Railway operation safety assessment method and system

Abstract

The invention provides a railway operation safety assessment method and device, and relates to the technical field of data processing, and the method comprises the steps: collecting operation data of a target railway; generating a railway safety knowledge graph based on the operation data; performing feature extraction based on the railway safety knowledge graph to generate a text feature vector; performing risk calculation based on the railway safety knowledge graph and the text feature vector to generate a risk list; the risk list is disassembled, and a corresponding check item list is generated; and based on the inspection item list and a preset inspection standard, performing security assessment to obtain an assessment result. According to the railway operation safety assessment method provided by the invention, credible assessment of railway operation safety is realized, the traceability and the real-time performance of an assessment result are ensured, and the efficiency and the precision of railway operation safety assessment of different regions are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for assessing railway operation safety. Background Technology

[0002] Since railway operations in different regions typically follow different rules or technical specifications, safety assessments of railway operations in different regions also require reference to different assessment standards or historical operational cases.

[0003] Current technologies for safety assessment of railway operations in different regions mainly rely on expert teams to check or sample massive amounts of data item by item. At the same time, the lack of unified assessment standards due to differences in rules in different regions leads to low efficiency and accuracy in railway operation safety assessment. Summary of the Invention

[0004] This invention provides a railway operation safety assessment method and system to solve the technical problems of low efficiency and accuracy in existing railway operation safety assessments for different regions.

[0005] This invention provides a method for assessing railway operation safety, comprising the following steps: Collect operational data for the target railway; Based on the operational data, a railway safety knowledge graph is generated; Based on the railway safety knowledge graph, feature extraction is performed to generate text feature vectors; Based on the railway safety knowledge graph and the text feature vector, risk calculation is performed to generate a risk list; The risk list is broken down to generate a corresponding list of inspection items; Based on the list of inspection items and preset inspection standards, a security assessment is conducted to obtain the assessment results.

[0006] According to a railway operation safety assessment method provided by the present invention, the step of generating a railway safety knowledge graph based on the operation data includes: The operational data is sequentially filtered for stop words and tagged with parts of speech to obtain standard data; Entity recognition is performed on the standard data using a BiLSTM-CRF model to obtain multiple entities; The railway safety knowledge graph is generated based on the attributes of each entity and the relationships between the multiple entities.

[0007] According to the railway operation safety assessment method provided by the present invention, the step of calculating risks and generating a risk list based on the railway safety knowledge graph and the text feature vector includes: Determine the cosine similarity between the text feature vectors; Based on the cosine similarity, the semantic difference between the text feature vectors is determined; Based on the semantic difference degree and the first preset threshold, a first risk category is determined; Based on the numerical parameters and the second preset threshold in the railway safety knowledge graph, a second risk category is determined; Based on the entities in the railway safety knowledge graph and the preset reasoning rules, a third risk category is determined; The risk score is obtained by weighted summation of the first risk category, the second risk category, and the third risk category. Based on the risk score, risk areas in the railway safety knowledge graph are determined; Based on the risk areas, the risk list is generated.

[0008] According to a railway operation safety assessment method provided by the present invention, the step of decomposing the risk list to generate a corresponding inspection item list includes: The risk list is broken down into risk chains using the Bowtie model; each risk chain includes a top event, a threat, and a barrier. Based on the type of the barrier, the risk link is mapped to a preset XML template to generate corresponding inspection item statements; the XML template is generated based on historical railway safety-specific inspection items. Perform syntax validation on the checked statement to generate the checked list.

[0009] According to a railway operation safety assessment method provided by the present invention, the step of conducting a safety assessment based on the inspection item list and preset inspection standards to obtain assessment results includes: Based on the checklist, the status data of the barrier is repeatedly collected within a preset time period; Based on the status data and the level of the threat, the probability of the occurrence of the top event is determined; The evaluation result is generated based on the probability of occurrence.

[0010] According to a railway operation safety assessment method provided by the present invention, after conducting a safety assessment based on the checklist and preset check standards to obtain the assessment result, the method further includes: Based on the evaluation results, a problem description feature vector is generated using the TF-IDF algorithm. The KNN algorithm is used to match multiple historical rectification cases from the historical rectification case library based on the problem description feature vector; the historical rectification cases include problem description information, rectification measures, and rectification effect scores. Calculate the case similarity between each problem description and the problem description feature vector; The rectification effect score and the case similarity are weighted and summed to obtain the recommended score for each historical rectification case. The historical rectification case with the highest recommended score was selected as the target rectification case. Based on the rectification measures described in the target rectification case, safety rectification was carried out.

[0011] This invention also provides a railway operation safety assessment system, comprising the following modules: The data acquisition module collects operational data from the target railway. The graph generation module is used to generate a railway safety knowledge graph based on the operational data. The feature extraction module is used to extract features based on the railway safety knowledge graph and generate text feature vectors; The risk calculation module is used to perform risk calculations and generate a risk list based on the railway safety knowledge graph and the text feature vector. The decomposition module is used to decompose the risk list and generate a corresponding list of inspection items; The evaluation module is used to perform a security evaluation based on the list of inspection items and preset inspection standards, and obtain the evaluation results.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the railway operation safety assessment method as described above.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the railway operation safety assessment method as described above.

[0014] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the railway operation safety assessment method as described above.

[0015] The railway operation safety assessment method and apparatus provided by this invention, by collecting operational data of the target railway, provides a complete and accurate data foundation for subsequent analysis, solving the problem of data isolation; based on the operational data, a railway safety knowledge graph is generated, thereby establishing the relationship between key elements in the operational data using the knowledge graph, realizing the visual identification of hidden risks, and improving the accuracy and efficiency of risk identification; based on the railway safety knowledge graph, feature extraction is performed to generate text feature vectors, realizing the quantitative representation of text semantics, and providing a feature foundation for risk calculation; based on the railway safety knowledge graph and text feature vectors, risk calculation is performed to generate a risk list, realizing the automated quantitative classification of risks, reducing errors caused by human subjectivity, and improving the accuracy of risk calculation; the risk list is decomposed to generate a corresponding list of inspection items, improving the cross-project reuse rate and reliability of inspection items; based on the list of inspection items and preset inspection standards, a safety assessment is performed to obtain the assessment results, realizing a credible assessment of railway operation safety, ensuring the traceability and real-time nature of the assessment results, and improving the efficiency and accuracy of railway operation safety assessment in different regions. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the railway operation safety assessment method provided by the present invention.

[0018] Figure 2 This is a schematic diagram of the railway operation safety assessment device provided by the present invention.

[0019] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0021] Existing technologies for conducting safety assessments of railway operations in different regions have the following main shortcomings: (1) Lack of technical means to identify differentiated risks. Existing solutions are designed based on a "single stable environment" and lack multimodal data fusion or differential quantification algorithms, making it impossible to identify cross-regional specific risks. For example, manual comparison of standards in region A with the actual situation in region B lacks semantic parsing of the legal texts in region A and fails to use knowledge graphs to establish a "laws-standards-risk" association, making it easy to miss implicit risks such as "legal conflicts" and "cultural implementation deviations". The lack of algorithms such as feature vector similarity calculation and rule reasoning makes it impossible to quantify risks such as "the degree of conflict between standards in different regions" and "the probability of insufficient local rescue capacity", resulting in large deviations in the judgment of the same risk in different assessments.

[0022] (2) The assessment content is out of touch with the local context and lacks the ability to generate technically feasible inspection items. Existing solutions cannot use technical means to transform abstract rules into actionable inspection items. For example, the inspection items in region A depend on the regulations of region A (such as "whether the contact network voltage meets the 27.5kV standard"). If the project in region B uses 15kV voltage, the items will directly become invalid, and there is no template mapping technology to adapt to different regions. The "linkage mechanism-specific barrier-inspection item" is not broken down using Bowtie model digital technology, making it difficult for operators to understand and resulting in insufficient implementation rate of inspection items.

[0023] (3) The static assessment model is mismatched with dynamic risks and lacks a technical update mechanism. For example, the checklist method in each region is only manually assessed once before the project is launched. There is no timed task scheduling, and the risk list cannot be updated in real time according to factors such as "personnel evacuation progress" and "extreme weather", which is lagging. There is no data-driven risk level update logic, and the review list needs to be manually recompiled, which cannot cope with the high-frequency changing risk environment in different regions.

[0024] (4) Insufficient verification of the effectiveness of control measures and lack of technical verification methods. Existing plans only examine "whether measures exist" without verifying "whether the measures are effective". For example, the checklist method used in various regions to check "whether an emergency plan has been formulated" only looks at paper documents and lacks technologies such as mandatory data collection via mobile APP and recording of data from actual drills, making it impossible to verify "whether the plan is adapted to the local language" and "whether the response time in coordination with the local fire department meets the standards". In the existing plans, the review of "personnel training" only involves spot-checking training records and lacks technical verification such as simulator testing and operational data monitoring, resulting in insufficient probability of the actual effectiveness of the rectification measures.

[0025] To this end, the present invention provides a railway operation safety assessment method and device, which combines mature technologies such as multimodal data processing, rule reasoning, and blockchain, and realizes automated and reliable assessment of railway operation safety in different regions based on conventional computer hardware and software.

[0026] The following is combined with Figures 1 to 3 The present invention describes a railway operation safety assessment method and apparatus.

[0027] Figure 1 This is a flowchart illustrating the railway operation safety assessment method provided by the present invention, as shown below. Figure 1 As shown, the method includes the following steps: Step 101: Collect operational data of the target railway; Specifically, a combination of hardware and software methods is used to collect railway operation data from different regions. This operation data mainly includes railway laws and regulations, technical standards, line design standards, operation management systems, and maintenance data. In some embodiments, the data collection process specifically includes: (1) Hardware requirements (standard configuration, no special equipment required): Server: CPU ≥ 4 cores (such as Intel Xeon E3-1230 v5), memory ≥ 16GB (DDR4 2666MHz), hard drive ≥ 500GB (SSD); Industrial gateway: Supports RS485 / Ethernet interfaces for sensor data acquisition; Mobile devices: Android 10.0+ / iOS 14.0+ smartphones (camera ≥ 13 megapixels, GPS positioning supported).

[0028] (2) Data Acquisition Channel: Application Programming Interface (API) for data acquisition: Based on the RESTful protocol, it connects to the regulatory database of Region A and the railway sensor platform (such as track vibration, catenary current, etc.), sets a 30-second response timeout threshold and a 3-times automatic retry mechanism to ensure stable data acquisition; it supports switching data sources from different regions (such as the Ministry of Transport API of Region B and the Railway Bureau API of Region C) through parameter configuration. SFTP encrypted transmission: It uses the AES-256 encryption algorithm to receive technical standard documents (PDF / Word format) and historical case reports, and supports breakpoint resume (based on file size comparison) to avoid data loss due to network interruption; Web data entry system: Based on the Django framework, it implements a three-level permission system of "administrator-evaluator-auditor" (such as a role-based access control (RBAC) model). The administrator is responsible for permission configuration, the evaluator enters unstructured cases (such as unpublished minor fault records in region A), and the auditor verifies the authenticity of the data to ensure the accuracy of manually supplemented data.

[0029] In the embodiments of the present invention, a large amount of data on railway operations in different regions is collected through a variety of software and hardware combination methods, providing a complete and accurate data basis for subsequent analysis and solving the problem of data isolation.

[0030] Step 102: Generate a railway safety knowledge graph based on the operation data; Further, the generating a railway safety knowledge graph based on the operation data includes: Filter out stop words and perform词性标注 on the operation data in sequence to obtain standard data; Perform entity recognition on the standard data through a BiLSTM-CRF model to obtain multiple entities; Generate the railway safety knowledge graph based on the attributes of each entity and the relationships between the multiple entities.

[0031] Specifically, the text processing of the operation data includes: Stop word filtering: Use the Jieba tool (used to expand railway domain-specific vocabulary, such as "catenary anti-salt fog treatment", "signal system SIL4 level", etc.) to remove meaningless stop words (such as "of", "in", etc.); 词性标注: Perform词性标注 on the text through the spaCy tool (including the Chinese model zh_core_web_md, with a词性标注 accuracy rate ≥ 92%) to obtain standard data, providing a basis for subsequent entity recognition.

[0032] After obtaining the standard data, to obtain multiple entities, a Bidirectional Long Short-Term Memory-Conditional Random Field (BiLSTM-CRF) model is used for entity recognition. Multiple entities such as "speed limit value", "safety level (SIL4)", "standard number (TB / T 3553)" are extracted from the standard data and stored in JSON-LD format (for facilitating subsequent knowledge graph construction). Among them, the BiLSTM-CRF model is trained based on 500,000 pieces of railway safety domain corpus, and the entity recognition F1 value on the test set ≥ 0.91, which can accurately perform entity recognition.

[0033] It should be noted that the term "词性标注" in the original text seems to be a Chinese term that needs to be accurately translated according to the specific meaning in the context. Here, a placeholder is used for now. If there is a more accurate English term, it can be replaced accordingly.In some embodiments, for sensor data in the standard data (e.g., track vibration, catenary current, etc. with a sampling frequency of 1Hz), outliers can be removed using the 3σ principle (removing data that exceeds the mean ± 3 times the standard deviation), and noise can be smoothed using a 5s moving average window to ensure data stability. For standard parameters in the standard data, the parameter units are standardized (e.g., the unit for slope is "‰", and the unit for voltage is "kV"). For example, "0.025 (decimal slope)" is converted to "25‰" to avoid evaluation errors caused by unit confusion.

[0034] In this embodiment of the invention, based on the attributes of each entity and the relationships between multiple entities, a three-layer structure of "entity-relationship-attribute" is constructed using Neo4j 4.4.0 graph database (open source community version, single node supports 1 billion-level relationship storage), specifically including: Entity layer: Includes multiple entity types such as "Law", "Standard", "Equipment", "Risk", and "Case". Each entity type contains multiple attributes, such as "Law" which includes the IDs "country", "issueOrg", "effectiveDate", and "clauseCount". Relationship layer: Defines various relationships such as "conflict with (CONFLICTS_WITH)", "cause (CAUSES)", and "requirements (REQUIRES)". Relationship attributes include "association strength" (0-1, calculated by text similarity) and "confidence" (0-1, trained based on expert-annotated samples).

[0035] For example, the correlation strength between the "requirements" of the Railway Safety Management Regulations (Law) of Region A and the Signaling System Standard (Standard, SIL4) of Region A is 0.95, and the confidence level is 0.98; the correlation strength between the Signaling System Standard of Region A and the Signaling System Standard (SIL3) of Region C is 0.82, and the confidence level is 0.9.

[0036] Relationship strength calculation: Text vectors are extracted using the spaCy pre-trained model, and the cosine similarity formula is used to calculate the relationship strength between entities (range 0-1). For example, if the similarity between the standard text of the signal from region A and region C is 0.18, then the relationship strength of "conflict with" is 0.82 (1-0.18).

[0037] Graph integrity verification: The integrity of the association between "risk" entities and "case" entities is verified by using the depth-first search (DFS) algorithm to ensure that each "risk" entity is associated with at least one "case" entity, avoid risk entries without empirical support, and improve the credibility of risk identification.

[0038] This invention automatically and accurately extracts key information from massive amounts of unstructured text to construct a structured railway safety knowledge graph, avoiding the problems of data isolation and omission of hidden risks. It provides a high-quality, structured data foundation for subsequent risk quantification calculations, while also improving the processing efficiency of unstructured data.

[0039] Step 103: Based on the railway safety knowledge graph, perform feature extraction to generate text feature vectors; Specifically, the text content such as "standard clauses" and "regulatory clauses" in the railway safety knowledge graph is converted into 300-dimensional numerical vectors using a word vector (Word to Vector, Word2Vec) model (based on Skip-gram architecture) to generate text feature vectors, thereby realizing the quantitative representation of the text.

[0040] For example, the word vector model parameters are set as follows: word vector dimension 300, context window size 5, minimum word frequency 5, and number of iterations 10, to ensure coverage of core vocabulary in railway operations such as "signaling system", "SIL level", and "gradient limit". The long text processing method is as follows: for texts with more than 512 words (such as complete legal chapters), the vector dimension is unified by "segmentation (500 words / segment, 10 overlapping words) + mean pooling" to avoid loss of information in long texts.

[0041] Based on the above embodiments, the standard text of the signal system in region A ("must meet SIL4 level") can be converted into a 300-dimensional vector [0.123, -0.045, 0.217, ...], and the standard text of the signal system in region C ("must meet SIL3 level") can be converted into a vector [0.089, 0.156, -0.092, ...], thereby providing a quantitative basis for subsequent difference calculations.

[0042] Step 104: Based on the railway safety knowledge graph and the text feature vector, perform risk calculation and generate a risk list; Furthermore, the step of calculating risks and generating a risk list based on the railway safety knowledge graph and the text feature vector includes: Determine the cosine similarity between the text feature vectors; Based on the cosine similarity, the semantic difference between the text feature vectors is determined; Based on the semantic difference degree and the first preset threshold, a first risk category is determined; Based on the numerical parameters and the second preset threshold in the railway safety knowledge graph, a second risk category is determined; Based on the entities in the railway safety knowledge graph and the preset reasoning rules, a third risk category is determined; The risk score is obtained by weighted summation of the first risk category, the second risk category, and the third risk category. Based on the risk score, risk areas in the railway safety knowledge graph are determined; Based on the risk areas, the risk list is generated.

[0043] Specifically, the cosine similarity formula is used to calculate the cosine similarity between text feature vectors.

[0044] For example, the cosine similarity formula is used to calculate the cosine similarity between the baseline vector (A) of region A and the vector (B) of region B. The formula is: similarity=(A·B) / (||A||×||B||) In the formula, similarity represents cosine similarity, A represents the baseline vector of region A, B represents the vector of region B, ||A|| represents the magnitude of vector A, and ||B|| represents the magnitude of vector B.

[0045] Based on the calculated cosine similarity, the semantic difference between text feature vectors is defined as: semantic_diff=1-similarity Set the similarity threshold (i.e. the first preset threshold) α=0.7 (which can be adjusted within the range of 0.5-0.9 according to the actual situation). Based on the semantic difference and the first preset threshold, determine the first risk category (including potential risks and foreseeable risks).

[0046] For example, when similarity is less than α (i.e., semantic_diff is less than 0.3), it is marked as "potential risk"; when similarity is greater than or equal to α (i.e., semantic_diff is greater than or equal to 0.3), it is marked as "foreseeable risk".

[0047] For example, the cosine similarity between the standard text vectors of signals from region A and region C is 0.65 (<0.7), and the semantic difference is 0.35, which is marked as a potential risk.

[0048] For numerical parameters such as "maximum gradient", "wind speed limit", and "contact network voltage" in the railway safety knowledge graph, first define the benchmark interval (i.e., the second preset threshold) [X_min, X_max] for region A (e.g., gradient [-30‰, 30‰]), and set the tolerance coefficient Δ=5% (special parameters such as wind speed limit are set to 10%). If the target value Y for region C is not within the range of [X_min×(1-Δ), X_max×(1+Δ)], calculate the numerical difference degree diff_degree using the following formula to determine the second risk category (including risky and risk-free).

[0049] diff_degree=|Y-X_avg| / X_avg In the formula, X_avg is the average value of the baseline interval.

[0050] For example, if the maximum slope baseline range for region A is [-30‰, 30‰], and the target value for region C is 25‰ (within the tolerance range of [-28.5‰, 31.5‰]), and the numerical difference is 0, then the second risk category is no risk. If the target value for region C exceeds the tolerance range, then the second risk category is risky.

[0051] In this embodiment of the invention, based on the Drools 8.33.0 rule engine (which uses the rete algorithm to improve rule matching efficiency), inference rules are preset to build an inference rule library for the unique risks of railway operation in different regions.

[0052] For example, the "Signal System Safety Level Risk Rule" states that if the signal standard in region A is SIL4 and that in region B it is not SIL4, and the difference in the effective date of the standards is ≤5 years, then it is judged as medium risk (SIL3) or high risk (SIL2 and below); the "Preferred Standards of the Last 5 Years Rule" only compares standards that have been effective in the last 5 years to avoid misjudgments caused by outdated standards. Execution logic: Input the "standard" entity in the railway safety knowledge graph into the rule engine, automatically match the rules and output the rule risk level quantification value (i.e. the third risk category) (e.g., 1=low risk, 2=medium risk, 3=high risk), thereby realizing the automatic standard identification of risk category without human intervention.

[0053] Since the first risk category depends on semantic difference, the second risk category depends on numerical difference, and the third risk category depends on the quantified value of the rule risk level, a weighted scoring method is used to calculate the risk score (risk_score) based on the first, second, and third risk categories. The calculation formula can be defined as follows: risk_score = 0.4 × semantic difference + 0.3 × numerical difference + 0.3 × rule risk level quantification value In the formula, the weight of semantic difference is 0.4, the weight of numerical difference is 0.3, and the weight of the rule risk level quantification value is 0.3. It should be noted that the above weights can be adjusted according to the actual situation, and this invention does not limit them.

[0054] Based on the above embodiments, a risk heatmap (distinguished by red, yellow, and green according to risk level) is generated using ECharts to visually display the key areas of risk distribution in the railway safety knowledge graph, thereby identifying risk regions. A risk list corresponding to each risk region (including risk ID, type, semantic difference, numerical difference, score, level, and railway safety knowledge graph evidence links, etc.) is generated and stored using MySQL 8.0 (InnoDB engine, supporting transactions), achieving visualized storage.

[0055] Step 105: Decompose the risk list to generate a corresponding list of inspection items; Furthermore, the step of breaking down the risk list to generate a corresponding list of inspection items includes: The risk list is broken down into risk chains using the Bowtie model; each risk chain includes a top event, a threat, and a barrier. Based on the type of the barrier, the risk link is mapped to a preset XML template to generate corresponding inspection item statements; the XML template is generated based on historical railway safety-specific inspection items. Perform syntax validation on the checked statement to generate the checked list.

[0056] Specifically, using the Bowtie model, a drag-and-drop modeling interface was developed on the web based on the HTML5 Canvas + mxGraph open-source library. Based on the risk list, a risk chain was drawn from "top event (such as 'high-speed rail operation safety accident') - threat (such as 'signal system failure', 'personnel operation error') - barrier (such as 'regular inspection of signal system', 'personnel training')". The interface supports editing node attributes (such as setting the barrier effectiveness to 0.85) and adjusting relationship connections, and can be operated without professional drawing skills.

[0057] The model data is stored using a PostgreSQL 14 database, with three core tables designed as follows: The `event` table (top event) contains fields such as `event_id`, `event_name`, and `prob_init` (initial probability). The threat table contains fields such as threat_id, threat_name, threat_level (level 1-5), and event_id (related top event). The barrier table contains fields such as barrier_id, barrier_name, barrier_type (preventive / mitigation), effect (effectiveness 0-1), and threat_id (associated threat).

[0058] By using the Breadth-First Search (BFS) algorithm to verify the topological integrity of the model, we can ensure that each "threat" is associated with at least three "barriers" to avoid "risk getting out of control due to the failure of a single barrier" and improve the reliability of the model.

[0059] Pre-set railway safety-specific inspection item templates are stored in XML format, resulting in a pre-set XML template. Barriers are categorized according to their type (e.g., "Management Barriers" correspond to the "Personnel Training Inspection" template, "Technical Barriers" correspond to the "Equipment Inspection" template). Each template contains variable placeholders (e.g., {Position} {Cycle} {Quantity}). Risk links are then mapped to the pre-set XML template, specifically including: Template matching: The current barrier type is matched with templates in the template library using an O(n) complexity algorithm (n is the number of templates). The templates are sorted in descending order of barrier effectiveness and the top 3 best templates are selected. Variable population: Query contextual data from the railway safety knowledge graph (e.g., query {position} as "driver" through "project position configuration", query {cycle} as "half a year" through "training plan"), and replace variable placeholders in the template with regular expressions; if there is no corresponding data in the knowledge graph, obtain it through API connection to the project management system (timeout can be set to 15 seconds), thereby generating check item statements.

[0060] The Language Technology Platform (LTP) (with a fluency verification accuracy of ≥92%) is invoked to verify the fluency of the generated check item statements. For example, a pass rate of ≥95% is set to output the check item list.

[0061] For example, if the inspection item statement is "Randomly check the six-month training records of 5 drivers to verify whether they have completed the signal system operation training in Area A and passed the assessment (passing standard: score ≥ 80 points)", the LTP verification statement is fluent and can be output to the inspection item list for direct use in on-site assessment.

[0062] This invention addresses the problems of disconnect between inspection items and regional scenarios, as well as low compilation efficiency. Through Bowtie model-driven and XML template mapping, it can quickly and in batches generate targeted, uniformly formatted, and directly applicable inspection item lists, thereby improving the efficiency and reusability of cross-regional project evaluation.

[0063] Step 106: Based on the list of inspection items and the preset inspection standards, conduct a security assessment and obtain the assessment results.

[0064] Furthermore, the security assessment based on the checklist and preset check standards, and the resulting assessment, includes: Based on the checklist, the status data of the barrier is repeatedly collected within a preset time period; Based on the status data and the level of the threat, the probability of the occurrence of the top event is determined; The evaluation result is generated based on the probability of occurrence.

[0065] Specifically, this embodiment of the invention develops a mobile APP based on the Flutter 3.0 cross-platform framework (supporting Android 10.0+ / iOS 14.0+) to achieve mandatory evidence collection at the assessment site.

[0066] Collection constraints: The inspection items in the inspection item list are bound to the collection requirements. For example, "fire extinguisher inspection" requires taking 3 photos of the fire extinguisher from different angles (front, pressure gauge, and expiration date), and "signal system test" requires recording a video of the actual operation of the signal system (duration ≥ 30s). The evaluation results cannot be submitted if the collection is not completed, so as to avoid missing evidence. Evidence watermark: Photos / videos are automatically given an unalterable watermark, which includes the device ID (such as the Android ID for Android devices, the identifier for vendors for iOS devices), a second-level timestamp (such as "2024-05-20 14:30:05.123"), and GPS coordinates (error ≤ 10m). The watermark is fixed in the lower right corner to prevent later tampering. Data transmission: Evidence data is uploaded using the HTTPS protocol (TLS 1.3 encryption), with a 3-retry mechanism (5-second interval) to avoid data loss due to network fluctuations. The evidence files are Base64 encoded before uploading to ensure the integrity of data transmission.

[0067] In some embodiments, a three-tiered storage architecture of "structured data database storage + unstructured evidence object storage + hash value blockchain notarization" can be adopted to ensure the immutability and traceability of evidence. Specifically, this includes: Data storage: Structured data (inspection item ID, assessment result, risk level) is stored in MySQL 8.0; Unstructured evidence (photos, videos) is stored in MinIO 2023.11.0 (an open-source distributed object storage system that supports 3-replica redundancy to ensure no data loss). Hash calculation and on-chain evidence storage: Calculate the SHA-256 hash value of the evidence file (a unique identifier; the hash value will change if the file content is tampered with). The Hyperledger Fabric 2.5 consortium blockchain smart contract is invoked to write the hash value into the block and generate a unique evidence chain ID (i.e., a blockchain transaction ID). Multi-party query: Supports regulatory agencies, supervisors, and operating units in various regions to query hash values ​​through the evidence chain ID, compare them with the hash values ​​of local evidence files, verify the authenticity of evidence, and meet the audit needs of multiple parties.

[0068] Based on the above embodiments, in order to further improve the real-time performance of security assessments, a scheduled task can be configured based on APScheduler (Python) to repeatedly collect barrier status data within a preset time period. For example, a risk reassessment is triggered at 2:00 AM every day by default (the execution time can be adjusted through the configuration file). The data collection task is triggered by the server hardware clock to ensure accurate execution. After the task starts, the latest data is automatically queried from the MySQL assessment result table to update the status and failure probability of the "barrier" in the Bowtie model.

[0069] The qualitative "compliant / non-compliant" assessment conclusions in traditional evaluations are transformed into quantitative failure probability values ​​P according to a pre-defined mapping rule. Then, Fault Tree Analysis (FTA) is used for precise calculation to obtain the probability of the top event occurring. The mapping rule is jointly formulated based on industry standards, historical statistical data, and expert experience, as detailed below: Base Mapping: When the evaluation conclusion is "compliant," the barrier is considered to be in an "effective" state. Its failure probability P_failure is the product of the industry-recognized base failure rate λ_base and the inspection period T. For example, based on the daily re-evaluation cycle (T=1 day) and the requirements for high-risk safety systems in standards such as IEC 61508, the following settings can be made: P_failure = λ_base × T ≈ 0.001 In the formula, 0.001 indicates that even if the barrier is currently effective, it still has an extremely low inherent failure rate in the next evaluation period.

[0070] When the assessment conclusion is "non-compliant," the barrier is considered to be in a "failed" state. Its failure probability, P_failure, is directly set to 0.999. This value indicates that the barrier has been confirmed to be failed, almost inevitably leading to the loss of control over its associated threats.

[0071] Dynamic Adjustment Factor (K): To more accurately reflect dynamic changes in risk, this embodiment of the invention introduces an adjustment factor K (range 0.5-2.0) to fine-tune the base probability value. The K value is calculated from the following dimensions: Historical compliance record (weight 0.5): The compliance rate of this barrier in the most recent 10 assessments is calculated. For every 10% decrease in the compliance rate, the K value increases by 0.2. Rectification verification status (weight 0.3): If the rectification measures associated with this "non-compliance" conclusion have exceeded the deadline and have not been completed, the K value is set to the maximum value of 2.0; if the rectification has been completed and verified, the K value is reset to 1.0; Strength of evidence (weight 0.2): The type of evidence that generates this assessment conclusion affects the credibility. For example, automatic sensor data (K=0.8) > evidence collected by manual APP (K=1.0) > conclusions entered purely manually (K=1.2).

[0072] Finally, the formula for calculating the failure probability description P_failure_adj can be expressed as: P_failure_adj=min(0.999, P_failure×K) This rule ensures that the failure probability description P_failure_adj can dynamically fluctuate with the actual performance of the barrier, making the risk assessment more realistic.

[0073] FTA Calculation: Construct a fault tree with "railway operation safety accident" as the top event, and use the uplink minimum cut set algorithm to calculate the probability P(T) of the top event. The formula is as follows: P(T)=1-∏(1-P(Ci)) In the formula, Ci is the minimal cut set, and P(Ci) is the probability of the cut set Ci occurring.

[0074] The probability of the top event is calculated by using the updated P_failure_adj as the input for the failure probability of each barrier.

[0075] Based on the probability of occurrence, an assessment result is generated, including whether an alert is needed or not. For example, if a risk threshold of 0.001 is set, and the probability of the top event exceeds the threshold, the assessment result indicates that an alert is needed. In this case, an alert will be automatically triggered via SMS (e.g., by calling the Alibaba Cloud SMS API), APP push (JPush), or email (SMTP protocol) to notify relevant personnel to take timely action.

[0076] For example, suppose the barrier "periodic signal system detection" is rated "compliant" in this assessment, with a base P_failure of 0.001. However, its recent compliance rate is 80% (a decrease of 20%, K_history = 1 + 0.2 × 2 = 1.4), and this conclusion is based on manual input (K_evidence = 1.2). The overall K value is (1.4 × 0.5 + 1.0 × 0.3 + 1.2 × 0.2) / (0.5 + 0.3 + 0.2) = 1.24. Therefore, its adjusted failure probability P_failure_adj = 0.001 × 1.24 = 0.00124.

[0077] The other barrier, "personnel training," was assessed as "non-compliant," with a P_failure value of 0.999.

[0078] If the top event is caused only by the minimal cut set formed by these two barriers, then the probability of the top event is 1-(1-0.00124)×(1-0.999)≈1-(0.99876×0.001)≈0.99999, which far exceeds the risk threshold. The assessment result is that an early warning is required, and the corresponding early warning operation should be triggered immediately.

[0079] This invention, by repeatedly collecting barrier status data within a preset time period, predicts the probability of top events, transforming railway operation safety assessment from static and lagging to dynamic and real-time. This effectively addresses the high-frequency changes in the risk environment during the initial stage of railway project operation, and improves the effectiveness and accuracy of assessing dynamic risks in railway operation.

[0080] The railway operation safety assessment method provided by this invention provides a complete and accurate data foundation for subsequent analysis by collecting operational data of the target railway, thus solving the problem of data isolation. Based on the operational data, a railway safety knowledge graph is generated, thereby establishing the relationship between key elements in the operational data and realizing the visual identification of hidden risks, improving the accuracy and efficiency of risk identification. Based on the railway safety knowledge graph, feature extraction is performed to generate text feature vectors, realizing the quantitative representation of text semantics and providing a feature basis for risk calculation. Based on the railway safety knowledge graph and text feature vectors, risk calculation is performed to generate a risk list, realizing the automated quantitative classification of risks, reducing errors caused by human subjectivity, and improving the accuracy of risk calculation. The risk list is decomposed to generate a corresponding list of inspection items, improving the cross-project reuse rate and reliability of inspection items. Based on the list of inspection items and preset inspection standards, a safety assessment is performed to obtain the assessment results, realizing a credible assessment of railway operation safety, ensuring the traceability and real-time nature of the assessment results, and improving the efficiency and accuracy of railway operation safety assessment in different regions.

[0081] Furthermore, after conducting a security assessment based on the checklist and preset check standards to obtain the assessment results, the method further includes: Based on the evaluation results, a problem description feature vector is generated using the TF-IDF algorithm. The KNN algorithm is used to match multiple historical rectification cases from the historical rectification case library based on the problem description feature vector; the historical rectification cases include problem description information, rectification measures, and rectification effect scores. Calculate the case similarity between each problem description and the problem description feature vector; The rectification effect score and the case similarity are weighted and summed to obtain the recommended score for each historical rectification case. The historical rectification case with the highest recommended score was selected as the target rectification case. Based on the rectification measures described in the target rectification case, safety rectification was carried out.

[0082] Specifically, based on the assessment results, the descriptive text of the inspection items that were assessed as "non-compliant" (such as "the driver did not know how to operate the Chinese signal system") is traced back to the barrier assessment. The term frequency-inverse document frequency (TF-IDF) algorithm is used to transform it into a 200-dimensional feature vector, generating a problem description feature vector, thereby quantifying the semantics of the text. The K-Nearest Neighbors (KNN) algorithm (K=5, Euclidean distance is used to calculate similarity) is used to match the 5 most similar historical rectification cases in the historical rectification case library based on the problem description feature vector (the number of cases can be set according to actual needs, and the embodiment of this invention uses 5 cases as an example for illustration). Each historical rectification case in the historical rectification case library contains fields such as "problem description information", "rectification measures" and "rectification effect score" (e.g., 1-5 points). Calculate the case similarity (e.g., cosine similarity) between each problem description and its feature vector. Then, perform a weighted sum of the rectification effect score (e.g., with a weight of 0.6) and the case similarity (e.g., with a weight of 0.4) to calculate the recommended score for each historical rectification case. Based on the recommended scores, sort the historical rectification cases and output the historical rectification case with the highest recommended score as the target rectification case (e.g., "Case ID: CASE2023001, Case Similarity: 0.92, Rectification Effect Score: 4.8 / 5.0"). Based on the rectification measures in the target rectification case, help the operating unit quickly formulate a precise rectification plan.

[0083] This invention uses the KNN algorithm to recommend corrective measures, thereby achieving dynamic risk management and improving the response speed and effectiveness of corrective measures. Furthermore, intelligent recommendations avoid repeating past mistakes and blind rectification, forming a complete intelligent management loop of "problem discovery - intelligent recommendation - measure implementation," thus improving practicality and reliability.

[0084] The railway operation safety assessment device provided by the present invention is described below. The railway operation safety assessment device described below can be referred to in correspondence with the railway operation safety assessment method described above.

[0085] Figure 2 This is a schematic diagram of the railway operation safety assessment device provided by the present invention, as shown below. Figure 2 As shown. An embodiment of the present invention provides a railway operation safety assessment device, comprising a data acquisition module 201, a map generation module 202, a feature extraction module 203, a risk calculation module 204, a disassembly module 205, and an assessment module 206, wherein: The data acquisition module 201 acquires operational data of the target railway; the graph generation module 202 generates a railway safety knowledge graph based on the operational data; the feature extraction module 203 extracts features from the railway safety knowledge graph to generate text feature vectors; the risk calculation module 204 calculates risks based on the railway safety knowledge graph and the text feature vectors to generate a risk list; the decomposition module 205 decomposes the risk list to generate a corresponding list of inspection items; and the evaluation module 206 performs a safety evaluation based on the list of inspection items and preset inspection standards to obtain the evaluation results.

[0086] The railway operation safety assessment device provided by this invention collects operational data of the target railway, providing a complete and accurate data foundation for subsequent analysis and solving the problem of data isolation. Based on the operational data, it generates a railway safety knowledge graph, thereby establishing connections between key elements in the operational data and realizing the visual identification of hidden risks, improving the accuracy and efficiency of risk identification. Based on the railway safety knowledge graph, it extracts features and generates text feature vectors, realizing the quantitative representation of text semantics and providing a feature basis for risk calculation. Based on the railway safety knowledge graph and text feature vectors, it performs risk calculation and generates a risk list, realizing automated quantitative classification of risks, reducing errors caused by human subjectivity, and improving the accuracy of risk calculation. It decomposes the risk list to generate a corresponding list of inspection items, improving the cross-project reuse rate and reliability of inspection items. Based on the list of inspection items and preset inspection standards, it performs safety assessments and obtains assessment results, realizing a credible assessment of railway operation safety, ensuring the traceability and real-time nature of the assessment results, and improving the efficiency and accuracy of railway operation safety assessments in different regions.

[0087] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, the communications interface 320, and the memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute a railway operation safety assessment method, which includes: Collect operational data for the target railway; Based on the operational data, a railway safety knowledge graph is generated; Based on the railway safety knowledge graph, feature extraction is performed to generate text feature vectors; Based on the railway safety knowledge graph and the text feature vector, risk calculation is performed to generate a risk list; The risk list is broken down to generate a corresponding list of inspection items; Based on the list of inspection items and preset inspection standards, a security assessment is conducted to obtain the assessment results.

[0088] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0089] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the railway operation safety assessment method provided by the above methods, the method comprising: Collect operational data for the target railway; Based on the operational data, a railway safety knowledge graph is generated; Based on the railway safety knowledge graph, feature extraction is performed to generate text feature vectors; Based on the railway safety knowledge graph and the text feature vector, risk calculation is performed to generate a risk list; The risk list is broken down to generate a corresponding list of inspection items; Based on the list of inspection items and preset inspection standards, a security assessment is conducted to obtain the assessment results.

[0090] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the railway operation safety assessment method provided by the methods described above, the method comprising: Collect operational data for the target railway; Based on the operational data, a railway safety knowledge graph is generated; Based on the railway safety knowledge graph, feature extraction is performed to generate text feature vectors; Based on the railway safety knowledge graph and the text feature vector, risk calculation is performed to generate a risk list; The risk list is broken down to generate a corresponding list of inspection items; Based on the list of inspection items and preset inspection standards, a security assessment is conducted to obtain the assessment results.

[0091] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0092] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0093] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0094] In this application's embodiments, "determine B based on A" means that factor A must be considered when determining B. It is not limited to "B can be determined based solely on A," but should also include: "determine B based on A and C," "determine B based on A, C, and E," "determine C based on A, and further determine B based on C," etc. Additionally, it can include using A as a condition for determining B, for example, "when A meets the first condition, determine B using the first method"; another example, "when A meets the second condition, determine B," etc.; another example, "when A meets the third condition, determine B based on the first parameter," etc. Of course, it can also be a condition where A is a factor in determining B, for example, "when A meets the first condition, determine C using the first method, and further determine B based on C," etc.

[0095] It should also be noted that the terms "target," "first," and "second" in this invention are used to distinguish similar objects, not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, without limiting the number of objects; for example, the first object can be one or more.

[0096] In this invention, the term "multiple" refers to two or more kinds, and other quantifiers are similar.

[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing railway operation safety, characterized in that, include: Collect operational data for the target railway; Based on the operational data, a railway safety knowledge graph is generated; Based on the railway safety knowledge graph, feature extraction is performed to generate text feature vectors; Based on the railway safety knowledge graph and the text feature vector, risk calculation is performed to generate a risk list; The risk list is broken down to generate a corresponding list of inspection items; Based on the list of inspection items and preset inspection standards, a security assessment is conducted to obtain the assessment results.

2. The railway operation safety assessment method according to claim 1, characterized in that, The generation of a railway safety knowledge graph based on the operational data includes: The operational data is sequentially filtered for stop words and tagged with parts of speech to obtain standard data; Entity recognition is performed on the standard data using a BiLSTM-CRF model to obtain multiple entities; The railway safety knowledge graph is generated based on the attributes of each entity and the relationships between the multiple entities.

3. The railway operation safety assessment method according to claim 2, characterized in that, The process of calculating risks and generating a risk list based on the railway safety knowledge graph and the text feature vector includes: Determine the cosine similarity between the text feature vectors; Based on the cosine similarity, the semantic difference between the text feature vectors is determined; Based on the semantic difference degree and the first preset threshold, a first risk category is determined; Based on the numerical parameters and the second preset threshold in the railway safety knowledge graph, a second risk category is determined; Based on the entities in the railway safety knowledge graph and the preset reasoning rules, a third risk category is determined; The risk score is obtained by weighted summation of the first risk category, the second risk category, and the third risk category. Based on the risk score, risk areas in the railway safety knowledge graph are determined; Based on the risk areas, the risk list is generated.

4. The railway operation safety assessment method according to claim 1, characterized in that, The process of breaking down the risk list to generate a corresponding list of inspection items includes: The risk list is broken down into risk chains using the Bowtie model; each risk chain includes a top event, a threat, and a barrier. Based on the type of the barrier, the risk link is mapped to a preset XML template to generate corresponding inspection item statements; the XML template is generated based on historical railway safety-specific inspection items. Perform syntax validation on the checked statement to generate the checked list.

5. The railway operation safety assessment method according to claim 4, characterized in that, The process of conducting a security assessment based on the checklist and preset check standards to obtain assessment results includes: Based on the checklist, the status data of the barrier is repeatedly collected within a preset time period; Based on the status data and the level of the threat, the probability of the occurrence of the top event is determined; The evaluation result is generated based on the probability of occurrence.

6. The railway operation safety assessment method according to claim 1, characterized in that, After conducting a security assessment based on the checklist and preset check standards, and obtaining the assessment results, the method further includes: Based on the evaluation results, a problem description feature vector is generated using the TF-IDF algorithm. The KNN algorithm is used to match multiple historical rectification cases from the historical rectification case library based on the problem description feature vector; the historical rectification cases include problem description information, rectification measures, and rectification effect scores. Calculate the case similarity between each problem description and the problem description feature vector; The rectification effect score and the case similarity are weighted and summed to obtain the recommended score for each historical rectification case. The historical rectification case with the highest recommended score was selected as the target rectification case. Based on the rectification measures described in the target rectification case, safety rectification was carried out.

7. A railway operation safety assessment system, characterized in that, include: The data acquisition module collects operational data from the target railway. The graph generation module is used to generate a railway safety knowledge graph based on the operational data. The feature extraction module is used to extract features based on the railway safety knowledge graph and generate text feature vectors; The risk calculation module is used to perform risk calculations and generate a risk list based on the railway safety knowledge graph and the text feature vector. The decomposition module is used to decompose the risk list and generate a corresponding list of inspection items; The evaluation module is used to perform a security evaluation based on the list of inspection items and preset inspection standards, and obtain the evaluation results.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the railway operation safety assessment method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the railway operation safety assessment method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the railway operation safety assessment method as described in any one of claims 1 to 6.

Images