Leaky coaxial cable alarm trend analysis system based on multi-mode storage

Abstract

The invention discloses a leaky coaxial cable alarm trend analysis system based on multi-modal storage. The system comprises the following steps: S1, confirming a system core target; s2, a multi-mode storage core support; s3, analyzing the alarm trend of the leaky coaxial cable; and S4, outputting a trend analysis result. On the fault early warning level, a 50-meter geofence constructed by a space-time alignment engine and a GPS clock synchronization mechanism are combined with a dual-channel degradation detection model, the false alarm rate is reduced by more than 40%, and subway tunnel false alarm in a typical scene is reduced from 15 times per day to 9 times per day. In an operation and maintenance decision-making level, a fault root cause probability matrix is fused with a multi-source evidence chain to realize three-minute quick positioning, so that the response efficiency of a railway external damage accident is improved, and in a system efficiency level, a cold and hot data layering strategy and radio frequency parameter piecewise linear compression cooperate to reduce the storage requirement, original data is compressed, and the system efficiency is improved. And the edge side lightweight reasoning module realizes 200ms-level real-time detection through a pruning model and I NT8 quantization, and the alarm response delay is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of G05B, in particular to a leaky cable alarm trend analysis system based on multi-modal storage. BACKGROUND

[0002] In the field of communication infrastructure, leaky cable is the core transmission medium of closed spaces such as tunnels and subways, and its health status is directly related to public safety and service quality.

[0003] The current industry generally uses an alarm monitoring system based on a time series database, which has the following technical bottlenecks: 1. Analysis is one-sided due to fragmented data The existing system independently processes data such as radio frequency performance, environmental sensing, and operation and maintenance records, and lacks a unified spatio-temporal correlation mechanism. For example, in a water leakage fault of a certain railway leaky cable, the vibration sensor triggered an alarm, but the temperature and humidity data were not correlated in real time because they were stored in different platforms.

[0004] 2. Early warning mechanism is lagging and has a high false positive rate Traditional threshold alarms rely on fixed thresholds and cannot adapt to dynamic changes in the environment.

[0005] Deterioration trend prediction only relies on a single radio frequency parameter, ignoring the synergistic effects of vibration, temperature, and humidity.

[0006] 3. Low storage and computing efficiency Storing the full amount of raw radio frequency parameters wastes resources.

[0007] Centralized cloud computing cannot meet real-time requirements, so there is a need for a leaky cable alarm trend analysis system based on multi-modal storage. SUMMARY

[0008] The purpose of the present application is to provide a leaky cable alarm trend analysis system based on multi-modal storage. Through the synergistic innovation of multi-modal storage architecture and cross-modal analysis engine, the present application achieves a breakthrough in the health management of communication leaky cables, thereby solving the technical problems mentioned in the background art.

[0009] To achieve the above purpose, the present application provides the following technical solution: a leaky cable alarm trend analysis system based on multi-modal storage, comprising the following steps: S1, system core target confirmation: Precise early warning: identify potential leaky cable faults in advance and change passive response to active defense; Trend insight: analyze alarm history, environmental parameters, and operation and maintenance records to discover fault patterns, deterioration rules, and external causes; Decision support: optimize inspection plans, spare parts reserves, and network reinforcement strategies to reduce costs and increase efficiency; Knowledge accumulation: build a leaky cable health state model and accumulate a fault diagnosis knowledge base.

[0010] S2, multi-modal storage core support: Confirm data types, solve the unified management problem of massive heterogeneous data, and obtain leaky cable monitoring data in real time through heterogeneous data acquisition interfaces. The data includes: Time series data: radio frequency performance parameters, environmental sensor readings; Event data: device alarm logs and level labels; Spatial data: leaky cable GIS coordinates and topological relationships; Document data: operation and maintenance records and external associated information; S3, leaky cable alarm trend analysis: based on multi-modal storage data lake and data warehouse, deep analysis is carried out, which includes: Multi-modal data fusion analysis: accurately map alarm events to GIS maps, and analyze propagation path and impact range in combination with time line; Time series analysis: using statistical methods and machine learning to identify index abnormalities in real time, using time series model to predict the deterioration trend of key indicators, to predict future failure points, and to find periodic fluctuations and stage deterioration characteristics.

[0011] S4, output trend analysis results: Generate leaky cable health degree space-time heat map and mark high-risk sections; Predict the remaining effective life of the specified leaky cable in the future time window; Output fault probability weighted report associated with external factors.

[0012] Preferably, in S2, the The core monitoring indicators used by the time series data include: VSWR, RF power, insertion loss, environmental temperature and humidity, and vibration amplitude. The monitoring data is stored in a time series database; Alarm events include: real-time alarm logs generated by devices, time, location, type, and level; Spatial geographic data includes: accurate location of leaky cable, creating GIS coordinates, establishing tunnel intervals, setting pole road information, and fault point positioning; Operation and maintenance documents include: record patrol reports, maintenance records, pictures, video evidence, and design drawings; External associated information includes: device account, maintenance personnel information, work order system record, and weather data.

[0013] Preferably, in S2, the data is divided into an optimized storage module according to the modal type: Time series data is written into a time series database, and an index with device ID and timestamp as primary key is established; Event data is written into a log database, and a joint index is established based on alarm type and geographic location hash value; Spatial data is injected into a spatial database to construct a topological overlay layer; Document data is stored in object storage and associated with the corresponding device through metadata tag chain.

[0014] Preferably, the database in S2 requires unified metadata management to establish a global data directory and clearly define the mode, location, and association of each modal data; To improve index efficiency, establish targeted indexes for each data type, set timestamps, device IDs, geographic locations, alarm types, and text keywords, and establish independent data pipelines for reliable data collection, cleaning, conversion, and loading processes to ensure data quality and consistency; At the same time, the data is divided into hot and cold layers, and the historical data is archived and compressed to reduce costs.

[0015] Preferably, the response trend analysis request in S3 is executed by cross-modal data fusion: Based on the spatio-temporal alignment engine, the time series data, event data, and spatial data within the same geographic fence are synchronized in coordinate-time; Triggering the multi-modal correlation analyzer to perform the following operations: Performing adaptive threshold anomaly detection on time series data to dynamically generate a degradation baseline; Aggregating event data according to the causal graph model to extract alarm propagation paths; Fusing environmental parameters and operation and maintenance records to construct a fault root cause probability matrix.

[0016] Preferably, it includes: a multi-modal collection gateway: a data access adapter supporting RS485, IoT protocol, and API; Hierarchical storage cluster: a physically isolated storage pool composed of time series database, spatial database, and document database; Cross-modal analysis engine: integrating spatio-temporal alignment module, multi-source feature extractor, and survival analysis predictor; Visual decision terminal: providing alarm propagation animation simulation and maintenance strategy simulation interface.

[0017] Preferably, it includes: the hierarchical storage cluster deploys a hot and cold data layering strategy: High-frequency access data within 30 days is stored in the SSD storage pool; Enable columnar compression storage for historical archive data and migrate to the HDD pool; The cross-modal analysis engine has a built-in lightweight inference module that supports real-time anomaly scoring calculation on edge nodes.

[0018] Preferably, the spatio-temporal alignment engine comprises Geofence generator: automatically create a 50-meter radius circular analysis area according to the leak cable GIS coordinates; Time synchronization device: configure GPS clock source to unify the time stamps of each modal data; Coordinate mapping table: store the correspondence between leak cable physical location and time series data collection point.

[0019] Preferably, the physical architecture of the hierarchical storage cluster is: Time series storage node: adopt SSD disk array, deploy RAID10 redundancy configuration; Spatial storage node: equipped with GPU acceleration card for geographic data processing server; Document storage node: build a distributed object storage cluster, and the minimum storage unit is 1MB data block.

[0020] Compared with the prior art, the beneficial effects of the present application are: In the fault early warning layer, the 50-meter geofence constructed by the spatio-temporal alignment engine and the GPS clock synchronization mechanism, combined with the double-channel degradation detection model, reduces the false alarm rate by more than 40%, and in a typical scenario, the false alarm of the subway tunnel is reduced from 15 times per day to 9 times. At the same time, the multivariate time series prediction makes the early warning of key indicator anomalies up to 72 hours, and the prediction accuracy of VSWR degradation trend is improved to 92%. In the operation and decision-making layer, the fault root cause probability matrix integrates multi-source evidence chain to realize 3-minute rapid positioning, and the response efficiency of railway external damage accidents is improved; In the system efficiency layer, the cold and hot data layering strategy and the segmented linear compression of radio frequency parameters cooperatively reduce the storage demand, the original data compression, and the edge side lightweight inference module realizes 200ms level real-time detection through pruning model and INT8 quantization, and the alarm response delay is reduced.

[0021] In the long-term value layer, the knowledge graph constructed by the multi-modal correlation analysis accumulates fault cases, so that the accuracy rate of complex fault diagnosis such as water ingress is greatly improved; the preventive maintenance strategy guided by the health degree spatio-temporal heat map significantly delays the degradation, and the life of the leak cable in the high humidity area of the tunnel is prolonged. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 The figure is a structural schematic diagram of the present application. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0024] The application provides a cable leakage alarm trend analysis system based on multi-modal storage, as shown in the following steps: Figure 1 S1, system core target confirmation: Precise early warning: early identification of potential cable leakage faults, from passive response to active defense; Trend insight: analyze alarm history, environmental parameters, and operation and maintenance records to find fault patterns, deterioration rules, and external causes; Decision support: optimize inspection plan, spare parts reserve, and network reinforcement strategy to reduce cost and increase efficiency; Knowledge sedimentation: build a cable health state model and accumulate a fault diagnosis knowledge base.

[0025] S2, multi-modal storage core support: Confirm the data type, solve the unified management problem of massive heterogeneous data, and obtain cable monitoring data in real time through a heterogeneous data acquisition interface, including: Time series data: radio frequency performance parameters and environmental sensor readings, the core monitoring indicators of the time series data include: VSWR, radio frequency power, insertion loss, environmental temperature and humidity, and vibration amplitude, and the monitoring data is stored in a time series database; Event data: device alarm logs and level labels, the alarm events include: real-time alarm logs generated by the device, time, location, type, and level; Spatial data: cable GIS coordinates and topological relationships, the spatial geographic data includes: accurate cable location, GIS coordinate creation, tunnel interval establishment, pole road information setting, and fault point positioning; Document data: operation and maintenance records and external associated information, the operation and maintenance documents include: record inspection reports, maintenance records, pictures, video evidence, and design drawings; the external associated information includes: device account, maintenance personnel information, work order system record, and weather data; The data is divided into an optimization storage module according to the modal type: time series data is written into a time series database, an index is established with device ID and timestamp as the primary key; event data is written into a log database, a joint index is established with alarm type and geographic location hash value; spatial data is injected into a spatial database to construct a topological overlay layer; document data is stored in an object storage and is associated to the corresponding device through a metadata tag chain; the database needs unified metadata management, a global data directory is established to clearly show the mode, location, and association of each modal data; the index efficiency is improved, specific indexes are established for each data type, time stamp, device ID, geographic location, alarm type, and text keyword are set, an independent data pipeline is established, reliable data collection, cleaning, conversion, and loading processes are ensured to ensure data quality and consistency; ​Meanwhile, the data is hot and cold layered, the historical data is archived and compressed, and the cost is reduced; S3, leak cable alarm trend analysis: based on multi-modal storage data lake, data warehouse, deep analysis, the deep analysis includes: Multi-modal data fusion analysis: accurately map alarm events to GIS maps, combine timeline analysis to spread path, impact range, multi-modal collection gateway supports RS485, IoT protocol and API data access adapter; Time series analysis: using statistical methods, machine learning to identify index anomalies in real time, using time series model to predict the degradation trend of key indicators, to predict future failure points, to find periodic fluctuations and stage degradation characteristics; Respond to trend analysis request, perform cross-modal data fusion: based on space-time alignment engine, coordinate-time synchronization of time series data, event data and spatial data within the same geographic fence; Trigger multi-modal correlation analyzer to perform the following operations: adaptive threshold anomaly detection on time series data, dynamically generate degradation baseline; aggregate event data according to causal graph model, extract alarm propagation path; fuse environment parameters and operation and maintenance records to build fault root cause probability matrix; Causal graph model aggregates event data: Build a causal graph model based on Bayesian network, nodes are alarm types (such as water ingress alarm, vibration overrun), edge weights are calculated by conditional probability of historical data:

[0026] Propagation path extraction uses D-S evidence theory to fuse multi-source alarms, and generates propagation chain when confidence CF>0.8 (example: construction alarm→vibration overrun→VSWR anomaly); Among them, adaptive threshold anomaly detection and degradation baseline generation: In the time series database, the sliding window standard deviation method is used to dynamically calculate the threshold value of core indicators such as VSWR and RF power. The window size is 15 minutes, and when 3 consecutive data points exceed μ±3σ range, an exception is triggered; The generation of degradation baseline is realized through the exponential weighted moving average (EWMA) model: ; Among them, the smoothing factor α=0.2, dynamically adapts to the baseline drift in the aging process of the leak cable; Baseline t and Baseline t-1 is a floating point number, Baseline t is the current floating point number, Baseline t-1 is the previous time point floating point number, and Xt is the sensor data.

[0027] Further, the fault root cause probability matrix is constructed: Define matrix M[RxC]: row R is the fault type (joint degradation / external damage, etc.), column C is the evidence factor (sudden change of temperature and humidity, alarm frequency); Output root cause probability distribution P(joint degradation)=0.92); Hierarchical storage cluster: a physically isolated storage pool composed of time series database, spatial database and document database; Cross-modal analysis engine: integrated time-space alignment module, multi-source feature extractor and survival analysis predictor; Visual decision terminal: provides alarm propagation dynamic simulation and maintenance strategy simulation interface; The hierarchical storage cluster deploys a cold and hot data layering strategy: high-frequency access data within 30 days are stored in the SSD storage pool; historical archive data are enabled for columnar compression storage and migrated to the HDD pool; The cross-modal analysis engine has a built-in lightweight inference module that supports real-time exception score calculation on edge nodes; Lightweight inference module (edge computing): Model pruning technology is used to remove weight connections less than 0.001 in the neural network, and the compressed model size is reduced by 68%; INT8 quantization maps floating-point weights to the [-128, 127] integer interval, and the inference speed is increased to 187ms / time (test platform: Jetson Nano); The space-time alignment engine includes a geofencing generator: automatically creates a 50-meter radius circular analysis area based on the cable GIS coordinates; Time synchronization device: configure GPS clock source to unify the time stamp of each modal data; Coordinate mapping table: stores the correspondence between the physical location of the cable and the time series data collection point.

[0028] The physical architecture of the hierarchical storage cluster is: time series storage node: uses SSD disk array, deploys RAID10 redundancy configuration; Spatial storage node: equipped with GPU acceleration card for geographic data processing server; Document storage node: build distributed object storage cluster, the minimum storage unit is 1MB data block.

[0029] S4, output trend analysis result: Generate a space-time heat map of cable health, and label high-risk sections; Predict the remaining effective life of the specified cable in the future time window; Remaining effective life prediction (survival analysis predictor): Based on the Weibull proportional risk model:

[0030] Where, the shape parameter k=1.8, and the scale parameter λ is dynamically adjusted according to the environmental corrosion rate.

[0031] RUL(t) is the predicted remaining useful life at time t, S(t) is the Weibull survival function, which describes the theoretical life distribution under no additional stress, β1 is the VSWR degradation weight, which quantifies the influence intensity of standing wave ratio fluctuation on life, VSWR t t is the measured value of standing wave ratio at time t, which represents the impedance matching state of the radio frequency system, β2 is the vibration damage weight, which quantifies the acceleration effect of mechanical vibration on life, Iᵥ is the vibration integral intensity, which reflects the cumulative mechanical fatigue damage; Output the failure probability weighted report associated with external factors; Failure probability weighted report generation method: Introduce external influencing factors: weather data (rainfall > 50mm / h weight + 0.3), construction record (within 500 meters of construction weight + 0.4);

[0032] Where, P final is the weighted comprehensive failure probability, reflecting the actual failure risk under the influence of external environment, P root is the initial probability of failure root cause, calculated by the failure root cause probability matrix, ωi is the influence coefficient of environmental factor, which quantifies the influence intensity of environmental factors on the current failure mode, which changes with external conditions, E i is the i-th external environmental factor, the quantitative index after normalization processing of objective monitoring value, n is a constant.

[0033] According to the above, there are the following specific scene embodiments: Scene description: Cable leakage positioning: Beijing-Shanghai high-speed railway K12+500 Basic probability: Proot=0.65P_{\text{root}}=0.65Proot=0.65 (corrosion degradation) Environmental state: Heavy rain 83mm / h for 2 hours → E1=0.92E_1=0.92E1=0.92 There are 3 excavators working 300m away → E2=0.6E_2=0.6E2=0.6 Temperature drops 15℃ in 3 hours → E3=0.5E_3=0.5E3=0.5 Dynamic weight calculation:

[0034] Final failure probability:

[0035] The specific execution code is: { "Risk level": "Red alert", "Combined probability": 0.99, "Primary cause contribution": { "Corrosion degradation": 65.0%, "Rainfall weighting": 33.7%, "Construction vibration": 9.6%, "Thermal stress": 6.3%}, "Treatment recommendation": "Immediately perform waterproof reinforcement + 2 hours of vibration monitoring" Therefore, the treatment recommendation is obtained: immediately perform waterproof reinforcement + 2 hours of vibration monitoring. In summary, the beneficial effects of the present application are: At the fault early warning level, the 50-meter geographic fence constructed by the space-time alignment engine is synchronized with the GPS clock mechanism, combined with the double-channel degradation detection model, which reduces the false positive rate by more than 40%, and in typical scenarios, the subway tunnel false alarm is reduced from 15 times per day to 9 times. At the same time, the multivariate time series prediction makes the key indicator anomaly early warning advance by 72 hours, and the VSWR degradation trend prediction accuracy is improved to 92%. At the operation and decision level, the fault root cause probability matrix integrates multi-source evidence chain to realize 3-minute rapid positioning, which improves the response efficiency of railway external damage accidents; At the system efficiency level, the cold and hot data layering strategy and the segmented linear compression of radio frequency parameters cooperate to reduce the storage demand, the original data compression, and the edge side lightweight inference module realizes 200ms level real-time detection through pruning model and INT8 quantization, and the alarm response delay is reduced.

[0036] At the long-term value level, the knowledge graph constructed by multi-modal correlation analysis accumulates fault cases, which makes the accuracy of complex fault diagnosis such as water ingress jump; the preventive maintenance strategy guided by the health degree space-time thermal map significantly delays the degradation, and the cable life in the high humidity area of the tunnel is extended.

[0037] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A multi-modal storage based cable leak warning trend analysis system, characterized in that: Comprising the following steps: S1, system core target confirmation: Precise early warning: identify potential cable leakage faults in advance, change passive response to active defense; Trend insight: analyze alarm history, environmental parameters, and operation and maintenance records to find fault patterns, deterioration rules, and external causes; Decision support: optimize inspection plans, spare parts reserves, and network reinforcement strategies to reduce costs and increase efficiency; Knowledge sedimentation: build a cable leakage health status model and accumulate a fault diagnosis knowledge base. S2, multi-modal storage core support: Confirm data types and solve the unified management problem of massive heterogeneous data. Real-time acquisition of cable leakage monitoring data through heterogeneous data acquisition interfaces, including: Time series data: radio frequency performance parameters and environmental sensor readings; Event data: device alarm logs and level labels; Spatial data: cable leakage GIS coordinates and topological relationships; Document data: operation and maintenance records and external associated information; S3, cable leakage alarm trend analysis: based on the data lake and data warehouse of multi-modal storage, deep analysis is conducted, including: Multi-modal data fusion analysis: accurately map alarm events to GIS maps, analyze propagation paths and impact ranges using time series analysis; Time series analysis: use statistical methods and machine learning to identify index abnormalities in real time, use time series models to predict the deterioration trend of key indicators, predict future fault points, and find periodic fluctuations and stage deterioration characteristics. S4, output trend analysis results: Generate a cable leakage health degree space-time heat map and label high-risk sections; Predict the remaining useful life of a specified cable in a future time window; Output a fault probability weighted report associated with external factors.

2. The multi-modal storage based cable leak warning trend analysis system of claim 1, wherein: In S2, the Core monitoring indicators used for time series data include: VSWR, RF power, insertion loss, environmental temperature and humidity, and vibration amplitude. Monitoring data is stored in a time series database; Alarm events include: real-time alarm logs generated by devices, time, location, type, and level; Spatial geographic data includes: accurate cable leakage location, creating GIS coordinates, establishing tunnel intervals, setting pole road information, and fault point positioning; Operation and maintenance documents include: patrol reports, maintenance records, pictures, video evidence, and design drawings; External associated information includes: device account, maintenance personnel information, work order system records, and weather data.

3. The multi-modal storage based leaky cable alert trend analysis system of claim 1, wherein: In S2, the data is divided into optimization storage modules according to modal types: Time series data is written to a time series database, establishing an index with device ID and timestamp as the primary key; Event data is written to a log database, establishing a joint index with alarm type and geographic location hash value; Spatial data is injected into a spatial database to construct a topological overlay layer; Document data is stored in object storage and associated with the corresponding device through metadata tag chains.

4. The multi-modal storage based leaky cable alert trend analysis system of claim 1, wherein: In S2, the database needs unified metadata management to establish a global data directory and clearly define the mode, location, and association of each modal data; Improve index efficiency by establishing targeted indexes for each data type, setting timestamps, device IDs, geographic locations, alarm types, and text keywords, and establishing independent data pipelines with reliable data collection, cleaning, conversion, and loading processes to ensure data quality and consistency; At the same time, the data is divided into hot and cold layers, historical data is archived and compressed to reduce costs.

5. The multi-modal storage based leaky cable alert trend analysis system of claim 1, wherein: The S3 responds to the trend analysis request, performs cross-modal data fusion: Based on the spatio-temporal alignment engine, the time series data, event data and spatial data within the same geofence are synchronized in coordinates-time. Trigger the multi-modal correlation analyzer to perform the following operations: Adaptive threshold anomaly detection is performed on time series data, and a degradation baseline is dynamically generated. Event data is aggregated according to the causal graph model to extract alarm propagation paths. Fusion of environmental parameters and operation and maintenance records to construct a fault root cause probability matrix.

6. The multi-modal storage based leaky cable alert trend analysis system of claim 1, wherein: Including: Multi-modal acquisition gateway: supports RS485, IoT protocol and API data access adapter; Hierarchical storage cluster: composed of time series database, spatial database and document database physical isolation storage pool; Cross-modal analysis engine: integrated time-space alignment module, multi-source feature extractor, survival analysis predictor; Visual decision terminal: provides alarm propagation dynamic effect simulation and maintenance strategy simulation interface.

7. The multi-modal storage based leaky cable alert trend analysis system of claim 6, wherein: Including: The hierarchical storage cluster deploys a hot and cold data layering strategy: High-frequency access data within 30 days are stored in the SSD storage pool; Enable columnar compression storage for historical archive data and migrate to the HDD pool; The cross-modal analysis engine has a built-in lightweight inference module that supports real-time anomaly score calculation on edge nodes.

8. The multi-modal storage based leaky cable alert trend analysis system of claim 5, wherein: The spatio-temporal alignment engine includes Geofence generator: automatically creates a 50-meter radius circular analysis area based on the cable leakage GIS coordinates; Time synchronization device: configure GPS clock source to unify the time stamp of each modal data; Coordinate mapping table: stores the correspondence between cable leakage physical location and time series data collection point.

9. The multi-modal storage based leaky cable alert trend analysis system of claim 6, wherein: The physical architecture of the hierarchical storage cluster is: Time series storage node: uses SSD disk array, deploys RAID10 redundancy configuration; Spatial storage node: equipped with GPU acceleration card for geographic data processing server; Document storage node: build a distributed object storage cluster, the minimum storage unit is 1MB data block.

Images