Method and system for identifying electrified railway detection data based on artificial intelligence

Abstract

The invention provides an electrified railway detection data identification method and system based on artificial intelligence, and the method comprises the steps: obtaining fault element identifier construction basic data through integrating multiple types of scarce labeled fault sample data and calling a railway fault element identifier library based on railway industry safety technical specifications; a railway fault exclusive fault element identification space covering all equipment fault dimensions of a railway is constructed, training optimization of a fault element identification extractor is implemented, the fault element identification extractor is obtained, electrified railway multi-equipment collection data to be detected are integrated and input into the fault element identification extractor, a final fault element identification set to be detected is obtained, and the fault element identification set to be detected is obtained. And mapping the to-be-detected fault element identifier to a railway fault exclusive fault element identifier space to obtain a final to-be-detected fault element identifier matching result, and generating an electrified railway fault identification result in combination with the calibrated fault type hierarchy mapping relationship in the railway fault element identifier library. According to the invention, the accuracy of fault identification can be improved.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to a method and system for identifying electrified railway inspection data based on artificial intelligence. Background Technology

[0002] With the continuous development of electrified railway operation and maintenance technology, electrified railway inspection data identification technology has emerged. This technology intelligently processes operational inspection data from various facilities such as railway catenary, track lines, and traction substations to pinpoint the specific attributes of faults and their associated impact range, providing data support for railway operation and maintenance decisions. Currently, data collected from different equipment is typically processed separately, relying on general artificial intelligence identification models or directly referring to railway industry safety technical specifications for fault identification. Some technologies use labeled fault samples to train the identification model, but the training process mostly uses samples as basic data input, failing to fully integrate the structured rules of industry specifications to effectively constrain the model's core identification logic. Existing technologies generally suffer from insufficient alignment between the model's identification logic and the actual fault correlation patterns in the railway system. When data from different equipment is processed independently, the potential correlations between faults are easily overlooked, making it difficult for the identification results to accurately correspond to the hierarchical mapping requirements of fault types in industry specifications. Summary of the Invention

[0003] This invention provides a method and system for identifying electrified railway inspection data based on artificial intelligence.

[0004] In a first aspect, embodiments of the present invention provide an artificial intelligence-based method for identifying electrified railway inspection data. The method includes: integrating multi-type scarce labeled fault sample data and retrieving a railway fault meta-identifier library based on railway industry safety technical specifications; combining the multi-type scarce labeled fault sample data and the railway fault meta-identifier library to obtain basic data for constructing fault meta-identifiers; based on the basic data for constructing fault meta-identifiers, constructing a railway fault-specific meta-identifier space covering all fault dimensions of railway equipment; and implementing training and optimization of a fault meta-identifier extractor. The training process uses the fault association mapping relationship within the railway fault-specific meta-identifier space as a constraint benchmark, continuously adjusting the internal association logic of the extractor to obtain a final fault meta-identifier extractor adapted for electrified railway fault identification. The system integrates data collected from multiple devices of the electrified railway to be tested. This data is then input into a fault identifier extractor specifically designed for electrified railway fault identification, resulting in a final set of fault identifiers to be detected. This final set of fault identifiers is mapped to a dedicated fault identifier space for railway faults. Based on pre-defined fault generalization content within this space, cross-dimensional combination matching of the fault identifiers is performed to obtain the final matching result. Finally, based on this matching result and the calibrated fault identifier library, a final electrified railway fault identification result is generated, including the precise fault type, associated equipment, and associated data dimensions.

[0005] Secondly, embodiments of the present invention provide a computer system, including: a memory storing a computer program; and a processor for loading the computer program to implement the above-described artificial intelligence-based method for identifying electrified railway inspection data.

[0006] This invention deeply integrates scarce labeled fault samples with a structured mapping framework under railway industry standards, ensuring that the logical basis for subsequent fault identification fully aligns with the fault association rules of the railway industry. This avoids the problem of general identification logic being disconnected from actual industry rules. By constructing a dedicated identifier space covering all equipment fault dimensions and using it to constrain the training of the extractor, the extractor's identification logic accurately adapts to the inherent association logic of faults across all railway equipment. This avoids the problem of general extractors being unable to capture the unique fault propagation patterns of the railway. By integrating the data to be detected from multiple devices into a unified input to the extractor, collaborative identification of fault information from multiple devices is achieved, avoiding the problem of fragmented fault association information caused by independently processing data from multiple devices. Cross-dimensional combination matching is implemented through a fault generalization method based on the dedicated space, allowing the fault information to be detected to form a deep correspondence with the preset fault logic. By combining hierarchical mapping relationships to generate identification results, accurate correspondence between fault type, associated equipment, and data dimensions is achieved, avoiding the problem of scattered and unrelated output information. Attached Figure Description

[0007] Figure 1 This is a flowchart of an artificial intelligence-based method for identifying electrified railway inspection data, provided in an embodiment of the present invention.

[0008] Figure 2 This is a schematic diagram of the composition of a computer system provided in an embodiment of the present invention. Detailed Implementation

[0009] Please see Figure 1 The method for identifying electrified railway inspection data based on artificial intelligence provided in this embodiment of the invention includes the following steps: Step S100: Integrate multi-type scarce labeled fault sample data and retrieve the railway fault meta-identifier library based on railway industry safety technical specifications. Combine the multi-type scarce labeled fault sample data with the railway fault meta-identifier library to obtain the basic data for constructing fault meta-identifiers.

[0010] The multi-type scarce labeled fault sample data includes fault association collection data and corresponding fault type labeling information for railway catenary, track lines, and traction substation equipment. The railway fault element identifier library contains a structured classification framework and association mapping relationships for fault association data. The multi-type scarce labeled fault sample data originates from railway catenary, track lines, and traction substation equipment. For the railway catenary, fault association collection data includes contact wire temperature, height, and pull-out value. This data is acquired through various sensors installed on the catenary, such as temperature sensors measuring contact wire temperature and laser rangefinders measuring contact wire height and pull-out value. The corresponding fault type labeling information clearly indicates the fault type; for example, excessively high contact wire temperature may correspond to "contact wire overheating fault," and abnormal height may correspond to "contact wire height exceeding standard fault." Fault association collection data for track lines includes geometric parameters such as track gauge, level, and triangular grooves, collected by track geometry measuring instruments. Fault type labeling information includes terms such as "track gauge exceeding limit fault" and "leveling failure fault." The fault correlation data collected for traction substation equipment includes the equipment's voltage, current, power factor, etc., and is obtained from equipment such as voltage transformers and current transformers. The fault type is labeled as "overvoltage fault" or "low power factor fault".

[0011] In one implementation, step S100 may specifically include the following steps S110 to S160: Step S110: Receive fault association collection data, corresponding fault type labeling information, and associated equipment operating time and collection location information pushed by the respective storage nodes of the railway catenary, track line, and traction substation in the multi-type scarce labeled fault sample data. Complete the data binding according to the physical connection relationship of the equipment to obtain the physically associated fault sample dataset.

[0012] Storage nodes for railway overhead contact lines, track lines, and traction substation equipment are used to store various fault-related data. These storage nodes push fault-related collected data, corresponding fault type annotations, equipment operating time, and collection location information to the data processing terminal. Track line storage nodes push abnormal track geometric parameter data collected at adjacent locations within the same time period, along with corresponding fault type annotations. Traction substation equipment storage nodes send abnormal equipment operating parameter data and fault type annotations for the corresponding time period and location. The physical connection relationships of the equipment are reflected in the actual railway system: traction substation equipment supplies power to the overhead contact line via cables, and trains obtain power from the overhead contact line through pantographs and run on the track. Based on these physical connection relationships, fault data from different devices are bound together to form a physically correlated fault sample dataset.

[0013] Step S120: For each data point in the physically associated fault sample dataset, analyze the transmission correspondence between the fault association data and the fault type labeling information, record the node sequence and time difference from the occurrence to the manifestation of the fault, and obtain the pre-constructed fault transmission path result.

[0014] Data mining and machine learning algorithms, such as decision tree algorithms, can be used to construct a decision tree model by using fault association data as input features and fault type labeling information as output labels. Through learning from a large amount of data, the decision tree model generates a decision path from input features to output labels, thereby clarifying the transmission correspondence between fault association data and fault type labeling information. When recording the sequence and time difference of fault occurrence from occurrence to manifestation, the operating logic of the equipment and the time series of data acquisition are combined. For example, in an electrical fault, the traction substation's protection device may first detect an abnormal current and issue a warning signal; this is the starting point of the fault. After a period of time, the voltage of the contact network fluctuates; this is the second point. After another period of time, the train's pantograph trips; this is the point where the fault manifests. By recording the occurrence time of these points and calculating the time difference between adjacent points, the time series information of the fault from occurrence to manifestation can be obtained.

[0015] Step S130: Receive a structured document based on railway industry safety technical specifications, complete machine-readable conversion according to fault propagation classification rules, embed the association mapping framework of physical device connections, and obtain a railway fault meta-identifier library with standardized propagation logic.

[0016] To enable computers to process and utilize the information in these documents, machine-readable transformation can be achieved through natural language processing (NLP) techniques. Text extraction tools are used to extract key fault-related information from structured documents, such as fault names, fault characteristics, and fault propagation rules. Then, this information undergoes semantic analysis, converting it into a computer-understandable format, such as a knowledge graph or database table. After machine-readable transformation, this information is embedded into an association mapping framework for physical connections of equipment. This framework describes the actual physical connections and data interaction relationships between the railway catenary, track lines, and traction substation equipment. By combining fault propagation information with physical connection information, the propagation and impact of faults throughout the railway system can be described more accurately.

[0017] Step S140: Perform transmission matching between the pre-constructed fault transmission path results and the association mapping framework of the railway fault meta-identifier library with standardized transmission logic, record the correspondence between the matched links and multi-type scarce labeled fault sample data, and obtain the fault sample dataset of transmission matching.

[0018] During transmission matching, each node and transmission relationship in the pre-constructed fault transmission path is compared with the association mapping framework. A graph matching algorithm can be used, treating the pre-constructed fault transmission path as a directed graph and the association mapping framework as a directed graph, and determining the degree of matching by comparing the similarity of nodes and edges in the two graphs.

[0019] For example, the pre-constructed fault propagation path results show that the fault starts from a transformer fault in the traction substation and propagates to an insulator flashover fault in the overhead contact line after a period of time. In the association mapping framework, if a propagation path from the transformer fault to the insulator flashover fault exists, and the time difference and fault characteristics also match, then it is considered a matched propagation link.

[0020] When recording the correspondence between the matched links and the multi-type scarce labeled fault sample data, each matched link is associated with the corresponding fault-related collection data, fault type labeling information, and equipment runtime and collection location information.

[0021] Step S150: For the fault sample dataset of transmission matching, complete the cross-device transmission links between faults of different equipment categories, and synchronously update the corresponding association mapping framework of the railway fault meta-identifier library with standardized transmission logic to obtain the meta-identifier library with cross-device transmission logic calibration.

[0022] The fault sample dataset for transmission matching contains the transmission paths of faults across different devices and the corresponding fault data. However, in actual railway systems, there may be some cases where fault transmission between different device categories is not fully captured or clearly defined. Therefore, it is necessary to complete the cross-device transmission links between faults of different device categories.

[0023] To complete the cross-device fault propagation chain, a combination of data analysis and expert experience can be used. First, a thorough analysis of the fault sample dataset for propagation matching is conducted to identify potential cross-device fault propagation relationships. Simultaneously, the experience of railway experts is leveraged to assist in identifying and refining these cross-device propagation chains. Experts can verify and supplement the potential propagation relationships discovered in the data based on their in-depth understanding of the railway system and their practical experience. After completing the cross-device propagation chain, the corresponding association mapping framework of the standardized railway fault meta-identifier library is updated synchronously. This is because the association mapping framework is a crucial basis for describing fault propagation logic and needs to reflect the latest fault propagation information. During the update, newly discovered cross-device propagation chains are added to the association mapping framework, and the relevant association relationships and parameters are adjusted.

[0024] Step S160: Fill the fault sample dataset of the transmission matching into the corresponding structured classification framework of the meta-identifier library of cross-device transmission logic calibration, complete the temporal correlation description of the transmission link, and obtain the basic data for constructing fault meta-identifiers.

[0025] Populating the fault sample dataset with transmission matching into the structured classification framework involves placing each fault data point in the corresponding category of the meta-identifier library based on its equipment type, fault type, and other information. For example, a fault data point from the overhead contact line is placed under the "overhead contact line fault" category in the meta-identifier library; a cross-equipment fault data point involving traction substation equipment and the overhead contact line is placed under the corresponding cross-equipment fault category based on its transmission path and correlation. During the population process, it is also necessary to complete the temporal correlation description of the transmission links. The temporal correlation description of the transmission links refers to a detailed description of the time sequence and time intervals of fault transmission between different devices. This can be achieved by analyzing the equipment runtime and acquisition time information in the fault sample dataset with transmission matching. For example, by comparing the acquisition times of fault data from different devices, the time taken for a fault to transmit from one device to another, as well as the temporal order between each transmission node, can be determined.

[0026] Step S200: Based on the fault meta-identifier, construct basic data, build a fault meta-identifier space for railway faults that covers all fault dimensions of railway equipment, and implement training and optimization of the fault meta-identifier extractor. The training process uses the fault association mapping relationship in the fault meta-identifier space for railway faults as the constraint benchmark, and continuously adjusts the internal association logic of the extractor to obtain the final fault meta-identifier extractor adapted to the identification of electrified railway faults.

[0027] In one implementation, step S200 may specifically include the following steps S210 to S2120: Step S210: Receive the full content of the basic data for constructing fault element identifiers, and complete the splitting according to the dimensions of railway catenary, track line, and traction substation equipment. Sort out the node order and correlation strength of the fault propagation link within the same dimension to obtain a single-dimensional fault propagation link network.

[0028] Within the same dimension, we analyze the node sequence and correlation strength of fault propagation links. For the overhead contact system, nodes in a fault propagation link may include components such as the contact wire, insulators, and cantilever arms. Node sequence refers to the order in which the fault propagates between these components; for example, when a fault occurs in the contact wire, it may propagate to the insulator, and then affect the cantilever arm. Correlation strength indicates the likelihood and degree of impact of fault propagation between different nodes. Correlation strength can be determined by analyzing historical fault data and pre-constructed fault propagation paths.

[0029] Step S220: For the single-dimensional fault propagation link network, connect the propagation link nodes of different device dimensions, sort out the node correspondence and triggering conditions of cross-device fault propagation, generate cross-dimensional fault propagation association entries, and obtain the cross-dimensional propagation logic network.

[0030] When connecting conductive link nodes, identify potentially related fault nodes across different equipment dimensions. For example, an abnormal output voltage node of traction substation equipment may be related to a flashover node of the overhead contact line insulator, because overvoltage in the traction substation equipment may cause the overhead contact line insulator to experience excessively high voltage, resulting in flashover.

[0031] The process involves identifying the node correspondences and triggering conditions for cross-equipment fault propagation. Node correspondences clarify which nodes across different equipment dimensions are linked by fault propagation, such as the correspondence between the overvoltage node of the traction substation and the flashover node of the overhead contact line insulator. Triggering conditions define the circumstances under which a fault will propagate from one device to another. For example, in the case of overvoltage propagation from the traction substation to flashover of the overhead contact line insulator, the triggering condition might be that the output voltage of the traction substation exceeds the withstand voltage threshold of the overhead contact line insulator. By analyzing these node correspondences and triggering conditions, cross-dimensional fault propagation association entries are generated. Each association entry includes information such as the starting node, target node, and triggering conditions for fault propagation.

[0032] Step S230: Integrate the single-dimensional fault propagation link network and the cross-dimensional propagation logic network to build a multi-dimensional framework covering the fault propagation logic of all equipment. The link nodes in this framework are associated with their corresponding spatiotemporal attributes, thus obtaining the initial railway fault-specific fault element identifier space.

[0033] During integration, it is crucial to ensure the correctness of node and edge connections between the two networks. For example, if a faulty node in the overhead contact line of a single-dimensional fault propagation link network is connected to a propagation edge pointing to the track line in a cross-dimensional propagation logic network, the logical rationality and consistency of the connection must be guaranteed. A multi-dimensional framework encompassing the fault propagation logic of all equipment should be constructed. This framework includes not only the fault propagation logic within different devices but also the fault propagation logic between devices. For instance, when a traction substation fails, it can affect the overhead contact line and track line through cross-device propagation links, and this propagation process can be clearly demonstrated in the multi-dimensional framework. To more accurately describe the fault propagation, the link nodes in the multi-dimensional framework are associated with corresponding spatiotemporal attributes. These attributes include the time and location information of the node's failure. For a faulty node in the overhead contact line, its specific time and location of failure can be associated.

[0034] Step S240: Dynamically simulate the transmission links within the initial railway fault-specific fault element identifier space, simulate the transmission path and evolution rate of the fault at different nodes, record the changes in the links and the node status after simulation, and obtain the dynamic simulation results of the transmission links.

[0035] During dynamic simulation, the propagation path and evolution rate of faults at different nodes are simulated. For example, assuming a fault occurs at a node of the traction substation, according to the propagation logic in the initial railway fault-specific fault element identifier space, the fault may propagate to a node in the overhead contact line with a certain probability and rate. This process is simulated by establishing a fault propagation model, which can consider various factors such as the physical characteristics of the equipment, its operating status, and environmental conditions. For a certain insulator node in the overhead contact line, considering high ambient humidity, the rate at which the fault propagates from adjacent nodes may be accelerated. During the simulation, the node status and link information are continuously updated. As the fault propagates, the node status may change from a normal state to a fault state, and the propagation probability and rate of the link may also change. The changes in the link and the node status after the simulation are recorded. The changes in the link include whether the propagation direction has changed and whether the propagation probability has changed. The node status record can include information such as the fault type and severity of the node.

[0036] Step S250: For the new transmission path in the dynamic simulation result of the transmission link, update the corresponding multi-dimensional framework content of the initial railway fault-specific fault element identifier space, supplement the temporal correlation logic of the new transmission path, and obtain the self-evolved railway fault-specific fault element identifier space.

[0037] For new transmission paths, the corresponding multi-dimensional framework content of the initial railway fault-specific fault identifier space is updated. This means adding new nodes and edges to the multi-dimensional framework to represent the new transmission path. For example, if dynamic simulation reveals that a fault in the traction substation can be transmitted to a specific node on the track in a new way, the corresponding connection is added to the multi-dimensional framework. Simultaneously, the temporal correlation logic for the new transmission path is supplemented. This temporal correlation logic describes the temporal order and time intervals of the fault during transmission. For the new transmission path, the required transmission time is determined based on the actual situation or further simulation analysis. For example, the transmission time from the fault node in the traction substation to the target node on the track is determined, as well as the fault occurrence time of each intermediate node during this process.

[0038] Step S260: Traverse all propagation links in the self-evolved railway fault-specific fault element identifier space, confirm the continuity and consistency of the link propagation logic, record the continuous link content and node relationships, and obtain the verified railway fault-specific fault element identifier space.

[0039] During the traversal, the logical relationships between the starting, intermediate, and target nodes of each transmission link are checked for rationality. For example, a transmission link might start from a faulty node in the traction substation, pass through a node in the overhead contact line, and finally reach a node on the track. This transmission process is checked to ensure it conforms to physical laws and actual operating conditions. If a link is found to have a fault that propagates from a low-voltage device to a high-voltage device without a reasonable explanation, then the transmission logic of this link may be problematic. Simultaneously, the consistency of the link transmission logic is checked. Consistency includes whether the logic between different links is coordinated, and whether the transmission rules and timing relationships of the links are consistent with the overall space settings. For example, in different transmission links, the transmission probability and time interval for the same type of fault node should be relatively consistent.

[0040] Step S270: Receive the content of the basic data for constructing fault element identifiers, and filter out fault sample data that can reflect the fault propagation link. The fault sample data includes node propagation relationship and timing information, forming a training dataset for propagation logic association.

[0041] The screening process involves analyzing basic data to determine its relevance to fault propagation chains based on its content and characteristics. For example, data containing fault information from multiple devices with a clear propagation relationship between these faults can be used as fault sample data. Fault sample data includes node propagation relationships and timing information. Node propagation relationships describe the path of fault propagation between different nodes, such as from a faulty node in the traction substation to a node in the overhead contact line. Timing information records the order and time intervals of fault occurrence at each node, which is crucial for understanding the fault development process.

[0042] Step S280: Receive the transmission link logic output by the verified railway fault-specific fault meta-identifier space, embed the link nodes of the training dataset associated with the transmission logic into the initial logic structure of the fault meta-identifier extractor, and build the initial association network driven by the transmission logic.

[0043] The initial logical structure of the fault identifier extractor embeds the link nodes of the training dataset associated with the fault propagation logic into the fault identifier extractor. The initial logical structure of the fault identifier extractor can be the basic architecture of a neural network model, such as the initial structure of a multilayer perceptron (MLP) or convolutional neural network (CNN). Link nodes represent various nodes in the fault propagation process, such as fault nodes in traction substations or overhead contact lines. Embedding these nodes into the extractor's initial logical structure adds the node information and features as part of the input layer to the neural network.

[0044] Step S290: Input the training dataset associated with the transmission logic into the initial association network driven by the transmission logic to complete the first round of fault element identification. Simultaneously record the correspondence between the identification results and the transmission links in the training dataset associated with the transmission logic to obtain the first round of transmission-driven identification results.

[0045] The training dataset for the transmission logic association contains fault sample data that reflects the fault transmission chain. Each sample data has a clear node transmission relationship and timing information. This dataset is input into the initial association network driven by the transmission logic. The initial association network processes the input data according to its existing logical structure and embedded link node information.

[0046] During processing, the initial association network attempts to extract fault metadata from the input data. Fault metadata represents information about the characteristics and state of a fault, such as its type, severity, and potential impact. Through a series of calculations and transformations, the initial association network outputs the fault metadata identification results from the input data.

[0047] Step S2100: Compare the first round of transmission drive identification results with the transmission link logic of the railway fault-specific fault element identifier space after verification, locate the mismatched transmission nodes and timing deviations in the results, record the logical deviation content and impact range of the nodes, and obtain the transmission logic deviation location results.

[0048] The first round of transmission-driven identification results are obtained by the initial association network driven by transmission logic to identify the training dataset, while the transmission link logic of the railway fault-specific fault meta-identifier space after verification is a verified and accurate fault transmission rule.

[0049] By comparing these two, check whether there are any discrepancies between the initial conduction-driven identification results and the conduction link logic. Discrepancies may include incorrect conduction nodes, such as the faulty conduction node identified in the initial association network being inconsistent with the actual conduction node; or timing deviations, meaning that the identified fault conduction time sequence and time interval do not match the actual conduction link logic.

[0050] When mismatched propagation nodes and timing deviations are found, record the content and scope of the logical deviation. The logical deviation includes instances where the node should have propagated to a specific location, but it actually propagated to another node, or instances where the propagation should have occurred at a specific time, but the actual propagation time differed from the identified time. The scope of impact refers to the potential effect of this logical deviation on the overall fault propagation analysis and fault identification results.

[0051] Step S2110: For the deviation nodes in the propagation logic deviation location results, adjust the corresponding propagation logic of the initial association network driven by the propagation logic, update the spatiotemporal association description of the link nodes, and obtain the association network after propagation logic correction.

[0052] The results of the propagation logic deviation localization clarified the propagation nodes and timing deviations existing in the initial association network driven by the propagation logic. For the deviation nodes, the corresponding propagation logic of the association network was adjusted.

[0053] Adjusting the transmission logic can be achieved by modifying the weights and parameters in the interconnected network. For example, if a malfunctioning node has an error in its transmission logic, causing the fault transmission path to deviate from reality, the error can be calculated using the backpropagation algorithm, and the weights associated with that node in the interconnected network can be adjusted accordingly. This changes the connection strength and transmission rules between nodes, making them more consistent with the actual fault transmission logic. Simultaneously, the spatiotemporal association descriptions of the link nodes are updated. These descriptions include the node's time and location information. When the transmission logic changes, the spatiotemporal associations of the nodes may also need to be updated accordingly.

[0054] Step S2120: Input the training dataset associated with the transmission logic into the association network after the transmission logic is corrected multiple times, repeat the identification and logic adjustment until the identification result completely matches the transmission link of the railway fault-specific fault element identifier space after verification, and obtain the final fault element identifier extractor adapted to electrified railway fault identification.

[0055] In one implementation, step S2120 may specifically include the following steps S2121~S2126: Step S2121: Receive the associated network content output by the fault element identifier extractor that is finally adapted to electrified railway fault identification, split all transmission link nodes, record the input-output transmission relationship and timing constraints of each node, and obtain the extraction node splitting result.

[0056] Deconstructing all transmission link nodes involves analyzing each transmission node in the interconnected network independently. Each node has a specific role and function, playing a different role in the fault transmission process. For example, a faulty node in a traction substation may transmit fault information to multiple nodes in the overhead contact line.

[0057] Record the input-output propagation relationships and timing constraints for each node. The input propagation relationship describes which node(s) will propagate the fault information to the current node, while the output propagation relationship indicates which nodes(s) the current node will propagate the fault information to. Timing constraints specify the time order and time interval for the propagation of fault information between nodes.

[0058] Step S2122: For each node in the splitting result of the extraction node, the fault propagation scenario and evolution trajectory of the corresponding node in the training dataset associated with the propagation logic are generated to produce the scenario association description and triggering conditions of the node, and the scenario association result of the propagation node is obtained.

[0059] The extractor's node decomposition results record in detail the input-output transmission relationships and timing constraints of each node. For each node, the fault transmission scenarios and evolution trajectories of the corresponding nodes in the training dataset associated with the transmission logic are also recorded.

[0060] The training dataset for the fault propagation logic contains a large number of fault samples, each with a specific fault propagation scenario and evolution trajectory. The fault propagation scenario describes how faults propagate between nodes under different conditions, such as propagation under equipment overload or harsh environments. The evolution trajectory records the development process of the fault from the starting node to the current node.

[0061] By associating nodes in the extractor's node splitting results with the fault propagation scenarios and evolution trajectories of corresponding nodes in the training dataset, scenario association descriptions and triggering conditions for nodes can be generated. The scenario association description details the fault propagation of nodes in different scenarios, while the triggering conditions specify under what conditions a node will begin fault propagation.

[0062] Step S2123: Embed the scene association results of the transmission nodes into the association network of the fault element identifier extractor that is finally adapted to electrified railway fault identification, update the scene association logic of each transmission node, and synchronously record the link transmission relationship after scene association to obtain the association network driven by scene association.

[0063] During the embedding process, the scenario association logic of each transmission node is updated. The scenario association logic specifies how nodes process and transmit input fault information under different scenarios. For example, for a node of a traction substation, its handling of fault information may differ under normal operation and overload scenarios; updating the scenario association logic adjusts the node's processing rules for different scenarios. The link transmission relationships after scenario association are recorded synchronously. The link transmission relationships describe the propagation path and method of fault information between different nodes. When a node's scenario association logic is updated, the link transmission relationships may also change.

[0064] Step S2124: Input the newly added fault sample data into the association network driven by the scene association, complete the fault element identification, and synchronously record the correspondence between the identification result and the transmission link in the newly added fault sample data to obtain the identification result driven by the transmission of the new data.

[0065] The scene-related network processes and analyzes newly added fault sample data based on its updated scene association logic and link propagation relationships. During processing, the network attempts to extract fault metadata from the input data to complete fault metadata identification.

[0066] The correspondence between the identification results and the transmission links in the newly added fault sample data is recorded synchronously. This is similar to recording the results of the first round of transmission-driven identification. Recording this correspondence is for subsequent evaluation of the network's ability to process and identify new data.

[0067] Step S2125: Compare the newly added data transmission-driven identification results with the transmission link logic of the railway fault-specific fault element identifier space after verification, locate the transmission scenarios and timing vulnerabilities not covered in the results, record the missing content and impact range of the scenarios, and obtain the transmission scenario missing location results.

[0068] The newly added data transmission-driven identification result is obtained by the association network driven by the scene association to identify the newly added fault sample data, while the transmission link logic of the railway fault-specific fault element identifier space after verification is a verified and accurate fault transmission rule.

[0069] By comparing these two methods, we can check whether there are any uncovered transmission scenarios and temporal vulnerabilities in the newly added data-driven identification results. Uncovered transmission scenarios may refer to scenarios appearing in the newly added fault sample data that the scenario-related network did not consider during previous training and adjustments, resulting in the inability to accurately identify fault meta-identifiers. Temporal vulnerabilities refer to discrepancies between the fault transmission time sequence and time interval in the identification results and the actual transmission chain logic.

[0070] When uncovered propagation scenarios and timing vulnerabilities are discovered, the missing details and scope of impact of the scenario are recorded. The missing details can describe in detail the specific circumstances of the uncovered scenario, such as the fault propagation under certain special environmental conditions. The scope of impact refers to the potential effect of such missing scenarios and timing vulnerabilities on the accuracy and reliability of fault identification results.

[0071] Step S2126: For the missing scenarios in the missing location results of the transmission scenario, supplement the corresponding node logic of the association network driven by the scenario association, update the link transmission relationship and timing constraints, and obtain the adaptively optimized fault element identifier extractor.

[0072] The missing fault propagation scenario localization results revealed uncovered propagation scenarios and timing vulnerabilities in the scenario-related network when processing newly added fault sample data. For these missing scenarios, the corresponding node logic of the scenario-related network was supplemented. This supplementation can be achieved by adding new rules and conditions. For example, if a fault propagation scenario in a humid environment is found to be uncovered, then logic rules related to the humid environment can be added to the corresponding node in the network, specifying how the node should process and propagate fault information in this environment.

[0073] Simultaneously, the link propagation relationships and timing constraints are updated. When new node logic is added, the propagation path and timing order of fault information between nodes may change. For example, in a new humid environment scenario, a node may propagate fault information to another node that was not previously connected, or the propagation time may change. Therefore, the link propagation relationships and timing constraints are updated to ensure that the associated network can accurately reflect the fault propagation situation.

[0074] Step S300: Integrate the multi-equipment data collected by the electrified railway to be tested. The multi-equipment data collected by the electrified railway to be tested includes online monitoring data of the overhead contact system, vibration data of the track line, and operation data of the traction substation. Input the multi-equipment data collected by the electrified railway to be tested into the fault element identifier extractor that is finally adapted to electrified railway fault identification to obtain the final set of fault element identifiers to be tested.

[0075] In one implementation, step S300 may specifically include the following steps S310 to S360: Step S310: Receive real-time data from the online monitoring system for overhead contact lines, the vibration acquisition device for track lines, and the traction substation operation monitoring device in the multi-equipment data collection data of the electrified railway to be tested. Complete real-time binding according to the physical connection relationship of the equipment, and synchronously embed the association description between the current equipment operating status and the acquisition environment to obtain the physically associated dataset to be tested.

[0076] The overhead contact line online monitoring system, track vibration acquisition device, and traction substation operation monitoring device are responsible for collecting real-time data from the overhead contact line, track line, and traction substation equipment, respectively. The overhead contact line online monitoring system collects real-time operating parameters of the overhead contact line through various sensors installed on it, such as temperature sensors and displacement sensors. The track vibration acquisition device typically uses vibration sensors installed on the track to collect vibration information of the track during train operation. The traction substation operation monitoring device uses voltage transformers, current transformers, and other equipment to monitor the electrical parameters of the traction substation equipment.

[0077] After receiving the real-time data collected from these devices, real-time binding is completed according to the physical connection relationships of the equipment. In actual railway systems, traction substations supply power to the overhead contact line via cables, and trains obtain power from the contact line through pantographs and travel on the track. These physical connections determine the data correlation between the devices. For example, when the output voltage of the traction substation changes, it directly affects the voltage state of the contact line, thereby affecting the current collection of the train and the electrical characteristics of the track. Therefore, binding the real-time data collected from different devices according to this physical connection relationship can more accurately reflect the operating status of the entire railway system. Simultaneously, a correlation description between the current equipment operating status and the data collection environment is embedded. The equipment operating status includes whether the equipment is operating normally, overloaded, or faulty, while the data collection environment involves factors such as ambient temperature, humidity, and wind speed.

[0078] Step S320: For the physically associated dataset to be detected, sort it according to the runtime phase, and synchronously associate the running status descriptions and environmental parameters of different devices in the same time period to obtain the time-associated dataset to be detected.

[0079] The physically associated dataset contains real-time data collected from the overhead contact system, track lines, and traction substations, and this data has been bound according to the physical connections of the equipment. To better analyze the development process and transmission mechanism of faults, this data is sorted chronologically by operating segment.

[0080] Sorting by runtime sequence means arranging the data according to the order in which it was collected. Based on the timestamp information in the data, data collected by different devices at the same or adjacent times can be grouped together.

[0081] Simultaneously, the operational status descriptions and environmental parameters of different devices within the same time period are correlated. The operational status descriptions explain the equipment's operating condition during that period, such as whether it is operating normally or whether any abnormalities have occurred. Environmental parameters include environmental factors such as temperature, humidity, and wind speed. For example, at a specific time, the overhead contact line may be operating normally, the track vibration frequency may be slightly higher than normal, and the ambient temperature may be higher. Correlating these equipment operational statuses and environmental parameters within the same time period allows for a more comprehensive understanding of the railway system's operational status during that period.

[0082] Step S330: Input the temporally correlated dataset to be detected into the adaptively optimized fault meta-identifier extractor to complete real-time fault meta-identifier recognition, and synchronously record the correspondence between the recognition results and the propagation nodes in the temporally correlated dataset to be detected, thereby obtaining a real-time fault meta-identifier set.

[0083] The adaptively optimized fault identifier extractor is a model trained and optimized in the previous steps. It can accurately extract identifier information reflecting fault characteristics based on the input data. The temporally correlated dataset to be detected is input into this extractor, which then processes and analyzes the data.

[0084] During processing, the extractor identifies potential fault identifiers from the input data based on its internal logical structure and trained rules. For example, when the input time-series correlated dataset contains information such as abnormally high temperature of the overhead contact line, abnormally high vibration frequency of the track line, and current fluctuations in the traction substation, the extractor identifies the corresponding fault identifiers based on these data characteristics, such as "overheating fault of the overhead contact line," "abnormal vibration fault of the track line," and "unstable current fault of the traction substation."

[0085] The system synchronously records the correspondence between the identification results and the transit nodes in the time-series-related dataset to be detected. These transit nodes represent key locations on the propagation path of a fault across different devices and systems. Recording this correspondence clarifies which transit nodes generated each fault identifier, aiding in subsequent analysis of fault propagation and development.

[0086] Step S340: For the real-time fault meta-identifier set, associate the transmission link logic of the railway fault-specific fault meta-identifier space after verification, generate the fault candidate transmission link and temporal evolution trajectory corresponding to the current identification result, and obtain the real-time fault candidate transmission link set.

[0087] The real-time fault metadata set contains fault metadata identified from the temporally correlated dataset to be detected, along with their correspondence with propagation nodes. The propagation link logic of the validated railway fault-specific fault metadata space is a validated and accurate fault propagation rule that describes the propagation path and temporal sequence of faults between different devices and systems.

[0088] For the set of real-time fault identifiers, the associated transmission link logic is implemented. This involves matching each fault identifier in the real-time fault identifier set with the transmission links in the verified railway fault-specific fault identifier space to find possible corresponding fault transmission paths. For example, for the fault identifier "overheating of the overhead contact line," all possible transmission paths that could lead to overheating of the overhead contact line are searched in the verified railway fault-specific fault identifier space, such as overvoltage from traction substation equipment being conducted to the overhead contact line, causing the contact wire temperature to rise.

[0089] Based on the matched propagation paths, candidate fault propagation paths and temporal evolution trajectories corresponding to the current identification result are generated. Candidate fault propagation paths refer to fault propagation paths that may lead to the occurrence of the currently identified fault meta-identifier. The temporal evolution trajectory describes the temporal order of fault development along these propagation paths.

[0090] Step S350: For the real-time fault candidate propagation link set, track the real-time data changes and state transitions of each node in the link, and synchronously update the propagation state description and temporal constraints of the link to obtain a dynamically updated fault candidate propagation link set. The real-time fault candidate propagation link set contains the fault candidate propagation link and temporal evolution trajectory corresponding to the current identification result. However, in the actual operation of the railway system, the state of each device is constantly changing, and faults may also develop and evolve over time. Therefore, each link in the real-time fault candidate propagation link set is tracked in real time.

[0091] For each node in the link, its real-time data changes and state transitions must be closely monitored. Taking a node in the overhead contact system as an example, real-time data changes may include real-time fluctuations in parameters such as the temperature and tension of the contact wire, while state transitions may involve changing from a normal state to a fault state, or from a minor fault state to a severe fault state. Real-time data tracking of nodes can be achieved by continuously receiving real-time data from the online monitoring system of the overhead contact system, and by combining this with preset state judgment rules to determine the state transition of the node. For example, when the temperature of the contact wire exceeds a preset safety threshold, the node is determined to have entered a fault state.

[0092] During the tracking process, the transmission state description and timing constraints of the link are updated synchronously. The transmission state description describes the propagation of the fault in the link, such as whether the fault has propagated from one node to the next, or whether propagation has temporarily stopped at a certain node. The timing constraints specify the time requirements and order of fault propagation between nodes. As the real-time data of the nodes changes and their states transition, the transmission state description and timing constraints also need to be adjusted accordingly.

[0093] Step S360: Integrate the real-time fault meta-identifier set with the dynamically updated fault candidate propagation link set, complete the spatiotemporal correlation description of the propagation link, and obtain the final fault meta-identifier set to be detected.

[0094] In one implementation, step S360 may specifically include the following steps S361 to S366: Step S361: Receive the full content of the final set of fault identifiers to be detected, split it according to the collection period, and synchronously associate the fault candidate propagation links and time-series evolution trajectories of each period to obtain the fault identifiers and link sets of each period.

[0095] The final set of fault identifiers to be detected contains fault information after integration and completion of spatiotemporal correlation descriptions, covering fault identifiers, transmission links, and spatiotemporal correlations. After receiving the full content of this set, it is split according to the collection period. The collection period refers to the time interval for data collection, which can be divided according to time intervals such as every hour or every half hour.

[0096] During the data splitting process, the data in the final set of fault identifiers to be detected is grouped according to the collection time period. The data in each time period includes the fault identifiers within that time period, as well as the corresponding fault candidate transmission links and time-series evolution trajectories. For example, for the collection time period from 9:00 AM to 10:00 AM, the fault identifiers of the overhead contact system, track lines, and traction substations within that time period, along with the related transmission links and time-series evolution trajectories, are grouped together.

[0097] The fault candidate propagation links and temporal evolution trajectories for each time period are synchronously correlated. The fault candidate propagation links describe the propagation path of the fault between different devices and nodes, while the temporal evolution trajectories record the time sequence of the fault's development on these links.

[0098] Step S362: For the fault identifiers and adjacent time period content of the link set in the time period segmentation, match the node correspondence and time sequence connection point of the transmission link, generate cross-time period fault transmission link continuation entries and state transition descriptions, and obtain cross-time period transmission link continuation results.

[0099] The time-segmented fault identifiers and link sets divide fault information according to the collection time period, and the fault propagation links and temporal evolution trajectories within each time period have their own characteristics. However, in the actual fault development process, faults often continue to propagate and evolve across different collection time periods. Therefore, it is necessary to analyze and process the content of adjacent time periods.

[0100] Matching the node correspondences and timing transition points of the transmission links is a crucial step. For two adjacent data acquisition periods, the correspondence between the transmission link nodes in the previous period and those in the next period must be determined. For example, in the previous period, a node in the overhead contact system might be faulty, and a transmission link might point to a node on the track. In the next period, it's necessary to check whether this track node received the fault information and whether a new transmission link originates from that node. Simultaneously, timing transition points must be identified, i.e., the time transition points between adjacent periods, such as the fault state at the end of the previous period and the initial state of the fault at the beginning of the next period.

[0101] Based on the matching results, cross-time period fault propagation link continuation entries and state transition descriptions are generated. Cross-time period fault propagation link continuation entries describe the propagation path and continuation of the fault between adjacent time periods, such as the specific link through which a fault continues from a node in one time period to another node in the next. State transition descriptions illustrate the state changes of the fault between adjacent time periods, such as from a minor fault state to a severe fault state, or from a fault state back to a normal state.

[0102] Step S363: Embed the cross-time period transmission link continuation result into the fault identifier and link set of time period split, update the fault candidate transmission link and linkage constraint of each time period, and obtain the fault identifier and link set associated across time periods.

[0103] The cross-time period propagation link continuation results contain information on the propagation path and state transitions of the fault between adjacent time periods. This information is crucial for understanding the overall development process of the fault. Embedding the cross-time period propagation link continuation results into the fault identifiers and link sets segmented by time period means integrating this cross-time period information into the fault information of each data collection period.

[0104] During the embedding process, the candidate fault propagation links and linkage constraints for each time period are updated. The candidate fault propagation links describe the propagation path of a fault within a time period. As the results of cross-time period propagation link continuation are embedded, these propagation links may need to be adjusted and supplemented. For example, if the cross-time period propagation link continuation results show that a faulty node from the previous time period continues to propagate the fault to another node in the current time period, then the corresponding link information needs to be added to the candidate fault propagation links for the current time period.

[0105] Interconnection constraints define the relationships and limitations between different devices and nodes during fault propagation. As the cross-time propagation link continues, the interconnection constraints may need to be updated.

[0106] Step S364: For fault identifiers and link sets associated across time periods, track the cross-time period data changes and status trends of link nodes, synchronously update the transmission trend description and evolution rate of the links, and obtain the cross-time period transmission trend tracking results.

[0107] The cross-time-period associated fault identifiers and link sets have already divided fault information according to the collection period and considered the continuity of faults between adjacent time periods. However, the development of faults is not static across different time periods; their data and status may exhibit certain trends over time. Therefore, cross-time-period tracking is performed on the link nodes in the cross-time-period associated fault identifiers and link sets.

[0108] To track the changes and trends in the data of link nodes across different time periods, it is necessary to observe the changes in the real-time data of the nodes over multiple collection periods, as well as the evolution trend of the node's status between different time periods. For example, for a node in the overhead contact system, it is necessary to observe whether its contact wire temperature continuously rises, falls, or remains stable over several consecutive collection periods, and whether the node's fault status gradually worsens, alleviates, or remains unchanged. This tracking can be achieved by performing statistical analysis and trend prediction algorithms on the node data for each collection period.

[0109] The transmission trend description and evolution rate of the link are updated synchronously. The transmission trend description describes the direction and development of the fault in the link, such as whether the fault is propagating at an accelerated, decelerated, or at a constant speed. The evolution rate represents the speed at which the fault propagates between nodes. As node data changes and state trends are determined, the transmission trend description and evolution rate also need to be adjusted accordingly.

[0110] Step S365: Integrate the cross-time period transmission trend tracking results into the cross-time period associated fault identifier and link set, complete the spatiotemporal association description of the linkage status, and obtain the linkage associated fault identifier and link set.

[0111] Cross-period transmission trend tracking results reflect the development trend and changes of faults across time periods, while the cross-period associated fault identifiers and link sets contain information on the continuity and correlation of faults across different data collection periods. Integrating these two can further improve the completeness and accuracy of fault information.

[0112] During the integration process, the spatiotemporal correlation description of the linkage status is completed. The linkage status describes the mutual influence and synergy between different devices and nodes during fault propagation, while the spatiotemporal correlation description combines spatial location and temporal sequence information. For example, during cross-time period propagation, a fault in traction substation equipment may interact with faults in the overhead contact line and track, and this influence manifests differently at different spatial locations and time points. By combining the trend and time information from cross-time period propagation trend tracking results with the fault identifiers associated with cross-time periods and the spatial location and device association information in the link set, the spatiotemporal correlation description of the linkage status can be completed.

[0113] Specifically, based on the fault propagation trend and evolution rate in the cross-time period propagation trend tracking results, the linkage status of different equipment and nodes in different times and spaces can be determined. For example, when a fault in traction substation equipment accelerates propagation within a certain time period, its impact range and timing on the contact network and track line in adjacent time periods can be analyzed, and this information can be correlated with the corresponding spatial locations.

[0114] Step S366: Integrate the fault identifiers and link sets associated with the linkage, clarify the cross-device transmission links and their linkage logic contained therein, and obtain the final set of fault element identifiers to be detected that includes cross-device linkage logic.

[0115] In one implementation, step S366 may specifically include the following steps S3661 to S3666: Step S3661: Receive the full content of the final set of fault element identifiers to be detected, which contains the cross-time period transmission trend, and complete the splitting according to the dimensions of railway catenary, track line, and traction substation equipment. Synchronously associate the transmission link and time-series evolution trajectory of each dimension to obtain the fault identifier and link set of the dimension split.

[0116] During the data breakdown process, based on the data source and related attributes, fault identifiers, transmission links, and time-series evolution trajectories related to the overhead contact system are categorized into overhead contact system data, relevant content for the track lines is categorized into track line data, and data related to traction substation equipment is categorized into traction substation equipment data. For example, for the overhead contact system dimension, information such as fault identifiers of the contact wire, transmission links of overhead contact system faults, and the time-series evolution trajectories of the faults at different time periods are integrated together.

[0117] Synchronously link the transmission links and temporal evolution trajectories of each dimension to ensure that fault identifiers in each dimension match their corresponding transmission links and temporal evolution trajectories. For example, in the catenary dimension, clarify which specific transmission link corresponds to a certain catenary fault identifier, and how the fault evolves over time, such as the specific time points from the onset of the fault to the expansion of its impact range.

[0118] Step S3662: For the different dimensions of the fault identifier and link set in the dimension split, match the node correspondence and linkage conditions of the transmission link, generate cross-device fault transmission linkage entries and state transition descriptions, and obtain cross-device transmission linkage results.

[0119] The node correspondence in a transmission link refers to the association between nodes in transmission links across different equipment dimensions. For example, a node in the traction substation transmission link (such as a transformer fault node) may correspond to a node in the overhead contact line transmission link (such as an abnormal overhead contact line voltage node), because a fault in the traction substation may cause an abnormal overhead contact line voltage. The linkage condition refers to the circumstances under which these nodes will link together. For example, when the output voltage of the traction substation exceeds a certain threshold, it will trigger a linkage with the overhead contact line node.

[0120] By analyzing and comparing transmission links in different dimensions, the corresponding relationships and linkage conditions between nodes can be identified. For example, by analyzing historical fault data and real-time monitoring data, it can be determined that when a certain key parameter (such as current) of the traction substation reaches a specific value, a certain related parameter (such as voltage) of the contact network will change accordingly, thereby establishing the corresponding relationship and linkage conditions between these two nodes.

[0121] Based on the matching results, cross-device fault propagation linkage entries and state transition descriptions are generated. The cross-device fault propagation linkage entries record in detail the specific circumstances of fault propagation between different device dimensions, such as the fault propagation path and method from traction substation equipment to the overhead contact line. The state transition descriptions illustrate the state changes of each node during the linkage process, such as the transition process of an overhead contact line node from a normal state to a voltage abnormal state.

[0122] Step S3663: Embed the cross-device transmission and linkage results into the fault identifier and link set of the dimension split, update the fault candidate transmission link and linkage constraint of each dimension, and obtain the cross-device associated fault identifier and link set.

[0123] During the embedding process, the fault candidate propagation links and linkage constraints for each dimension are updated. For fault candidate propagation links, the original propagation links are supplemented or adjusted based on the cross-device propagation linkage results.

[0124] Interlocking constraints define the limitations and interrelationships of fault propagation between different equipment dimensions. Based on the results of cross-equipment propagation, the interlocking constraints are updated to better reflect actual fault propagation conditions. For example, the time interval constraints for fault propagation between traction substations and the overhead contact line can be adjusted, or interlocking relationships that only occur under certain specific conditions can be defined.

[0125] Through the above embedding and update operations, a set of fault identifiers and links associated across devices is obtained. This set integrates fault information from different device dimensions and the linkage relationship between devices.

[0126] Step S3664: For the fault identifiers and link sets associated across devices, track the cross-device data changes and linkage trends of the link nodes, synchronously update the linkage status description and evolution rate of the links, and obtain the cross-device linkage status tracking results.

[0127] The cross-device associated fault identifiers and link sets have integrated fault information from different device dimensions and cross-device linkage relationships. For this set, cross-device data changes and linkage trends of link nodes are tracked.

[0128] Cross-device data changes in a transmission link refer to the changes in relevant node data over time within the transmission link across different device dimensions. For example, changes in voltage and current data of traction substation equipment may affect relevant parameters of the overhead contact line. By continuously monitoring these data changes, it is possible to understand the propagation of faults across devices. Interconnection trends refer to the development direction and changing trend of fault transmission across different device dimensions, such as whether the fault is accelerating, decelerating, or remaining stable.

[0129] During the tracking process, the linkage status description and evolution rate of the link are updated synchronously. The linkage status description describes the current state of the fault during cross-device propagation, such as whether a node has been affected by the faults of other devices, and to what extent. The evolution rate indicates the speed at which the fault propagates between devices, such as the change in the time required for the fault to propagate from the traction substation to the overhead contact line. Real-time monitoring systems and data analysis technologies can be used to track cross-device data changes and linkage trends of link nodes.

[0130] Step S3665: Integrate the cross-device linkage status tracking results into the cross-device associated fault identifier and link set, complete the spatiotemporal association description of the linkage status, and obtain the linkage associated fault identifier and link set.

[0131] Cross-device linkage status tracking results record cross-device data changes, linkage trends, updated linkage status descriptions, and evolution rates of link nodes. Meanwhile, the cross-device associated fault identifiers and link sets contain fault information from different device dimensions and cross-device linkage relationships. Integrating the cross-device linkage status tracking results into the cross-device associated fault identifiers and link sets means incorporating the latest tracking information into the existing set.

[0132] During the integration process, the spatiotemporal correlation description of the linkage status is completed. This description combines information from both spatial and temporal dimensions, enabling a more comprehensive depiction of the linkage status during fault propagation across equipment. For example, it clarifies the specific spatial location of the fault propagation between different devices and the changes in linkage status at different points in time. For fault propagation from traction substations to the overhead contact line, it records the time when the fault begins at its specific location on the traction substation (e.g., a substation), the time it takes to propagate to its specific location on the overhead contact line (e.g., a section of contact wire), and describes the changes in the linkage status of each node during this process.

[0133] By integrating the cross-device linkage status tracking results with the cross-device associated fault identifiers and link sets, and supplementing the spatiotemporal association description of the linkage status, the linkage-associated fault identifiers and link sets are obtained.

[0134] Step S3666: Integrate the fault identifiers and link sets associated with the linkage, clarify the cross-device transmission links and their linkage logic contained therein, and obtain the final set of fault element identifiers to be detected that includes cross-device linkage logic.

[0135] The linked fault identifiers and link sets have integrated fault information from different device dimensions, cross-device linkage relationships, and the latest linkage status tracking results and spatiotemporal correlation descriptions. Integrating this set involves further sorting out and refining the key information to clarify the cross-device transmission links and their linkage logic.

[0136] Cross-device transmission links describe the propagation path of faults between different devices (such as railway overhead contact lines, track lines, and traction substations). During the integration process, all transmission links in the interconnected fault identifiers and link sets are analyzed and organized to identify those transmission paths that cross different device dimensions. For example, this determines the specific path from a fault node in the traction substation, through which intermediate links, ultimately propagating to a specific node in the overhead contact line or track line.

[0137] The linkage logic explains the interaction and coordination between nodes in these cross-device transmission links. By analyzing the transmission process and linkage status of faults between different devices, linkage logic rules are summarized. For example, when the output voltage of traction substation equipment rises abnormally, it will trigger certain faults in the overhead contact line according to a certain probability and time sequence, and this triggering relationship will be affected by other factors (such as environmental conditions, equipment operating status, etc.).

[0138] Step S400: Map the final set of fault meta-identifiers to be detected to the railway fault-specific fault meta-identifier space. Based on the preset fault generalization content in the railway fault-specific fault meta-identifier space, perform cross-dimensional combination matching of the fault meta-identifiers to be detected. The fault generalization content includes the association matching logic of fault meta-identifiers and the hierarchical mapping relationship of fault types to obtain the final matching result of the fault meta-identifiers to be detected.

[0139] In one implementation, step S400 may specifically include the following steps S410 to S460: Step S410: Receive the full content of the final set of fault element identifiers to be detected, which includes cross-device linkage logic, map each fault element identifier to the corresponding transmission link node in the verified railway fault-specific fault element identifier space, record the mapped node position and timing constraints, and obtain the node mapping result of the identifier space.

[0140] The validated railway fault-specific fault identifier space has undergone rigorous verification and optimization. The propagation link nodes within it represent different fault states and propagation paths. Mapping a fault identifier to its corresponding propagation link node involves finding the best-matching propagation link node within the validated railway fault-specific fault identifier space based on the fault identifier's characteristic information, such as fault type, location, and time. For example, for a fault identifier representing "traction substation overvoltage fault," the corresponding propagation link node for the traction substation overvoltage fault is found within the validated railway fault-specific fault identifier space based on its fault type and related spatiotemporal information.

[0141] During the mapping process, the mapped node positions and timing constraints are recorded. The node position clarifies the specific location of the fault element identifier in the verified railway fault-specific fault element identifier space, which helps in subsequent fault localization and analysis. Timing constraints specify the time requirements and order of the fault at that node, such as the time range of the fault occurrence and the time limit for fault propagation to that node. By recording this information, the state and propagation of the fault in the verified railway fault-specific fault element identifier space can be understood more accurately.

[0142] Step S420: For each node in the node mapping result of the identifier space, associate the cross-dimensional association logic of the corresponding transmission link of the railway fault-specific fault element identifier space after verification, generate the cross-dimensional transmission link and temporal evolution trajectory corresponding to the identifier, and obtain the cross-dimensional transmission link generation result.

[0143] The node mapping result maps the fault element identifiers in the final set of fault element identifiers to the transmission link nodes in the verified railway fault-specific fault element identifier space. For each mapped node, its cross-dimensional fault transmission information is further mined.

[0144] For each node in the identifier space node mapping result, the cross-dimensional association logic of the corresponding transmission link is established. By analyzing and verifying the cross-dimensional association rules in the railway fault-specific fault element identifier space, the transmission links of other dimensions related to that node are identified. For example, for a node located in the catenary dimension, the transmission links of the associated traction substation dimension and track line dimension are identified according to the cross-dimensional association logic.

[0145] Based on the association results, cross-dimensional transmission links and temporal evolution trajectories corresponding to the identifiers are generated. Cross-dimensional transmission links describe the propagation path of faults between different dimensions, such as the transmission path from the traction substation dimension to the catenary dimension and then to the track line dimension. The temporal evolution trajectory records the development time sequence of faults on these cross-dimensional transmission links, such as the specific time points and time intervals at which the fault propagates from the traction substation fault node to the catenary fault node.

[0146] Step S430: Embed the cross-dimensional transmission link generation result into the identifier space node mapping result, update the cross-dimensional association description and temporal constraints of each identifier, and obtain the set of cross-dimensional associated identifiers and links.

[0147] During the embedding process, the cross-dimensional association description and timing constraints of each identifier are updated. The cross-dimensional association description explains the relationship between the fault element identifier and faults in other dimensions; for example, how a fault element identifier of a traction substation might be related to faults in the overhead contact line and track. The timing constraints specify the time requirements and order of fault propagation in the cross-dimensional process. By adding information from the cross-dimensional propagation link generation results to the identifier space node mapping results, the cross-dimensional association description of each identifier can be made more detailed and accurate, while also updating its corresponding timing constraints.

[0148] For example, for a traction substation fault identifier that has been mapped to the verified railway fault-specific fault identifier space, after embedding the cross-dimensional transmission link generation result, its cross-dimensional association description can be updated to "the fault may be transmitted to the contact network through the power supply line, causing abnormal contact network voltage, which in turn affects the operation of the train on the track line", and the timing constraint can be updated to "the time interval from the fault transmission from the traction substation to the contact network is g minutes, and the time interval from the contact network affecting the track line is h minutes".

[0149] Step S440: For the cross-dimensional associated identifiers and link sets, deduce the probability of fault propagation and the direction of evolution of each node in the link, and simultaneously record the propagation status and time series prediction after the deduction to obtain the cross-dimensional propagation deduction results.

[0150] To extrapolate the probability of fault propagation, various factors are considered, such as the current state of a node, the state of adjacent nodes, and environmental conditions. For example, for a contact network node, if it is currently in a state of minor fault and an adjacent traction substation node also has a fault, then the probability of fault propagation for that contact network node is relatively high. A probabilistic model can be established, combining historical fault data and real-time monitoring data, to calculate the probability of fault propagation for each node.

[0151] Deducing the evolution direction involves determining the direction of fault propagation between nodes. This can be determined based on the propagation link logic and cross-dimensional association rules in the verified railway fault-specific fault element identifier space. For example, based on the power supply relationship and fault propagation logic between traction substations and the overhead contact line, when a fault occurs in the traction substation, the fault is likely to propagate towards the overhead contact line.

[0152] During the simulation, the propagation status and timing predictions are recorded simultaneously. The propagation status describes the current state and propagation of the fault at each node, such as whether the fault has propagated to that node and the severity of the fault at that node. The timing predictions estimate the time it takes for the fault to propagate between nodes, such as predicting the time required for the fault to propagate from one node to the next.

[0153] Step S450: Integrate the cross-dimensional transmission inference results with the cross-dimensional associated identifier and link set to form an identifier and link set containing the transmission status and prediction after inference, thus obtaining the inference associated identifier and link set.

[0154] During the integration process, the simulated transmission status and predicted information are added to the description of each identifier and link. For each fault element identifier, its cross-dimensional association description is updated to include the simulated transmission status and timing prediction information. For example, for a fault element identifier of a traction substation, the original cross-dimensional association description might only indicate its potential association with the overhead contact line and track. After integrating the cross-dimensional transmission simulation results, its cross-dimensional association description can be updated to "This fault has an a% probability of being transmitted to the overhead contact line within the next b minutes, causing a voltage fluctuation fault in the overhead contact line."

[0155] The link information is also updated accordingly. The simulated conduction status and timing prediction information are added to the link description to make the conduction situation of the link clearer and more accurate. For example, for a conduction link from traction substation to the overhead contact line, its description is updated to "This link is currently in a pending conduction state, and it is expected to start conducting faults in Z minutes, with a probability of W% of successful conduction".

[0156] Step S460: Integrate the set of identifiers and links associated with the inference with the final set of identifiers of the fault elements to be detected that includes cross-device linkage logic, supplement the inference details of the cross-dimensional transmission links, and obtain the matching result of the final identifiers of the fault elements to be detected that includes cross-dimensional transmission inference.

[0157] In one implementation, step S460 may specifically include the following steps S461 to S466: Step S461: Receive the full content containing the final detectable fault element identifier matching results of cross-dimensional propagation inference, split all cross-dimensional propagation links, record the inference state description and time series prediction of each link, and obtain the inference link splitting results.

[0158] Cross-dimensional transmission links may contain multiple nodes and complex transmission paths. Decomposing these links allows for a more detailed analysis of each link. For each cross-dimensional transmission link, it is broken down according to nodes and transmission order, resulting in multiple independent sub-links. For example, a cross-dimensional transmission link from traction substation to overhead contact line and then to track line can be broken down into a sub-link from traction substation to overhead contact line and a sub-link from overhead contact line to track line.

[0159] During the decomposition process, the projected state description and time series prediction for each link are recorded. The projected state description illustrates the current state and propagation of the fault in that link, such as whether the fault has started propagating and the progress of propagation. The time series prediction estimates the propagation time of the fault in that link, such as predicting the time required for the fault to propagate from the starting node to the ending node of the link. By recording this information, a clearer understanding of the specific situation of each cross-dimensional propagation link can be obtained.

[0160] Step S462: For each link in the deduced link splitting result, associate the corresponding link logic with the verified railway fault-specific fault element identifier space, calibrate the link nodes and timing constraints of the logic deviation in the deduced result, and obtain the deduced link logic calibration result.

[0161] The derivation link splitting result includes the cross-dimensional propagation links in the final fault element identifier matching result of the cross-dimensional propagation derivation, which are then split and recorded. For each split link, it is logically associated with the corresponding link in the verified railway fault-specific fault element identifier space.

[0162] The link logic in the verified railway fault-specific fault identifier space has undergone rigorous verification and optimization, defining the propagation rules and timing requirements of faults between different nodes. Each link in the deduced link decomposition results is compared with its corresponding link logic to check for logical deviations. Logical deviations may include incorrect fault propagation direction, unreasonable propagation probabilities, and timing constraints that do not conform to reality.

[0163] For example, in the deduced link breakdown results, there might be a transmission link from the overhead contact line to the traction substation. However, according to the verified link logic of the railway fault-specific fault element identifier space, under normal circumstances, the fault should be transmitted from the traction substation to the overhead contact line, which presents a logical deviation. To address this, the link nodes and timing constraints that cause logical deviations in the deduced results are calibrated.

[0164] Calibrling link nodes can be achieved by adjusting the start and end nodes of the link to conform to the link logic of the verified railway fault-specific fault identifier space. Calibrating timing constraints can adjust the time requirements for fault propagation between nodes based on the verified link logic. For example, if the propagation time of a fault between a certain node is too short or too long in the simulation results, which does not conform to the verified link logic, that time needs to be adjusted.

[0165] Step S463: Embed the logical calibration results of the inference links into the inference link splitting results, update the inference state description and timing prediction of each link, and obtain the logically calibrated inference link set.

[0166] During the embedding process, the inferred state description and timing prediction for each link are updated. Since the link's logic is calibrated, its inferred state and timing may change. For example, when the link's starting node or propagation time is adjusted, the fault's inferred state (e.g., whether propagation has started, the progress of propagation, etc.) and timing prediction (e.g., the time it takes for the fault to propagate to the next node) in that link need to be updated accordingly.

[0167] By adding the calibrated link information to the original simulated link breakdown results, the description of each link can be made more accurate and complete. For example, for a logically calibrated transmission link from traction substation to the overhead contact line, its simulated status description is updated to "The fault has begun to propagate to the overhead contact line, and the current propagation progress is c%", and its timing prediction is updated to "The fault is expected to be fully propagated to the overhead contact line in d minutes".

[0168] Step S464: For the set of simulated links after logical calibration, associate the real-time transmission status of the data collected by multiple devices of the electrified railway to be tested, confirm the real-time adaptability and evolution trend of the link simulation results, and obtain the real-time adaptability confirmation result of the simulated links.

[0169] During the correlation process, it is checked whether the simulation results of each link match the real-time transmission status. For example, for a transmission link from traction substation to the overhead contact line, the real-time status of the traction substation and the overhead contact line in the multi-equipment data collected by the electrified railway to be tested is examined to determine whether the fault is indeed being transmitted as simulated. If the simulation results show that the fault has begun to transmit to the overhead contact line, but the actual data collected shows that the status of the overhead contact line is normal, then it means that the simulation results of this link do not match the real-time situation.

[0170] Confirm the real-time adaptability and evolution trend of the link simulation results. Real-time adaptability refers to the degree to which the link simulation results match the current actual situation, while the evolution trend refers to the future direction of the fault. By analyzing the changing trends of real-time acquired data and the link simulation results, the evolution trend of the fault can be predicted. For example, if real-time acquired data shows that the fault in the traction substation is worsening, and the link simulation results show that the fault will continue to propagate to the overhead contact line, then it can be determined that the evolution trend of the fault is towards the overhead contact line.

[0171] Step S465: Integrate the real-time adaptation confirmation results of the inference link into the logically calibrated inference link set, complete the spatiotemporal correlation description of real-time adaptation, and obtain the real-time adapted inference link set.

[0172] During the integration process, the spatiotemporal correlation description for real-time adaptation is supplemented. This description, combining spatial location and temporal sequence information, provides a more comprehensive picture of fault propagation within the actual railway system. For example, for a transmission link from traction substation to the overhead contact line, after integrating and demonstrating the real-time adaptation confirmation results, the link's location information in the actual space (such as the specific geographical locations of the traction substation and the overhead contact line) and the specific time information of fault propagation within that link (such as the time when fault propagation begins and the expected time when propagation ends) can be supplemented.

[0173] By combining real-time adaptability and evolution trend information with the original set of links, the description of each link can be made more accurate and detailed. For example, for a transmission link from traction substation to catenary that has been confirmed through real-time adaptability, its description can be updated to "A fault was transmitted from a traction substation (specific location) in a certain railway section to the catenary (specific location) at point p in the morning. The current transmission progress is n%, and it is expected to be fully transmitted to the catenary by point q in the morning."

[0174] Step S466: Integrate the real-time adapted simulation link set with the final detectable fault element identifier set containing cross-device linkage logic, supplement the real-time adaptation details of the simulation link, and obtain a logically coherent final detectable fault element identifier matching result.

[0175] In one implementation, step S466 may specifically include the following steps S4661 to S4666: Step S4661: Receive the full content of the logically coherent final fault element identifier matching result, associate it with all propagation scenarios of the corresponding fault type in the calibrated railway fault element identifier library, generate a multi-scenario association description and evolution trajectory of the link, and obtain the inferred link multi-scenario association result.

[0176] The calibrated railway fault element identification library is calibrated and optimized, containing detailed information on various fault types and corresponding conduction scenarios. For each fault element identification and conduction link in the logically coherent final to-be-detected fault element identification matching result, associate all the conduction scenarios of the corresponding fault type in the calibrated railway fault element identification library. For example, for a fault element identification representing "overvoltage fault of traction substation equipment", search in the calibrated railway fault element identification library for all conduction scenarios related to the overvoltage fault of traction substation equipment, such as the catenary fault that may be caused by overvoltage, the impact on the track line, etc. According to the association result, generate the multi-scenario association description and evolution trajectory of the link. The multi-scenario association description illustrates the conduction of the fault under different scenarios. For example, in normal operation scenarios, overload scenarios, harsh environment scenarios, etc., the conduction path and mode of the fault may be different. The evolution trajectory records the chronological order of the development of the fault under these different scenarios. For example, in the normal operation scenario, the overvoltage fault of traction substation equipment may conduct to the catenary after a certain time, resulting in voltage fluctuations of the catenary; while in the overload scenario, the fault may conduct to the catenary faster and the affected range may be wider.

[0177] Step S4662: Embed the multi-scenario association result of the deduction link into the logically coherent final to-be-detected fault element identification matching result, update the scenario association description and timing constraint of each link, and obtain the multi-scenario association matching result set.

[0178] During the embedding process, update the scenario association description and timing constraint of each link. The scenario association description illustrates the conduction of the fault under different scenarios. By adding the information in the multi-scenario association result of the deduction link to the logically coherent final to-be-detected fault element identification matching result, the scenario association description of each link can be made more detailed and accurate. For example, for a conduction link from traction substation equipment to the catenary, update its scenario association description to "In the normal operation scenario, the fault may conduct to the catenary after X minutes; in the overload scenario, the fault may conduct to the catenary after Y minutes (Y < X)". The timing constraint stipulates the time requirements and order of the fault under different scenarios. According to the evolution trajectory in the multi-scenario association result, update the timing constraint of each link. For example, if the conduction speed of the fault increases in a certain scenario, then correspondingly shorten the conduction time constraint of the fault in that scenario.

[0179] Step S4663: For the multi-scenario association matching result set, associate the full-time operation status of the multi-equipment acquisition data of the to-be-detected electrified railway, confirm the global adaptability and evolution trend of the link scenario association, and obtain the global adaptability confirmation result of the matching result.

[0180] During the association process, it is checked whether the scenario association description of each link matches the operational status throughout the entire time period. For example, for a transmission link from traction substation to catenary in a multi-scenario association matching result set, the operational status of traction substation and catenary in different time periods in the multi-equipment data of the electrified railway to be tested is examined to determine whether the transmission of faults in different scenarios matches the actual situation. If, in a certain scenario, the scenario association description of the link indicates that the fault should be transmitted to the catenary, but the actual data collected shows that the catenary is in normal condition during that time period, then it means that the scenario association of that link does not match the actual situation in that scenario.

[0181] Confirm the global adaptability and evolution trend of link scenario associations. Global adaptability refers to the degree to which link scenario associations conform throughout the entire data collection period and under various scenarios, while evolution trend refers to the future development direction of the fault. By analyzing the changing trends of the data collected throughout the entire time period and the link scenario association descriptions, the evolution trend of the fault can be predicted. For example, if the data collected throughout the entire time period shows that the fault in the traction substation has a worsening trend, and the link scenario association description shows that the fault will continue to propagate to the overhead contact line in a certain scenario, then it can be determined that the evolution trend of the fault is towards the overhead contact line.

[0182] Step S4664: Integrate the global adaptation confirmation results into the multi-scene related matching result set, complete the spatiotemporal association description of global adaptation, and obtain the global adaptation matching result set.

[0183] During the integration process, the spatiotemporal correlation description for global adaptation is supplemented. This description, combining spatial location and temporal sequence information, provides a more comprehensive picture of fault propagation within the actual railway system. For example, for a transmission link from traction substation to the overhead contact line, after integrating the matching results and confirming global adaptation, the spatial location information of this link throughout the entire data collection period (such as the specific geographical locations of the traction substation and the overhead contact line) and the specific temporal information of fault propagation under different scenarios (such as the time when the fault begins propagation under normal operating conditions and the time when it propagates to the overhead contact line under overload conditions) can be supplemented.

[0184] By adding information on global adaptability and evolution trends to the original matching result set, the description of each link can be made more accurate and complete. For example, for a transmission link from traction substation to catenary that has been confirmed through global adaptability, its description can be updated to "During the entire data collection period, this link is located between the traction substation (specific location) and the catenary (specific location) of a certain railway section. Under normal operating conditions, the fault begins to be transmitted to the catenary at point e in the morning, and under overload conditions, the fault is fully transmitted to the catenary at point f in the afternoon."

[0185] Step S4665: For the matching result set of the full-domain adaptation, integrate the full content of the final set of fault meta-identifiers to be detected, which includes cross-device linkage logic, to generate a full-domain coverage description and time-series evolution trajectory of the fault association, and obtain the full-domain coverage result of the fault association.

[0186] During the integration process, a comprehensive description of the fault's impact and a temporal evolution trajectory are generated. The comprehensive description illustrates the fault's influence and correlations throughout the entire railway system, including the connections between the fault and various parts of the overhead contact system, track, and traction substation. The temporal evolution trajectory records the development sequence of the fault under different scenarios and at different times. For example, for a fault involving traction substation, overhead contact system, and track, the generated comprehensive description can explain how the fault, starting from the traction substation, affects various parts of the overhead contact system and track through cross-equipment linkage logic, and the degree of impact under different scenarios. The temporal evolution trajectory can record the development time nodes and sequence of the fault under normal operation scenarios, overload scenarios, etc.

[0187] By combining the information from the matching result set of the global adaptation with the original information from the final set of fault identifiers to be detected, which includes cross-device linkage logic, the global coverage description and temporal evolution trajectory of fault associations can be made more accurate and detailed. For example, based on the fault propagation in different scenarios in the matching result set of the global adaptation and the linkage relationship between devices in the final set of fault identifiers to be detected, which includes cross-device linkage logic, a more comprehensive fault association description and temporal evolution trajectory can be generated.

[0188] Step S4666: Integrate the fault association full-domain coverage result with the final set of fault element identifiers to be detected that includes cross-device linkage logic to form the final fault element identifier matching result with full-domain coverage.

[0189] During the integration process, the full-domain coverage description and temporal evolution trajectory from the fault correlation full-domain coverage results are incorporated into the final set of fault element identifiers to be detected, which includes cross-device linkage logic. For each fault element identifier and propagation link, its description is updated to include full-domain coverage information. For example, for a fault element identifier representing "traction substation overvoltage fault," after integration, its description can be updated to "This fault has a wide impact throughout the railway system. Through cross-device linkage logic, it may affect multiple parts of the overhead contact line and track in different scenarios. The specific temporal evolution trajectory is...". By integrating the fault correlation full-domain coverage results with the final set of fault element identifiers to be detected, which includes cross-device linkage logic, fault information can be made more comprehensive and accurate, covering all correlations and developments of the fault across different scenarios, times, and devices.

[0190] Step S500: Based on the final fault element identifier matching result and combined with the fault type hierarchy mapping relationship in the calibrated railway fault element identifier library, generate the final electrified railway fault identification result, which includes the precise fault type, associated equipment, and fault-related collected data dimensions.

[0191] In one implementation, step S500 may specifically include the following steps S510 to S560: Step S510: Receive the full content of the final detected fault element identifier matching result with full coverage, track the starting node and triggering conditions of the cross-dimensional transmission link, locate the initial source and timing start point of the fault, and obtain the fault initial source location result.

[0192] Track the starting node and triggering conditions of cross-dimensional transmission links. A cross-dimensional transmission link describes the propagation path of a fault between different devices (such as the overhead contact system, track line, and traction substation equipment). The starting node is where the fault begins to propagate, and the triggering conditions are the causes or factors that lead to the fault. For example, for a cross-dimensional transmission link from the traction substation equipment to the overhead contact system and then to the track line, determine which specific component of the traction substation equipment (such as the transformer, circuit breaker, etc.) failed as the starting node, and what caused the fault (such as equipment overload, short circuit, etc.).

[0193] Based on the tracking results, the initial source and timing start point of the fault are located. The initial source is the specific location and equipment where the fault first appeared, while the timing start point is the time when the fault began. By analyzing the time information and propagation link information in the final detected fault element identifier matching results of the full-coverage system, the initial source and timing start point of the fault can be determined. For example, if the analysis reveals that a short-circuit fault occurred in the transformer of the traction substation at a specific moment, thereby triggering subsequent cross-dimensional propagation, then the transformer is the initial source of the fault, and that moment is the timing start point.

[0194] Step S520: For the initial node in the fault initial source location result, associate it with the full-time operation status of the multi-equipment data of the electrified railway to be tested, trace back the transmission path and state transition of the fault from the initial to the present, and obtain the fault transmission path tracing result.

[0195] Tracing the transmission path and state transitions of a fault from its initial stage to the present involves analyzing how the fault, based on associated real-time operational data, develops from its initial node, through which devices and nodes, to its current state. For example, if the initial node is a transformer fault in the traction substation, reviewing the real-time operational data reveals that after the fault occurred, the output voltage of the traction substation became abnormal, leading to voltage fluctuations in the overhead contact line. These abnormalities in the contact line then affect the current collection of trains on the track, causing changes in the electrical parameters of the track. Furthermore, it allows determining the state transitions of each node during the fault transmission process, such as from a normal state to a fault warning state, and then to a fault occurrence state.

[0196] Time series analysis and data mining techniques can be used to trace the fault propagation path. Based on the timestamps of equipment operating status data, a time series model of fault propagation is constructed. By analyzing the data change trends and correlations in the model, the fault propagation path and state transition process can be determined.

[0197] Step S530: Based on the transmission path in the fault transmission path backtracking results, delineate all devices, data collection dimensions, and environmental parameters involved in the path to obtain the fault impact range delineation results.

[0198] The equipment involved includes both those directly involved in fault propagation and those indirectly affected. For example, in a fault propagation path, traction substations, overhead contact lines, and related equipment on the track may be directly affected, while equipment such as the train's pantograph may also be indirectly affected by the fault propagation. All of these devices need to be identified.

[0199] Data acquisition dimensions refer to the various data types used to monitor and analyze equipment status during fault propagation. For traction substation equipment, data acquisition dimensions may include voltage, current, power factor, etc.; for overhead contact lines, they may include contact wire temperature, tension, pull-out value, etc.; for track lines, they may include geometric parameters such as track gauge, level, triangular grooves, and vibration frequency, etc.

[0200] Environmental parameters are also important factors affecting fault propagation and equipment operation, and should be clearly defined. Environmental parameters include ambient temperature, humidity, wind speed, and atmospheric pressure. For example, in high-temperature environments, the resistance of the contact wires increases, potentially leading to a rise in the contact wire temperature, which in turn affects fault propagation and equipment operating conditions.

[0201] Step S540: Integrate the initial source location results, fault propagation path tracing results, and fault impact range delineation results to generate a hierarchical description and temporal evolution trajectory of the precise fault type, and obtain the integrated result of the precise fault type.

[0202] Generating hierarchical descriptions of precise fault types involves classifying and describing faults in detail based on the hierarchical mapping relationship of fault types in the calibrated railway fault metadata library. For example, for a fault involving traction substation equipment and the overhead contact line, it is first determined to belong to the general category of electrical faults, then further subdivided into the subcategories of traction substation faults and overhead contact line faults. Finally, based on the specific fault manifestations and transmission patterns, a more precise fault type description is determined, such as "overvoltage fault in traction substation equipment triggers flashover fault in overhead contact line insulators."

[0203] The temporal evolution trajectory records the time sequence and development process of a fault from its initial occurrence to its current state. By combining the temporal starting point in the fault initial source location results and the state transition time in the fault propagation path backtracking results, the temporal evolution trajectory of the fault is constructed. For example, it records when the fault started from the transformer of the traction substation, how long it took to propagate to the overhead contact line, and the state changes of the overhead contact line at different times.

[0204] Step S550: The results of integrating the accurate fault type, the description of the associated equipment, the dimension description of the fault association data, and the spatiotemporal attribute information of the fault propagation path are organized in a structured format to obtain the preliminary fault identification results of the electrified railway.

[0205] Organizing this information in a structured format means arranging and organizing it according to certain rules and formats to facilitate storage, transmission, and analysis. This can be done using database tables, storing the precise fault type integration results, associated device descriptions, fault association data dimension descriptions, and spatiotemporal attribute information of the fault propagation path as different fields.

[0206] Step S560: Integrate all the contents of the preliminary electrified railway fault identification results, supplement the node details and timing constraints of the fault propagation path, and obtain the final electrified railway fault identification results.

[0207] In one implementation, step S560 may specifically include the following steps S561 to S566: Step S561: Receive the full content of the final electrified railway fault identification result, split the fault precise type, associated equipment, fault associated data collection dimensions and transmission path description, record the spatiotemporal association constraints of each part, and obtain the identification result field splitting result.

[0208] The fault is broken down into precise types, and then decomposed into different categories and subcategories according to a hierarchical structure, with a clear description for each level. For example, "overvoltage fault of traction substation equipment causing flashover fault of contact wire insulator" is broken down into "electrical fault - traction substation equipment fault - overvoltage fault" and "electrical fault - contact wire fault - insulator flashover fault".

[0209] Break down the related equipment, list all equipment associated with the fault, and provide a detailed description of each piece of equipment. For example, for faults involving traction substations, overhead contact lines, and track lines, list the specific equipment such as transformers and circuit breakers for traction substations, contact wires and insulators for overhead contact lines, and rails and fasteners for track lines.

[0210] The fault-related data collection dimensions are broken down, and various data types are categorized and organized. For example, data such as voltage, current, and power factor of traction substations are grouped into one category, while data such as contact wire temperature and tension of the overhead contact system are grouped into another. The transmission path description is also broken down, decomposing the fault transmission path according to nodes and sequence, clarifying the location and function of each node. For example, the path of a fault transmission from the traction substation to the overhead contact system and then to the track is broken down into "traction substation node - overhead contact system node - track line node".

[0211] During the breakdown process, the spatiotemporal correlation constraints of each part are recorded. These constraints define the temporal and spatial relationships between different parts. For example, the time required for a fault to propagate from the traction substation to the overhead contact line is recorded, as well as the relative spatial positions of the traction substation and the overhead contact line.

[0212] Step S562: Based on the transmission path description in the identification result field splitting result, generate a structured node map of the transmission link, mark the transmission order, state transition and temporal constraints of the nodes, and obtain the fault transmission link map generation result.

[0213] The transmission path description in the identified results' field splitting details the transmission process of the fault between different devices and nodes. A structured node map of the transmission links is generated based on this description. The transmission order of nodes is marked, clarifying the sequence of fault propagation between nodes. For example, in a transmission path from traction substation to the overhead contact line and then to the track, the traction substation node is marked as the first node, the overhead contact line node as the second node, and the track node as the third node. State transitions are marked, illustrating the state changes of each node during fault transmission. For example, for the traction substation node, the process of its transition from a normal state to a fault state is marked, such as "normal operation - overvoltage warning - overvoltage fault". Timing constraints are marked, specifying the time requirements and sequence of fault transmission between nodes. For example, the time for fault transmission from the traction substation node to the overhead contact line node is marked as 5 minutes, and the time for transmission from the overhead contact line node to the track node is marked as 3 minutes.

[0214] Step S563: Embed the fault propagation link map generation result into the identification result field splitting result, update the structured description and visualization presentation logic of the final electrified railway fault identification result, and obtain the enhanced visualization fault identification result.

[0215] Update the structured description of the final electrified railway fault identification results to include relevant information from the fault propagation path graph. For example, in the fault precision type hierarchical description, add explanations related to nodes and propagation paths in the graph; in the associated equipment description, combine the state transition information of nodes in the graph to describe the equipment fault situation in more detail. Update the visualization logic so that the final electrified railway fault identification results can be displayed in a more intuitive and clear way. A graphical interface tool can be used to integrate the fault propagation path graph with information from other parts to form a complete visualization interface. For example, while displaying the fault precision type hierarchical description in the interface, the fault propagation path can be displayed graphically, and detailed state transition and timing constraint information can be viewed by clicking on nodes.

[0216] Step S564: Based on the visualized and enhanced fault identification results, associate the corresponding fault handling related content in the calibrated railway fault meta-identifier library, generate the fault handling association description and timing suggestions, and obtain the fault handling association generation results.

[0217] The calibrated railway fault identifier library contains detailed information on various fault types and corresponding handling associations. The handling associations include processing measures, maintenance methods, required equipment and personnel, etc., for different fault types. For example, for "traction substation overvoltage fault," the handling associations might include processing measures such as checking the transformer's protection devices, adjusting the output voltage, and replacing damaged components, as well as the required tools and technicians. Based on the association results, a fault handling association description and timing recommendations are generated. The association description details the specific handling measures and steps to be taken for the current fault. For example, for a fault involving traction substation and overhead contact line, the association description could be: "First, investigate the overvoltage fault in the traction substation and check if the transformer's protection devices are functioning correctly; then, inspect the insulators of the overhead contact line for flashover and replace them if necessary; finally, debug and test the entire system to ensure the equipment returns to normal operation." The timing recommendations specify the time requirements and sequence for each handling step.

[0218] Step S565: Embed the fault handling association generation result into the visualized and enhanced fault identification result, complete the spatiotemporal association description of fault handling, and obtain the fault identification result with enhanced handling association.

[0219] During the embedding process, the spatiotemporal correlation description of fault handling is completed. This description clarifies the temporal and spatial requirements and constraints of fault handling. For example, it specifies the exact implementation location of each handling step and the travel time between different locations. For a fault involving traction substations and the overhead contact line, it stipulates that overvoltage fault investigation should be conducted at the location of the traction substation, and insulator inspection and replacement should be performed at a specific location on the overhead contact line. The travel time from the location of the traction substation to the location of the overhead contact line is calculated. Simultaneously, the spatiotemporal correlation description of fault handling is further refined by incorporating the spatiotemporal attribute information of the fault propagation path. For example, based on the temporal and spatial relationship of the fault propagation from the traction substation to the overhead contact line, the temporal sequence of handling steps is rationally arranged to ensure that the handling work can be carried out in a timely and effective manner.

[0220] Step S566: Organize and encapsulate the fault identification results with enhanced handling association according to a standardized output format to form the final outputtable electrified railway fault identification results containing complete fault information and handling suggestions.

[0221] Standardized output formats can be a unified file format, such as XML or JSON, or a specific database table structure. During the processing, the various parts of the fault identification results with enhanced correlation are arranged and organized according to the requirements of the standardized format. The encapsulation process packages the processed information into a complete file or data record. For example, the XML file can be saved as a separate file, or the data can be inserted into a record in a database table.

[0222] Please see Figure 2 , Figure 2This is a schematic diagram of a computer system provided in an embodiment of the present invention. The computer system, for example, is a distributed railway system, and includes at least a processor 101, a communication interface 102, and a memory 103. The processor 101, communication interface 102, and memory 103 can be connected via a bus or other means. The processor 101 (or Central Processing Unit, CPU) is the computing and control core of the computer system, capable of parsing various instructions and processing various data within the computer system. The communication interface 102 may optionally include standard wired interfaces or wireless interfaces (such as Wi-Fi, mobile communication interfaces, etc.), and can be used to send and receive data under the control of the processor 101; the communication interface 102 can also be used for data transmission and interaction within the computer system. The memory 103 is a storage device in the computer system used to store programs and data. It is understood that the memory 103 here can include the computer system's built-in memory, or it can include extended memory supported by the computer system. The memory 103 provides storage space, which stores the computer system's operating system; this invention does not limit the storage space. In one embodiment, the processor 101 executes the artificial intelligence-based electrified railway inspection data identification method provided above in the embodiments of the present invention by running a computer program in the memory 103.

Claims

1. A method for identifying electrified railway inspection data based on artificial intelligence, characterized in that, The method includes: By integrating multi-type scarce labeled fault sample data and retrieving the railway fault meta-identifier library based on railway industry safety technical specifications, and combining multi-type scarce labeled fault sample data with the railway fault meta-identifier library, the basic data for constructing fault meta-identifiers is obtained. Based on the fault meta-identifier, a fault meta-identifier space for railway faults covering all fault dimensions of railway equipment is constructed. The fault meta-identifier extractor is trained and optimized. The training process uses the fault association mapping relationship in the fault meta-identifier space for railway faults as the constraint benchmark and continuously adjusts the internal association logic of the extractor to obtain the final fault meta-identifier extractor adapted to the identification of electrified railway faults. The data collected from multiple devices of the electrified railway to be tested are integrated, and the data collected from multiple devices of the electrified railway to be tested are input into the fault element identifier extractor that is finally adapted to the fault identification of the electrified railway to obtain the final set of fault element identifiers to be tested. The final set of fault meta-identifiers to be detected is mapped to the railway fault-specific fault meta-identifier space. Based on the preset fault generalization content in the railway fault-specific fault meta-identifier space, cross-dimensional combination matching of the fault meta-identifiers to be detected is performed to obtain the final fault meta-identifier matching result. Based on the final fault identifier matching results and combined with the fault type hierarchy mapping relationship in the calibrated railway fault identifier library, the final electrified railway fault identification results are generated, which include the precise fault type, associated equipment, and fault-related collected data dimensions.

2. The method according to claim 1, characterized in that, The process involves integrating multi-type scarce labeled fault sample data and retrieving a railway fault meta-identifier library based on railway industry safety technical specifications. Combining this multi-type scarce labeled fault sample data with the railway fault meta-identifier library yields the basic data for constructing fault meta-identifiers, including: The system receives fault association collection data, corresponding fault type labeling information, and associated equipment operating time and collection location information pushed by the respective storage nodes of the railway catenary, track line, and traction substation in the multi-type scarce labeled fault sample data. The system completes data binding according to the physical connection relationship of the equipment to obtain a physically associated fault sample dataset. For each data point in the physically associated fault sample dataset, the transmission correspondence between the fault association collection data and the fault type labeling information is sorted out, and the node sequence and time difference from the occurrence to the manifestation of the fault are recorded to obtain the pre-constructed fault transmission path result. Receive structured documents based on railway industry safety technical specifications, perform machine-readable conversion according to fault propagation classification rules, embed the association mapping framework of physical equipment connections, and obtain a railway fault meta-identifier library with standardized propagation logic; The pre-constructed fault propagation path results are matched with the association mapping framework of the railway fault meta-identifier library with the standardized propagation logic. The correspondence between the matched links and the multi-type scarce labeled fault sample data is recorded to obtain the fault sample dataset of propagation matching. For the fault sample dataset of the transmission matching, the cross-device transmission links between faults of different equipment categories are completed, and the corresponding association mapping framework of the railway fault meta-identifier library of the transmission logic is updated synchronously to obtain the meta-identifier library of cross-device transmission logic calibration. The fault sample dataset of the conduction matching is filled into the corresponding structured classification framework of the meta-identifier library of the cross-device conduction logic calibration, and the temporal correlation description of the conduction link is completed to obtain the basic data for constructing the fault meta-identifier.

3. The method according to claim 1, characterized in that, The basic data constructed based on fault metadata includes building a railway-specific fault metadata space covering all fault dimensions of railway equipment, including: The system receives the full content of the basic data for constructing the fault element identifier, and then splits it according to the dimensions of railway catenary, track line, and traction substation equipment. The system sorts out the node order and association strength of the fault propagation link within the same dimension to obtain a single-dimensional fault propagation link network. For the single-dimensional fault propagation link network, the propagation link nodes of different device dimensions are connected, the node correspondence and triggering conditions of cross-device fault propagation are sorted out, and cross-dimensional fault propagation association entries are generated to obtain the cross-dimensional propagation logic network. By integrating the single-dimensional fault propagation link network and the cross-dimensional propagation logic network, a multi-dimensional framework covering all equipment fault propagation logic is built. The link nodes in this framework are associated with their corresponding spatiotemporal attributes, thus obtaining the initial railway fault-specific fault element identifier space. The propagation links within the initial railway fault-specific fault element identifier space are dynamically simulated to deduce the propagation path and evolution rate of the fault at different nodes. The changes in the links and the node states after the simulation are recorded to obtain the dynamic simulation results of the propagation links. For the new transmission path in the dynamic simulation result of the transmission link, update the corresponding multi-dimensional framework content of the initial railway fault-specific fault element identifier space, supplement the temporal association logic of the new transmission path, and obtain the self-evolved railway fault-specific fault element identifier space. Traverse all propagation links in the self-evolved railway fault-specific fault element identifier space, confirm the continuity and consistency of the link propagation logic, record the continuous link content and node relationships, and obtain the verified railway fault-specific fault element identifier space.

4. The method according to claim 3, characterized in that, The training and optimization of the fault element identifier extractor involves a training process that uses the fault association mapping relationship within the railway fault-specific fault element identifier space as a constraint benchmark, continuously adjusting the internal association logic of the extractor to obtain a final fault element identifier extractor adapted for electrified railway fault identification, including: The system receives the content of the basic data constructed by the fault element identifier, filters out the fault sample data that can reflect the fault propagation link, and the fault sample data includes node propagation relationship and time sequence information to form a training dataset of propagation logic association. The transmission link logic output by the verified railway fault-specific fault meta-identifier space is received, and the link nodes of the training dataset associated with the transmission logic are embedded into the initial logic structure of the fault meta-identifier extractor to build the initial association network driven by the transmission logic. The training dataset associated with the transmission logic is input into the initial association network driven by the transmission logic to complete the first round of fault element identification. The correspondence between the identification results and the transmission links in the training dataset associated with the transmission logic is recorded simultaneously to obtain the first round of transmission-driven identification results. By comparing the first round of transmission-driven identification results with the transmission link logic of the verified railway fault-specific fault element identifier space, the mismatched transmission nodes and timing deviations in the location results are identified. The logical deviation content and impact range of the nodes are recorded to obtain the transmission logic deviation location results. For the deviation nodes in the propagation logic deviation location results, adjust the corresponding propagation logic of the initial association network driven by the propagation logic, update the spatiotemporal association description of the link nodes, and obtain the association network after propagation logic correction. The training dataset associated with the transmission logic is repeatedly input into the association network after the transmission logic is corrected, and the identification and logic adjustment are repeated until the identification result completely matches the transmission link of the verified railway fault-specific fault element identifier space, thus obtaining the final fault element identifier extractor adapted for electrified railway fault identification.

5. The method according to claim 4, characterized in that, The process continues until the identification result completely matches the transmission link of the verified railway fault-specific fault element identifier space, resulting in the final fault element identifier extractor adapted for electrified railway fault identification, including: Receive the associated network content output by the fault element identifier extractor that is finally adapted to electrified railway fault identification, split all transmission link nodes, record the input-output transmission relationship and timing constraints of each node, and obtain the extraction node splitting result; For each node in the extraction node splitting result, the fault propagation scenario and evolution trajectory of the corresponding node in the training dataset associated with the propagation logic are associated, and the scenario association description and triggering conditions of the node are generated to obtain the propagation node scenario association result. The scenario association results of the transmission nodes are embedded into the association network of the fault element identifier extractor that is finally adapted to electrified railway fault identification. The scenario association logic of each transmission node is updated, and the link transmission relationship after scenario association is recorded synchronously to obtain the scenario association-driven association network. Input the newly added fault sample data into the association network driven by the scenario association, complete the fault element identification, and synchronously record the correspondence between the identification result and the transmission link in the newly added fault sample data to obtain the identification result driven by the transmission of the new data. By comparing the newly added data transmission-driven identification results with the transmission link logic of the verified railway fault-specific fault element identifier space, the transmission scenarios and timing vulnerabilities not covered in the location results are identified, the missing content and impact range of the scenarios are recorded, and the missing transmission scenario location results are obtained. For the missing scenarios in the missing location results of the transmission scenario, the corresponding node logic of the association network driven by the scenario is supplemented, the link transmission relationship and timing constraints are updated, and an adaptively optimized fault element identifier extractor is obtained.

6. The method according to claim 5, characterized in that, The process involves integrating the collected data from multiple electrified railway devices to be tested, inputting this data into a fault identifier extractor specifically designed for electrified railway fault identification, to obtain the final set of fault identifiers to be detected, including: The system receives real-time data from the online monitoring system for overhead contact lines, the vibration acquisition device for track lines, and the operation monitoring device for traction substations, which are part of the multi-equipment data collection data of electrified railways to be tested. It completes real-time binding according to the physical connection relationship of the equipment and synchronously embeds the association description between the current equipment operation status and the acquisition environment to obtain the physically associated dataset to be tested. For the physically associated dataset to be detected, sort it according to the time sequence of the runtime segment, and synchronously associate the running status descriptions and environmental parameters of different devices in the same time period to obtain the time-series associated dataset to be detected. The time-series associated dataset to be detected is input into the adaptively optimized fault meta-identifier extractor to complete real-time fault meta-identifier recognition. The correspondence between the recognition results and the propagation nodes in the time-series associated dataset to be detected is recorded synchronously to obtain a real-time fault meta-identifier set. For the real-time fault meta-identifier set, the transmission link logic associated with the verified railway fault-specific fault meta-identifier space is used to generate the fault candidate transmission link and temporal evolution trajectory corresponding to the current identification result, thereby obtaining the real-time fault candidate transmission link set. For the set of real-time fault candidate propagation links, the real-time data changes and state transitions of each node in the link are tracked, and the propagation state description and timing constraints of the link are updated synchronously to obtain a dynamically updated set of fault candidate propagation links. By integrating the real-time fault metadata set with the dynamically updated fault candidate propagation link set, and completing the spatiotemporal correlation description of the propagation link, the final fault metadata set to be detected is obtained.

7. The method according to claim 6, characterized in that, The process of integrating the real-time fault metadata set with the dynamically updated fault candidate propagation link set, and completing the spatiotemporal correlation description of the propagation links, yields the final set of fault metadata to be detected, including: Receive the full content of the final set of fault identifiers to be detected, split it according to the collection period, and synchronously associate the fault candidate propagation links and time-series evolution trajectories of each period to obtain the fault identifiers and link sets split by period. For the fault identifiers and adjacent time period content of the link set in the time period segmentation, the node correspondence and time sequence connection point of the transmission link are matched to generate cross-time period fault transmission link continuation entries and state transition descriptions to obtain cross-time period transmission link continuation results. The cross-time period propagation link continuation result is embedded into the fault identifier and link set of the time period segmentation, and the fault candidate propagation link and timing constraints of each time period are updated to obtain the cross-time period associated fault identifier and link set. For the fault identifiers and link sets associated across time periods, track the data changes and status trends of link nodes across time periods, and synchronously update the transmission trend description and evolution rate of the links to obtain the cross-time period transmission trend tracking results. The cross-time period transmission trend tracking results are integrated with the cross-time period associated fault identifiers and link sets to update the transmission trend description of the links, thereby obtaining the trend-associated fault identifiers and link sets. By integrating the full content of the fault identifiers and link sets associated with the trend, and supplementing the state transition details of the cross-time period transmission links, the final set of fault meta-identifiers to be detected, which includes the cross-time period transmission trend, is obtained.

8. The method according to claim 7, characterized in that, The process integrates the full content of the fault identifiers and link sets associated with the trend, supplements the state transition details of the cross-time period transmission links, and obtains the final set of fault meta-identifiers to be detected that includes the cross-time period transmission trend, including: Receive the full content of the final set of fault element identifiers to be detected, which includes the cross-time period transmission trend, and complete the splitting according to the dimensions of railway catenary, track line, and traction substation equipment. Simultaneously associate the transmission link and time-series evolution trajectory of each dimension to obtain the fault identifier and link set of dimension splitting. For the different dimensions of the fault identifiers and link sets in the aforementioned dimension splitting, the node correspondence and linkage conditions of the transmission links are matched to generate cross-device fault transmission linkage entries and state transition descriptions, and the cross-device transmission linkage results are obtained. The cross-device transmission and linkage results are embedded into the fault identifier and link set of the dimension split, and the fault candidate transmission links and linkage constraints of each dimension are updated to obtain the cross-device associated fault identifier and link set. For the cross-device associated fault identifiers and link sets, track the cross-device data changes and linkage trends of link nodes, synchronously update the linkage status description and evolution rate of the links, and obtain the cross-device linkage status tracking results. The cross-device linkage status tracking results are integrated into the cross-device associated fault identifier and link set to complete the spatiotemporal association description of the linkage status, thus obtaining the linkage associated fault identifier and link set. The fault identifiers and link sets associated with the linkage are integrated to clarify the cross-device transmission links and their linkage logic, thereby obtaining the final set of fault element identifiers to be detected that includes cross-device linkage logic.

9. The method according to any one of claims 3 to 7, characterized in that, The process involves mapping the final set of fault metadata to be detected to a railway-specific fault metadata space, and performing cross-dimensional combination matching of the fault metadata based on the preset fault generalization content within the railway-specific fault metadata space to obtain the final fault metadata matching result, including: Receive the full content of the final set of fault element identifiers to be detected, which includes cross-device linkage logic, map each fault element identifier to the corresponding transmission link node in the verified railway fault-specific fault element identifier space, record the mapped node position and timing constraints, and obtain the identifier space node mapping result. For each node in the node mapping result of the identifier space, the cross-dimensional association logic of the corresponding transmission link of the verified railway fault-specific fault element identifier space is associated to generate the cross-dimensional transmission link and temporal evolution trajectory corresponding to the identifier, and the cross-dimensional transmission link generation result is obtained. The cross-dimensional transmission link generation result is embedded into the identifier space node mapping result, and the cross-dimensional association description and temporal constraints of each identifier are updated to obtain a set of cross-dimensional associated identifiers and links; For the set of cross-dimensional associated identifiers and links, the probability of fault propagation and the direction of evolution of each node in the link are deduced, and the propagation status and time series prediction after the deduction are recorded simultaneously to obtain the cross-dimensional propagation deduction results. The cross-dimensional transmission inference results are integrated with the cross-dimensional associated identifier and link set to form an identifier and link set that includes the transmission state after inference and the prediction, thus obtaining the inference associated identifier and link set. By integrating the set of identifiers and links associated with the inference with the set of final fault identifiers to be detected that includes cross-device linkage logic, and supplementing the inference details of cross-dimensional transmission links, the matching result of the final fault identifier to be detected that includes cross-dimensional transmission inference is obtained.

10. A computer system, characterized in that, include: A memory, wherein a computer program is stored; A processor is configured to load the computer program to implement the artificial intelligence-based electrified railway inspection data identification method as described in any one of claims 1-9.

Images