Railway public safety risk dynamic assessment and early warning method

Abstract

The invention relates to the technical field of railway safety, and discloses a railway public safety risk dynamic assessment and early warning method, which comprises the following steps: S1, collecting multi-dimensional big data of a railway operation full scene, the multi-dimensional big data at least comprises equipment operation data, environment perception data, personnel behavior data, road network topology data and public security event historical data; s2, after standardized preprocessing is conducted on the multi-dimensional big data, a multi-source data fusion model is constructed, and space-time correlation fusion of data of different dimensions is achieved through a weighted Bayesian network algorithm; and S3, based on the fused data set, a dynamic risk assessment index system is established, and the index system comprises five first-level indexes including the equipment health degree, the environmental influence degree, the personnel safety compliance degree, the road network bearing capacity and the event associated risk degree. According to the method, multi-dimensional data fusion is achieved, risk assessment is accurate, early warning response is rapid, and the railway public safety guarantee capacity is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of railway safety technology, specifically to a method for dynamic assessment and early warning of railway public safety risks. Background Technology

[0002] Against the backdrop of rapid development in railway transportation, its operating scenarios are becoming increasingly complex, covering various types such as conventional railway trunk lines, high-speed railway hubs, and mountain railways, and facing multi-dimensional safety risks such as aging equipment, extreme weather, personnel violations, passenger congestion, and geological disasters.

[0003] Existing railway safety risk assessment methods mostly rely on single-dimensional data or static assessment models, which have significant limitations: First, the data collection dimensions are one-sided, making it difficult to integrate information from all scenarios, such as equipment operation, environmental perception, and personnel behavior, resulting in incomplete risk identification. Second, data processing lacks a scientific fusion mechanism, failing to effectively address issues such as noise interference and spatiotemporal misalignment from multiple data sources, thus affecting data quality. Third, the assessment indicator system is rigid, and the weight allocation lacks dynamic adjustment capabilities, making it unable to adapt to real-time changing operational scenarios. Fourth, early warning responses are lagging, risk level classifications are vague, and response recommendations are not targeted enough. Furthermore, the coordination of multi-channel push notifications is poor, making it difficult to quickly link multiple platforms such as railway operation management, onboard terminals, and passenger services, thus failing to meet the high-precision and high-efficiency requirements for railway public safety assurance.

[0004] To address this issue, we propose a dynamic assessment and early warning method for railway public safety risks. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] In view of the shortcomings of the prior art, the present invention provides a solution to the problems mentioned in the background art, which relates to the field of railway safety technology.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention provides the following technical solution: a method for dynamic assessment and early warning of railway public safety risks, comprising the following steps:

[0009] S1: Collect multi-dimensional big data from the entire railway operation scenario. The multi-dimensional big data includes at least equipment operation data, environmental perception data, personnel behavior data, railway network topology data, and historical data of public safety incidents.

[0010] S2: After standardizing and preprocessing the multidimensional big data, a multi-source data fusion model is constructed, and the spatiotemporal correlation fusion of data from different dimensions is achieved through a weighted Bayesian network algorithm;

[0011] S3: Based on the fused dataset, establish a dynamic risk assessment indicator system. The indicator system includes five primary indicators: equipment health, environmental impact, personnel safety compliance, road network carrying capacity, and event-related risk. Each primary indicator has at least three secondary quantitative indicators.

[0012] S4: The weights of each indicator are determined by the improved analytic hierarchy process, and the weight coefficients are dynamically adjusted in combination with the real-time data update frequency. The comprehensive evaluation value of railway public safety risk is calculated by the fuzzy comprehensive evaluation method.

[0013] S5: Based on the comprehensive risk assessment value and the preset risk level threshold, generate graded early warning information, which includes the risk level, scope of impact, key risk sources, and targeted handling suggestions.

[0014] Preferably, the equipment operation data includes track geometry parameters, catenary voltage and current data, train braking system status data, and signal equipment transmission delay data, with a data acquisition frequency of 1-5 minutes per time.

[0015] Preferably, the environmental perception data includes meteorological data along the route, geological disaster monitoring data, surrounding obstacle intrusion data, and electromagnetic interference data, wherein the meteorological data includes four core parameters: wind speed, rainfall, visibility, and temperature.

[0016] Preferably, the personnel behavior data includes records of violations by operators, passenger density data, and activity trajectory data of key personnel along the route, which are obtained through video surveillance identification, RFID positioning, and ticketing data correlation analysis.

[0017] Preferably, in the process of constructing the multi-source data fusion model, Kalman filtering is used to denoise the equipment operation data, sliding window mean method is used to smooth the environmental perception data, and outlier detection algorithm is used to remove invalid data from the personnel behavior data.

[0018] Preferably, the improved analytic hierarchy process (AHP) corrects subjective weights by introducing an entropy weight coefficient, which is calculated based on the information entropy of each indicator data. The information entropy reflects the dispersion of the indicator data.

[0019] Preferably, among the secondary indicators of the dynamic risk assessment index system, equipment health includes component wear rate, failure frequency, and maintenance cycle compliance rate, while environmental impact includes the probability of natural disasters, duration of extreme weather, and intensity of surrounding environmental interference.

[0020] Preferably, in the fuzzy comprehensive evaluation method, a five-level fuzzy evaluation matrix is ​​established. The elements of the evaluation matrix are determined by expert scoring combined with historical data statistical analysis, and the risk level of a single indicator is determined by the principle of maximum membership.

[0021] Preferably, the preset risk level threshold is divided into four levels: low risk, medium risk, high risk, and extremely high risk. The threshold for each level is obtained by ROC curve analysis combined with historical data of railway safety operation.

[0022] Preferably, the tiered early warning information is simultaneously pushed through multiple channels, including the railway operation management platform, onboard terminals, early warning broadcasting systems along the railway line, and passenger service APP, with a push delay of no more than 30 seconds.

[0023] (III) Beneficial Effects

[0024] Compared with existing technologies, this invention provides a method for dynamic assessment and early warning of railway public safety risks, which has the following beneficial effects:

[0025] The multi-dimensional data fusion in this invention enables accurate risk assessment, rapid early warning response, and significantly improves the railway public safety assurance capabilities. Detailed Implementation

[0026] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0027] Example 1: Dynamic Assessment and Early Warning of Public Safety Risks on Conventional Railway Trunk Lines

[0028] Data Acquisition: For a conventional railway trunk line (620 km long), multi-dimensional big data was collected at a frequency of 1-5 minutes per acquisition. This included: equipment operation data such as track gauge and level; 25kV voltage and load current data for the overhead contact system; air pressure and braking response time data for the train braking system; and signal equipment transmission delay data; environmental perception data covering wind speed, rainfall, visibility, and temperature data from 12 meteorological monitoring points along the line; displacement and settlement data from 5 potential geological disaster sites; intrusion monitoring data from the protective netting along the line; and electromagnetic interference intensity data; personnel behavior data obtained through recording violations by workers identified by 380 video surveillance cameras along the line, combined with station ticketing data for passenger density analysis, and RFID positioning to obtain the activity trajectories of patrol personnel along the line; and simultaneous collection of network topology data (including bridge, tunnel, and turnout distribution information) and historical data on public safety incidents over the past 10 years (including records of safety incidents caused by equipment failures, natural disasters, and personnel violations).

[0029] Data preprocessing and fusion: Kalman filtering is used to denoise equipment operation data, eliminating random interference in signal transmission; sliding window averaging (window size set to 3 acquisition cycles) is used to smooth environmental perception data, reducing instantaneous fluctuations in meteorological and geological data; outlier detection algorithms are used to remove invalid data (such as misidentification records caused by blurry video surveillance, and abnormal trajectories caused by lost RFID positioning signals) from personnel behavior data. A multi-source data fusion model is constructed based on standardized multi-dimensional data, and a weighted Bayesian network algorithm is used to achieve spatiotemporal correlation fusion of data from different dimensions. For example, geological settlement data of a tunnel section is correlated and matched with track geometry parameters and train frequency data in that area.

[0030] Risk Assessment: Based on the fused dataset, a dynamic risk assessment indicator system is established. Primary indicators include equipment health, environmental impact, personnel safety compliance, network carrying capacity, and event-related risk. Secondary indicators for equipment health include component wear rate, failure frequency, and maintenance cycle compliance rate. Secondary indicators for environmental impact include the probability of natural disasters, duration of extreme weather, and intensity of surrounding environmental interference. An improved analytic hierarchy process (AHP) is used to determine the weights of each indicator. Subjective weights are corrected by introducing an entropy weight coefficient (calculated based on the information dispersion of each indicator's data). This is combined with the real-time data update frequency (e.g., during rainfall, the environmental perception data update frequency increases to 1 minute / time, and the corresponding indicator weights are dynamically adjusted). A five-level fuzzy evaluation matrix is ​​established using the fuzzy comprehensive evaluation method (evaluation matrix elements are determined by combining scores from five railway safety experts and historical data statistical analysis). The maximum membership principle is used to determine the risk level of each indicator, and finally, the comprehensive evaluation value of railway public safety risk is calculated.

[0031] Early Warning Information Generation and Push: Based on the preset four-level risk level thresholds (calibrated by ROC curve analysis combined with nearly 10 years of safe operation history data of the main line: low risk ≤ 0.3, 0.3 < medium risk ≤ 0.6, 0.6 < high risk ≤ 0.8, and extremely high risk > 0.8), when a tunnel section experiences abnormal geological subsidence data due to continuous rainfall, after integrating changes in track geometry parameters and train traffic data, the calculated comprehensive risk evaluation value is 0.72, classifying it as high risk. An early warning message is generated, including the risk level (high risk), the affected area (the tunnel and 2 km sections before and after it), the key risk source (geological subsidence caused by continuous rainfall), and targeted handling suggestions (speed limit of 60 km / h, suspension of non-essential operations in the section, and additional personnel for on-site monitoring). This message is simultaneously pushed through multiple channels, including the railway operation management platform, train onboard terminals, the line-side early warning broadcast system, and the passenger service APP, with a push delay controlled within 22 seconds.

[0032] Example 2: Dynamic Assessment and Early Warning of Public Safety Risks in High-Speed ​​Railway Hubs

[0033] Data Acquisition: For a high-speed railway hub (including 3 passenger stations, 4 access lines, and 1 EMU depot), multi-dimensional big data was collected at a high frequency of 1 minute per acquisition. Equipment operation data focused on high-speed rail track smoothness parameters, catenary contact pressure data, EMU traction system current and voltage data, and CTCS-3 level train control system transmission delay data. Environmental perception data included refined meteorological data (wind speed, rainfall, visibility, and temperature) from 8 meteorological stations in the hub area, structural health monitoring data for bridges and platform canopies within the hub, obstacle intrusion monitoring data for the station area and along the line, and electromagnetic interference data generated by power and communication equipment within the hub. Personnel behavior data was collected through 520 high-definition video surveillance cameras in the station area and along the line. The system identifies records of violations by operational personnel (such as overhead contact line maintenance and signal maintenance personnel), analyzes passenger density by combining real-name ticketing data with station passenger flow monitoring equipment (focusing on areas such as waiting halls, ticket gates, and transfer passages), and obtains the activity trajectories of EMU depot maintenance personnel and station patrol personnel through RFID positioning. Simultaneously, it collects hub network topology data (including station track layout, number and distribution of turnouts, and overhead contact line power supply zones) and historical data on public safety incidents over the past 8 years (including safety incidents caused by high-speed rail equipment failures, passenger congestion, and extreme weather).

[0034] Data Preprocessing and Fusion: The same preprocessing method as in Example 1 is used to denoise the equipment operation data using Kalman filtering, smooth the environmental perception data using the sliding window mean method, and remove outliers from the personnel behavior data. A multi-source data fusion model is constructed using a weighted Bayesian network algorithm to achieve spatiotemporal correlation fusion. For example, passenger density data from a passenger station waiting hall is correlated with personnel violation data identified by video surveillance in the area and the station's ventilation equipment operation status data for analysis.

[0035] Risk Assessment: The dynamic risk assessment indicator system from Example 1 is adopted, using an improved analytic hierarchy process (AHP) to determine indicator weights and dynamically adjusting them based on data update frequency (e.g., during peak travel periods like holidays, the update frequency of personnel behavior data increases, correspondingly raising the weight of personnel safety compliance indicators). A fuzzy comprehensive evaluation method is used to calculate the comprehensive risk assessment value. For example, during the Spring Festival travel peak, if the passenger density in a passenger station's waiting hall exceeds the standard, and some passengers violate regulations by climbing over railings, after integrating ventilation equipment operation data, the calculated comprehensive risk assessment value is 0.85, classifying it as extremely high risk.

[0036] Early warning information generation and push: Based on the calibrated four-level risk level threshold, early warning information is generated, specifying the risk level (extremely high risk), the scope of impact (Waiting Hall 2 of the passenger station and the corresponding ticket gates and transfer channels), key risk sources (excessive passenger gathering + personnel violations), and targeted handling suggestions (opening emergency channels to divert passengers, increasing security personnel to maintain order, and suspending the release speed of some ticket gates in the area). The information is simultaneously pushed through the railway operation management platform, EMU onboard terminals, station broadcasting system, passenger service APP, and electronic display screens in the station, with a push delay of 18 seconds.

[0037] Example 3: Dynamic Assessment and Early Warning of Public Safety Risks on Mountain Railways

[0038] Data collection: For a mountain railway (480 kilometers long, including 28 tunnels and 56 bridges, with complex terrain along the line), multi-dimensional big data was collected at a frequency of 2-5 minutes per time. The equipment operation data collection focuses on track geometry parameters inside the tunnel, bridge bearing settlement data, data related to the wind resistance stability of the overhead contact system (such as overhead contact system amplitude and conductor height changes), and the working status data of the train braking system on long downhill sections. Environmental perception data includes wind speed (with a focus on monitoring instantaneous strong winds in canyon sections), rainfall, visibility, and temperature data from 15 meteorological monitoring points along the line, monitoring data from 12 geological disaster hazard points (including landslide and debris flow prone areas), smoke and harmful gas concentration data inside the tunnel, and intrusion monitoring data of the protective netting along the line. Personnel behavior data is collected by identifying records of violations by workers through video surveillance inside the tunnel and around the bridges, analyzing passenger density by combining station ticketing data, and obtaining the activity trajectory of patrol personnel in mountainous sections through RFID positioning. Simultaneously, road network topology data (including tunnel length, bridge span, and slope distribution information) and historical data of public safety incidents over the past 12 years (including safety incidents caused by geological disasters unique to mountainous areas and extreme winds) are collected.

[0039] Data preprocessing and fusion: Kalman filtering is used to denoise equipment operation data (focusing on handling abnormal track parameters caused by signal transmission interference in tunnels). A sliding window mean method (with a window size set to 5 acquisition cycles) is used to smooth environmental perception data (mountainous meteorological and geological data fluctuate significantly). An outlier detection algorithm is used to remove invalid information from personnel behavior data. A multi-source data fusion model is constructed using a weighted Bayesian network algorithm to achieve spatiotemporal correlation fusion, for example, correlating and matching instantaneous strong wind data from a mountain canyon section with catenary stability data and train speed data.

[0040] Risk Assessment: Based on the fused dataset, the dynamic risk assessment index system of Example 1 is adopted. The weights of the indicators are determined by improving the analytic hierarchy process (mountain railways have a higher risk of geological disasters, and the weight of the environmental impact indicator is higher than that of conventional railway trunk lines). The weights are dynamically adjusted in conjunction with the real-time data update frequency (e.g., during heavy rain, the update frequency of geological disaster monitoring data is increased to 2 minutes / time, and the corresponding indicator weights are adjusted upward). The comprehensive risk assessment value is calculated using the fuzzy comprehensive evaluation method. For example, during a certain period of heavy rain, the displacement data of a landslide hazard point exceeds the standard. After integrating the track geometry parameters and train traffic data of the area, the comprehensive risk assessment value is calculated to be 0.68, which is judged as high risk.

[0041] Early warning information generation and push: Based on the calibrated risk level threshold, early warning information is generated, specifying the risk level (high risk), the scope of impact (a 3-kilometer section around the landslide hazard point and the corresponding tunnels and bridges), the key risk source (landslide risk caused by rainstorms), and targeted disposal suggestions (closing the line in this section, organizing trains to turn back or detour, and increasing the number of geological monitoring personnel on-site). The information is pushed simultaneously through the railway operation management platform, on-board terminals, the early warning broadcast system along the line, and the passenger service APP, with a push delay controlled within 25 seconds.

[0042] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A railway public safety risk dynamic assessment and early warning method, characterized in that, The method comprises the following steps: S1: Collecting multi-dimensional big data of the whole scene of railway operation, which at least includes equipment operation data, environment perception data, personnel behavior data, road network topology data and public safety event history data; S2: After standardizing and preprocessing the multi-dimensional big data, a multi-source data fusion model is constructed, and the spatio-temporal correlation fusion of different dimensional data is realized through a weighted Bayesian network algorithm; S3: Based on the fused data set, a dynamic risk assessment index system is established, which includes five first-level indexes of equipment health degree, environmental influence degree, personnel safety compliance degree, road network carrying capacity and event correlation risk degree, and at least three secondary quantitative indexes are set under each first-level index; S4: The improved analytic hierarchy process is used to determine the weight of each index, the weight coefficient is dynamically adjusted combined with the real-time data update frequency, and the fuzzy comprehensive evaluation method is used to calculate the comprehensive evaluation value of railway public safety risk; S5: According to the risk comprehensive evaluation value and the preset risk level threshold, a graded early warning information is generated, which includes risk level, influence range, key risk source and targeted disposal suggestion.

2. The railway public safety risk dynamic assessment and early warning method according to claim 1, characterized in that, The equipment operation data includes track geometry parameters, contact net voltage and current data, train braking system state data, signal equipment transmission delay data, and the data acquisition frequency is 1-5 minutes / time.

3. The railway public safety risk dynamic assessment and early warning method according to claim 1, characterized in that, The environment perception data includes along-line meteorological data, geological disaster monitoring data, surrounding obstacle intrusion data and electromagnetic interference data, wherein the meteorological data includes four core parameters of wind speed, rainfall, visibility and temperature.

4. The railway public safety risk dynamic assessment and early warning method according to claim 1, characterized in that, The personnel behavior data includes operation personnel violation operation record, passenger gathering density data and along-line key personnel activity trajectory data, which are obtained through video monitoring identification, RFID positioning and ticket data correlation analysis.

5. The railway public safety risk dynamic assessment and early warning method according to claim 1, characterized in that, In the construction process of the multi-source data fusion model, Kalman filter denoising processing is adopted for the equipment operation data, sliding window mean value method is adopted for the environment perception data, and abnormal value detection algorithm is adopted for the personnel behavior data to eliminate invalid data.

6. The railway public safety risk dynamic assessment and early warning method according to claim 1, characterized in that, The improved analytic hierarchy process corrects the subjective weight by introducing entropy weight coefficient, which is calculated according to the information entropy of each index data, and the information entropy reflects the dispersion degree of index data.

7. The railway public safety risk dynamic assessment and early warning method according to claim 1, characterized in that, In the secondary indexes of the dynamic risk assessment index system, the equipment health degree includes component wear rate, fault occurrence frequency and maintenance cycle compliance rate, and the environmental influence degree includes natural disaster occurrence probability, extreme weather duration and surrounding environment interference intensity.

8. The railway public safety risk dynamic assessment and early warning method according to claim 1, characterized in that, In the fuzzy comprehensive evaluation method, a five-level fuzzy evaluation matrix is established, the elements of the evaluation matrix are determined through expert scoring combined with historical data statistical analysis, and the maximum membership degree principle is adopted to determine the single-index risk level.

9. The railway public safety risk dynamic assessment and early warning method according to claim 1, characterized in that, The preset risk level threshold is divided into four levels, namely low risk, medium risk, high risk and extremely high risk, and each level threshold is calibrated through ROC curve analysis combined with railway safety operation historical data.

10. The railway public safety risk dynamic assessment and early warning method according to claim 1, characterized in that, The graded early warning information is pushed through the railway operation management platform, the vehicle-mounted terminal, the along-line early warning broadcast system and the passenger service APP, and the pushing delay is not more than 30 seconds.