and cut off the evolution chain of abnormal phenomena.

[0086] 3.3 Training Plan Development

[0087] Combining the probability value and proportion value of abnormal phenomena (corresponding to) Figure 9 (Table of Probability / Percentage of Abnormal Phenomena)), Risk Value, Potential Hazard Value (corresponding to) Figure 7 (Assignment diagram of risk values ​​and hidden danger values ​​for abnormal phenomena), and the human-machine-environment-management vehicle data model database, targeting high-risk and high-hidden-danger abnormal phenomena (corresponding to Figure 8 (Diagram of abnormal phenomenon evolution (partial))), combined with different vehicle types, stations, sections, passenger flow periods (off-peak and peak periods), passenger carrying status, etc., refer to Figure 11 (Training Plan Template) Develop a detailed training plan to avoid wasting resources on "all-encompassing" training, improve training efficiency, ensure that drivers are proficient in emergency response methods for key abnormal phenomena, and reduce the risk of human error.

[0088] Example 1: Application of Urban Rail Transit Operation Risk Management Based on Basic Algorithms (Shenzhen Metro Line 1)

[0089] This embodiment uses Shenzhen Metro Line 1 as the application scenario. The main line has an operating mileage of approximately 42 kilometers, 30 stations, and five types of operating trains. The line has stable passenger flow and diverse train configurations, making it suitable for conducting basic algorithm implementation verification. Focusing on the full-process application of core algorithms, it achieves standardized collection, digital analysis, and basic safety control of train operation events. The specific implementation process is as follows:

[0090] 1. Building a driving event data collection model

[0091] Combining Kantian epistemology and set theory, this paper clarifies the core definitions of train operation events and anomalies on Shenzhen Metro Line 1: Anomalies are defined as sample points in set theory, with a single train operation event corresponding to a complete sample space. That is, multiple related or independent anomalies occurring in a scenario together constitute a complete train operation event. Based on this definition, a five-level data acquisition model is constructed, comprising a phenomenon layer, a handling layer, an attribute layer, an evaluation layer, and relational operators. Each level has a clear division of labor and is interconnected, forming a complete data acquisition system. The specific level settings are as follows:

[0092] - Phenomenon layer: As the basic data source of the entire analysis system, it focuses on recording common abnormal phenomena that can be observed and traced during train operation, covering core contents such as train idling and coasting, running failure, doors and platform doors not interlocked and opened, abnormal operating conditions of traction system, and abnormal response of braking system. It also records the real-time status information of related equipment to ensure that the phenomenon description is complete and without omission.

[0093] - Response Layer: Standardized emergency response record specifications are established for various abnormal phenomena. These specifications uniformly record the response measures taken by the driver, such as changing driving mode, reporting to the dispatching department, manually aligning with the target, triggering emergency braking, and restarting and resetting the equipment. This eliminates the problem of chaotic and non-standard descriptions in the response records and ensures that the data of the entire response process is traceable and reviewable.

[0094] - Attribute layer: Refine the supporting attribute information of abnormal phenomena, accurately record the incident time (accurate to the minute level), the station or operating section where the incident occurred, the vehicle model number involved, the driver's name and employee number, the faulty equipment number and other key information, improve the data tags from multiple dimensions, and provide complete support for subsequent multi-dimensional data analysis and source tracing.

[0095] - Evaluation layer: Establish qualitative evaluation standards for the safety level of driving events, classify driving events into three safety levels: general, major, and serious, and complete an objective qualitative evaluation for each driving event to provide core basis for subsequent data analysis, event classification, and risk assessment;

[0096] - Relationship Symbols: Two core relationship symbols, evolution and parallel, are selected to clearly associate each abnormal phenomenon with the corresponding handling measures and clarify the logical relationship between the phenomena. For example, "traction system abnormality causes idling and coasting" is defined as an evolution relationship, and "car door not open, platform door not open" is defined as a parallel relationship. This clarifies the data logic and lays the foundation for subsequent quantitative analysis.

[0097] 2. Basic Data Preparation

[0098] Based on the established five-level data acquisition model, a dedicated basic data template was created and pre-filled with core parameters for the five types of train models on Shenzhen Metro Line 1, basic information for 30 stations and the entire line, and standardized train operation procedures. This reduces the workload of subsequent data entry and ensures the consistency of basic information. Simultaneously, a dual-platform event entry interface was developed for both web and mobile devices. The interface design is tailored to drivers' working habits and includes four core data collection modules: incident time, incident location, people / objects involved, incident phenomena, and handling. The interface is simple and convenient to use, allowing drivers to quickly complete data entry during breaks, achieving standardized and diversified data collection, and covering all operating teams and trains along the entire line.

[0099] 3. Design and Verification of Bayesian Analysis Algorithm

[0100] Based on Bayes' theorem as the core computational logic, and deeply integrated with Heinrich's law's core safety logic of "accumulated anomalies leading to accidents," a dedicated driving event data analysis algorithm was designed. It abandons a single analytical dimension and adopts a dual calculation method of probability and proportion values ​​to accurately quantify the evolutionary correlation between abnormal phenomena. The specific computational logic is clear and standardized, as follows:

[0101] First, define the core calculation parameters: Within the statistical period, set the total number of occurrences of anomaly B as Z, the total number of occurrences of anomaly A as Y, and the number of times anomaly A immediately follows anomaly B within the same period as X. Based on these parameters, calculate two types of core indicators: the probability of anomaly B transforming into anomaly A = X / Z, which directly reflects the likelihood of anomaly A following the occurrence of anomaly B; and the percentage of anomaly A caused by anomaly B = X / Y, which directly reflects the percentage of all occurrences of anomaly A directly caused by anomaly B.

[0102] Complete data from 500 historical train operation events on Shenzhen Metro Line 1 over the past three years were selected. The algorithm was applied to the historical data to complete the entire process verification. The probability and proportion of various abnormal phenomena were calculated one by one according to the algorithm logic, and the data were compiled into a standardized table. The calculation results were compared with the actual accident evolution and on-site handling. The algorithm parameters were adjusted and the calculation logic was optimized to ensure that the accuracy of the algorithm calculation results was not less than 90%. After passing multiple rounds of verification, the system officially entered the subsequent development stage to ensure the practicality and accuracy of the algorithm.

[0103] 4. Development and testing of data acquisition and analysis system

[0104] Utilizing conventional computer programming techniques and combining them with the Graphviz third-party open-source visualization module, a basic version of a driving event data acquisition and analysis system was developed. Focusing on core functionalities, the system enables rapid event information entry, automatic data analysis, and visualized event evolution. The system interface is simple, clear, and easy to use. A dedicated event display interface was also developed, supporting a global view of the entire route's abnormal phenomenon evolution, as well as a micro-view to examine the correlation between individual abnormal phenomena, balancing overall control with detailed analysis.

[0105] The system inputs 1,000 train operation events from a single year on Shenzhen Metro Line 1. The entire data is uploaded via a dual-port input interface. The system undergoes continuous stability and data accuracy testing, generating a complete panoramic view of the event evolution and tables of probability and percentage values. The system runs smoothly without any lag, data loss, or calculation errors. All functions operate stably and the data calculations are accurate, meeting the needs of actual operation.

[0106] 5. System Application and Implementation Results

[0107] The tested and qualified system will be officially applied to the daily operation safety management of Shenzhen Metro Line 1, focusing on two core application scenarios to effectively leverage data-driven safety control. Specific application details are as follows:

[0108] - Safety control settings: Based on the panoramic event evolution map generated by the system, the system focuses on frequently occurring abnormal phenomena such as insufficient marking and coasting during idle driving. It deeply analyzes the correlation between these phenomena and identifies the core high-risk evolution path of insufficient marking caused by traction inverter failure and coasting during idle driving caused by braking system abnormality. For traction inverter failure, real-time dynamic monitoring and control measures are set up. Once an abnormality is detected, the system will promptly remind the driver to intervene quickly. For braking system abnormality, the system will regularly notify the professional technical department to carry out targeted inspection and maintenance, optimize the entire process control procedure, and block the risk evolution chain from the source.

[0109] - Basic training plan development: Based on the probability and percentage values ​​of abnormal phenomena calculated by the algorithm, the core high-risk evolution paths are identified, and a special driver training plan is developed. The focus is on training drivers on standardized emergency response methods in scenarios where the traction and braking systems malfunction. The training content covers all vehicle types and scenarios at all stations along the entire line, improving the relevance and practicality of the training.

[0110] Implementation Results: After six months of continuous application of the system, the stability and control effectiveness of the system operation were continuously monitored. The incidence of high-frequency abnormal phenomena such as under-standard operation and idle coasting on Shenzhen Metro Line 1 decreased by 32%, the accuracy of drivers' emergency response operations increased by 38%, and the number of driving incidents caused by human error decreased by 25%. The system effectively improved the overall driving safety control level of the line and reduced the risk of safe operation.

[0111] Example 2: Application of Smart Card Control for Urban Rail Transit Using Computer Vision (Guangzhou Metro Line 3)

[0112] This embodiment uses Guangzhou Metro Line 3 as an application scenario. During peak hours, this line experiences high passenger flow and drivers work continuously at high intensity. With a mainline operating length of approximately 64 kilometers and 29 stations, it is a core urban backbone line, demanding higher precision in train safety management. Building upon the core algorithm and basic system functions of Embodiment 1, this embodiment adds computer vision data acquisition and intelligent analysis functions to achieve intelligent and automated monitoring of driver work behavior and the on-site operating environment, addressing the shortcomings of manual supervision. The specific implementation process is as follows:

[0113] 1. Data acquisition model optimization and vision device deployment

[0114] Based on the five-level data acquisition model established in Example 1, and combined with the application requirements of computer vision, the content of the phenomenon layer and attribute layer was specifically optimized. A dedicated module for acquiring driver physiological indicators, behavioral characteristics, and on-site environmental data was added, expanding the dimensions of data acquisition. Simultaneously, high-definition cameras, microcontrollers, stable power supplies, and 5G high-speed network cards were deployed throughout the driver's cabs of all operating trains on Guangzhou Metro Line 3, enabling automatic real-time capture, local storage, and high-speed uploading of visual data without manual intervention.

[0115] The system focuses on collecting two types of core visual data: first, driver physiological and behavioral data, including key indicators such as continuous eye closure duration, number of eye closures per unit time, lookout interval duration, and time away from the driver's cab; and second, on-site environmental data, covering the operating status of the driver's cab control panel and the track environment outside the window, comprehensively capturing the driver's working status and on-site operating environment, providing a complete data source for subsequent intelligent analysis.

[0116] 2. Bayesian Algorithm Optimization and Validation

[0117] Based on the Bayesian analysis algorithm framework of Example 1, and combined with the characteristics of the newly added visual data, the algorithm logic was optimized and upgraded, and the visual data was formally included in the full-process analysis scope. The new logic for correlation analysis between visual data and abnormal driving phenomena was added, and the evolutionary correlation between abnormal driver behavior and driving accidents and abnormal phenomena was accurately calculated. For example, the probability and proportion of high-frequency risk paths such as the train overshooting due to the driver closing his eyes for too long or the failure to detect foreign objects on the track in time due to the excessively long lookout interval were accurately identified. High-risk nodes caused by human factors such as fatigue driving and operational negligence were accurately identified, and full-dimensional control from equipment abnormalities to human risks was achieved.

[0118] Data from 400 historical train operation events accumulated over the past two years on Guangzhou Metro Line 3 were selected, along with 1,000 hours of complete visual monitoring data. The optimized and upgraded algorithm underwent comprehensive verification. The calculated probability and percentage values ​​were compiled and compared with actual cases one by one to ensure that the accuracy of the correlation analysis between visual data and anomalies was no less than 88%. After passing multiple rounds of testing and verification, the system function upgrade was officially completed, ensuring that the algorithm is adapted to the needs of visual data analysis.

[0119] 3. System upgrade and testing

[0120] The existing basic data acquisition and analysis system has been upgraded with three new core modules: an automatic visual data acquisition module, an intelligent visual data analysis module, and an intelligent early warning module for abnormal behavior. This enables seamless linkage and integrated analysis of visual data and routine driving event data. The panoramic display of event evolution has been optimized, and visual data association annotations have been added. When viewing any abnormal phenomenon, driver behavior and on-site environmental data for the corresponding time period can be retrieved simultaneously, enabling data linkage and traceability. A new intelligent early warning function has been added. By setting standardized thresholds, the system monitors the driver's operating status in real time. Once abnormalities such as excessive eye-closing time or excessively long lookout intervals are detected, an automatic audible and visual warning is issued to remind the driver to correct their behavior promptly.

[0121] Data from 500 train operation events and 500 hours of visual monitoring data for a complete quarter of Guangzhou Metro Line 3 were entered. A full-process test was conducted on the upgraded system, generating a panoramic map of event evolution including visual data correlation, a probability percentage data table, and a special analysis report on driver behavior. The test results showed that the system operated stably throughout the process, the abnormal behavior warning response time was no more than 1 second, and the data linkage analysis was accurate and error-free, meeting the intelligent management and control requirements of peak lines.

[0122] 4. System Application and Implementation Results

[0123] The upgraded intelligent system has been officially applied to the train operation safety management of Guangzhou Metro Line 3. Building upon the existing safety control and basic training applications of Implementation Example 1, two new intelligent application scenarios have been added to further improve the control system. Specific application details are as follows:

[0124] - Intelligent driver behavior control: Relying on visual data collection and intelligent analysis functions, the system automatically identifies abnormal behaviors such as driver fatigue, operational negligence, and inadequate observation, and issues early warnings and interventions in the first instance. At the same time, it conducts in-depth analysis on the intrinsic relationship between abnormal driver behavior and abnormal driving phenomena, and generates special analysis reports. This provides objective data support for the daily management of driver behavior and performance evaluation, replacing the traditional manual spot check mode and improving supervision efficiency.

[0125] - Refined training plan development: By combining quantitative data on the probability, proportion, and risk values ​​of abnormal phenomena with driver behavior analysis data, we can accurately identify high-frequency violations such as fatigued driving and failure to keep a timely lookout. We can then develop personalized and differentiated training plans to address these issues, abandoning the broad-based, all-encompassing training model and focusing on strengthening practical training in weak areas, thereby significantly improving the relevance and efficiency of training.

[0126] Implementation Results: After six months of continuous application, the incidence of abnormal driving phenomena caused by human factors such as driver fatigue and operational negligence on Guangzhou Metro Line 3 decreased by 45%. The accuracy rate of the system's intelligent early warning was no less than 90%, and the rate of standardized operation by drivers increased by 50%. This further reduced the risk of train safety on the line and significantly reduced the daily supervision pressure on front-line safety management personnel, achieving intelligent and lightweight management.

[0127] Example 3: Digital Implementation of a Dual Prevention Mechanism (Shanghai Metro Line 10)

[0128] This embodiment uses Shanghai Metro Line 10 as an application scenario. This line covers various operating sections, including underground, ground-level, and elevated sections, resulting in a complex and variable operating environment. The main line is approximately 36 kilometers long with 28 stations, placing extremely high demands on the systematization and refinement of safety management. This embodiment focuses on the core objective of digitally implementing a dual-prevention mechanism for safe production. Building upon the core algorithm and basic system of Embodiment 1, it prioritizes optimizing the quantitative analysis functions of risk and hazard values. This achieves full-process digitalization and systematization of risk classification management and hazard investigation and control, completely transforming the traditional experience-based management model. The specific implementation process is as follows:

[0129] 1. Data acquisition model optimization and basic data improvement

[0130] Based on the five-level data acquisition model in Example 1, the evaluation layer content has been specifically optimized, and a special evaluation and calculation module for risk values ​​and hidden danger values ​​has been added to adapt to the digital implementation needs of the dual prevention mechanism. Simultaneously, the original basic data template has been improved, and two core standards have been added: one is the standard for setting the difficulty coefficient of abnormal phenomenon rectification, and the other is the standard for assigning values ​​to the severity of subsequent phenomena. The assignment standards are formulated in conjunction with industry-standard risk matrix tables to ensure compliance and practicality.

[0131] Clear assignment rules: Only terminal anomalies that fall under the category of safety incidents and accidents are assigned a severity level, which is divided into levels 1-10, with higher levels indicating greater risks; the rectification difficulty coefficient is divided into the range of 0.1-1.0, and each specialty such as equipment, crew, and engineering sets a corresponding coefficient based on the actual situation of on-site rectification workload, technical difficulty, time and cost, etc., to ensure that the quantitative assignment is objective and feasible.

[0132] 2. Optimization by combining Bayesian algorithm with dual prevention mechanism

[0133] Based on the Bayesian algorithm framework of Example 1, the core requirements of the dual prevention mechanism are deeply integrated, the core logic of the algorithm is optimized, and a new function for quantitative calculation of risk and hidden danger values ​​is added. This enables the digital assignment and classification of all abnormal phenomena, completely breaking down the barriers between the algorithm and the dual prevention mechanism. The specific calculation formulas are standardized and clear, as follows:

[0134] Risk value = Probability value × Severity of aftereffect / Risk value

[0135] Hazard value = Percentage value × Severity of subsequent phenomena / Risk value × Difficulty coefficient of rectification

[0136] The formula explicitly states that the severity and risk value of subsequent phenomena can be flexibly interchanged according to actual analysis needs, adapting to different control scenarios. Through this formula, all abnormal phenomena within the line are fully covered by risk and hazard values, enabling precise screening and identification of high-risk and high-hazard anomalies. This provides accurate quantitative basis for subsequent targeted rectification of hazards and risk-level control. Data from 600 historical train operation events accumulated over the past three years on Shanghai Metro Line 10 were used to conduct full-process verification of the optimized fusion algorithm, ensuring that the accuracy rate of risk and hazard value calculations is no less than 92%, guaranteeing that the quantitative results align with actual control needs.

[0137] 3. System function upgrades and testing

[0138] The existing basic system has been upgraded with three new core modules: a risk intelligent classification module, a hidden danger closed-loop investigation module, and a rectification progress tracking module. This enables full-process digital closed-loop management of risk classification and control, hidden danger investigation and management, and rectification tracking and closure. The panoramic display function of event evolution has also been optimized, with real-time labeling of risk and hidden danger values. It supports quick filtering of abnormal phenomena according to risk and hidden danger levels, intuitively displaying high-risk and high-hidden danger core nodes within the line, facilitating managers to quickly identify key control points.

[0139] A dedicated interface for tracking hazard rectification was developed concurrently, enabling full online management of the entire process, from initiating hazard rectification tasks online to real-time progress updates and closed-loop cancellation upon completion. The entire process is traceable and documented. Data from 800 operational incidents on Shanghai Metro Line 10 over a full year was entered, and the upgraded system underwent comprehensive testing. This resulted in the generation of risk classification reports, hazard investigation reports, and a panoramic map of event evolution. Test results showed that the system can accurately quantify and assign risk and hazard values ​​for all abnormal phenomena, achieving a hazard rectification tracking closure rate of no less than 95%. The system is stable, fully functional, and meets the requirements for the digital implementation of the dual prevention mechanism.

[0140] 4. System Application and Implementation Results

[0141] The upgraded digital system has been officially applied to the train safety management of Shanghai Metro Line 10, focusing on the digital implementation of the dual prevention mechanism and comprehensively carrying out three core applications to improve the quality and efficiency of safety management. The specific application content is as follows:

[0142] - Risk classification and control: Based on the risk value of abnormal phenomena calculated by the system, a four-level risk classification standard is established, which divides risks into four levels: major, relatively large, general and low risk. Differentiated control measures are formulated for different levels of risk, and superior resources are concentrated to focus on controlling major and relatively large risk nodes, optimize the allocation of control resources, avoid extensive control, and achieve precise risk control.

[0143] - Hazard identification and mitigation: Based on the hazard values ​​of abnormal phenomena calculated by the system, a hazard classification and rectification mechanism is established. High-hazard abnormal phenomena are prioritized for identification and rectification within a specified time limit. Special hazard rectification tasks are initiated directly through the system, and the rectification progress is tracked in real time. After the rectification is completed, the task is reviewed and closed online, realizing a closed-loop management of hazard rectification. At the same time, the root causes of hazards are analyzed in depth, and targeted prevention and control measures are formulated to fundamentally reduce the recurrence of similar hazards.

[0144] - Training plan optimization: By combining four core data categories of abnormal phenomena—risk value, hidden danger value, probability value, and proportion value—high-risk and high-hidden danger abnormal phenomena can be accurately identified. Based on the characteristics of different operating sections of the line (underground, ground, and elevated) and different passenger flow periods (off-peak and peak), a refined and scenario-based training plan can be developed to fit actual operating scenarios, strengthen the ability to handle weak links, and improve training effectiveness.

[0145] Implementation Results: After eight months of continuous application, the incidence of high-risk anomalies on Shanghai Metro Line 10 decreased by 40%, the closed-loop rectification rate of high-risk anomalies reached 98%, and the incidence of various traffic safety accidents decreased by 30%. The system successfully achieved a core transformation of the dual prevention mechanism from traditional experience-driven to data-driven, and comprehensively improved the systematization, refinement, and intelligence of line traffic safety management.

[0146] The examples provided in this invention are not intended to limit the implementation. Those skilled in the art will recognize that various variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations, and any obvious variations or modifications derived therefrom are still within the scope of this invention.

Claims

1. A city rail transit traffic event analysis algorithm based on Heinrich and Bayesian methods, characterized in that, Combining the core logic of Heinrich's Law, Kant's epistemology, and Bayes' theorem, this study achieves precise analysis of the evolution patterns of abnormal phenomena in urban rail transit events by constructing a standardized data collection model for train operation events, designing quantitative analysis algorithms, and developing a data collection and analysis system. The specific steps are as follows: (1) Construction of driving event data collection model: Combining Kantian epistemology and set theory, the definitions of driving events and abnormal phenomena are clarified. Abnormal phenomena correspond to sample points in set theory, and driving events correspond to sample space. That is, a driving event is a collection of multiple abnormal phenomena in a certain scenario. A five-level data collection model including phenomenon layer, handling layer, attribute layer, evaluation layer and relation symbol is constructed. (2) Basic data preparation: Based on the established data collection model, create a basic data template and fill in the basic information related to urban rail transit operation, including vehicle type, station, section, operation process, etc. Develop an event information collection function to support the input of driving event data from multiple terminals; (3) Bayesian analysis algorithm design: Based on Bayes' theorem and combined with the logic of Heinrich's law that "accumulation of anomalies leads to accidents", a driving event data analysis algorithm is designed. Two calculation methods, probability value and proportion value, are used to quantify the evolutionary relationship between abnormal phenomena. (4) Algorithm verification and optimization: Apply the developed algorithm to historical driving event data. If the verification fails, redesign the algorithm. If the verification succeeds, proceed to the next step. (5) Data acquisition and analysis system development: Using computer programming technology and combined with third-party open source modules, develop a driving event data acquisition and analysis system to realize event information input, automatic data analysis, visualization display functions, and generate a panoramic view of the evolution of driving events; (6) System testing and application: Input train operation event data to test the system. If the test fails, rebuild the data acquisition model. If the test passes, apply it to urban rail train operation safety management.

2. The urban rail transit event analysis algorithm based on Heinrich and Bayes as described in claim 1, characterized in that, In the five-level data acquisition model: the phenomenon layer records abnormal phenomena and related information observed during driving; the handling layer records standardized emergency response measures for abnormal phenomena; the attribute layer records the time, location, personnel, and equipment attribute information corresponding to each abnormal phenomenon; the evaluation layer performs qualitative evaluation of the safety attributes of driving events; and the relational symbols adopt parallel, evolution, addition, subtraction, right, and wrong relationships to associate elements within and between each level.

3. The urban rail transit event analysis algorithm based on Heinrich and Bayes as described in claim 1, characterized in that, The specific calculation logic of the Bayesian analysis algorithm is as follows: Let Z be the total number of occurrences of anomaly B within a period, Y be the total number of occurrences of anomaly A, and X be the number of occurrences of anomaly A immediately following anomaly B within a period; the probability value of anomaly B transforming into anomaly A is X / Z, and the proportion of anomaly A caused by anomaly B is X / Y.

4. The urban rail transit event analysis algorithm based on Heinrich and Bayes as described in claim 1, characterized in that, The algorithm can also be combined with computer vision technology. By installing cameras, microcontrollers, power supplies, network cards and other devices in the driver's cab, it can automatically collect drivers' physiological indicators, behavioral characteristics and on-site environmental data. Combined with Bayesian algorithm analysis of data correlation, it can provide data support for safety management and training.

5. The urban rail transit event analysis algorithm based on Heinrich and Bayes as described in claim 1, characterized in that, The visualization display function of the data acquisition and analysis system includes an event evolution panorama and a probability / proportion table. The event evolution panorama allows for a macroscopic view of the overall risk of the line and a microscopic view of the correlation between specific abnormal phenomena. The probability / proportion table can separately present the probability value and proportion value of two related phenomena.

6. An application of an urban rail transit event analysis algorithm based on Heinrich and Bayesian methods, characterized in that, The analysis algorithm described in any one of claims 1-5 can be applied to urban rail transit safety management. Specific application scenarios include the implementation of a dual prevention mechanism, the setting of safety control measures, and the formulation of training plans, so as to achieve precise control of urban rail transit risks, optimize the allocation of training resources, and reduce the probability of accidents.

7. The application of the urban rail transit event analysis algorithm based on Heinrich and Bayes as described in claim 6, characterized in that, The specific application of the dual prevention mechanism is as follows: combining the calculation formulas for risk value and hidden danger value, using the probability value and proportion value calculated by the algorithm, digitally assigning values ​​to abnormal phenomena, and realizing the digitalization and systematization of risk classification control and hidden danger investigation and management; The risk value = probability value × severity of subsequent phenomenon / risk value; the hidden danger value = proportion value × severity of subsequent phenomenon / risk value × rectification difficulty coefficient.

8. The application of the urban rail transit event analysis algorithm based on Heinrich and Bayes as described in claim 6, characterized in that, The specific application of the safety control settings is as follows: by generating a panoramic view of the evolution of driving events through algorithms, the evolution relationship of abnormal phenomena on the line is macroscopically displayed, focusing on a single abnormal phenomenon, analyzing the proportion of its precursor sources and the probability of subsequent evolution, locating the risk control position, and setting targeted control measures.

9. The application of the urban rail transit event analysis algorithm based on Heinrich and Bayes as described in claim 6, characterized in that, The specific application method of the training plan is as follows: combining the probability value and proportion value of abnormal phenomena, as well as the human-machine-environment-management driving data associated with the algorithm, a refined training plan is formulated for high-risk abnormal phenomena, taking into account different vehicle types, stations, passenger flow periods, and passenger-carrying status scenarios.

Images