Railway potential safety hazard analysis method and system under cooperation of main control and branch control

Abstract

The invention discloses a railway potential safety hazard analysis method and system under main and branch control cooperation, and relates to the technical field of data processing. The method comprises the following steps: configuring a cooperative control linkage response gradient based on a node association relationship between main control and sub-control, and constructing a cooperative sensing network; the sensing devices connected with the sub-control nodes are used for extracting sensing monitoring features, projecting the sensing monitoring features to the collaborative sensing network, identifying risk events and positioning a collaborative control linkage response link; performing event factor analysis positioning of each coordinated control linkage response node, and generating prevention and control early warning information; and feedback response information is generated according to the prevention and control early warning information, prevention and control link reconstruction is carried out in the collaborative awareness network, a risk joint control trend is determined, and a potential safety hazard broadcast is sent to carry out main and branch control node early warning reminding. The technical problems that in the prior art, railway potential safety hazard recognition efficiency is low, and early warning is not timely are solved, and the technical effects that a cooperative sensing network is constructed through main control and branch control cooperation to achieve rapid potential safety hazard recognition, and therefore early warning response timeliness is improved are achieved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for analyzing railway safety hazards under master-slave control collaboration. Background Technology

[0002] With the continuous expansion of the railway transportation network, a multi-level collaborative operation system has been formed among the train control center, section control stations, and field equipment, leading to a continuously increasing demand for real-time monitoring of railway operation status. Current railway safety hazard analysis largely relies on data from single monitoring nodes or local sensors, making it difficult to correlate and analyze monitoring information scattered across different control nodes, resulting in a lag in hazard identification. Furthermore, risk events exhibit chain-like propagation characteristics within the railway network. When an anomaly occurs at a single node, there is a lack of timely assessment of its impact range, propagation path, and the linkage relationships between related nodes, preventing hierarchical and coordinated safety warnings. In addition, existing technologies generally lack mechanisms for unified modeling and fusion reasoning of multi-node monitoring data, hindering the formation of a holistic risk situation awareness capability. This results in low efficiency in railway safety hazard analysis and untimely warnings, limiting further improvements in railway operational safety. Summary of the Invention

[0003] This application provides a method and system for analyzing railway safety hazards under master-slave control collaboration, which solves the technical problems of low efficiency in identifying railway safety hazards and untimely early warning in the prior art.

[0004] The first aspect of this application provides a method for analyzing railway safety hazards under a master-slave control collaborative system, the method comprising:

[0005] Based on the node association relationship between the main control and sub-control, a collaborative control and linkage response gradient is configured to construct a collaborative sensing network. Sensing devices connected to the sub-control nodes extract sensing and monitoring features and project them onto the collaborative sensing network to identify risk events and locate collaborative control and linkage response links. Based on these links, event factors of each collaborative control and linkage response node are analyzed and located to generate prevention and control early warning information, which is used for locating prevention and control factors and decomposing prevention and control instructions at the collaborative control and linkage response nodes. Feedback response information is generated based on the early warning information, and the prevention and control link is reconstructed in the collaborative sensing network to determine the risk control trend and send a safety hazard broadcast to provide early warning reminders to the main and sub-control nodes.

[0006] A second aspect of this application provides a railway safety hazard analysis system under master-slave control collaboration, the system comprising:

[0007] The system comprises the following modules: a perception network construction module, a link localization module, and a warning module. The perception network construction module configures collaborative control response gradients based on the node associations between the main control and sub-control nodes, and constructs a collaborative perception network. The link localization module connects the sensing devices of the sub-control nodes, extracts sensing and monitoring features, projects them onto the collaborative perception network, identifies risk events, and locates the collaborative control response links. The analysis module analyzes and locates event factors of each collaborative control response node based on the collaborative control response links, generates prevention and control early warning information, and uses this information to locate prevention and control factors and decompose prevention and control instructions for the collaborative control response nodes. The warning module generates feedback response information based on the prevention and control early warning information, reconstructs the prevention and control links in the collaborative perception network, determines the risk control trend, and sends a safety hazard broadcast to provide early warning reminders to the main and sub-control nodes.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] First, based on the node association relationship between the main control and sub-control nodes, a collaborative control and linkage response gradient is configured to construct a collaborative sensing network. Next, sensing devices connected to the sub-control nodes extract sensor monitoring features and project them onto the collaborative sensing network to identify risk events and locate the collaborative control and linkage response links. Then, based on the collaborative control and linkage response links, event factor analysis and location are performed on each collaborative control and linkage response node to generate prevention and control early warning information, which is used for locating prevention and control factors and decomposing prevention and control instructions at the collaborative control and linkage response nodes. Finally, feedback response information is generated based on the prevention and control early warning information, and the prevention and control link is reconstructed in the collaborative sensing network to determine the risk control trend and send a safety hazard broadcast to provide early warning reminders to the main and sub-control nodes. This solves the technical problems of low efficiency in railway safety hazard identification and untimely early warning in existing technologies, achieving the technical effect of rapid identification of safety hazards through the collaborative sensing network constructed by the main and sub-control nodes, thereby improving the timeliness of early warning response. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A schematic diagram of the railway safety hazard analysis method under master-slave control collaboration provided in the embodiments of this application;

[0012] Figure 2 A schematic diagram of the railway safety hazard analysis system under master-slave control collaboration provided in this application embodiment.

[0013] Figure labeling: Perception network construction module 11, link localization module 12, parsing module 13, early warning module 14. Detailed Implementation

[0014] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0015] Example 1, as Figure 1 As shown, this application provides a method for analyzing railway safety hazards under a master-slave control collaborative system, wherein the method includes:

[0016] Based on the node association relationship between the master controller and the sub-controllers, a collaborative control linkage response gradient is configured to construct a collaborative sensing network.

[0017] Furthermore, based on the node association relationship between the master controller and the sub-controllers, a collaborative control and linkage response gradient is configured to construct a collaborative sensing network, including:

[0018] Using the master control node and sub-control nodes as centers, the physical location, path function association, and data flow dependency analysis of each node are performed to determine the association relationships in each dimension. Based on the association relationships in each dimension, multi-gradient partitioning is performed to determine the gradient region distribution of each center. Using the master control node and sub-control nodes as nodes, connection relationship edges between nodes are established according to the association relationships in each dimension, and the association gradient circles of each node are labeled according to the gradient region distribution of each center to construct the collaborative perception network.

[0019] First, using the main control node and each sub-control node as the association center, a multi-dimensional association analysis is conducted on all business nodes within the railway line area. By acquiring the physical location relationship of each node, the path function distribution in the line topology, and the data flow interaction dependency between nodes, association characteristics such as geographical proximity, functional coupling, and data dependency strength are calculated to form an association relationship description in each dimension. Subsequently, based on the above association relationships, nodes are divided into multiple gradients. Nodes that are physically close to the main control node or sub-control node, have a close business function association, or a high data flow dependency are classified into high gradient regions. Nodes with moderate correlation are classified into medium gradient regions, and peripheral nodes with weak correlation are classified into low gradient regions, thus forming a gradient region distribution centered on the main control and sub-control nodes. After completing the gradient region division, with the master control node and sub-control nodes as the network core, connection edges are established between nodes based on the aforementioned multi-dimensional correlation to construct a network topology that reflects the sensing, control, and data exchange capabilities between nodes. At the same time, each node is labeled with an associated gradient circle according to the gradient region to clarify the importance level and interaction strategy priority of each node in the collaborative control and linkage response. Finally, a collaborative sensing network structure is formed that can be used for subsequent sensor feature projection, risk event identification, and joint control link simulation.

[0020] Furthermore, the collaborative sensing network is constructed by labeling the associated gradient circles of each node according to the gradient region distribution of each center, including:

[0021] Establish a mapping relationship between the associated gradient and the data transmission frequency, which includes at least high frequency, medium frequency, and low frequency, with the transmission frequency corresponding to the associated gradient; based on the mapping relationship, set the transmission rules for each node and its corresponding associated gradient circles, and construct a transmission channel, wherein a high-frequency data channel is established between nodes within the same associated gradient circle, and a low-frequency heartbeat interaction is maintained between different associated gradient circles; establish the data flow rules for the master control node, and construct the data sensing transmission channel between the master control node and each sub-control node according to the data flow rules.

[0022] After labeling network nodes with associated gradient circles according to the gradient region distribution of each center, a communicable structure for the collaborative sensing network is constructed by establishing a mapping relationship between the associated gradient and the data transmission frequency. High gradient regions correspond to high-frequency data transmission, medium gradient regions to medium-frequency data transmission, and low gradient regions to low-frequency data transmission. Based on this mapping relationship, corresponding data interaction rules are configured for nodes within different gradient circles: high-frequency or medium-frequency data channels are established between nodes within the same associated gradient circle to transmit real-time data such as operating status, sensing features, or risk events; while nodes across gradient circles do not engage in continuous business data interaction, but only maintain low-frequency heartbeat interaction to reduce communication redundancy and ensure network stability. Heartbeat interaction refers to the periodic sending of heartbeat data frames containing node identifier, current operating status, communication status, and timestamps to its associated nodes within a preset heartbeat period, indicating that the node is in a normal online state and maintaining connection validity; when no heartbeat data frames are received from a certain node within several consecutive heartbeat periods, the node is determined to have entered a communication abnormal state, triggering the corresponding link status marking or abnormal reporting process. Based on the aforementioned heartbeat interaction mechanism, only necessary online status confirmation and link liveness detection are maintained between different associated gradient circles to avoid network load caused by frequent data transmission, while ensuring that the master control node and sub-control nodes can perceive the connection status of cross-gradient nodes in real time. On this basis, data flow rules for the master control node are constructed. According to the gradient circle where each sub-control node is located and its associated characteristics, corresponding data aggregation frequencies and transmission scheduling strategies are allocated to the master control node. Data perception transmission channels are established between the master control node and each sub-control node according to the aforementioned data flow rules, enabling the master control node to collect the perception data uploaded by each sub-control node in a hierarchical, frequency-based, and priority-based manner, thereby completing the construction of the collaborative perception network.

[0023] Furthermore, establishing data flow rules for the master node includes:

[0024] Set the interaction frequency baseline for the main and sub-control nodes; configure the frequency adjustment amplitude of each node according to the interaction association characteristics of each sub-control node; adjust the interaction frequency baseline based on the frequency adjustment amplitude of each node to determine the self-interaction frequency of each sub-control node; configure the data flow rules of the main control node according to the self-interaction frequency of each sub-control node.

[0025] Specifically, firstly, based on the overall communication load capacity of the railway line, the network structure scale, and the business collaboration requirements of the main control node and its sub-control nodes, a baseline for the interaction frequency between the main control node and each sub-control node is established as a default data exchange frequency reference for each node. Then, combining previously acquired interaction correlation characteristics of the sub-control nodes, including indicators such as the node's functional criticality in the network, data dependency strength, and risk propagation sensitivity, corresponding frequency adjustment amplitudes are configured for each sub-control node, allowing different nodes to obtain different interaction frequency adjustment capabilities based on their business importance. On this basis, the frequency adjustment amplitudes are used to correct the established interaction frequency baseline, calculating the self-interaction frequency of each sub-control node, giving higher data interaction frequencies to sub-control nodes with higher importance or stronger correlation. Finally, based on the self-interaction frequencies of each sub-control node, the data flow rules for the main control node are configured, including the main control node's data retrieval cycle for different nodes, data priority sorting strategies, and the rhythm of sensing data integration, thereby forming a data flow system that supports multi-gradient, multi-priority collaborative sensing.

[0026] Furthermore, based on the interaction and correlation characteristics of each sub-control node, the frequency adjustment amplitude of each node is configured, including:

[0027] A four-dimensional weighted evaluation model is constructed, including traffic flow dimension, connection density dimension, functional criticality dimension, and isolation risk dimension. The evaluation results of each dimension are carried out through the four-dimensional weighted evaluation model, and the comprehensive weight of each node is calculated by weighted geometric average. Based on the difference between the comprehensive weight and the benchmark evaluation quantity, the control node level is configured, and the frequency adjustment amplitude of each node is configured according to the node level.

[0028] When configuring frequency adjustment amplitudes based on the interactive characteristics of each sub-control node, this embodiment constructs a four-dimensional weighted evaluation model consisting of traffic flow, connection density, functional criticality, and isolation risk dimensions. This model comprehensively evaluates the operational importance of sub-control nodes in the collaborative sensing network and their impact on the overall security situation. Specifically, the traffic flow dimension reflects the train operation density and traffic load level of the node's section; the connection density dimension measures the degree of topological coupling between the node and other sub-control nodes; the functional criticality dimension characterizes the strength of the node's role in signal control, dispatch command transmission, or critical business links; and the isolation risk dimension reflects the independent risk diffusion effect that the node may cause to the surrounding area under abnormal conditions.

[0029] A four-dimensional weighted evaluation model is used to calculate the evaluation value of each dimension for each node, and a weighted geometric average is used to fuse the four-dimensional evaluation results to obtain a comprehensive weight that can comprehensively represent the importance of the node's business. Subsequently, the difference between the comprehensive weight and the preset benchmark evaluation value is used as the basis for node level classification, dividing the sub-control nodes into different control levels, and configuring a corresponding frequency adjustment amplitude for each sub-control node according to the adjustment range corresponding to the control level. This allows nodes with higher comprehensive weights to obtain a greater increase in interaction frequency, thereby maintaining the ability to prioritize the monitoring of high-value nodes and high-frequency data interaction in the collaborative sensing network.

[0030] The weighted geometric mean method is used to calculate the overall node weight. The calculation formula is as follows:

[0031] Where T is the standardized value of the traffic flow dimension, C is the value of the connectivity density dimension, F is the score of the functional criticality dimension, and R is the isolation risk coefficient, with weighting coefficients of 0.3 for traffic flow, 0.25 for connectivity density, 0.25 for functional criticality, and 0.2 for isolation risk. A power of 1 / 4 is used to normalize the four-dimensional weighted geometric mean, ensuring comparability and scale consistency of the comprehensive weight. Based on the comprehensive weight W, the sub-control nodes are divided into four levels: critical nodes, important nodes, ordinary nodes, and edge nodes, for subsequent configuration of corresponding frequency adjustment amplitudes and interaction priorities.

[0032] The sensing devices connected to the sub-control nodes extract sensing and monitoring features and project them onto the collaborative sensing network to identify risk events and locate the collaborative control and linkage response links.

[0033] Furthermore, the sensing devices connected to the sub-control nodes extract sensing and monitoring features and project them onto the collaborative sensing network to identify risk events and locate the collaborative control and response links, including:

[0034] Multimodal sensing devices connected to sub-control nodes collect multi-dimensional monitoring data on environment, structure, and operating status. Sensing features, including time-domain features, frequency-domain features, statistical features, and anomaly features, are extracted from the multi-dimensional monitoring data. These extracted features are projected into the collaborative sensing network. Risk events are identified based on the sensing features of each sub-control node, determining the identified risk events and their probabilities. A graph neural network is used to perform feature propagation and influence range calculation based on the node relationships within the collaborative sensing network. The collaborative control response link is located based on feature propagation and influence range, determining the influence propagation path, key response nodes, and associated gradient circles, thus constructing the response chain topology.

[0035] In the process of connecting the sensing devices at the sub-control nodes and projecting the sensing monitoring features onto the collaborative sensing network, multi-modal sensing devices are first deployed at each sub-control node to synchronously collect multi-dimensional monitoring data such as environmental parameters, structural status, and operational status. This includes real-time data of types such as temperature and humidity, wind speed, vibration response, track deformation, electrical signals, power load, and signal equipment status. After preprocessing the above multi-dimensional monitoring data, monitoring features reflecting changes in node status are extracted through techniques such as time-domain analysis, frequency-domain transformation, statistical modeling, and anomaly detection. This forms a feature vector set composed of time-domain features, frequency-domain features, statistical features, and anomaly features. Specifically, trend changes, fluctuation amplitudes, abrupt change points, and delay features are extracted in the time domain; energy spectrum distribution, harmonic components, and periodicity features are extracted in the frequency domain; statistical indicators such as mean, standard deviation, and coefficient of variation are extracted in the statistical dimension; and outliers, anomaly distribution patterns, and multi-dimensional anomaly combinations are extracted in the anomaly dimension. Subsequently, the extracted feature vectors are projected onto the corresponding sub-control nodes according to the topology of the collaborative sensing network. That is, based on the connection edges and gradient circle annotations between the aforementioned nodes, each type of sensing and monitoring feature is mapped to its corresponding network node, allowing the node's state in the network to be dynamically described by its real-time sensing features. Based on the mapped node feature vectors, a pre-defined risk event identification model is invoked to jointly analyze the operational status of each node. The risk event identification model comprehensively utilizes information such as temporal abrupt changes, frequency domain energy anomalies, statistical offsets, and abnormal feature distributions to identify potential risk event categories, including equipment failure risks, environmental impact risks, and signal anomaly risks. Simultaneously, the probability of risk occurrence is calculated based on the degree of feature deviation and the persistence of anomalies, and the specific node location where the risk event occurs is identified. On this basis, a graph neural network is used as the propagation mechanism to perform cross-node propagation and influence diffusion calculations of risk features using the node relationships already constructed in the collaborative sensing network, thereby assessing the propagation trend and impact range of risk events in the network. Based on the propagation results and the distribution of influence, we further locate the collaborative control and response links, clarify the possible propagation paths of risks, key response nodes and their associated gradient circles, and then construct a response chain topology that represents the scope of linkage and response priority.

[0036] The probability of a risk event is calculated based on the deviation of node feature vectors, the persistence of anomalies, and the coupling degree of multidimensional features. Specifically, firstly, a difference analysis is performed between the current features of a node and its historical steady-state features to obtain a feature deviation index reflecting the magnitude of the deviation; secondly, the frequency of occurrence and duration of anomalous features are statistically analyzed within a preset time sliding window to obtain an anomaly persistence index; simultaneously, the feature coupling anomaly degree is calculated based on the correlation between time-domain, frequency-domain, and statistical features. These three indices are then normalized and weighted according to preset weights to obtain the risk probability value of the node at the current time, which characterizes the likelihood of a risk event occurring at that node.

[0037] Based on the node connection relationships of the collaborative sensing network, feature vectors are input into a graph neural network model. Features are propagated across nodes through graph convolution and feature aggregation mechanisms. The influence diffusion range of the features in the network is calculated to reflect the potential paths and influence range of an abnormal state that may propagate to neighboring nodes. Specifically, the connection relationships of each control node in the collaborative sensing network are first represented as an adjacency matrix. The time-domain, frequency-domain, statistical, and abnormal features extracted from each node are combined into a node feature matrix, and both are input into the graph neural network model. The graph neural network performs neighborhood aggregation and weighted propagation of node features by executing multi-layer graph convolution operations, so that the state representation of each node not only includes its own features but also incorporates the feature information of its neighboring nodes. During propagation, the graph convolutional layers perform weighted summation or feature aggregation of neighboring features according to the weights of the connecting edges, thereby forming a high-order feature representation that reflects the coupling relationship between nodes. As the graph convolutional layers iterate layer by layer, the abnormal features gradually spread along different paths in the network. The model calculates the response sensitivity of each node to the abnormal source based on the magnitude of feature changes after propagation, the gradient response strength, and the degree of influence on downstream nodes, thus obtaining the influence diffusion range of the features in the network. Specifically, after the graph neural network completes multi-layer feature propagation, the updated feature vectors of each node are compared dimension by dimension. By calculating the magnitude change of the feature difference before and after propagation, the direct response magnitude of the node to the abnormal disturbance is obtained. Simultaneously, based on the cumulative gradient value of the propagation weights during graph convolution, the gradient response intensity of the node in the propagation chain is extracted to characterize the amplification or attenuation trend of the anomaly's impact on the node. Furthermore, combined with the node's position in the topology, the feature changes of its downstream nodes are weighted statistically to assess the potential impact of the node's anomaly propagating to more distant levels. The above three types of indicators are standardized, and the node's response sensitivity score to the anomaly source is calculated using weighted summation or weighted geometric mean. The higher the response sensitivity, the more critical the node is in the anomaly propagation chain, and the greater the likelihood of being affected. Threshold screening, sensitivity hierarchical clustering, or exponential decay analysis are performed on the response sensitivity of all nodes to delineate the diffusion boundary of the abnormal state in the collaborative sensing network, including: a set of highly sensitive nodes (the main affected area), a set of moderately sensitive nodes (the secondary diffusion area), and a set of low-sensitive nodes (the marginal affected area). The resulting set of highly sensitive, moderately sensitive, and lowly sensitive nodes constitutes the range of influence diffusion of the anomaly in the network.

[0038] Based on the aforementioned collaborative control and response link, the event factors of each collaborative control and response node are analyzed and located to generate prevention and control early warning information, which is used to locate the prevention and control factors of the collaborative control and response node and decompose the prevention and control instructions.

[0039] Furthermore, based on the aforementioned collaborative control and response link, event factor analysis and location are performed on each collaborative control and response node to generate prevention and control early warning information, including:

[0040] Based on the collaborative control and response link, event factors are analyzed and located at each response node to identify direct causal factors, propagation path factors, and node vulnerability factors. Node self-check instructions are generated based on the event factor analysis results to guide nodes in checking for similar problems or potential risks. The cascading effects of related risk events are monitored, including signal transmission interruptions, power supply impacts, and data transmission anomalies. Based on the cascading effects and node self-check instructions, prevention and control early warning information is generated, prevention and control factors are located and broken down into specific prevention and control instructions, which are fed back to the corresponding sub-control nodes and the main control node. Dynamic monitoring and blocking strategies are implemented to logically block and physically isolate key propagation paths.

[0041] Based on the collaborative control and response link, a comprehensive analysis is performed on the operational data, characteristic propagation information, and historical status of all response nodes in the link. Event factor analysis is then performed on each node to pinpoint its location. Event factor analysis includes identifying direct causal factors (such as equipment component wear and tear, sudden environmental changes, and abnormal signal fluctuations that directly trigger risks), propagation path factors (such as the transmission effect caused by anomalies in adjacent nodes and the amplification effect caused by link coupling), and node vulnerability factors (such as structural weaknesses that may escalate risks, such as insufficient node redundancy, a single dependency path, and excessive load on critical resources). Through the quantification and positioning of these factors, the functional role of each response node in the risk propagation chain and its potential risk sources can be clearly defined. Subsequently, based on the event factor analysis results, a corresponding node self-check instruction is generated for each response node. This self-check instruction guides the node to proactively check the status of its internal components, the health of its data channels, the working status of its sensors, and whether there are similar hidden dangers or potential risks in its redundant resources. This achieves bottom-up risk confirmation and status verification within the link.

[0042] Based on this, the cascading effects related to risk events are monitored. By analyzing key status indicators such as whether signal transmission is interrupted or unstable, whether power supply fluctuates or experiences momentary drops, and whether data transmission is abnormal, it is determined whether the risk has developed from a single node to a multi-node chain effect. If cascading signs are found, the execution results and abnormal characteristics of node self-inspection commands are further combined to construct the link diffusion situation. Based on the cascading effect and the execution feedback of node self-inspection commands, prevention and control early warning information for regional joint prevention and control is generated. The prevention and control early warning information not only locates the prevention and control factors that may cause risks, but also breaks down the warning content into specific executable prevention and control commands, including equipment self-inspection commands, link logical blocking commands, node physical isolation commands, upstream status verification commands, and downstream risk warning commands. The prevention and control commands will be synchronously fed back to the corresponding sub-control nodes and master control nodes to implement dynamic monitoring and blocking strategies, logically blocking and physically isolating key propagation paths in the response link, thereby effectively blocking the spread trend of risks.

[0043] Based on the prevention and control early warning information, feedback response information is generated, the prevention and control link is reconstructed in the collaborative sensing network, the risk joint control trend is determined, and a security hazard broadcast is sent to provide early warning reminders to the main and sub-control nodes.

[0044] After generating prevention and control early warning information, the master control node comprehensively summarizes the operational status of nodes in the link based on the self-inspection status of each response node, cascade effect monitoring data, and the execution results of various prevention and control instructions. This results in feedback response information that includes node risk level, link blockage status, state stability, and remaining risk factors. This feedback response information, serving as reverse verification data for the coordinated control and response link, will be synchronously input into the collaborative sensing network to update the state vectors of nodes and the weight attributes of connecting edges in the network.

[0045] Based on feedback response information, the collaborative sensing network performs prevention and control link reconstruction processing. By reassessing the risk propagation path, node importance, and current robustness of the response link, it unfreezes frozen edges, reduces the weight of high-risk edges or disconnects them, and enhances redundant paths around important nodes, thereby generating a dynamically adjusted prevention and control link topology. The reconstructed prevention and control link reflects the latest risk propagation structure, providing an accurate structural basis for subsequent risk prediction.

[0046] After link reconstruction, based on the updated link topology and node status information, the potential diffusion trend of risk in different gradient circles and regions is calculated using a feature propagation model or diffusion trend prediction algorithm. This determines risk control trend indicators, including the scope of risk impact, the speed of risk outward diffusion, the probability of secondary nodes being affected, and potential multi-linked risk paths. The risk control trend is used to guide the master node in deciding whether to expand the scope of risk control or upgrade the emergency response level.

[0047] Finally, based on the risk control trends and the reconstruction results of the prevention and control links, a safety hazard broadcast is generated and sent to the main control node and all relevant sub-control nodes, including the risk source node, secondary affected nodes, and peripheral nodes that may be affected by the joint control. The safety hazard broadcast is used to trigger a unified early warning within the system, enabling each node to promptly implement localized monitoring enhancement, resource reservation, emergency response, and risk avoidance measures according to its own gradient region, thereby achieving global collaborative early warning and handling of railway safety hazards.

[0048] Furthermore, based on the aforementioned prevention and control early warning information, feedback response information is generated, and the prevention and control link is reconstructed in the collaborative sensing network to determine the risk joint control trend, including:

[0049] Based on the node self-inspection instructions and cascading effects in the prevention and control early warning information, the system integrates the node self-inspection execution status, cascading effect monitoring data, safety risk event evolution status, and monitoring and blocking strategy effectiveness evaluation data to generate the feedback response information. This feedback response information is then sent to the master control node and the associated gradient circles corresponding to the sub-control nodes. Real-time risk identification and assessment are performed based on the feedback response information. When the risk level exceeds a preset threshold, a dynamic expansion mechanism for the risk joint control range is activated, adjusting the boundaries and node connection relationships of the collaborative sensing network. Based on the adjusted risk joint control range, the prevention and control link topology is reconstructed, optimizing the collaborative relationships and resource allocation strategies between nodes. Based on the reconstructed prevention and control link, risk feature propagation model simulation and diffusion trend prediction analysis are performed to determine risk joint control trend indicators.

[0050] First, according to the node self-check instructions and cascade effect monitoring results in the prevention and control early warning information, integrate and analyze the self-check execution status of each response node in the link, cascade effect monitoring data, evolution status of security risk events, and effectiveness evaluation data of the previously implemented monitoring isolation strategies, to form structured feedback response information describing node health, link stability, and risk propagation blocking effect. The feedback response information may include content such as node self-check pass / fail status, cascade event triggering situation, risk event development stage, isolation strategy success rate, and remaining risk intensity. Subsequently, send the feedback response information to the master node and all slave nodes in its corresponding associated gradient circle, so that nodes in different gradient circles can timely receive the latest link status according to their positions. After receiving the feedback response information, the master node and slave nodes respectively perform real-time risk identification and risk level assessment based on the network policy model. When the risk level of a certain node or a certain link reaches or exceeds the preset threshold, the system automatically activates the dynamic expansion mechanism of the risk joint control scope, and adjusts the structure of the collaborative perception network by expanding the boundary of the collaborative perception network, increasing the weight of relevant nodes, increasing the number of node connections, or updating the data transmission weight value, etc., to include potential affected areas in the joint control scope.

[0051] After completing the adjustment of the joint control scope, reconstruct the prevention and control link topology according to the updated risk joint control area structure, reconfigure the node connection method, connection direction, connection weight, and node association degree in the link, optimize the collaborative response relationship between each node, and adjust the allocation strategy of monitoring resources and computing resources, so that the new prevention and control link can more effectively reflect the current risk propagation path structure and support rapid linkage response. Finally, based on the reconstructed prevention and control link, call the risk feature propagation model to simulate the risk diffusion trend in the link, by simulating the conduction, amplification, or attenuation behavior of risk features between different nodes, and combining with the diffusion equation or graph convolution prediction model for trend prediction, to obtain the joint control trend indicators of the risk in the spatial and temporal domains in the future, including risk diffusion range, diffusion speed, probability of secondary nodes being affected, potential cascade paths, and high-risk node aggregation areas, etc., for guiding subsequent early warning broadcast and emergency control decisions.

[0052] Furthermore, it also includes:

[0053] The system monitors the heartbeat signals of each sub-control node. When a preset number of heartbeat signals are lost consecutively, the corresponding sub-control node is determined to have entered a communication interruption state. A final data transmission analysis mechanism is activated to extract the last transmitted data of the interrupted node within a preset time window before the communication interruption from the collaborative sensing network. The last transmitted data undergoes integrity verification, abnormal pattern identification, and situation reconstruction analysis. Based on the analysis results, the cause of the interruption, the risk evolution path, and the impact range on the associated gradient circle are inferred. A communication interruption emergency response plan is generated to guide adjacent nodes in adjusting their monitoring strategies and resource allocation. Furthermore, after activating the final data transmission analysis mechanism, the system also includes: establishing an interruption recovery monitoring mechanism to quickly rebuild node connections and data synchronization after communication is restored.

[0054] This application embodiment also includes a real-time monitoring and emergency handling mechanism for the communication status of sub-control nodes. Specifically, a heartbeat signal mechanism is configured for each sub-control node in the collaborative sensing network. Each sub-control node periodically sends heartbeat signals to the master control node and its adjacent nodes within its associated gradient circle according to a preset heartbeat cycle. The heartbeat signal is a lightweight status message periodically sent by the sub-control node, which includes at least node identification information, a timestamp, and a communication liveness flag to characterize whether the communication connection between the sub-control node and the collaborative sensing network is valid. The master control node performs time continuity and integrity checks on the received heartbeat signals and statistically analyzes the heartbeat reception status of each sub-control node based on a sliding time window. When a preset number of heartbeat signals are lost consecutively within the sliding time window, or when a heartbeat signal is received but its timestamp offset exceeds an allowable threshold, the corresponding sub-control node is determined to have entered a communication interruption state. After determining that a sub-control node has entered a communication interruption state, the system immediately starts a last data transmission analysis mechanism to extract the last transmitted data of the sub-control node within the preset time window before the communication interruption from the data buffer of the collaborative sensing network. For the extracted data, the system first performs data integrity verification to identify potential problems such as missing data, abnormal jumps, and time sequence disorders. Then, it uses an anomaly pattern recognition model to analyze feature deviations in the data and conducts situational reconstruction analysis based on the node's topological location and event link structure. Based on the integrity verification, anomaly pattern, and situational reconstruction results, it infers that communication interruptions may be caused by equipment failure, link anomalies, power supply fluctuations, or upstream risk diffusion. Simultaneously, it predicts the possible evolution path and impact range of risks within the associated gradient circle and generates a communication interruption emergency response plan to guide adjacent nodes in adjusting monitoring strategies, increasing data sampling frequency, or reallocating resources to ensure that local network areas still have basic monitoring capabilities during node interruptions. Building on this, the system further establishes an interruption recovery monitoring mechanism. Specifically, when it detects that a sub-control node has retransmitted a valid heartbeat signal and continuously meets a preset number of stable signals, the interruption recovery monitoring mechanism is triggered, automatically reconstructing the node connection relationship and executing the data synchronization process. This allows the node to be reintegrated into the normal monitoring and control system of the collaborative sensing network, thereby ensuring that the entire collaborative sensing network maintains stable operation and effective risk perception capabilities in the face of sudden communication anomalies.

[0055] In summary, the embodiments of this application have at least the following technical effects:

[0056] First, based on the node association relationship between the main control and sub-control nodes, a collaborative control and linkage response gradient is configured to construct a collaborative sensing network. Next, sensing devices connected to the sub-control nodes extract sensor monitoring features and project them onto the collaborative sensing network to identify risk events and locate the collaborative control and linkage response links. Then, based on the collaborative control and linkage response links, event factor analysis and location are performed on each collaborative control and linkage response node to generate prevention and control early warning information, which is used for locating prevention and control factors and decomposing prevention and control instructions at the collaborative control and linkage response nodes. Finally, feedback response information is generated based on the prevention and control early warning information, and the prevention and control link is reconstructed in the collaborative sensing network to determine the risk control trend and send a safety hazard broadcast to provide early warning reminders to the main and sub-control nodes. This solves the technical problems of low efficiency in railway safety hazard identification and untimely early warning in existing technologies, achieving the technical effect of rapid identification of safety hazards through the collaborative sensing network constructed by the main and sub-control nodes, thereby improving the timeliness of early warning response.

[0057] Example 2, based on the same inventive concept as the railway safety hazard analysis method under master-slave control coordination in the aforementioned examples, such as... Figure 2 As shown, this application provides a railway safety hazard analysis system under master-slave control collaboration, wherein the system includes:

[0058] The sensing network construction module 11 configures the collaborative control and linkage response gradient based on the node association relationship between the main control and sub-control, and constructs a collaborative sensing network; the link positioning module 12 connects the sensing devices of the sub-control nodes, extracts the sensing monitoring features and projects them onto the collaborative sensing network, identifies risk events and locates the collaborative control and linkage response links; the parsing module 13 performs event factor parsing and positioning of each collaborative control and linkage response node based on the collaborative control and linkage response links, generates prevention and control early warning information, and uses it to locate the prevention and control factors of the collaborative control and linkage response nodes and decompose prevention and control instructions; the early warning module 14 generates feedback response information based on the prevention and control early warning information, reconstructs the prevention and control links in the collaborative sensing network, determines the risk joint control trend, and sends a safety hazard broadcast to provide early warning reminders to the main and sub-control nodes.

[0059] Furthermore, the perception network construction module 11 is used to perform the following methods:

[0060] Using the master control node and sub-control nodes as centers, the physical location, path function association, and data flow dependency analysis of each node are performed to determine the association relationships in each dimension. Based on the association relationships in each dimension, multi-gradient partitioning is performed to determine the gradient region distribution of each center. Using the master control node and sub-control nodes as nodes, connection relationship edges between nodes are established according to the association relationships in each dimension, and the association gradient circles of each node are labeled according to the gradient region distribution of each center to construct the collaborative perception network.

[0061] Furthermore, the perception network construction module 11 is used to perform the following methods:

[0062] Establish a mapping relationship between the associated gradient and the data transmission frequency, which includes at least high frequency, medium frequency, and low frequency, with the transmission frequency corresponding to the associated gradient; based on the mapping relationship, set the transmission rules for each node and its corresponding associated gradient circles, and construct a transmission channel, wherein a high-frequency data channel is established between nodes within the same associated gradient circle, and a low-frequency heartbeat interaction is maintained between different associated gradient circles; establish the data flow rules for the master control node, and construct the data sensing transmission channel between the master control node and each sub-control node according to the data flow rules.

[0063] Furthermore, the perception network construction module 11 is used to perform the following methods:

[0064] Set the interaction frequency baseline for the main and sub-control nodes; configure the frequency adjustment amplitude of each node according to the interaction association characteristics of each sub-control node; adjust the interaction frequency baseline based on the frequency adjustment amplitude of each node to determine the self-interaction frequency of each sub-control node; configure the data flow rules of the main control node according to the self-interaction frequency of each sub-control node.

[0065] Furthermore, the perception network construction module 11 is used to perform the following methods:

[0066] A four-dimensional weighted evaluation model is constructed, including traffic flow dimension, connection density dimension, functional criticality dimension, and isolation risk dimension. The evaluation results of each dimension are carried out through the four-dimensional weighted evaluation model, and the comprehensive weight of each node is calculated by weighted geometric average. Based on the difference between the comprehensive weight and the benchmark evaluation quantity, the control node level is configured, and the frequency adjustment amplitude of each node is configured according to the node level.

[0067] Furthermore, the link location module 12 is used to perform the following method:

[0068] Multimodal sensing devices connected to sub-control nodes collect multi-dimensional monitoring data on environment, structure, and operating status. Sensing features, including time-domain features, frequency-domain features, statistical features, and anomaly features, are extracted from the multi-dimensional monitoring data. These extracted features are projected into the collaborative sensing network. Risk events are identified based on the sensing features of each sub-control node, determining the identified risk events and their probabilities. A graph neural network is used to perform feature propagation and influence range calculation based on the node relationships within the collaborative sensing network. The collaborative control response link is located based on feature propagation and influence range, determining the influence propagation path, key response nodes, and associated gradient circles, thus constructing the response chain topology.

[0069] Furthermore, the parsing module 13 is used to perform the following methods:

[0070] Based on the collaborative control and response link, event factors are analyzed and located at each response node to identify direct causal factors, propagation path factors, and node vulnerability factors. Node self-check instructions are generated based on the event factor analysis results to guide nodes in checking for similar problems or potential risks. The cascading effects of related risk events are monitored, including signal transmission interruptions, power supply impacts, and data transmission anomalies. Based on the cascading effects and node self-check instructions, prevention and control early warning information is generated, prevention and control factors are located and broken down into specific prevention and control instructions, which are fed back to the corresponding sub-control nodes and the main control node. Dynamic monitoring and blocking strategies are implemented to logically block and physically isolate key propagation paths.

[0071] Furthermore, the early warning module 14 is used to perform the following methods:

[0072] Based on the node self-inspection instructions and cascading effects in the prevention and control early warning information, the system integrates the node self-inspection execution status, cascading effect monitoring data, safety risk event evolution status, and monitoring and blocking strategy effectiveness evaluation data to generate the feedback response information. This feedback response information is then sent to the master control node and the associated gradient circles corresponding to the sub-control nodes. Real-time risk identification and assessment are performed based on the feedback response information. When the risk level exceeds a preset threshold, a dynamic expansion mechanism for the risk joint control range is activated, adjusting the boundaries and node connection relationships of the collaborative sensing network. Based on the adjusted risk joint control range, the prevention and control link topology is reconstructed, optimizing the collaborative relationships and resource allocation strategies between nodes. Based on the reconstructed prevention and control link, risk feature propagation model simulation and diffusion trend prediction analysis are performed to determine risk joint control trend indicators.

[0073] Furthermore, the early warning module 14 is used to perform the following methods:

[0074] The system monitors the heartbeat signals of each sub-control node. When a preset number of heartbeat signals are lost consecutively, the corresponding sub-control node is determined to have entered a communication interruption state. A final data transmission analysis mechanism is activated to extract the last transmitted data of the interrupted node within a preset time window before the communication interruption from the collaborative sensing network. The last transmitted data undergoes integrity verification, abnormal pattern identification, and situation reconstruction analysis. Based on the analysis results, the cause of the interruption, the risk evolution path, and the impact range on the associated gradient circle are inferred. A communication interruption emergency response plan is generated to guide adjacent nodes in adjusting their monitoring strategies and resource allocation. Furthermore, after activating the final data transmission analysis mechanism, the system also includes: establishing an interruption recovery monitoring mechanism to quickly rebuild node connections and data synchronization after communication is restored.

[0075] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for analyzing railway safety hazards under a master-slave control system, characterized in that, The method includes: Based on the node association relationship between the master control and the sub-control, configure the collaborative control linkage response gradient and build a collaborative sensing network; The sensing devices connected to the sub-control nodes extract sensing and monitoring features and project them onto the collaborative sensing network to identify risk events and locate the collaborative control and linkage response links. Based on the aforementioned collaborative control and response link, the event factors of each collaborative control and response node are analyzed and located to generate prevention and control early warning information, which is used to locate the prevention and control factors of the collaborative control and response node and decompose the prevention and control instructions. Based on the prevention and control early warning information, feedback response information is generated, the prevention and control link is reconstructed in the collaborative sensing network, the risk joint control trend is determined, and a safety hazard broadcast is sent to provide early warning reminders to the main and sub-control nodes.

2. The railway safety hazard analysis method under master-slave control coordination according to claim 1, characterized in that, Based on the node association relationship between the master controller and the sub-controllers, a collaborative control and linkage response gradient is configured to construct a collaborative sensing network, including: By taking the master control node and sub-control nodes as the center, we can analyze the physical location, path function association, and data flow dependency of each node to determine the relationship in each dimension. Based on the correlation between the dimensions, multi-gradient partitioning is performed to determine the gradient region distribution of each center; Using the master control node and sub-control nodes as nodes, establish connection edges between nodes according to the correlation of each dimension, and label the correlation gradient circles of each node according to the gradient region distribution of each center to construct the collaborative perception network.

3. The railway safety hazard analysis method under master-slave control coordination according to claim 2, characterized in that, The collaborative sensing network is constructed by labeling the associated gradient circles of each node according to the gradient region distribution of each center, including: Establish a mapping relationship between the correlation gradient and the data transmission frequency, which includes at least high frequency, medium frequency and low frequency, with the transmission frequency corresponding to the correlation gradient; Based on the mapping relationship, the transmission rules of each node and its corresponding associated gradient circles are set, and a transmission channel is constructed. Among them, a high-frequency data channel is established between nodes within the same associated gradient circle, and a low-frequency heartbeat interaction is maintained between different associated gradient circles. Establish data flow rules for the master control node, and construct data perception and transmission channels between the master control node and each sub-control node according to the data flow rules.

4. The railway safety hazard analysis method under master-slave control coordination according to claim 3, characterized in that, Establish data flow rules for the master node, including: Set the baseline for the interaction frequency of the main and sub-control nodes; Configure the frequency adjustment amplitude of each node based on the interaction and correlation characteristics of each sub-control node; Based on the frequency adjustment amplitude of each node, the frequency of the interaction frequency baseline is adjusted to determine the self-interaction frequency of each sub-control node. Configure the data flow rules of the master control node according to the self-interaction frequency of each sub-control node.

5. The railway safety hazard analysis method under master-slave control coordination according to claim 4, characterized in that, Based on the interaction and correlation characteristics of each sub-control node, configure the frequency adjustment amplitude of each node, including: A four-dimensional weighted evaluation model is constructed, including traffic flow dimension, connectivity density dimension, functional criticality dimension, and isolation risk dimension; The evaluation results of each dimension are obtained through the four-dimensional weight evaluation model, and the comprehensive weight of each node is calculated by weighted geometric average. Based on the difference between the comprehensive weight and the benchmark evaluation quantity, the control node level is configured, and the frequency adjustment amplitude of each node is configured according to the node level.

6. The railway safety hazard analysis method under master-slave control coordination according to claim 2, characterized in that, The sensing devices connected to the sub-control nodes extract sensing and monitoring features and project them onto the collaborative sensing network to identify risk events and locate the collaborative control and response links, including: Multimodal sensing devices connected to sub-control nodes collect multi-dimensional monitoring data on environment, structure, and operating status; From the multi-dimensional monitoring data, sensor monitoring features are extracted, including time-domain features, frequency-domain features, statistical features, and anomaly features; The extracted sensor features are projected into the collaborative sensing network. Risk events are identified based on the sensor features of each sub-control node, the risk events and risk probabilities are determined, and the graph neural network is used to perform feature propagation and influence range calculation based on the node association relationship in the collaborative sensing network. Based on the characteristics of propagation and the scope of influence, the collaborative control and response links are located, the influence propagation path, key response nodes and associated gradient circles are determined, and the response chain topology is constructed.

7. The railway safety hazard analysis method under master-slave control coordination according to claim 6, characterized in that, Based on the aforementioned coordinated control and response link, event factor analysis and location are performed on each coordinated control and response node to generate prevention and control early warning information, including: Based on the collaborative control and response link, the event factors of each response node are analyzed and located to identify direct causal factors, propagation path factors and node vulnerability factors. Based on the results of event factor analysis, node self-check instructions are generated to guide nodes to check whether they have similar problems or potential risks. Monitor the cascading effects of related risk events, including signal transmission interruptions, power supply disruptions, and data transmission anomalies; Based on the cascading effect and node self-inspection instructions, the prevention and control early warning information is generated, the prevention and control factors are located and broken down into specific prevention and control instructions, which are fed back to the corresponding sub-control nodes and the main control node. Dynamic monitoring and blocking strategies are implemented to logically block and physically isolate key transmission paths.

8. The railway safety hazard analysis method under master-slave control coordination according to claim 7, characterized in that, Based on the aforementioned prevention and control early warning information, feedback response information is generated, and the prevention and control link is reconstructed in the collaborative sensing network to determine the risk joint control trend, including: Based on the node self-inspection instructions and cascading effects in the prevention and control early warning information, the node self-inspection execution status, cascading effect monitoring data, safety risk event evolution status, and monitoring and blocking strategy effectiveness evaluation data are integrated to generate the feedback response information; The feedback response information is sent to the master control node and the associated gradient circle corresponding to the sub-control node. Real-time risk identification and assessment are performed based on the feedback response information. When the risk level exceeds the preset threshold, a dynamic expansion mechanism for the risk joint control range is activated to adjust the boundary and node connection relationship of the collaborative perception network. Based on the adjusted scope of joint risk control, the topology of the prevention and control link will be reconstructed, and the collaborative relationships and resource allocation strategies between nodes will be optimized. Based on the reconstructed prevention and control chain, we conduct risk characteristic propagation model simulation and diffusion trend prediction analysis to determine risk joint control trend indicators.

9. The railway safety hazard analysis method under master-slave control coordination according to claim 8, characterized in that, Also includes: Monitor the heartbeat signals of each sub-control node. When the heartbeat signals are lost for a preset number of consecutive times, the corresponding sub-control node is determined to enter a communication interruption state. The final data transmission analysis mechanism is activated to extract the last transmitted data of the interrupted node within a preset time window before the communication interruption from the collaborative sensing network. The last transmitted data is subjected to integrity verification, abnormal pattern identification, and situation reconstruction analysis. Based on the analysis results, the cause of the interruption, the risk evolution path, and the impact range on the associated gradient circle are inferred. An emergency response plan for communication interruption is generated to guide adjacent nodes to adjust their monitoring strategies and resource allocation. After initiating the final data transmission analysis mechanism, the following is also included: Establish an interruption recovery monitoring mechanism to quickly rebuild node connections and data synchronization after communication is restored.

10. A railway safety hazard analysis system under master-slave control collaboration, characterized in that, The system is used to implement the railway safety hazard analysis method under master-slave control coordination as described in any one of claims 1-9, and the system comprises: Perception network construction module: Based on the node association relationship between the master controller and the sub-controllers, configure the collaborative control linkage response gradient and construct a collaborative perception network; Link positioning module: Connects to the sensing devices of the sub-control nodes, extracts sensing and monitoring features and projects them onto the collaborative sensing network to identify risk events and locate the collaborative control and linkage response links; Analysis module: Based on the collaborative control and response link, analyze and locate the event factors of each collaborative control and response node, generate prevention and control early warning information, and use it to locate the prevention and control factors of the collaborative control and response node and decompose the prevention and control instructions; Early warning module: Generates feedback response information based on the prevention and control early warning information, reconstructs the prevention and control link in the collaborative sensing network, determines the risk joint control trend, and sends a safety hazard broadcast to provide early warning reminders to the main and sub-control nodes.

Images