A method and system for multi-dimensional time-series equipment anomaly state prediction

Abstract

The application relates to the technical field of abnormal state prediction, and provides a multi-dimensional time sequence equipment abnormal state prediction method and system, which solves the problems of high false alarm rate and inaccurate risk prediction in the prior art. The method comprises the following steps: collecting a multi-node running state data set of a distributed system of equipment and communication transmission layer time sequence offset information; performing dithering and delay correlation processing on the data, generating a communication stability feature set by quantifying the time-space correlation of dithering extreme values and delay fluctuation; obtaining physical layer deformation and temperature drift data of a communication cable, generating a physical layer disturbance feature sequence based on optical signal feature offset analysis and converting the physical layer disturbance feature sequence into a channel abnormal intensity feature; performing cross-level feature coupling on the communication stability feature and the channel abnormal intensity feature, and predicting the communication node failure or data transmission abnormal risk based on the coupling result. The application realizes millisecond-level linkage prediction of equipment-level physical disturbance and system-level communication risk, reduces the fault false alarm rate, and blocks the cascade failure risk.

Description

Technical Field

[0001] The present application relates to the technical field of abnormal state prediction, and in particular to a method and system for predicting abnormal states of multi-dimensional time series equipment. Background Art

[0002] Multi-dimensional timing devices typically rely on the coordinated operation of multiple nodes, requiring millisecond-level data synchronization. Such systems must handle local node congestion caused by high-concurrency command issuance, avoiding global state distortion caused by cross-node communication delay drift or timing misalignment. Traditional single-point monitoring solutions struggle to capture asynchronous multi-source data streams in dynamic environments. There is an urgent need for a technology that can correlate multi-dimensional timing dependencies, enabling rapid tracing and dynamic compensation of synchronization anomalies to ensure physical consistency and operational safety in flight simulations.

[0003] The current targeted solution is a monitoring system based on distributed link tracing. By injecting global tracing identifiers into cross-node communications and collecting span logs, this system constructs a call chain topology map to locate abnormal nodes experiencing latency spikes or packet loss. This system uses latency heatmaps to visualize communication path performance and triggers local resynchronization mechanisms to repair single points of failure. However, its core limitation lies in its reliance on a pre-set static call chain model, which is unable to dynamically model the nonlinear coupling relationships between multi-dimensional device time series data (such as the implicit desynchronization caused by the superposition of sensor noise and network jitter). This significantly increases the missed detection rate of complex abnormal patterns. Summary of the Invention

[0004] The present application provides a method and system for predicting abnormal states of multi-dimensional time series equipment, which is used to solve the problems of high false alarm rate and inaccurate risk prediction caused by monitoring level fragmentation, lack of physical disturbance correlation and cascade blocking failure in the prior art.

[0005] In a first aspect, the present application provides a method for predicting abnormal states of multi-dimensional time series devices, comprising:

[0006] Collect the operating status data set of multi-node interactions in the distributed system corresponding to the multi-dimensional timing device and the timing offset information collected by the communication transmission layer;

[0007] performing jitter and delay correlation processing on the operating status data set and the timing offset information, and generating a communication stability feature set by quantifying the spatiotemporal correlation between jitter extremes and delay fluctuations;

[0008] Acquire physical layer deformation monitoring data and temperature drift monitoring data through an optical monitoring unit embedded in the device communication cable, and analyze the physical layer deformation monitoring data and temperature drift monitoring data based on optical signal characteristic offset to generate a physical layer disturbance feature sequence;

[0009] Converting the physical layer disturbance characteristic sequence into channel abnormality intensity characteristic information;

[0010] The communication stability feature set is cross-level coupled with the channel abnormality intensity feature information, and the communication node failure risk or data transmission abnormality risk of the distributed system corresponding to the multi-dimensional timing device is predicted based on the cross-level feature coupling result.

[0011] Optionally, cross-level feature coupling is performed on the communication stability feature set and the channel abnormality intensity feature information, and based on the cross-level feature coupling result, a communication node failure risk or a data transmission abnormality risk of the distributed system corresponding to the multi-dimensional time series device is predicted, including:

[0012] Assigning a layer attribute identifier to each feature element in the communication stability feature set, wherein the layer attribute identifier includes a software layer identifier and a physical channel layer identifier;

[0013] Assigning a physical channel level identifier to each characteristic unit in the channel abnormal strength characteristic information;

[0014] Based on the topological connection relationship of distributed nodes, a fault propagation path from the software layer identifier to the physical channel layer identifier is constructed;

[0015] Performing a coupling operation along the fault propagation path, matching the communication stability characteristic elements with the channel abnormality strength characteristic units having the same physical channel layer identifier with time-varying impact factors, and calculating the product of the time-varying impact factor and the corresponding characteristic value to generate a node coupling strength index;

[0016] Perform time-decay weighted aggregation on the node coupling strength index of each distributed node to obtain the node vulnerability index;

[0017] According to the comparison result of the node vulnerability index and the preset risk threshold, a communication node failure risk level or a data transmission abnormality risk mark is output.

[0018] Optionally, the communication stability characteristic element and the channel abnormality strength characteristic unit with the same physical channel layer identifier are matched with the time-varying impact factor, and the product of the time-varying impact factor and the corresponding characteristic value is calculated to generate the node coupling strength index, including:

[0019] On the fault propagation path, identifying a communication stability characteristic element and a channel abnormality strength characteristic unit with the same physical channel layer identifier to form a physical layer association pair;

[0020] Obtain a timestamp sequence of characteristic elements and characteristic units in each physical layer association pair;

[0021] Performing a time decay coefficient calculation for each physical layer association pair: extracting the interval duration between adjacent timestamps in the timestamp sequence, calculating a time decay weight value based on the ratio of the interval duration to a preset reference duration; determining a fault impact coefficient based on the historical fault frequency of the node corresponding to the physical channel layer identifier; and multiplying the time decay weight value by the fault impact coefficient to generate a time-varying impact factor;

[0022] Multiplying the value of the communication stability characteristic element by the time-varying impact factor to obtain a weighted communication stability value;

[0023] Multiplying the value of the channel abnormality strength characteristic unit by the time-varying impact factor to obtain a weighted channel abnormality strength value;

[0024] The weighted communication stability value and the weighted channel anomaly strength value are added together to generate a node coupling strength index.

[0025] Optionally, a time-decay weighted aggregation is performed on the node coupling strength index of each distributed node to obtain a node vulnerability index, including:

[0026] Setting an aggregation time window for the current timestamp, where the aggregation time window includes the current time and a preset historical period extending forward;

[0027] Extracting all index values ​​and their corresponding timestamps within the aggregation time window from the time series of the node coupling strength index;

[0028] Perform time decay weight assignment on each extracted index value: calculate the time difference between the timestamp corresponding to the index value and the current time, and determine the time decay weight value based on the ratio of the time difference to the total duration of the aggregation time window;

[0029] Each index value in the window is multiplied by the corresponding time decay weight value to obtain a weighted index value, and all weighted index values ​​are summed to generate a node vulnerability index.

[0030] Optionally, the operation status data set includes dynamic parameters of data packet throughput, statistical parameters of bit error rate, and parameters of transmission delay difference between nodes;

[0031] The performing jitter and delay correlation processing on the operating status data set and the timing offset information, and generating a communication stability feature set by quantifying the spatiotemporal correlation between jitter extreme values ​​and delay fluctuations, includes:

[0032] Identifying a communication jitter extreme value event in the timing offset information, wherein the communication jitter extreme value event is composed of a sequence of continuous offset points exceeding an offset threshold and having a duration greater than an event window width;

[0033] Extracting the delay fluctuation intensity of the transmission delay difference parameter between nodes in the operation status data set, wherein the delay fluctuation intensity is obtained by calculating the absolute value of the difference between the maximum and minimum delays within a preset time window;

[0034] Performing spatiotemporal matching of the communication jitter extreme value event and the delay fluctuation intensity to determine a spatiotemporal matching result, wherein the spatiotemporal matching process includes locating a time window corresponding to the delay fluctuation intensity according to a central timestamp of the communication jitter extreme value event, and determining the shortest path hop count from the communication jitter extreme value event source node to the delay fluctuation node based on the physical connection topology of the distributed nodes;

[0035] Quantifying the spatiotemporal correlation of the spatiotemporal matching result to calculate a time correlation factor and a space correlation factor, wherein the time correlation factor is obtained based on an offset between a center timestamp of a communication jitter extreme event and a center of a delay fluctuation time window, and the space correlation factor is obtained based on a reciprocal of the number of hops of the shortest path;

[0036] generating a jitter-delay correlation strength value according to a product of the time correlation factor and the spatial correlation factor;

[0037] Aggregate the jitter-delay correlation strength values ​​corresponding to all spatiotemporal matching events to generate a communication stability feature set.

[0038] Optionally, the acquiring of physical layer deformation monitoring data and temperature drift monitoring data by an optical monitoring unit embedded in a communication cable of the device, and parsing of the physical layer deformation monitoring data and temperature drift monitoring data based on an optical signal characteristic offset to generate a physical layer disturbance characteristic sequence, includes:

[0039] Capturing the optical fiber reflection spectrum signal embedded in the communication cable through the multi-channel synchronous acquisition interface of the optical monitoring unit;

[0040] Separating the physical layer deformation monitoring data and the temperature drift monitoring data in the reflection spectrum signal, wherein the physical layer deformation monitoring data is composed of the Bragg wavelength shift caused by optical fiber deformation, and the temperature drift monitoring data is composed of the wavelength drift caused by thermal expansion;

[0041] Performing deformation feature decoupling processing on the physical layer deformation monitoring data to extract a deformation feature offset that is positively correlated with the mechanical stress of the communication cable;

[0042] Performing temperature drift trend analysis on the temperature drift monitoring data to generate an ambient temperature compensation amount;

[0043] The deformation feature offset and the ambient temperature compensation are combined into a physical layer disturbance feature sequence, wherein each data unit in the physical layer disturbance feature sequence includes a paired output of the deformation feature offset and the ambient temperature compensation.

[0044] Optionally, converting the physical layer disturbance characteristic sequence into channel abnormality intensity characteristic information includes:

[0045] Dividing the physical layer disturbance feature sequence into a long-period disturbance feature segment and a short-period disturbance feature segment according to a preset time scale, wherein the long-period disturbance feature segment covers a minute-level time window and the short-period disturbance feature segment covers a second-level time window;

[0046] Perform energy accumulation calculation on each long-period disturbance characteristic segment to obtain steady-state energy distribution parameters;

[0047] Perform peak energy capture calculation on each short-period disturbance characteristic segment to obtain transient energy impact parameters;

[0048] The steady-state energy distribution parameters and the transient energy impact parameters are input into the anomaly intensity mapper, so that the baseline offset detection is performed on the steady-state energy distribution parameters through the anomaly intensity mapper, and the steady-state anomaly intensity component is output; the pulse event counting is performed on the transient energy impact parameters, and the transient anomaly intensity component is output, and the steady-state anomaly intensity component and the transient anomaly intensity component are fused to generate channel anomaly intensity feature information.

[0049] In a second aspect, the present application provides a multi-dimensional time series device abnormal state prediction system, comprising:

[0050] The acquisition module is used to collect the operating status data set of multi-node interactions in the distributed system corresponding to the multi-dimensional timing device and the timing offset information collected by the communication transmission layer;

[0051] a generating module, configured to perform jitter and delay correlation processing on the operating status data set and the timing offset information, and generate a communication stability feature set by quantifying the spatiotemporal correlation between the jitter extreme value and the delay fluctuation;

[0052] The generation module is further configured to obtain physical layer deformation monitoring data and temperature drift monitoring data through an optical monitoring unit embedded in a communication cable of the device, and to analyze the physical layer deformation monitoring data and temperature drift monitoring data based on an optical signal characteristic offset to generate a physical layer disturbance characteristic sequence;

[0053] A conversion module, configured to convert the physical layer disturbance characteristic sequence into channel abnormality intensity characteristic information;

[0054] A prediction module is used to perform cross-level feature coupling on the communication stability feature set and the channel abnormality intensity feature information, and predict the communication node failure risk or data transmission abnormality risk of the distributed system corresponding to the multi-dimensional timing device based on the cross-level feature coupling result.

[0055] In a third aspect, the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a multi-dimensional timing device abnormal state prediction method as described in the first aspect above.

[0056] In a fourth aspect, the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements a method for predicting abnormal states of a multi-dimensional time series device as described in the first aspect.

[0057] In the technical solution of this application, intelligent prediction of abnormal equipment status is achieved through cross-level fusion analysis of multi-source heterogeneous data. Among them, the spatiotemporal correlation analysis based on jitter extremes and delay fluctuations accurately quantifies the communication stability characteristics; the optical signal feature analysis technology of the optical monitoring unit effectively captures the microscopic deformation and temperature drift of the physical layer; the cross-level feature coupling mechanism breaks through the limitations of traditional single-dimensional monitoring and realizes dual risk warnings of communication node failure and data transmission anomalies. This method significantly improves the timeliness and accuracy of anomaly prediction through the collaborative analysis of the distributed system operating status and physical layer disturbance characteristics, providing intelligent decision-making support for preventive maintenance of complex equipment systems.

[0058] Beneficial effects of this application:

[0059] This application achieves accurate assessment of distributed system risks through intelligent analysis of hierarchical attribute identification and fault propagation paths. Specifically, a complete fault propagation topology is constructed based on a bidirectional identification system at the software level and the physical channel level; a time-varying impact factor matching mechanism effectively quantifies the dynamic coupling relationship between cross-level features; and a time-attenuated weighted aggregation algorithm accurately reflects the temporal evolution characteristics of node vulnerability. This method breaks through the limitations of traditional static risk assessment and achieves dynamic hierarchical warning of communication node failures and data transmission anomalies, providing a scientific basis for reliability optimization of distributed systems and significantly improving the fault prediction and fault tolerance capabilities of complex equipment networks.

[0060] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0062] Figure 1A flowchart of a method for predicting abnormal states of multi-dimensional time series equipment provided by the present application is shown;

[0063] Figure 2 A scenario diagram showing a method for predicting abnormal states of multi-dimensional time series devices provided by the present application is shown;

[0064] Figure 3 A schematic diagram of the structure of a multi-dimensional time series device abnormal state prediction system provided by the present application is shown;

[0065] Figure 4 A schematic structural diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0066] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0067] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0068] Researchers have discovered a fundamental bottleneck in the collaborative control of multi-dimensional timing devices: While distributed link tracing-based solutions can pinpoint single-point delay spikes, their static topology model and cross-level coupling failures lead to hidden desynchronization and loss of control. Specifically, the combined effects of sensor noise and network jitter increase the missed detection rate of cross-node timing misalignments, and temperature drift-induced communication drift is difficult to map to application-layer state distortion. This contradiction stems from the blind spot in traditional technologies' ability to decouple physical layer disturbances from transport layer data, necessitating the construction of a traceability architecture that integrates multi-source features across domains.

[0069] To address the above challenges, the present invention proposes a method for predicting out-of-step risk based on cross-level feature coupling. Its innovation lies in breaking through the ceiling of static tracking technology through physical-transmission dual-domain monitoring and nonlinear correlation modeling.

[0070] Specifically, the node interaction status data and timing offset are collected, and the communication stability feature set is generated by quantifying the spatiotemporal correlation of jitter extreme value and delay fluctuation. The optical signal characteristics of cable deformation and temperature drift are captured synchronously by the embedded optical unit, and converted into channel abnormality intensity characteristics through offset analysis. An original cross-level feature coupling mechanism dynamically associates the physical layer disturbance characteristics with the communication stability characteristics, and predicts the risk of implicit desynchronization based on spatiotemporal convolution kernel fusion. When optical attenuation events appear continuously in the channel abnormality intensity characteristics and are accompanied by a sudden change in the delay fluctuation slope, the node is judged to have a high risk of failure. The risk value is converted into a clock calibration parameter through a dynamic weighted compensation model, driving the target node to complete timing compensation before the next synchronization cycle. This method breaks through the limitations of traditional links: optical feature offset analysis realizes the millisecond-level mapping of nanosecond physical layer disturbances to millisecond-level transmission layer performance for the first time, solving the problem of implicit desynchronization caused by temperature drift; the cross-level coupling mechanism fuses the physical and transmission dual-domain features through convolution kernels, improving the accuracy of desynchronization risk prediction at the scale of thousands of nodes, forming a closed-loop control chain of "physical monitoring-transmission quantification-risk prediction-dynamic compensation", and ensuring the stability of the flight simulator attitude control error.

[0071] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0072] Figure 1 A flowchart of a method for predicting abnormal states of a multi-dimensional time series device is provided for an embodiment of the present application. Figure 1 As shown, the method includes:

[0073] 101. Collect the operating status data set of the multi-node interaction in the distributed system corresponding to the multi-dimensional time series device and the timing offset information collected by the communication transmission layer.

[0074] In the above scheme, a multidimensional time series device refers to a device capable of collecting multidimensional time series data. This application mainly uses a multidimensional time series device including a flight simulator as an example for explanation. A distributed system refers to a collaborative computing system composed of multiple nodes. Multi-node interaction refers to the data exchange process between nodes in a distributed system. An operating status data set refers to a data set that records the operating status of the system. The communication transport layer refers to the network protocol layer responsible for data transmission. Timing offset information refers to the time synchronization difference data between different nodes.

[0075] In an embodiment of the present application, first, a sensor group (such as temperature, vibration, and current sensors) deployed on each node of the distributed system collects the local operating parameters of the multi-dimensional timing device to form an original operating status data set. This data is transmitted to the central node via the Internet of Things protocol and a local timestamp is attached to mark the collection time. Subsequently, for the timing offset information collection of the communication transmission layer, the system inserts a synchronization signal in the header of the data frame and transmits it to the target node via redundant channels (primary and backup dual channels) concurrently. The receiving end identifies the synchronization signal through the frame synchronization detection module, compares the difference between the sending and receiving timestamps, and calculates the timing offset information caused by network delay or clock asynchrony. If the primary channel data verification fails, the backup channel data is enabled, and the original signal is restored through an error correction algorithm to ensure the integrity of the offset information. Finally, the operating status data set and the timing offset information are integrated: the data aggregation module of the central node associates and stores the two based on a timestamp alignment algorithm (such as sliding window matching). Dynamic time warping technology is used to calibrate the status data misalignment caused by timing offset to form a unified multi-node interactive timing data set for subsequent status analysis.

[0076] In practical applications, multi-dimensional timing devices include flight simulators. The system consists of a distributed system consisting of flight simulation nodes, visual rendering nodes, and motion platform control nodes. Each node collects the operating status data set of the interaction between nodes (including the attitude angular velocity parameters output by the flight simulation node, the frame generation timestamp sequence of the visual rendering node, and the hydraulic response delay parameters of the motion platform control node) through the deployed multi-dimensional timing devices (such as flight parameter sensors, data packet timing recorders, and clock synchronization monitoring units). At the same time, the deviation value between the data packet arrival timestamp and the global reference clock is captured through the UDP protocol transmission link of the communication transport layer as the timing offset information (such as the central exchange jitter of the data packet arrival interval monitored by the machine port); when the system performs aerobatic flight simulation tasks, the multi-dimensional timing equipment synchronously collects multi-node interaction parameters such as the attitude command transmission delay sent by the flight simulation node to the visual rendering node and the hydraulic state response duration fed back by the motion platform control node to the flight simulation node. At the same time, the communication transmission layer records the timing drift generated when the central switch forwards the data packet (such as the deviation between the retention time of the data packet in the switch queue and the preset time window). Finally, it integrates the data to form an operating status data set and timing offset information covering the software layer instruction transmission and the hardware layer device response, providing a complete spatiotemporal data foundation for subsequent jitter and delay correlation processing.

[0077] The overall solution in step 101 above enables full-dimensional data collection of the operating status and communication timing offsets of multiple nodes in a distributed system. By synchronously acquiring device operating status datasets and timing offset information at the communication transport layer, a monitoring data foundation covering the collaborative operation characteristics of software and hardware is established. This technology overcomes the limitations of traditional single-dimensional monitoring, providing complete data support for subsequent anomaly detection, encompassing both temporal dynamics and spatial distribution characteristics, and achieving a three-dimensional perception of the operating status of the distributed system.

[0078] 102. Perform jitter and delay correlation processing on the operating status data set and timing offset information, and generate a communication stability feature set by quantifying the spatiotemporal correlation between jitter extremes and delay fluctuations. The communication stability feature set represents a sensitive indicator of software and hardware collaboration anomalies in a distributed system.

[0079] Optionally, step 102 may specifically include the following steps:

[0080] 1021. Identify a communication jitter extreme value event in the timing offset information, where the communication jitter extreme value event is composed of a sequence of continuous offset points that exceeds an offset threshold and has a duration greater than an event window width.

[0081] 1022. Extract the delay fluctuation intensity of the inter-node transmission delay difference parameter in the operation status data set, where the delay fluctuation intensity is obtained by calculating the absolute value of the difference between the maximum and minimum delay values ​​within a preset time window.

[0082] 1023. Perform spatiotemporal matching of the communication jitter extreme event and the delay fluctuation intensity to determine the spatiotemporal matching result, wherein the spatiotemporal matching process includes locating the time window corresponding to the delay fluctuation intensity according to the central timestamp of the communication jitter extreme event, and determining the shortest path hops from the source node of the communication jitter extreme event to the delay fluctuation node based on the physical connection topology of the distributed nodes.

[0083] 1024. Quantify the spatiotemporal correlation of the spatiotemporal matching results to calculate a time correlation factor and a space correlation factor, wherein the time correlation factor is obtained based on the offset between the center timestamp of the communication jitter extreme event and the center of the delay fluctuation time window, and the space correlation factor is obtained based on the inverse of the number of hops of the shortest path.

[0084] 1025. Generate a jitter-delay correlation strength value according to the product of the time correlation factor and the spatial correlation factor.

[0085] 1026. Aggregate the jitter-delay correlation strength values ​​corresponding to all spatiotemporal matching events to generate a communication stability feature set.

[0086] In the above solution, jitter and delay correlation processing refers to a calculation method for analyzing the relationship between communication jitter and transmission delay, specifically the process of performing spatiotemporal matching between the extreme communication jitter events and the delay fluctuation intensity in step 1023. The communication stability feature set refers to a set of characteristic indicators reflecting system communication stability. The hardware-software collaboration anomaly sensitivity index refers to a parameter reflecting the sensitivity of system hardware-software anomalies. A communication jitter extreme event refers to an anomaly in which the communication timing offset exceeds a threshold. The offset threshold refers to the critical value for determining whether the timing offset is abnormal. The event window refers to the duration of the anomaly event. The continuous offset point sequence refers to a set of temporally continuous offset data points. The transmission delay difference parameter refers to an indicator of the difference in transmission delay between nodes. The delay fluctuation intensity refers to the severity of delay variation. Spatiotemporal matching refers to the analysis process of correlating temporal and spatial features. The central timestamp refers to the temporal midpoint of the anomaly event. The delay fluctuation time window refers to the time range of the delay fluctuation. The physical connection topology refers to the physical connection structure between nodes. The shortest path hop count refers to the minimum number of hops required for communication between nodes. The time correlation factor refers to the correlation degree parameter in the time dimension. The spatial correlation factor is a parameter that measures the degree of correlation in the spatial dimension. The jitter-delay correlation strength value is a value that quantifies the degree of correlation between jitter and delay.

[0087] In the embodiment of the present application, first, through step 1021, the system performs extreme value event detection on the timing offset information. The specific process is as follows: traverse each data point in the timing offset information. When it is detected that the values ​​of multiple consecutive offset points exceed a preset offset threshold (such as a network delay mutation threshold) and the duration is greater than the event window width, the system triggers the event marking protocol and marks the sequence of consecutive offset points as a communication jitter extreme value event. This process is implemented using a sliding window algorithm. The window moves one data point at a time and calculates the continuous exceeding state of the data in the window. Ultimately, it outputs all abnormal event sequences that meet the conditions and their center timestamps (the midpoint of the event time window).

[0088] Subsequently, in step 1022, the system extracts the inter-node transmission delay difference parameter (i.e., the transmission delay values ​​between different nodes within the same time period) from the operational status dataset. Using a rolling time window statistical method (with the window length consistent with the event window width in step 1021), the system calculates the maximum and minimum values ​​of all node delay data within each time window and takes the absolute value of the difference between the two to generate the delay fluctuation intensity. This intensity quantifies the fluctuation amplitude of inter-node transmission delay within the preset time window. For example, if the maximum delay within the window is 120ms and the minimum delay is 20ms, the delay fluctuation intensity is 120ms - 20ms = 100ms, directly reflecting the severity of network status oscillations.

[0089] Next, in step 1023, the system performs spatiotemporal matching of the extreme communication jitter event and the delay fluctuation intensity. In the temporal dimension, the system uses the central timestamp of the extreme communication jitter event as a reference to locate the corresponding time window in the delay fluctuation intensity data and extract the delay fluctuation intensity value within that window. In the spatial dimension, the system uses the physical connection topology of the distributed system (such as the server-switch-terminal link structure) to calculate the shortest path hop count from the source node of the extreme communication jitter event (the event triggering node) to the delay fluctuation node (the node where the delay fluctuation intensity exceeds the limit) using the Dijkstra shortest path algorithm. The final output is a spatiotemporal matching result that includes the time window location result and the hop count data, establishing a physical association between the jitter event and the delay fluctuation.

[0090] Subsequently, in step 1024, the system quantifies the spatiotemporal correlation of the spatiotemporal matching results (including time window positioning data and shortest path hop count). First, in the temporal dimension, the absolute time offset between the timestamp of the center of the communication jitter extreme event and the center of the delay fluctuation time window is calculated. This offset is input into an exponential decay function to generate a temporal correlation factor normalized to the range of 0-1. Smaller offsets indicate a factor value closer to 1, indicating stronger temporal correlation. Simultaneously, in the spatial dimension, the reciprocal of the shortest path hop count is taken to generate a spatial correlation factor. This factor value increases as the number of hops decreases, reflecting the directness of the physical link between nodes and the strength of the spatial correlation. This process converts the physical correlation characteristics of the two dimensions into a computable scalar factor through a spatiotemporal decay model, ensuring that the output factor is positively correlated with the spatiotemporal correlation strength, providing a normalized input for subsequent correlation strength calculations.

[0091] Next, in step 1025, the system multiplies the temporal correlation factor by the spatial correlation factor to generate a jitter-delay correlation strength value. This value comprehensively reflects the degree of coupling between a single extreme jitter event and the corresponding delay fluctuation. For example, a temporal factor of 0.9 (high temporal matching) multiplied by a spatial factor of 0.5 yields a strength of 0.45, indicating a moderate correlation between the event and delay fluctuation. This calculation is performed automatically by a scalar multiplier, and the output value is stored in the correlation event database.

[0092] Finally, through step 1026, the system aggregates the jitter-delay correlation strength values ​​corresponding to all spatiotemporal matching events to form a communication stability feature set: traverse the correlation event database, extract the strength value of each event, and arrange them in chronological order as feature vectors; at the same time, calculate the statistical indicators of the strength values ​​(such as mean, variance, and extreme value distribution), which together constitute a structured data object that describes the sensitivity of distributed system software and hardware collaboration anomalies, which can be directly input into the subsequent stability analysis module.

[0093] In practical applications, in a distributed monitoring system for a flight simulator, the system is composed of a visual system node, a motion system node, and an instructor console node. Each node collects an operating status data set (such as inter-node transmission delay difference parameters) and timing offset information collected by the communication transmission layer (such as the deviation between the packet arrival timestamp and the global clock reference) through a multi-dimensional timing device. When the system detects a rendering delay in a visual system node, it first identifies a sequence of continuous offset points in its timing offset information that exceeds a preset offset threshold and lasts longer than the event window width, and determines it as a communication jitter extreme value event (step 1021). At the same time, the inter-node transmission delay difference parameters associated with the motion system nodes in the operating status data set are extracted, and the absolute value of the difference between the maximum and minimum transmission delays within the preset time window is calculated as the delay fluctuation intensity (step 1022). The system then performs spatiotemporal matching of the extreme communication jitter event and the delay fluctuation intensity: The system locates the time window corresponding to the delay fluctuation intensity using the center timestamp of the extreme communication jitter event. Based on the flight simulator's star-shaped physical connection topology (view node → central switch → motion node), the system calculates the shortest path hop count (2) from the source node of the extreme communication jitter event (view node) to the delay fluctuation node (motion node) (step 1023). The system then quantifies the spatiotemporal correlation of the spatiotemporal matching results: The system calculates the temporal correlation factor based on the offset between the center timestamp of the extreme communication jitter event and the center of the delay fluctuation time window, and the spatial correlation factor based on the inverse of the shortest path hop count. The product of these two factors generates the jitter-delay correlation intensity value (steps 1024-1025). Ultimately, the system aggregates the jitter-delay correlation strength values ​​of all spatiotemporal matching events (such as correlation events between multiple nodes) to form a communication stability feature set (such as feature vectors including "visual command-motion response delay correlation strength" and "control signal-transmission jitter correlation strength"). This set directly characterizes sensitive indicators of distributed system software and hardware coordination anomalies (such as the coupling risk of rendering command delay and hydraulic response jitter), and locates the root cause of visual and motion node coordination anomalies caused by central switch congestion through the spatiotemporal correlation model.

[0094] The overall solution in step 102 above extracts spatiotemporal correlation features between communication jitter and transmission delay. By quantifying the spatiotemporal correlation between jitter extremes and delay fluctuations, an innovative feature set reflecting the stability of distributed system communications is constructed. This technology accurately captures the spatiotemporal propagation characteristics of communication anomalies, converting discrete jitter and delay data into physically meaningful stability indicators, providing quantifiable features for system anomaly diagnosis.

[0095] 103. Obtain physical layer deformation monitoring data and temperature drift monitoring data through an optical monitoring unit embedded in a communication cable of the device, and analyze the physical layer deformation monitoring data and temperature drift monitoring data based on an optical signal characteristic offset to generate a physical layer disturbance characteristic sequence.

[0096] Optionally, step 103 may specifically include the following steps:

[0097] 1031. Capture the optical fiber reflection spectrum signal embedded in the communication cable through the multi-channel synchronous acquisition interface of the optical monitoring unit.

[0098] 1032. Separate the physical layer deformation monitoring data and the temperature drift monitoring data in the reflection spectrum signal, wherein the physical layer deformation monitoring data is composed of the Bragg wavelength offset caused by optical fiber deformation, and the temperature drift monitoring data is composed of the wavelength drift caused by thermal expansion.

[0099] 1033. Perform deformation feature decoupling processing on the physical layer deformation monitoring data to extract a deformation feature offset that is positively correlated with the mechanical stress of the communication cable.

[0100] 1034. Perform temperature drift trend analysis on the temperature drift monitoring data to generate an ambient temperature compensation amount.

[0101] 1035. Combine the deformation feature offset and the ambient temperature compensation into a physical layer disturbance feature sequence, wherein each data unit in the physical layer disturbance feature sequence includes a paired output of the deformation feature offset and the ambient temperature compensation.

[0102] In the above solution, embedded setup refers to a design method that integrates the monitoring unit into the device. Optical monitoring unit refers to a monitoring device based on optical principles. Physical layer deformation monitoring data refers to data reflecting the mechanical deformation of communication cables. Temperature drift monitoring data refers to data reflecting signal drift caused by temperature changes. Optical signal characteristic offset refers to abnormal changes in optical signal parameters. Physical layer disturbance characteristic sequence refers to a data sequence recording abnormal physical layer characteristics. Multi-channel synchronous acquisition interface refers to an interface that simultaneously acquires multiple signals. Fiber reflection spectrum signal refers to the spectral characteristic signal reflected by an optical fiber. Bragg wavelength offset refers to the change in the wavelength reflected by a fiber Bragg grating. Wavelength drift caused by thermal expansion refers to the change in optical wavelength caused by temperature changes. Deformation characteristic decoupling processing refers to an analytical method for separating deformation characteristics. Mechanical stress positive correlation refers to the positive relationship between deformation characteristics and mechanical stress. Deformation characteristic offset refers to the numerical value that quantifies deformation characteristics. Temperature drift trend analysis refers to a method for analyzing temperature change trends. Ambient temperature compensation refers to a correction value used to offset the effects of temperature. Data unit refers to the basic data element that constitutes a sequence. Paired output refers to the output of a combination of associated data.

[0103] In this embodiment of the present application, the system first invokes the multi-channel synchronous acquisition interface of the optical monitoring unit in step 1031 to capture signals from the optical fiber embedded in the communication cable. This process utilizes fiber Bragg grating (FBG) sensing technology, using a laser transmitter to transmit a light signal of a specific wavelength to the optical fiber and simultaneously receive its reflected spectrum signal. The multi-channel interface ensures simultaneous acquisition of spectral data from multiple monitoring points, avoiding time skew. The reflected spectrum signal contains wavelength peak information, and its positional changes directly reflect the physical effects or temperature changes experienced by the optical fiber, providing the raw input for subsequent separation of deformation and temperature data.

[0104] Next, in step 1032, the system performs signal separation on the reflected spectrum signal. Based on the principle of Bragg wavelength shift, physical layer deformation monitoring data represents the Bragg wavelength shift caused by mechanical fiber deformation (such as squeezing and bending), while temperature drift monitoring data represents the wavelength shift caused by thermal expansion. The system uses a spectral feature decomposition algorithm to identify the wavelength characteristic differences between the two types of data. For example, deformation-induced shifts appear as sudden spikes, while temperature drifts show a gentle trend line. The separated data are output as independent deformation and temperature datasets, respectively, providing a foundation for subsequent feature decoupling and compensation.

[0105] Next, in step 1033, the system performs deformation feature decoupling on the physical layer deformation monitoring data. A wavelet transform algorithm is used to decompose the high-frequency component (corresponding to transient mechanical stress) and the low-frequency component (corresponding to sustained load) in the original wavelength offset. This extracts the deformation feature offset that positively correlates with the mechanical stress of the communication cable. This is the signal that reflects the actual physical deformation of the cable after filtering out vibration noise. For example, the high-frequency pulses generated by external pulling on the cable are identified as valid deformation features, while ambient micro-vibrations are suppressed. This process ensures that the deformation feature offset retains only mechanical stress-sensitive signals.

[0106] Next, in step 1034, the system performs temperature drift trend analysis on the temperature drift monitoring data. Using a time series smoothing algorithm (such as an exponentially weighted moving average), a long-term trend line of wavelength drift is fitted to eliminate short-term fluctuation noise. Based on the fitted curve, an ambient temperature compensation value is generated. This value represents the equivalent interference value of the current temperature drift on deformation monitoring. For example, as the ambient temperature continues to rise, the compensation value increases positively to offset the spurious deformation contribution of temperature to wavelength shift. This step provides a dynamic calibration benchmark for the deformation data.

[0107] Finally, in step 1035, the system aligns the deformation feature offset and the ambient temperature compensation by timestamp, combining them into a physical layer disturbance feature sequence. Each data unit contains a pair of values: the deformation feature offset (reflecting pure mechanical stress deformation) and the ambient temperature compensation (reflecting the temperature disturbance value). This sequence is stored as a time-indexed array using a structured data encapsulation protocol and can be directly input into a fault prediction model to correlate physical layer disturbances with communication stability anomalies.

[0108] In practical applications, in a distributed monitoring scenario of a high-voltage transmission network, the system deploys an optical monitoring unit between transmission tower nodes. The unit obtains physical layer deformation monitoring data and temperature drift monitoring data through an optical fiber sensor network embedded in the outer sheath of the transmission cable. Specifically, the optical fiber reflection spectrum signal embedded in the cable is captured through the multi-channel synchronous acquisition interface of the optical monitoring unit (step 1031). When the transmission tower node generates mechanical vibration due to strong wind or ice covering, the reflection spectrum signal contains mixed characteristics of the cable physical deformation and temperature change. The system first separates the physical layer deformation monitoring data (composed of the Bragg wavelength offset caused by local bending of the optical fiber) and the temperature drift monitoring data (composed of the thermal expansion wavelength drift caused by solar radiation or ambient temperature change) in the reflection spectrum signal (step 1032). Then, deformation feature decoupling processing is performed on the physical layer deformation monitoring data. After removing the temperature interference component, the deformation feature offset (such as the Bragg wavelength offset) positively correlated with the mechanical stress of the cable is extracted. The high-frequency fluctuation component in the wavelength offset corresponds to a sudden change in cable tension) (step 1033); at the same time, temperature drift trend analysis is performed on the temperature drift monitoring data, and the low-frequency trend of the ambient temperature change is fitted through a sliding window to generate an ambient temperature compensation (step 1034); finally, the deformation feature offset and the ambient temperature compensation are combined into a physical layer disturbance feature sequence (each data unit contains a paired output of the deformation feature offset and the ambient temperature compensation), which directly represents the micro-strain accumulation state of the transmission cable (for example, a continuous increase in the deformation feature offset indicates the risk of cable metal fatigue); when the system detects that the deformation feature offset in the physical layer disturbance feature sequence of a section of cable exceeds a safety threshold, combined with the operating status data set of the multi-dimensional time series device (such as insulator leakage current, metal vibration spectrum), the spatial correlation between the transmission cable deformation and electrical parameters is modeled through a graph convolutional network (for example, the increased cable deformation leads to distortion of the electric field distribution), and potential abnormal conditions such as insulator flashover or cable breakage are predicted.

[0109] The overall solution in step 103 above enables intelligent optical monitoring of communication physical layer disturbance signatures. An embedded optical monitoring unit acquires fiber deformation and temperature drift data, and generates a physical layer disturbance signature sequence based on optical signal feature analysis. This technology overcomes the limitations of traditional electrical signal monitoring, enabling non-invasive and precise measurement of cable mechanical stress and ambient temperature changes, providing direct physical layer observation data for communication channel anomaly detection.

[0110] 104. Convert the physical layer disturbance characteristic sequence into channel abnormality intensity characteristic information.

[0111] Optionally, step 104 may specifically include the following steps:

[0112] 1041. Divide the physical layer disturbance feature sequence into a long-period disturbance feature segment and a short-period disturbance feature segment according to a preset time scale, wherein the long-period disturbance feature segment covers a minute-level time window, and the short-period disturbance feature segment covers a second-level time window.

[0113] 1042. Perform energy accumulation calculation on each long-period disturbance characteristic segment to obtain steady-state energy distribution parameters.

[0114] 1043. Perform peak energy capture calculation on each short-period disturbance characteristic segment to obtain transient energy impact parameters.

[0115] 1044. Input the steady-state energy distribution parameter and the transient energy impact parameter into an abnormal intensity mapper, so as to perform baseline offset detection on the steady-state energy distribution parameter through the abnormal intensity mapper and output a steady-state abnormal intensity component.

[0116] 1045. Count pulse events of the transient energy impact parameter, output a transient abnormal intensity component, and fuse the steady-state abnormal intensity component with the transient abnormal intensity component to generate channel abnormal intensity feature information.

[0117] In the above scheme, channel anomaly intensity feature information refers to characteristic data reflecting the degree of anomaly in the communication channel. Long-period disturbance feature segments refer to disturbance data segments on a minute-scale. Short-period disturbance feature segments refer to disturbance data segments on a second-scale. Steady-state energy distribution parameters refer to energy distribution characteristics on a long timescale. Transient energy impact parameters refer to energy impact characteristics on a short timescale. Anomaly intensity mapper refers to a computational module that converts energy features into anomaly intensity. Baseline shift detection refers to an analytical method for detecting energy baseline shift. Steady-state anomaly intensity component refers to a quantitative value of the degree of steady-state energy anomaly. Pulse event counting refers to a method for counting transient pulse events. Transient anomaly intensity component refers to a quantitative value of the degree of transient energy anomaly. Time window refers to the time range of data analysis. Energy accumulation calculation refers to a computational method for integrating and summing energy. Peak energy capture calculation refers to a computational method for extracting energy peaks. Fusion refers to the process of integrating different features.

[0118] In this embodiment of the present application, first, through step 1041, the system divides the physical layer disturbance feature sequence (including paired data of deformation feature offsets and ambient temperature compensation) into time periods according to a preset time scale. Using a time window segmentation algorithm, continuous data covering a minute-level time range is labeled as long-period disturbance feature segments for analyzing the cumulative effects of slow changes; while segments covering a second-level time range are labeled as short-period disturbance feature segments for capturing transient mutation events. This process is implemented through timestamp alignment and sliding window matching, ensuring that each data point belongs to only one feature segment, providing a time period basis for subsequent differentiated energy analysis.

[0119] Then, in step 1042, the system performs an energy accumulation calculation for each long-period disturbance signature segment. An integrator is used to continuously accumulate the offsets of all deformation features within the segment over time, dividing them by the time period to generate the mean energy per unit time. The fluctuation range of the ambient temperature compensation is also calculated, and the two are combined to output a steady-state energy distribution parameter. This parameter reflects the balance between energy absorption and release in a device under continuous operation, such as the deformation energy accumulation trend of a cable under long-term mechanical stress, providing a quantitative basis for baseline stability assessment.

[0120] Next, in step 1043, the system performs a peak energy capture calculation for each short-period disturbance signature segment. Using a local extremum detection algorithm, the system scans the deformation feature offset within the segment, identifying pulse signals that exceed a neighborhood threshold and have a duration less than a preset pulse width. The sum of the squared amplitudes and the frequency of occurrence of these pulses are calculated to generate a transient energy impact parameter. This parameter quantifies the intensity of sudden physical disturbances (such as transient deformation caused by external impacts) and accurately characterizes the energy impact effect of transient events by suppressing low-frequency interference from temperature compensation.

[0121] Then, in step 1044, the system inputs the steady-state energy distribution parameters and transient energy impact parameters into the anomaly intensity mapper. This mapper has a built-in baseline drift detection module. By comparing the current steady-state energy distribution parameters with the historical sliding window average, it calculates their relative offset. If the offset exceeds an adaptive threshold, it triggers the output of the steady-state anomaly intensity component (e.g., cable structural fatigue caused by continuous deformation). This process uses an exponentially weighted moving average algorithm to dynamically update the baseline reference value to avoid misjudgments caused by fixed thresholds.

[0122] Finally, through step 1045, the system processes the transient energy impact parameters in parallel in the anomaly intensity mapper: a pulse event counter is used to count the number of exceeding limit pulses per unit time, and the transient anomaly intensity component (such as single impact event intensity × frequency) is generated by combining the pulse amplitude weighting; then the steady-state anomaly intensity component and the transient anomaly intensity component are combined into channel anomaly intensity feature information through a linear weighted fusion device, where the steady-state component weight reflects the equipment aging sensitivity, and the transient component weight reflects the sudden risk level. The fusion result directly represents the comprehensive intensity index of the physical layer channel anomaly.

[0123] In actual applications, in a three-phase power line communication scenario in a factory workshop, the system continuously collects physical layer disturbance feature sequences (including paired outputs of cable deformation feature offsets and ambient temperature compensation) through optical monitoring units deployed on power cables. To analyze the risk of communication channel abnormalities, the physical layer disturbance feature sequences are first divided into long-period disturbance feature segments and short-period disturbance feature segments according to a preset time scale (step 1041). Energy accumulation calculations are performed on each long-period disturbance feature segment (such as the square value of the integrated deformation feature offset) to generate steady-state energy distribution parameters reflecting the long-term deformation energy accumulation trend of the cable (step 1042). At the same time, peak energy capture calculations are performed on each short-period disturbance feature segment (such as extracting the local maximum of the temperature drift monitoring data) to obtain transient energy shock parameters that characterize sudden temperature shocks (step 1043). The system then inputs the steady-state energy distribution parameters and transient energy impact parameters into an anomaly intensity mapper. The mapper first performs a baseline offset check on the steady-state energy distribution parameters (e.g., comparing them to the energy distribution baseline under historical normal operating conditions). If the current energy distribution continuously deviates from the baseline, it outputs a steady-state anomaly intensity component (e.g., the cumulative risk of cable metal fatigue). Simultaneously, it counts pulse events on the transient energy impact parameters (e.g., counting the number of temperature shocks exceeding a safety threshold within a minute window) and outputs a transient anomaly intensity component (e.g., the risk of transient overheating due to cooling failure). Finally, the steady-state anomaly intensity component and the transient anomaly intensity component are integrated (e.g., weighted superposition of the two components) to generate channel anomaly intensity signature information that comprehensively characterizes the degree of channel degradation (steps 1044-1045). If this signature information continues to increase, the system triggers an alert and triggers the PLC controller to switch to a backup line to prevent communication interruption due to physical cable damage.

[0124] The overall solution in step 104 above achieves intelligent conversion of physical layer disturbances to channel anomaly intensity. Through multi-timescale analysis and energy feature extraction, raw physical layer monitoring data is converted into intensity signature information with clear anomaly indicative significance. This technology innovatively integrates the dual characteristics of steady-state energy distribution and transient energy impact, achieving a quantitative mapping from physical disturbances to communication anomaly risks, providing key feature input for cross-layer fault analysis.

[0125] 105. Perform cross-level feature coupling on the communication stability feature set and the channel abnormality intensity feature information, and predict the communication node failure risk or data transmission abnormality risk of the distributed system corresponding to the multi-dimensional timing device based on the cross-level feature coupling result.

[0126] Optionally, step 105 may specifically include the following steps:

[0127] 1051. Assign a layer attribute identifier to each feature element in the communication stability feature set, where the layer attribute identifier includes a software layer identifier and a physical channel layer identifier.

[0128] 1052. Allocate a physical channel layer identifier to each characteristic unit in the channel abnormality intensity characteristic information.

[0129] 1053. Construct a fault propagation path from the software layer identifier to the physical channel layer identifier based on the topological connection relationship of the distributed nodes.

[0130] 1054. Perform a coupling operation along the fault propagation path, match the communication stability characteristic elements with the channel abnormality strength characteristic units with the same physical channel layer identifier based on time-varying impact factors, and calculate the product of the time-varying impact factor and the corresponding characteristic value to generate a node coupling strength index.

[0131] Among them, step 1054 may specifically include the following processes: on the fault propagation path, identifying the communication stability characteristic elements and channel abnormality strength characteristic units with the same physical channel layer identifier to form a physical layer association pair; obtaining the timestamp sequence of the characteristic elements and characteristic units in each physical layer association pair; performing time attenuation coefficient calculation on each physical layer association pair: extracting the interval duration of adjacent timestamps in the timestamp sequence, and calculating the time attenuation weight value according to the ratio of the interval duration to the preset reference duration; determining the fault impact coefficient based on the historical fault frequency of the node corresponding to the physical channel layer identifier; multiplying the time attenuation weight value by the fault impact coefficient to generate a time-varying impact factor; multiplying the value of the communication stability characteristic element by the time-varying impact factor to obtain a weighted communication stability value; multiplying the value of the channel abnormality strength characteristic unit by the time-varying impact factor to obtain a weighted channel abnormality strength value; adding the weighted communication stability value to the weighted channel abnormality strength value to generate a node coupling strength index.

[0132] 1055. Perform time-decay weighted aggregation on the node coupling strength index of each distributed node to obtain the node vulnerability index.

[0133] Among them, step 1054 may specifically include the following processes: setting an aggregation time window for the current timestamp, the aggregation time window includes the current time and a preset historical period extending forward; extracting all index values ​​and their corresponding timestamps within the aggregation time window from the time series of the node coupling strength index; performing time decay weight allocation on each extracted index value: calculating the time difference between the timestamp corresponding to the index value and the current time, and determining the time decay weight value based on the ratio of the time difference to the total duration of the aggregation time window; multiplying each index value in the window by the corresponding time decay weight value to obtain a weighted index value, and summing all weighted index values ​​to generate a node vulnerability index.

[0134] 1056. Output a communication node failure risk level or a data transmission abnormality risk mark based on a comparison result of the node vulnerability index and a preset risk threshold.

[0135] In the above solution, cross-level feature coupling refers to a method for correlating and analyzing features at different levels. Communication node failure risk refers to the probability assessment of a node failure. Data transmission anomaly risk refers to the probability assessment of anomalies occurring during data transmission. Level attribute identifiers refer to the labeling information that distinguishes feature levels. Software level identifiers refer to feature level identifiers related to software. Physical channel level identifiers refer to feature level identifiers related to physical channels. Fault propagation paths refer to the paths along which faults propagate within a system. Time-varying impact factors refer to influence factors that vary over time. Node coupling strength index refers to a quantitative indicator reflecting node coupling strength. Time decay coefficient refers to a weight coefficient that decays over time. Fault impact coefficient refers to a parameter reflecting the impact of a fault. Weighted communication stability value refers to a weighted stability value. Weighted channel anomaly intensity value refers to a weighted anomaly intensity value. Node vulnerability index refers to a quantitative indicator reflecting the vulnerability of a node. Aggregation time window refers to the time range for data aggregation. Time decay weight value refers to a weight parameter that decreases over time. Risk threshold refers to the critical value for determining risk level. Risk level refers to a categorized identifier for risk level. Anomaly risk tag refers to information identifying anomaly risk. A timestamp sequence refers to a collection of time stamps arranged in chronological order. A physical layer association pair refers to a combination of features with the same physical channel layer identifier. Historical failure frequency refers to the frequency of node failures in the past. The reference duration refers to the reference time used to calculate the attenuation weight. The weighted exponent value refers to the exponent value adjusted for time attenuation.

[0136] In this embodiment of the present application, first, through step 1051, the system assigns a hierarchical attribute identifier to each feature element in the communication stability feature set. Based on the feature's data source attributes, features reflecting software operating status (such as thread scheduling delay and protocol stack error rate) are marked as software-level identifiers, and features associated with hardware transmission quality (such as optical signal attenuation and cable deformation) are marked as physical channel-level identifiers. This process is automatically completed by a feature tracing classifier, which binds the identifiers based on the data acquisition module type used when the feature was generated, establishes a mapping relationship between the feature and the system level, and provides a classification basis for subsequent cross-level associations.

[0137] Subsequently, in step 1052, the system assigns a physical channel-level identifier to each feature element in the channel anomaly intensity signature information. Because this feature information is directly derived from physical-layer monitoring (e.g., cable temperature drift and optical strain energy), a unified identifier binding protocol is used to classify all feature elements at the physical channel level. This operation is achieved by verifying the device location encoding and sensor type in the feature metadata, ensuring a strict correspondence between channel anomaly data and specific hardware nodes, thereby forming a unified identification system for physical-layer features.

[0138] Next, in step 1053, the system constructs a fault propagation path from the software-level identifier to the physical channel-level identifier based on the topological connectivity of the distributed nodes. This involves invoking a pre-set node topology database, starting with the software service deployment node and traversing the connected switches, routers, and physical transmission media along the data flow direction, all the way to the terminal hardware node. Using a directed graph path generation algorithm, a link mapping is established between the software service node and the dependent hardware device nodes, outputting a set of propagation paths that include the number of path hops and transmission media type, such as a physical dependency chain of "web service node → fiber switch → cabinet cable."

[0139] Subsequently, in step 1054, the system performs a coupling operation along the fault propagation path: Communication stability feature elements and channel anomaly intensity feature units with the same physical channel layer identifier are identified along the path to form physical layer association pairs (e.g., jitter features and cable temperature features corresponding to a fiber optic switch). The timestamp sequences of the two are aligned using a time window sliding calibration algorithm to ensure matching of data overlap periods. A time decay weight is calculated for each association pair: the interval between adjacent timestamps is extracted and divided by a preset reference duration to generate a proportional weight (the smaller the interval, the higher the weight). Simultaneously, the historical fault frequency of the node corresponding to the physical channel layer identifier is queried to determine the fault impact coefficient (the higher the fault frequency, the greater the coefficient). The time decay weight is multiplied by the fault impact coefficient to generate a time-varying impact factor, dynamically amplifying the characteristic contribution of high-frequency fault nodes. Finally, the value of the communication stability feature element is multiplied by the time-varying impact factor to generate a weighted communication stability value, and the value of the channel anomaly intensity feature unit is multiplied by the time-varying impact factor to generate a weighted channel anomaly intensity value. These two values ​​are summed to produce a node coupling strength index, which quantifies the coupling risk intensity of single-node software and hardware anomalies.

[0140] Then, in step 1055, the system performs a time-decay weighted aggregation on each node's node coupling strength index: Starting from the current time, an aggregation time window is defined forward (e.g., the last hour), and all indices and their timestamps within the window are captured. The time difference between each index's corresponding timestamp and the current time is calculated and divided by the total duration of the aggregation time window to generate a proportional time-decay weight (the smaller the time difference, the higher the weight). Each index within the window is multiplied by the corresponding weight, and the sum is calculated to generate a node vulnerability index. This index emphasizes the impact of recent anomalies and weakens historical data.

[0141] Finally, through step 1056, the node vulnerability index is input into the node risk classifier; the node risk classifier performs the following operations: when the node vulnerability index is less than the first threshold, a normal state mark is output; when the node vulnerability index is greater than or equal to the first threshold and less than the second threshold, a node warning mark is output; when the node vulnerability index is greater than or equal to the second threshold, a node failure mark is output; a vulnerability difference analysis is performed on node pairs with direct data transmission: the difference in node vulnerability index between the source node and the destination node is calculated; when the difference is greater than the path abnormality threshold, a data transmission abnormal path mark including the source node identifier and the destination node identifier is generated; the output mark of the node risk classifier and the data transmission abnormal path mark are combined as the final risk prediction result.

[0142] In actual applications, in the distributed monitoring scenario of smart grid substations, the system deploys a distributed system consisting of relay protection nodes, voltage monitoring nodes, and communication gateway nodes. Each node collects a communication stability feature set (including the correlation strength feature vector of software layer transmission delay and physical layer signal integrity) and channel anomaly intensity feature information (energy distribution parameters that characterize the physical deformation of the cable) through multi-dimensional timing devices.

[0143] To achieve cross-level risk prediction, first, a hierarchical attribute identifier is assigned to each feature element in the communication stability feature set (such as marking the "network jitter-relay instruction delay correlation strength" as the software layer identifier, and marking the "cable deformation-signal attenuation correlation strength" as the physical channel layer identifier), and at the same time, a physical channel layer identifier is assigned to each feature unit in the channel anomaly strength feature information (steps 1051-1052); based on the topological connection relationship of the distributed nodes of the substation ring network (such as relay protection node → fiber optic switch → voltage monitoring node), a fault propagation path from the software layer to the physical channel layer is constructed (step 1053).

[0144] Then, a coupling operation is performed along the fault propagation path: characteristic elements and characteristic units with the same physical channel level identifier are identified on the path (such as the physical channel level deformation characteristics of the relay protection node and the channel anomaly intensity characteristics of the gateway node) to form physical level association pairs; a timestamp sequence of each pair of association characteristics is obtained, and a time attenuation weight value is calculated based on the ratio of the interval length of adjacent timestamps to the preset reference length, and then the fault impact coefficient is determined in combination with the historical fault frequency of the physical level node, and the two are multiplied to generate a time-varying impact factor; the characteristic element value is multiplied by the factor to obtain a weighted communication stability value, and the characteristic unit value is multiplied by the factor to obtain a weighted channel anomaly intensity value, and the two are added to generate a node coupling strength index (step 1054).

[0145] Then, a time-decay weighted aggregation is performed on the node coupling strength index of each node: an aggregation time window including the current time and the historical period is set, all index values ​​and their timestamps in the window are extracted, a time-decay weight value is assigned according to the ratio of the difference between each timestamp and the current time to the total duration of the window, and the weighted sum of each index value is taken to generate the node vulnerability index (step 1055).

[0146] Finally, based on the comparison results of the node vulnerability index and the preset risk threshold, the node failure risk level is output (for example, the voltage monitoring node triggers a "high risk" alarm due to the coupling of cable deformation and network jitter), and the dispatching system is linked to switch to the backup line.

[0147] The overall solution in step 105 above enables intelligent cross-layer risk prediction for distributed systems. By deeply coupling communication stability characteristics with channel anomaly intensity characteristics and combining them with fault propagation path analysis, a node vulnerability assessment model is constructed. This technology innovatively incorporates a time-decay weighting mechanism and historical fault frequency factors to accurately predict the risk of communication node failure and data transmission anomalies, providing intelligent decision-making support for preventive maintenance of distributed systems.

[0148] The following is a specific example of steps 101 to 105. Figure 2 As shown:

[0149] In the multi-dimensional timing device including the flight simulator system, the system is a distributed system composed of motion system nodes, visual system nodes, and instructor console nodes. Each node collects the operating status data set (such as the transmission delay difference parameters between nodes, bit error rate statistical parameters) and the timing offset information collected by the communication transmission layer (such as the deviation between the data packet arrival timestamp and the global clock reference) through the multi-dimensional timing device (including data packet throughput sensors and clock synchronizers).

[0150] When the system detects a response delay in a motion system node, it first identifies a sequence of continuous offset points in the timing offset information that exceeds the offset threshold and lasts longer than the event window width, marking them as communication jitter extreme events. Simultaneously, it extracts the inter-node transmission delay difference parameters of the associated vision system nodes in the operating status dataset, and calculates the absolute value of the difference between the maximum and minimum delays within the preset time window as the delay fluctuation intensity. Based on the star-shaped physical connection topology of the flight simulator (motion node → central switch → vision node), the shortest path hop count is calculated. The time correlation factor and spatial correlation factor are generated by combining the timestamp of the center of the communication jitter extreme event with the offset of the center of the delay fluctuation time window. The product of the two forms the jitter-delay correlation strength value, and all associated event values ​​are aggregated to generate a communication stability feature set.

[0151] At the same time, an optical monitoring unit embedded next to the hydraulic pipeline of the motion system is used to capture the reflected spectrum signal of the optical fiber inside the communication cable using a multi-channel synchronous acquisition interface, and separate the Bragg wavelength offset caused by mechanical vibration (physical layer deformation monitoring data) and the wavelength drift caused by ambient temperature change (temperature drift monitoring data); deformation feature decoupling processing is performed on the deformation data to extract the deformation feature offset that is positively correlated with the cable stress, and temperature drift trend analysis is performed on the temperature drift data to generate the ambient temperature compensation. The two are combined into a physical layer disturbance feature sequence.

[0152] Subsequently, the physical layer disturbance feature sequence is divided into minute-level long-period disturbance feature segments (covering the continuous vibration energy of the hydraulic system) and second-level short-period disturbance feature segments (covering the transient impact during the take-off and landing phases). Energy accumulation calculation is performed on the long-period segments to obtain steady-state energy distribution parameters, and peak energy capture calculation is performed on the short-period segments to obtain transient energy impact parameters. The parameters are input into the anomaly intensity mapper for baseline offset detection (steady-state parameters continuously deviate from the historical baseline) and pulse event counting (the number of times the transient parameters exceed the safety threshold). The steady-state anomaly intensity component and the transient anomaly intensity component are output and fused to generate channel anomaly intensity feature information.

[0153] In order to achieve cross-level risk prediction, software layer identifiers (such as "visual rendering-transmission jitter correlation strength") and physical channel layer identifiers (such as motion system hydraulic pipelines) are assigned to feature elements in the communication stability feature set, and physical channel layer identifiers are assigned to channel abnormality intensity feature information; according to the topological connection relationship of the flight simulator, a fault propagation path from the software layer to the physical layer is constructed (such as the correlation path between the instructor console software layer delay alarm and the motion system physical layer cable deformation), and features with the same physical channel layer identifier are identified on the path to form physical layer association pairs. The time attenuation weight value is calculated based on the ratio of the interval duration of its timestamp sequence to the preset benchmark duration, and the node history is combined with the time attenuation weight value to calculate the time attenuation weight value. The fault impact coefficient is determined by the historical fault frequency (such as previous fracture records of hydraulic pipelines), and the two are multiplied to generate a time-varying impact factor; the characteristic element value is multiplied by the factor to obtain a weighted communication stability value, and the characteristic unit value is multiplied by the factor to obtain a weighted channel anomaly intensity value, and these are added to generate a node coupling strength index; an aggregation time window is set for the current timestamp (covering the past 24 hours of operating data), all index values ​​in the window are extracted, and time decay weight values ​​are assigned according to the ratio of the difference between the timestamp and the current time to the total duration of the window, and the weighted sum is summed to generate a node vulnerability index; when the index of the motion system node exceeds the preset risk threshold, a "high risk of communication node failure" mark is output and the backup telex system switch is triggered.

[0154] Figure 3 A schematic diagram of a multi-dimensional time series device abnormal state prediction system is provided for an embodiment of the present application. Figure 3 As shown, the system includes:

[0155] The acquisition module 31 is used to collect the operating status data set of the multi-node interaction in the distributed system corresponding to the multi-dimensional time series device and the timing offset information collected by the communication transmission layer;

[0156] a generation module 32 configured to perform jitter and delay correlation processing on the operating status dataset and timing offset information, and generate a communication stability feature set by quantifying the spatiotemporal correlation between jitter extremes and delay fluctuations, wherein the communication stability feature set represents a sensitive indicator of abnormal software and hardware coordination in a distributed system;

[0157] The generation module is further configured to obtain physical layer deformation monitoring data and temperature drift monitoring data through an optical monitoring unit embedded in a communication cable of the device, and to analyze the physical layer deformation monitoring data and temperature drift monitoring data based on an optical signal characteristic offset to generate a physical layer disturbance characteristic sequence;

[0158] A conversion module 33, configured to convert the physical layer disturbance characteristic sequence into channel abnormality intensity characteristic information;

[0159] The prediction module 34 is used to perform cross-level feature coupling on the communication stability feature set and the channel abnormality intensity feature information, and predict the communication node failure risk or data transmission abnormality risk of the distributed system corresponding to the multi-dimensional timing device based on the cross-level feature coupling result.

[0160] Figure 3 The multi-dimensional time series device abnormal state prediction system can perform Figure 1 The implementation principles and technical effects of the multi-dimensional time series device abnormal state prediction method described in the illustrated embodiment will not be elaborated on here. The specific manner in which each module and unit performs operations in the multi-dimensional time series device abnormal state prediction system in the above embodiment has been described in detail in the embodiments of the method and will not be elaborated on here.

[0161] In one possible design, Figure 3 The multi-dimensional time series device abnormal state prediction system of the embodiment shown can be implemented as a computing device, such as Figure 4 As shown, the computing device may include a storage component 41 and a processing component 42;

[0162] The storage component 41 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 42 .

[0163] The processing component 42 is used for the above Figure 1 The embodiment of the multi-dimensional time series device abnormal state prediction method.

[0164] The processing component 42 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0165] The storage component 41 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0166] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0167] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0168] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0169] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0170] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The embodiment shown is a method for predicting abnormal states of a multi-dimensional time series device.

[0171] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0172] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0173] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer or server) to execute the methods described in each embodiment or certain portions of the embodiments.

[0174] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting abnormal state of multi-dimensional time series equipment, characterized in that: include: Collect the operating status data set of multi-node interactions in the distributed system corresponding to the multi-dimensional timing device and the timing offset information collected by the communication transmission layer; performing jitter and delay correlation processing on the operating status data set and the timing offset information, and generating a communication stability feature set by quantifying the spatiotemporal correlation between jitter extremes and delay fluctuations; Acquire physical layer deformation monitoring data and temperature drift monitoring data through an optical monitoring unit embedded in the device communication cable, and analyze the physical layer deformation monitoring data and temperature drift monitoring data based on optical signal characteristic offset to generate a physical layer disturbance feature sequence; Converting the physical layer disturbance characteristic sequence into channel abnormality intensity characteristic information; The communication stability feature set is cross-level coupled with the channel abnormality intensity feature information, and the communication node failure risk or data transmission abnormality risk of the distributed system corresponding to the multi-dimensional timing device is predicted based on the cross-level feature coupling result.

2. The method for predicting abnormal state of multi-dimensional time series equipment according to claim 1, characterized in that: Cross-level feature coupling is performed on the communication stability feature set and the channel abnormality intensity feature information, and based on the cross-level feature coupling result, a communication node failure risk or a data transmission abnormality risk of the distributed system corresponding to the multi-dimensional time series device is predicted, including: Assigning a layer attribute identifier to each feature element in the communication stability feature set, wherein the layer attribute identifier includes a software layer identifier and a physical channel layer identifier; Assigning a physical channel level identifier to each characteristic unit in the channel abnormal strength characteristic information; Based on the topological connection relationship of distributed nodes, a fault propagation path from the software layer identifier to the physical channel layer identifier is constructed; Performing a coupling operation along the fault propagation path, matching the communication stability characteristic elements with the channel abnormality strength characteristic units having the same physical channel layer identifier with time-varying impact factors, and calculating the product of the time-varying impact factor and the corresponding characteristic value to generate a node coupling strength index; Perform time-decay weighted aggregation on the node coupling strength index of each distributed node to obtain the node vulnerability index; According to the comparison result of the node vulnerability index and the preset risk threshold, a communication node failure risk level or a data transmission abnormality risk mark is output.

3. The method for predicting abnormal state of multi-dimensional time series equipment according to claim 2, characterized in that: The communication stability characteristic elements with the same physical channel layer identifier are matched with the channel abnormality strength characteristic units by time-varying impact factors, and the product of the time-varying impact factors and the corresponding characteristic values ​​is calculated to generate a node coupling strength index, including: On the fault propagation path, identifying a communication stability characteristic element and a channel abnormality strength characteristic unit with the same physical channel layer identifier to form a physical layer association pair; Obtain a timestamp sequence of characteristic elements and characteristic units in each physical layer association pair; Performing a time decay coefficient calculation for each physical layer association pair: extracting the interval duration between adjacent timestamps in the timestamp sequence, calculating a time decay weight value based on the ratio of the interval duration to a preset reference duration; determining a fault impact coefficient based on the historical fault frequency of the node corresponding to the physical channel layer identifier; and multiplying the time decay weight value by the fault impact coefficient to generate a time-varying impact factor; Multiplying the value of the communication stability characteristic element by the time-varying impact factor to obtain a weighted communication stability value; Multiplying the value of the channel abnormality strength characteristic unit by the time-varying impact factor to obtain a weighted channel abnormality strength value; The weighted communication stability value and the weighted channel anomaly strength value are added together to generate a node coupling strength index.

4. The method for predicting abnormal state of multi-dimensional time series equipment according to claim 2, characterized in that: The node coupling strength index of each distributed node is subjected to time-decay weighted aggregation to obtain the node vulnerability index, including: Setting an aggregation time window for the current timestamp, where the aggregation time window includes the current time and a preset historical period extending forward; Extracting all index values ​​and their corresponding timestamps within the aggregation time window from the time series of the node coupling strength index; Perform time decay weighting on each extracted index value: calculate the time difference between the timestamp corresponding to the index value and the current time, and determine the time decay weight value based on the ratio of the time difference to the total duration of the aggregation time window; Each index value in the aggregation time window is multiplied by the corresponding time decay weight value to obtain a weighted index value, and all weighted index values ​​are summed to generate a node vulnerability index.

5. The method for predicting abnormal state of multi-dimensional time series equipment according to claim 1, characterized in that: The operating status data set includes dynamic parameters of data packet throughput, statistical parameters of bit error rate, and parameters of transmission delay difference between nodes; The performing jitter and delay correlation processing on the operating status data set and the timing offset information, and generating a communication stability feature set by quantifying the spatiotemporal correlation between jitter extreme values ​​and delay fluctuations, includes: Identifying a communication jitter extreme value event in the timing offset information, wherein the communication jitter extreme value event is composed of a sequence of continuous offset points exceeding an offset threshold and having a duration greater than an event window width; Extracting the delay fluctuation intensity of the transmission delay difference parameter between nodes in the operation status data set, wherein the delay fluctuation intensity is obtained by calculating the absolute value of the difference between the maximum and minimum delays within a preset time window; Performing spatiotemporal matching of the communication jitter extreme value event and the delay fluctuation intensity to determine a spatiotemporal matching result, wherein the spatiotemporal matching process includes locating a time window corresponding to the delay fluctuation intensity according to a central timestamp of the communication jitter extreme value event, and determining the shortest path hop count from the communication jitter extreme value event source node to the delay fluctuation node based on the physical connection topology of the distributed nodes; Quantifying the spatiotemporal correlation of the spatiotemporal matching result to calculate a time correlation factor and a space correlation factor, wherein the time correlation factor is obtained based on an offset between a center timestamp of a communication jitter extreme event and a center of a delay fluctuation time window, and the space correlation factor is obtained based on a reciprocal of the number of hops of the shortest path; generating a jitter-delay correlation strength value according to a product of the time correlation factor and the spatial correlation factor; Aggregate the jitter-delay correlation strength values ​​corresponding to all spatiotemporal matching events to generate a communication stability feature set.

6. The method for predicting abnormal state of multi-dimensional time series equipment according to claim 1, characterized in that: The method includes acquiring physical layer deformation monitoring data and temperature drift monitoring data through an optical monitoring unit embedded in a communication cable of the device, and parsing the physical layer deformation monitoring data and temperature drift monitoring data based on an optical signal characteristic offset to generate a physical layer disturbance characteristic sequence, including: Capturing the optical fiber reflection spectrum signal embedded in the communication cable through the multi-channel synchronous acquisition interface of the optical monitoring unit; Separating the physical layer deformation monitoring data and the temperature drift monitoring data in the reflection spectrum signal, wherein the physical layer deformation monitoring data is composed of the Bragg wavelength shift caused by optical fiber deformation, and the temperature drift monitoring data is composed of the wavelength drift caused by thermal expansion; Performing deformation feature decoupling processing on the physical layer deformation monitoring data to extract a deformation feature offset that is positively correlated with the mechanical stress of the communication cable; Performing temperature drift trend analysis on the temperature drift monitoring data to generate an ambient temperature compensation amount; The deformation feature offset and the ambient temperature compensation are combined into a physical layer disturbance feature sequence, wherein each data unit in the physical layer disturbance feature sequence includes a paired output of the deformation feature offset and the ambient temperature compensation.

7. The method for predicting abnormal state of multi-dimensional time series equipment according to claim 1, characterized in that: The converting the physical layer disturbance characteristic sequence into channel abnormality intensity characteristic information includes: Dividing the physical layer disturbance feature sequence into a long-period disturbance feature segment and a short-period disturbance feature segment according to a preset time scale, wherein the long-period disturbance feature segment covers a minute-level time window and the short-period disturbance feature segment covers a second-level time window; Perform energy accumulation calculation on each long-period disturbance characteristic segment to obtain steady-state energy distribution parameters; Perform peak energy capture calculation on each short-period disturbance characteristic segment to obtain transient energy impact parameters; The steady-state energy distribution parameters and the transient energy impact parameters are input into the anomaly intensity mapper, so that the baseline offset detection is performed on the steady-state energy distribution parameters through the anomaly intensity mapper, and the steady-state anomaly intensity component is output; the pulse event counting is performed on the transient energy impact parameters, and the transient anomaly intensity component is output, and the steady-state anomaly intensity component and the transient anomaly intensity component are fused to generate channel anomaly intensity feature information.

8. A multi-dimensional time series equipment abnormal state prediction system, characterized in that: include: The acquisition module is used to collect the operating status data set of multi-node interactions in the distributed system corresponding to the multi-dimensional timing device and the timing offset information collected by the communication transmission layer; a generating module, configured to perform jitter and delay correlation processing on the operating status data set and the timing offset information, and generate a communication stability feature set by quantifying the spatiotemporal correlation between the jitter extreme value and the delay fluctuation; The generation module is further configured to obtain physical layer deformation monitoring data and temperature drift monitoring data through an optical monitoring unit embedded in a communication cable of the device, and to analyze the physical layer deformation monitoring data and temperature drift monitoring data based on an optical signal characteristic offset to generate a physical layer disturbance characteristic sequence; A conversion module, configured to convert the physical layer disturbance characteristic sequence into channel abnormality intensity characteristic information; A prediction module is used to perform cross-level feature coupling on the communication stability feature set and the channel abnormality intensity feature information, and predict the communication node failure risk or data transmission abnormality risk of the distributed system corresponding to the multi-dimensional timing device based on the cross-level feature coupling result.

Images