Train bus digital twin fault inversion method and device based on layer theory and homology

Abstract

The application discloses a train bus digital twin fault inversion method and device based on layer theory and homology, and belongs to the technical field of rail transit vehicle communication and fault diagnosis. The method constructs a train bus physical topology category, collects data streams, equipment state variables and link state variables to form multi-source time sequence data, and obtains an information state category through TDA continuous homology noise reduction; a consistent criterion is constructed by using a continuous homology feature to constrain an inverse functor to satisfy a coherence axiom, and a digital twin mapping with structure preservation is established; the consistency is verified through a preset rule, the first sheaf cohomology group is calculated when the mapping degenerates into a pre-layer, the generator is extracted and used as a constraint to perform a minimum path backtracking, the fault source is determined and the type and position are output, and the node link weight is closed loop corrected. The application can realize decoupling of train bus coupling faults, filter out transient interference, reduce the influence of alarm storm, and has good positioning efficiency and anti-interference ability, and is suitable for high reliability requirements of train on-board operation and maintenance.

Description

Technical Field

[0001] This application relates to the field of communication and fault diagnosis technology for rail transit vehicles, and in particular to a method and apparatus for digital twin fault inversion of train bus based on layer theory and coherence. Background Technology

[0002] With the rapid development of high-speed railways and urban rail transit, train bus systems (such as Multifunctional Vehicle Bus (MVB) and Train Real-Time Data Network (TRDP)) are the core of data interaction for critical train systems, undertaking communication tasks for subsystems such as traction, braking, and door control. The rail transit sector has stringent requirements for system reliability, availability, maintainability, and safety. According to standards such as EN5012612017, if bus malfunctions or equipment disconnections cannot be quickly located and isolated, it will directly affect train operation safety and efficiency.

[0003] Modern train onboard monitoring and diagnostic hardware is evolving towards higher integration and lighter weight. This hardware typically uses integrated industrial chassis, housing the main control unit, power supply, and signal processing components, and directly connecting to the underlying sensor network and bus nodes via connectors. While this architecture improves system response speed and physical reliability, it also allows multi-source, heterogeneous operational data generated by bus faults to directly enter the main control unit, placing higher demands on the complex logic processing capabilities of fault diagnosis algorithms.

[0004] For vehicle bus fault diagnosis, existing technologies mainly fall into two categories: experience-based single-point threshold monitoring methods and machine learning and data-driven diagnostic methods based on statistical features. In the highly coupled train bus environment, existing technologies have the following significant drawbacks: First, they neglect the spatial structural coupling between system data and physical topology, easily leading to alarm storms. Bus networks have complex topologies; node or link anomalies can propagate along communication links. Traditional methods monitor nodes independently, resulting in a large number of concurrent alarms under coupled fault conditions, making it difficult to decouple and locate the true initial fault source. Second, they have weak resistance to electromagnetic interference and transient noise, leading to a high false alarm rate. During train operation, the bus is susceptible to transient electromagnetic interference, causing delays or packet loss. Traditional threshold methods or data-driven models relying solely on time / frequency domain features struggle to distinguish between interference signals and actual structural anomalies, resulting in false alarms. Furthermore, these models exhibit poor generalization performance under complex operating conditions outside of training samples. Third, existing digital twin models only achieve a shallow binding between physical devices and operational data, failing to establish a stable mapping and verification mechanism between physical entities and digital states, etc. When the bus experiences protocol anomalies, sudden changes in device status, or partial communication failures, the system cannot effectively determine the degree of matching between physical entities and digital states, making it difficult to reliably and accurately distinguish and locate fault sources, resulting in poor fault diagnosis.

[0005] Therefore, there is an urgent need for a train bus fault diagnosis method that can effectively characterize the consistency between physical topology and operational data and accurately locate fault sources under conditions of multi-source heterogeneous data and system physical topology. Summary of the Invention

[0006] To address the technical problems of existing train bus fault diagnosis methods, such as the tendency to generate alarm storms, weak resistance to electromagnetic interference and transient noise, lack of strict structural constraints in digital twin models, and susceptibility to misjudgment of fault location while failing to decouple node and link faults, this invention provides a train bus digital twin fault inversion method based on layer theory and coherence. The technical solution is as follows: On the one hand, a digital twin fault inversion method for train buses based on hierarchy theory and homology is provided, including: S1, Construct the physical twin network diagram of the train vehicle bus system. The physical twin network graph is then abstracted into a physical topology category. , where the set of nodes Corresponding to vehicle-mounted equipment, edge collection The corresponding communication link, the physical topology category The objects within correspond to physical nodes, and the morphisms correspond to physical links; S2, Acquiring data stream from the train bus Equipment state variables and link state variables The multi-source heterogeneous time-series data constitutes an information state category used to characterize the device operating state and link communication state. The information state category The formation process includes dividing the continuous time window The time-series data is mapped to a high-dimensional point cloud set X, and a Vitoris-Lipps complex is constructed through topological data analysis (TDA). And calculate the Betti number of the k-th homology group. Extracting persistent homology features that are stable across scales and filtering out transient noise; S3, Construct a constraint inverse functor based on the continuous homology feature. The mapping's topological consistency criterion, and the coercion of the contravariant functor by the topological consistency criterion. The mapping satisfies the adhesion axiom of layer theory in the physical topological domain. With respect to the aforementioned information state category Establish the inverse variable functional between them Mapping as a digital twin mapping ontology; S4, using train bus operation messages with unified timestamps and error code information to drive the inverse converter online. Mapping, making the information state category Synchronous evolution; S5, the inverse function is verified in real time using preset consistency judgment rules. The local data consistency of the mapping is determined. When the local data consistency is broken, the glue axiom is determined to be invalid, the digital twin mapping is degraded from a layer to a pre-layer and an initial set of anomaly candidates is determined. S6, when the inverse variable functor mapping degenerates into a pre-layer, calculate the first Cech cohomology group for the set of anomalous candidates. When the first Cech cohomology group Calculation results When the non-zero cohomology group is extracted, the generator is used as the fault topology feature; S7. Reverse map the generator back to the physical topology. Using the topological boundary defined by the generator as a constraint, perform the following: The objective function searches for the minimum propagation path of anomalous signals, uniquely determining the minimum topological subset that triggers consistency violations, and outputting the fault type and location (communication fault, equipment fault, or protocol anomaly). The weight coefficients of link ij are, Select variables for the path of link ij. =1 indicates that this link is selected. =0 indicates that no selection is made; S8. Feedback the fault inversion and location results to the physical twin network diagram, correct the reliability weights and health of the corresponding nodes or links, and complete the update of the fault diagnosis knowledge base.

[0007] Optionally, the physical twin network diagram for constructing the train vehicle bus system... In this architecture, a direct communication architecture with fewer intermediate processing nodes is adopted, and the sensor network is directly connected to the main control board. This is used to shorten the topology layers of abnormal signal propagation and reduce the spatial dimension of minimum propagation path search.

[0008] Optionally, the device state variables The link state variable is a variable used to characterize the operating state of the device. The data stream is a variable used to characterize the state of link communication. The information status category includes the size of data packets with a uniform timestamp, the sending frequency, and preset error codes. The local state features are represented as multidimensional tuples: ,in This is a set of exception error codes.

[0009] Optionally, in S2, the process of extracting persistent cohomology features that are stable across scales through topological data analysis (TDA) includes: Adjusting the observation scale Identify the linear, circular, and hole-like topological evolution processes of point cloud assemblies; Record the occurrence and disappearance of topological structures and calculate the lifetime of topological holes. When L It is identified as transient noise and filtered out. The topological structure that exists stably across multiple scales is taken as the true state feature of the system.

[0010] Optionally, the preset consistency judgment rule includes verifying whether the online status of the device matches the message output frequency, and whether the link connectivity status matches the latency and packet loss rate. If they do not match, it is determined that the local data consistency is broken. The inverse variable functor is forced by the topological consistency criterion. In the process of mapping to satisfy the adhesion axiom of layer theory, the physical twin network graph is defined. Partial open coverage For any adjacent subnet and When they intersect, they satisfy the following conditions. ,in, For local subnets, Let j be the set of subnet indices, and j be the index of the adjacent subnet. For subnet Local data section, For subnet Local data section, This indicates the constraints of the data cross-section within the intersection region; The first Cech homology group is calculated from the set of anomalous candidates. ,include: Coverage corresponding to the anomaly candidate set Calculate the first Cech cohomology group By calculating the previous loop Obtain the algebraic residual, where, This represents the residual of the upper cyclic algebra.

[0011] Optionally, the minimum propagation path search aims to minimize the propagation path of the abnormal signal, and performs topology backtracking along the communication links of the physical twin network graph to determine the starting point of the path as the physical fault source node.

[0012] Optionally, the method further includes a fault decoupling determination process, as follows: Fault decoupling is performed based on variable combination states to distinguish whether the fault source is located in the node object or the communication link. , and When the fault source is determined to be located in a node object; in, This refers to the device status variables of the main control unit. The values ​​in parentheses are the status values, with 0 indicating offline and 1 indicating online. For the device state variables of the door control unit, For data stream changes, The link state variables between the main control unit and the door control unit This is the link status variable between the door control unit and the main control unit. The value in parentheses is the status value, where 0 indicates disconnection and 1 indicates connection.

[0013] On the other hand, a train bus digital twin fault inversion device based on layer theory and coherence is provided, including a memory and one or more processors, wherein the memory stores executable code, which, when executed by the processor, implements the method described above.

[0014] This invention discloses a method and apparatus for fault inversion using digital twins of a train bus based on layer theory and cohomology. It constructs a physical topology domain for the train bus, collects data streams, device state variables, and link state variables to form multi-source time-series data, and performs continuous cohomology denoising using TDA to obtain an information state domain. A consistency criterion is constructed using continuous cohomology features to constrain inverse variable functors to satisfy the adhesion axiom, establishing a structure-preserving digital twin mapping. Consistency is verified through preset rules. When the mapping degenerates to a pre-layer, the first Cech cohomology group is calculated, generators are extracted, and minimum path backtracking is performed using these as constraints to determine the fault source and output its type and location. Simultaneously, the node link weights are corrected in a closed loop. This invention can decouple coupled faults in the train bus, filter out transient interference, reduce the impact of alarm storms, and provides good positioning efficiency and anti-interference capabilities, meeting the high reliability requirements of train onboard maintenance. Attached Figure Description

[0015] Figure 1 This diagram illustrates a process flow of a train bus fault inversion method based on digital twins and synchronization according to the present invention. Figure 2 This diagram illustrates a train bus digital twin system architecture. Figure 3 This diagram illustrates a structure-preserving mapping principle based on category hierarchy theory. Figure 4 A schematic diagram illustrating a fault inversion principle based on the Cech cohomology barrier is presented. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0017] In this article, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0018] The physical execution environment of this invention relies on a highly integrated, low-latency train communication bus monitoring system. In a preferred embodiment, the hardware carrier of this system is a standard 3U industrial chassis, which integrates a main control board, a power module, and a multi-channel signal processing card. To achieve real-time communication and a flatter topology mapping, the physical architecture of this system explicitly eliminates the traditional multi-level front-end processor nodes. Sensor networks and bus interfaces distributed throughout the train (such as doors, braking, and traction subsystems) are directly connected to the main control board via high-reliability, heavy-duty connectors (such as the HARTING Han series connectors suitable for the complex vibration environment of rail transit) on the front panel of the chassis. The analysis software engine is directly deployed within the main control board. This streamlined architecture significantly reduces the topological dimension of the subsequent fault algebra inversion space from the source.

[0019] Based on the aforementioned underlying hardware architecture, the analysis software within the main control board periodically or in real-time executes the steps of the train bus fault inversion method based on digital twins and coherence as described in the following embodiments. Example 1

[0020] like Figure 1 The diagram illustrates the overall flowchart of a train bus digital twin fault inversion method based on hierarchical theory and homology. It provides a train bus digital twin fault inversion method based on hierarchical theory and homology, including: S1, Construct the physical twin network diagram of the train vehicle bus system. Furthermore, the physical twin network graph is abstracted into a physical topology category. .

[0021] Among them, the node set Corresponding to vehicle-mounted equipment, edge collection Corresponding communication link, physical topology scope Objects within the physical twin network correspond to physical nodes, and morphisms correspond to physical links. Based on the actual equipment connection relationship of the train vehicle bus, the node set V includes onboard equipment such as door control unit, traction control unit, and main control unit, and the edge set E includes communication links such as MVB and TRDP.

[0022] Objects: The corresponding collection of nodes This refers to the specific physical equipment on the train, such as the insert sliding door control unit in the Fuxing bullet train (denoted as...). ), traction control unit ( ) and main control unit ( ).

[0023] Morphisms: the set of corresponding edges This refers to the MVB or TRDP communication links between devices and their data flow. Due to the elimination of the front-end processor, the composite path of the state (such as...) ) was simplified to direct mapping ( This significantly reduces the computational latency of signal transmission. .

[0024] The physical topology category C is used to structure the physical network. Objects directly correspond to on-board physical nodes, and states directly correspond to the communication links and data transmission relationships between nodes, providing a physical structural foundation for subsequent digital twin mapping.

[0025] In one example, the device connection relationships of the train vehicle bus system are obtained, and a physical twin network diagram of the vehicle bus is constructed. , where the set of nodes Represents in-vehicle equipment, side collection Represents communication links; abstracts network diagrams into physical topology categories. Objects within a scope correspond to physical nodes, and states correspond to physical links.

[0026] S2, Acquiring data stream from the train bus Equipment state variables and link state variables The multi-source heterogeneous time-series data constitutes an information state category used to characterize the device operating state and link communication state. .

[0027] Among them, the information status category The formation process includes dividing the continuous time window The temporal data within is mapped to a high-dimensional point cloud set X, and a Vitoris-Lipps complex is constructed through topological data analysis (TDA). And calculate the Betti number of the k-th homology group. Extract persistent homology features that are stable across scales and filter out transient noise.

[0028] Multi-source heterogeneous time-series data is collected in real time from the train bus, data stream Equipment state variables Link state variables Together, they reflect the bus's operating status; timing data is stitched together and mapped into a high-dimensional point cloud according to time windows; stable topology features are identified through topology data analysis (TDA); transient noise caused by electromagnetic interference and instantaneous jitter is eliminated, thus expanding the information state scope. Only the actual system state is retained to provide a clean data foundation for subsequent mapping.

[0029] In one possible implementation, the main control board captures multi-source heterogeneous data on the bus in real time via a front panel heavy-duty connector, constructing a fusion state space for the digital twin system. Its local state features are represented as multidimensional tuples: .

[0030] in, For data streams (including packet size and send / receive frequency); Specifically, it refers to the device's online / offline status; The specific normal / abnormal connectivity status of the link, i.e., the link status variable; This is a set of exception error codes.

[0031] In one example, an information behavior twin corresponding to the physical twin is created, and data streams are collected. Equipment state variables and link state variables Multi-source heterogeneous time-series data; treating the time-series data as high-dimensional point clouds, applying topological data analysis (TDA) to extract continuous cohomology features at different observation scales to filter out bus transient noise and form a pure information state category. .

[0032] Furthermore, to overcome transient electromagnetic interference generated by trains in diverse environments such as across climates and regions, the system will use continuous time windows... The data within is transformed into a point cloud set in a high-dimensional space. Applying Topological Data Analysis (TDA), given a distance threshold... Constructing the Vitoris-Ripple complex And calculate its first Betty number of homology group (like (Characterizing a topological "hole").

[0033] The system generates persistent coherence barcodes and performs topological denoising rules: if the lifetime of a certain topological hole... If the packet loss is determined to be a transient packet loss caused by occasional electromagnetic noise, it will be mapped to the information state category. The data is forcibly filtered out to ensure the purity of the twin data.

[0034] S3, Constructing a constraint inverse variable functor based on continuous homology features. The topological consistency criterion for mapping, and the coercion of the contravariant functor by the topological consistency criterion. The mapping satisfies the adhesion axiom of layer theory, within the realm of physical topology. With the category of information status Establish inverse variable functors between Mapping is used as the ontology of digital twin mapping.

[0035] Among them, the persistent homology feature is used as a topological invariant to construct a topological consistency criterion, which directly constrains the inverse functor. It forces the physical topology and information state to maintain structural consistency in overlapping regions, satisfying the layer-theoretic adhesion axiom; inverse variable functor As the core mapping of digital twins, it enables the physical entity and digital state to maintain a structural binding, which is different from traditional data mirror twins.

[0036] Specifically, concept art Partial open coverage .

[0037] The system forcibly sets the "glue axiom" as the normal operating condition benchmark for digital twins: for any two adjacent subnets and If their intersection Above, node With nodes The collected local data cross-sections satisfy: If this is the case, the train bus is considered to be operating normally, and the mapping structure between the entity and the twin is consistent. The following will demonstrate... Figure 3 Further explanation.

[0038] like Figure 3 The diagram illustrates a structure-preserving mapping principle based on category hierarchy theory. The left side represents the physical topological category. The right side represents the information status category. The two are connected by the inverse functor. Establish a structure-preserving digital twin mapping relationship.

[0039] In the realm of physical topology In this example, a local open cover is constructed based on the actual physical structure of the train bus system. The example is divided into a local open set U corresponding to the braking subsystem. 制动 Local open set U corresponding to the main control subsystem 主控 Their intersection region is U 制动 ∩U 主控Each local open set consists of node objects and their communication states. Nodes correspond to vehicle-mounted devices, and states correspond to communication links and their data transmission directions. In the diagram, M1, M2, and M3 represent the interaction relationships between different types of data streams or functional modules.

[0040] via the inverse functor The mapping effect maps each object x∈ in the physical topology. The mapping is performed on data sections F(x) within the information state domain, mapping each state vector f:x→y to the corresponding data constraint relation F(f):F(y), thereby achieving reverse dependency modeling from physical structure to information structure. This mapping process is performed within the information state domain. Specifically, this is manifested in the evolution of multidimensional state variables, including: data flow function D(t), device online state function Sv(t), and link state function Le(t), which together constitute the dynamic information expression of the system in the time dimension.

[0041] Furthermore, within the scope of information state In, for each local open set U i Define its local data section s i ∈F(U i This data section represents the state representation formed by the fusion of multi-source heterogeneous data within the corresponding subnet. According to the adhesion axiom of layer theory, when the system is under normal operating conditions, for any two adjacent open sets U... i and U j The data constraints in the intersection region should satisfy the consistency condition: This means that the observation data of different subsystems in the overlapping area remain consistent in the information scope, thereby ensuring the structural integrity of the digital twin mapping.

[0042] When the system malfunctions or malfunctions, such as Figure 3 As shown in the lower half, inconsistencies appear in the local data sections within the intersection region, namely: This indicates that the layer structure induced by the inverse variable functor has been disrupted, the mapping has degenerated from a layer to a pre-layer, and local information cannot be uniformly bonded globally. This inconsistency is the source of the topological barrier subsequently captured by Cech cohomology group calculations, providing a rigorous mathematical criterion for fault inversion.

[0043] thus, Figure 3 It not only depicts the mapping relationship between physical topology and information state, but also clearly describes the role mechanism of data flow, device state and link state in category mapping, as well as the judgment logic of layer structure consistency and failure under normal and fault conditions, providing a unified theoretical basis for consistency verification and fault location in subsequent embodiments.

[0044] S4 uses train bus operation messages with unified timestamps and error code information to drive the inverter functor online. Mapping makes the information state category Synchronous evolution.

[0045] Among them, unified timestamps ensure the timing alignment of messages, statuses, and error codes, and online drivers enable the information status scope. As the physical bus changes in real time, the digital twin and the physical entity are kept synchronized in real time to ensure that subsequent consistency verification is true and effective.

[0046] S5 verifies the inverse functor in real time using preset consistency judgment rules. The local data consistency of the mapping is determined. When the local data consistency is broken, the glue axiom is deemed invalid, the digital twin mapping degenerates from a layer to a pre-layer, and an initial set of anomaly candidates is determined.

[0047] The pre-defined consistency judgment rule is used to determine whether the physical state and digital state match. Once a mismatch occurs, it is determined that local consistency has been broken, the glue axiom has failed, and the digital twin has degenerated from a structurally stable "layer" to a structurally incomplete "pre-layer". The system then delineates the area where the anomaly is located, forming an anomaly candidate set. For example, if the protocol rules expect a non-empty acknowledgment packet but the receiving side remains at zero, or if the device status bit shows online but the message output frequency is abnormal, the digital twin degenerates from a strict "layer" to a "pre-layer", and the system then delineates this degenerated area as an anomaly candidate set.

[0048] Specifically, when the online status of a device node is found to be inconsistent with the message output frequency, or when the link connectivity status is inconsistent with the latency and packet loss rate, the glue axiom is deemed to be invalid, the digital twin degenerates from a layer to a pre-layer, and an initial set of anomaly candidates is generated accordingly.

[0049] S6, when the inverse variable functor map degenerates into a pre-layer, compute the first Cech cohomology group on the anomalous candidate set. When the first Cech upper harmonic group Calculation results When, the generators of non-zero cohomology groups are extracted as fault topological features.

[0050] Among them, the first Cech cophony group Computation is initiated only when the mapping degenerates to a pre-layer to avoid invalid operations; non-zero results indicate the existence of real topological barriers, i.e., real faults rather than disturbances, and the extracted generators can characterize the topological features corresponding to the faults and serve as the basis for fault inversion.

[0051] In one possible implementation, the coverage corresponding to the above-mentioned abnormal candidate set... Calculate the algebraic residual of its first Cech homology group, i.e., the upper cycle: If detected This leads to equivalence classes This mathematically confirms that the physical system has suffered substantial structural damage (transient errors have been filtered out in stage S2). The system then extracts... The non-zero generators are used as high-dimensional fault topology features.

[0052] S7 reverse maps the generator back to the physical topology. With the topological boundaries defined by the generators as constraints, perform the following: The objective function searches for the minimum propagation path of abnormal signals, uniquely determines the minimum topological subset that causes consistency failure, and outputs the fault type and location of communication failure, equipment failure, or protocol anomaly.

[0053] in, The weight coefficients of link ij are, Select variables for the path of link ij. =1 indicates that this link is selected. =0 indicates that no selection is made.

[0054] Among them, the generator acts as a topology boundary constraint, limiting the minimum path search range. By tracing back along the physical topology links, the fault origin can be located, the minimum topology subset that causes the anomaly can be determined, and the fault type and location can be output, which helps to reduce the interference of concurrent alarms on fault location.

[0055] In one possible implementation, taking an insertable sliding door embodiment as an example, the extracted generator is used as a boundary condition within the category. Perform a target optimization search based on minimum propagation path. Reverse mapping identifies the source of the fault.

[0056] Train insertion sliding door control unit ( ) and main control unit ( The decoupling is illustrated using a communication failure between the two sides as an example.

[0057] Normal operating conditions: Periodically send status messages such as door closure and anti-pinch, and the intersection section algebraic residuals. .

[0058] An anomaly occurred: The main control unit suddenly lost door data.

[0059] The system calculates the feature residuals of the intersection region in each dimension. The process includes: Equipment status: (The main controller is offline, but the underlying hardware heartbeat may still be active.)

[0060] Data flow dimension: (Packet size is zeroed, timeout error code is thrown).

[0061] Link connectivity dimension: (Level detection of conduction of underlying HARTING connector and physical wiring harness).

[0062] Algebraic inversion output: As can be seen from the above calculations, the upper loop barrier is entirely concentrated on the object representing the node. and Above, and in the representative state of shooting The upward projection is zero. The inverse mapping uniquely and definitively converges to the node object itself. The system accurately outputs a diagnostic conclusion: the internal core board of the insert sliding door control unit has crashed or the software process has collapsed (equipment failure), ruling out physical communication cable breakage (link failure). This process fundamentally avoids alarm storms caused by coupling effects from surrounding nodes.

[0063] S8 feeds back the fault inversion and location results to the physical twin network graph, corrects the reliability weights and health of the corresponding nodes or links, and completes the update of the fault diagnosis knowledge base.

[0064] Among them, feedback correction is used to update the health status of nodes and links, and store the fault characteristics and location logic in the knowledge base so that subsequent fault inversion has a reference basis, forming a continuously iterative closed-loop diagnostic mechanism.

[0065] After the system outputs the fault location results, within the physical topology scope... The health weight of the door control unit is reduced and corrected. Simultaneously, the TDA point cloud evolution mode and co-homogeneous residual characteristics corresponding to this equipment failure are synchronously stored in the fault diagnosis knowledge base. This closed-loop evolution mechanism unifies the underlying control logic and the top-level safety metrics, ensuring that the intelligent operation and maintenance and lifecycle management of the entire train bus system strictly comply with the stringent specifications regarding reliability, availability, maintainability, and safety of rail transit systems in standards such as EN50126-1-2017.

[0066] In summary, the embodiments of this application achieve decoupling analysis of train bus coupled faults by maintaining the structure mapping between physical topology and information state, continuous cohomology noise reduction, pre-layer degradation triggering cohomology calculation, and generator constraint positioning. This effectively filters out transient interference, reduces the impact of alarm storms, improves the accuracy, real-time performance, and anti-interference capability of fault diagnosis, and adapts to the operation monitoring requirements of train bus. Example 2

[0067] Figure 2This is a diagram of the digital twin system architecture for the train bus of the present invention. The diagram shows the hardware and software collaborative architecture for implementing the method of the present invention from the perspective of system hierarchy. It can be divided into five core layers: physical perception layer, data processing layer, twin mapping layer, fault inversion layer, and closed-loop correction layer.

[0068] The physical perception layer corresponds to the physical twin of the train bus and various onboard sensors and communication nodes, realizing the acquisition of physical device status and link data; the data processing layer completes TDA noise reduction and high-dimensional point cloud transformation of multi-source heterogeneous time-series data; the twin mapping layer establishes the inverse variable functor mapping between the physical topology category and the information state category and executes the adhesion axiom constraint; the fault inversion layer realizes consistency verification, cohomology calculation and fault source location; and the closed-loop correction layer completes fault result feedback and knowledge base update. The layers interact bidirectionally through a unified timestamp data stream. The architecture diagram clearly illustrates the functional division, data interaction relationships and hardware carrier adaptability of each layer, demonstrating the adaptation logic of the method of this invention with the simplified hardware architecture of the train, providing a clear system architecture reference for those skilled in the art to implement the engineering deployment of the method of this invention.

[0069] Based on Example 1, a physical twin network diagram of the train vehicle bus system is constructed. In this architecture, a direct communication architecture with fewer intermediate processing nodes is adopted, and the sensor network is directly connected to the main control board. This is used to shorten the topology layers of abnormal signal propagation and reduce the spatial dimension of minimum propagation path search.

[0070] Among them, the direct communication architecture reduces intermediate forwarding nodes such as front-end processors, enabling sensors to communicate directly with the main control board, reducing signal transmission latency. At the same time, it simplifies the topology hierarchy, making the computational workload of subsequent minimum propagation path search smaller and the positioning efficiency higher. It further simplifies the physical topology structure, reduces the computational complexity of fault inversion, improves the response speed of fault location, and adapts to the high real-time requirements of train onboard systems. Example 3

[0071] Based on Example 1, device state variables Link state variables are variables used to characterize the operating state of devices. Data flow is a variable used to characterize the state of link communication. Includes data packet size with uniform timestamp, transmission frequency, preset error codes, and information status range. The local state features are represented as multidimensional tuples: ,in This is a set of exception error codes.

[0072] Among them, device state variables Link state variables reflect the online, offline, and abnormal operating status of the device. Reflecting communication status such as link connectivity, latency, and packet loss, data flow It provides message-level quantitative features, and the three elements work together to form a complete state description of the train bus. This allows the multi-dimensional state data to be cross-verified, improving the reliability of consistency judgment and providing sufficient data support for the differentiation and location of fault types. Example 4

[0073] Based on Example 1, S2 includes the process of extracting persistent cohomology features that are stable across scales through topological data analysis (TDA), which includes: Adjusting the observation scale Identify the linear, circular, and hole-like topological evolution processes of point cloud assemblies; record the occurrence and disappearance of topological structures and calculate the lifecycle of topological holes. When L The transient noise is identified and filtered out; the topological structure that exists stably across multiple scales is taken as the true state feature of the system.

[0074] Different observation scales are used to capture global and local topological changes in point clouds. Temporarily existing topological structures are judged as invalid features caused by interference, while topological structures that exist stably across scales are judged as features of the true state of the system. This achieves effective filtering of transient noise, thereby distinguishing the topological change features corresponding to transient interference and real faults, reducing the impact of interference data on fault judgment, and improving the purity of data in digital twins. Example 5

[0075] Based on Example 1, a consistency determination rule is preset, including: Verify whether the online status of the device matches the message output frequency, and whether the link connectivity status matches the latency and packet loss rate. If they do not match, it is determined that the local data consistency is broken.

[0076] When the device is online, it should be accompanied by a stable message output frequency. When the link is in normal connectivity, it should be accompanied by low latency and packet loss rate. Through these two specific rules, it is possible to directly determine whether the physical entity status and the digital twin status match. Thus, the judgment rules are clear and executable, avoiding the bias caused by subjective judgment, and making the triggering conditions for pre-layer degradation clearer, more stable and repeatable. Example 6

[0077] Based on Example 1, the minimum propagation path search aims to minimize the propagation path of abnormal signals. It performs topology backtracking along the communication links of the physical twin network graph and determines the starting point of the path as the physical fault source node.

[0078] Among them, the shortest propagation path of the abnormal signal corresponds to the earliest location of the fault. By tracing back along the communication link in reverse topology, the initial location of the fault can be quickly locked. The positioning logic is clear and the calculation process is efficient. It can quickly find the location of the root cause fault from the concurrent abnormal signals of multiple nodes. Example 7

[0079] Based on Example 1, the method further includes a fault decoupling determination process, as follows: Fault decoupling is performed based on variable combination states to distinguish whether the fault source is located in the node object or the communication link. , and When the fault source is determined to be located in a node object. This refers to the device status variables of the main control unit. The values ​​in parentheses are the status values, with 0 indicating offline and 1 indicating online. For the device state variables of the door control unit, For data stream changes, The link state variables between the main control unit and the door control unit This is the link status variable between the door control unit and the main control unit. The value in parentheses is the status value, where 0 indicates disconnection and 1 indicates connection.

[0080] Among them, if the device state variable Abnormal link state variables If the link state variable remains normal, the fault source is determined to be located in the node object; Abnormal device state variables By maintaining normal operation and determining that the fault source is located in the communication link, the fault source can be accurately decoupled. This clearly distinguishes between node object faults and communication link faults, reduces misjudgments caused by coupled faults, and improves the practicality of fault diagnosis results.

[0081] like Figure 4 The diagram illustrates a fault inversion and decoupling determination principle based on the Cech cohomology barrier. This diagram corresponds to the fault decoupling determination process in this embodiment, illustrating how to determine the fault type and locate the fault source based on multi-source state residuals after detecting that the first Cech cohomology group is non-zero.

[0082] When the first Cech homology group corresponding to the abnormal candidate set satisfies When this occurs, it indicates that there is an inescapable topological consistency barrier in the corresponding coverage area, meaning that the physical system has experienced a real structural anomaly. Based on the generator of this non-zero cohomology group, a corresponding algebraic residual representation is constructed, and a decoupling decision matrix is ​​further formed to perform a unified analysis of multidimensional state variables.

[0083] The decoupling determination process is based on multi-source state residuals, including equipment state residuals. Data flow residual ΔD(t) and link state residual ΔL e .

[0084] Among them, when ≠0 and ≠0, at the same time When = 0, the source of the anomaly is determined to be located in the node object, corresponding to a device failure; when When the error is not equal to 0, the source of the anomaly is determined to be in the communication link, corresponding to a link failure; multidimensional residual combination can further distinguish between protocol anomalies or compound failures.

[0085] After determining the fault type, the generators of the non-zero cohomology group are used as topological constraints and inversely mapped to the physical topology. It performs a minimum propagation path search along the communication link to locate the minimum topology subset that causes the consistency failure, thereby uniquely identifying the fault source node or fault link.

[0086] exist Figure 4 In the illustrated embodiment, the door control node V is identified through joint analysis of the device state residuals and data stream residuals. door If an anomaly is detected while the link status remains normal, the fault source can be determined to be the device node itself, and communication link anomalies can be ruled out, thus achieving precise decoupling and location of the fault.

[0087] On the other hand, a train bus digital twin fault inversion device based on layer theory and coherence is provided, including a memory and one or more processors. The memory stores executable code, which, when executed by the processor, implements the method described above.

[0088] The device, deployed in the train's onboard main control unit, can receive multi-source heterogeneous data from the train bus in real time, perform fault inversion logic, output fault diagnosis results, and complete closed-loop updates of the knowledge base, adapting to the train's embedded operating environment. In some implementations, the executable code, after being called by the processor, is used to implement functions such as physical topology construction, time-series data processing, topology consistency determination, cohomology calculation, fault inversion location, and closed-loop correction and update. This transforms the train bus fault inversion method into an engineering-ready hardware device, facilitating deployment and integration into the train's onboard system and meeting the practical application needs of the rail transit field.

[0089] This application is not an improvement on a single module, but rather a closed-loop optimization mechanism formed through multi-module collaborative design at three levels: feature construction, matching reliability judgment, and feature enhancement. This is the overall technical path presented by this application compared to existing technologies.

[0090] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0091] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A digital twin fault inversion method for train buses based on layer theory and homology, characterized in that, include: S1, Construct the physical twin network diagram of the train vehicle bus system. The physical twin network graph is then abstracted into a physical topology category. , where the set of nodes Corresponding to vehicle-mounted equipment, edge collection The corresponding communication link, the physical topology category The objects within correspond to physical nodes, and the morphisms correspond to physical links; S2, Acquiring data stream from the train bus Equipment state variables and link state variables The multi-source heterogeneous time-series data constitutes an information state category used to characterize the device operating state and link communication state. The information state category The formation process includes dividing the continuous time window The time-series data is mapped to a high-dimensional point cloud set X, and a Vitoris-Lipps complex is constructed through topological data analysis (TDA). And calculate the Betti number of the k-th homology group. Extracting persistent homology features that are stable across scales and filtering out transient noise. The length of the time window; S3, Construct a constraint inverse functor based on the continuous homology feature. The mapping's topological consistency criterion, and the coercion of the contravariant functor by the topological consistency criterion. The mapping satisfies the adhesion axiom of layer theory in the physical topological domain. With respect to the aforementioned information state category Establish the inverse variable functional between them Mapping as a digital twin mapping ontology; S4, using train bus operation messages with unified timestamps and error code information to drive the inverse converter online. Mapping, making the information state category Synchronous evolution; S5, the inverse function is verified in real time using preset consistency judgment rules. The local data consistency of the mapping is determined. When the local data consistency is broken, the glue axiom is determined to be invalid, the digital twin mapping is degraded from a layer to a pre-layer and an initial set of anomaly candidates is determined. S6, when the inverse variable functor mapping degenerates into a pre-layer, calculate the first Cech cohomology group for the set of anomalous candidates. When the first Cech cohomology group Calculation results When the non-zero cohomology group is extracted, the generators are used as fault topological features; S7. Reverse map the generator back to the physical topology. Using the topological boundary defined by the generator as a constraint, perform the following: The objective function is to search for the minimum propagation path of abnormal signals, uniquely determine the minimum topological subset that causes consistency violations, and output the fault type and location of communication failures, equipment failures, or protocol anomalies. Here are the weight coefficients for link ij. Select variables for the path of link ij. =1 indicates that this link is selected. =0 indicates that no selection is made; S8. Feedback the fault inversion and location results to the physical twin network diagram, correct the reliability weights and health of the corresponding nodes or links, and complete the update of the fault diagnosis knowledge base.

2. The method according to claim 1, characterized in that, The physical twin network diagram for constructing the train vehicle bus system. In this architecture, a direct communication architecture with fewer intermediate processing nodes is adopted, and the sensor network is directly connected to the main control board. This is used to shorten the topology layers of abnormal signal propagation and reduce the spatial dimension of minimum propagation path search.

3. The method according to claim 1, characterized in that, The device state variables The link state variable is a variable used to characterize the operating state of the device. The data stream is a variable used to characterize the state of link communication. The information status category includes the size of data packets with a uniform timestamp, the sending frequency, and preset error codes. The local state features are represented as multidimensional tuples: ,in This is a set of exception error codes.

4. The method according to claim 1, characterized in that, In S2, the process of extracting persistent cohomological features that are stable across scales through topological data analysis (TDA) includes: Adjusting the observation scale Identify the linear, circular, and hole-like topological evolution processes of point cloud assemblies; Record the occurrence and disappearance of topological structures and calculate the lifetime of topological holes. When L The time is determined to be transient noise and filtered out, where, This is the observation scale at which the topological structure disappears. The observation scale at which the topological structure appears. This is the lifecycle threshold; The topological structure that exists stably across multiple scales is taken as the true state feature of the system.

5. The method according to claim 1, characterized in that, The preset consistency judgment rule includes verifying whether the device online status matches the message output frequency, and whether the link connectivity status matches the latency and packet loss rate. If they do not match, the local data consistency is determined to be broken. The inverse variable functor is forced by the topological consistency criterion. In the process of mapping to satisfy the adhesion axiom of layer theory, the physical twin network graph is defined. Partial open coverage For any adjacent subnet and When they intersect, they satisfy the following conditions. ,in, For local subnets, Let j be the set of subnet indices, and j be the index of the adjacent subnet. For subnet Local data section, For subnet Local data section, This indicates the constraints of the data cross-section within the intersection region; The first Cech homology group is calculated from the set of anomalous candidates. ,include: Coverage corresponding to the anomaly candidate set Calculate the first Cech cohomology group By calculating the previous loop Obtain the algebraic residual, where, This represents the residual of the upper cyclic algebra.

6. The method according to claim 1, characterized in that: The minimum propagation path search aims to minimize the propagation path of the abnormal signal. It performs topology backtracking along the communication links of the physical twin network graph and determines the starting point of the path as the physical fault source node.

7. The method according to claim 1, characterized in that, The method further includes a fault decoupling determination process, which is as follows: Fault decoupling is performed based on variable combination states to distinguish whether the fault source is located in the node object or the communication link. , and When the fault source is determined to be located in a node object; in, This refers to the device status variables of the main control unit. The values ​​in parentheses are the status values, with 0 indicating offline and 1 indicating online. For the device state variables of the door control unit, For data stream changes, The link state variables between the main control unit and the door control unit This is the link status variable between the door control unit and the main control unit. The value in parentheses is the status value, where 0 indicates disconnection and 1 indicates connection.

8. A digital twin fault inversion device for train buses based on layer theory and homology, characterized in that, The method includes a memory and one or more processors, wherein the memory stores executable code that, when executed by the processor, implements the method as described in any one of claims 1 to 7.

Images