An operator-oriented optical fiber communication line intelligent operation and maintenance platform

Abstract

The application belongs to the technical field of optical fiber communication line operation and maintenance, and discloses an intelligent operation and maintenance platform for optical fiber communication lines of operators, wherein three types of semantic libraries of equipment, environment and service are constructed through a data semantic fusion module, wherein the equipment semantic library converts OTDR curve characteristics and alarm codes into descriptions that can be directly understood by operation and maintenance personnel, such as attenuation abnormalities and joint faults; the environment semantic library is associated with soil settlement, temperature and humidity changes and optical cable extrusion risks, sheath aging acceleration; the service semantic library maps bandwidth utilization and service interruption records into bearing pressure levels and service influence degrees; meanwhile, a weight model is trained in combination with historical operation and maintenance data, semantic correlation coefficients are dynamically adjusted in real-time service scenes such as early peak of government special lines, and an edge and cloud collaborative semantic processing architecture is matched to reduce edge end computing power consumption; through a digital twin module, a service load heat map is embedded to quickly locate hot sections, and scattered data is converted into effective information supporting operation and maintenance decisions.

Description

Technical Field

[0001] This invention belongs to the field of optical fiber communication line operation and maintenance technology, specifically an intelligent operation and maintenance platform for optical fiber communication lines for operators. Background Technology

[0002] With the large-scale deployment of 5G networks, the accelerated construction of computing networks, and the rapid expansion of government and enterprise private line services, the scale of operators' fiber optic communication lines continues to expand, and they have now become an important infrastructure supporting the development of the digital economy. However, the current fiber optic operation and maintenance system is struggling to keep up with the demands of rapid business development, mainly due to the following technical problems:

[0003] Although the OTDR curves and alarm codes collected by the equipment, the temperature, humidity and soil settlement data obtained by environmental sensors, the fault descriptions recorded by manual inspections, and the bandwidth usage statistics of the business system have been stored in a unified format, they lack an effective semantic association mechanism. For example, when the equipment triggers a LOS (loss of signal) alarm, it is impossible to establish a causal mapping with the loose fiber optic cable joints found during inspections. When environmental data is abnormal, it is also difficult to identify its correlation with business bandwidth fluctuations. As a result, the value of this data cannot be deeply explored and can only remain at the basic storage level.

[0004] Most existing diagnostic solutions are developed based on urban pipeline optical cable scenarios and do not fully consider the differences in different laying environments. When applied to complex scenarios such as direct burial in mountainous areas and underwater laying, because they do not incorporate the characteristics of scenarios such as rock compression in mountainous areas and changes in underwater pressure, they not only have difficulty accurately identifying the type of fault, but also easily misjudge temporary attenuation caused by rainstorms as a real fault, thereby initiating unnecessary emergency repair procedures and increasing ineffective operation and maintenance costs.

[0005] Currently, the formulation of fiber optic cable replacement, expansion, and inspection plans relies solely on historical fault data without considering real-time service load conditions. Some backbone fiber optic cables carrying government dedicated lines are already operating at high bandwidth, yet expansion is still planned according to the original schedule, which can easily cause service disruptions during peak periods. Meanwhile, some access fiber optic cables have experienced a significant drop in load after service migration, yet they still maintain high-frequency inspections, resulting in a waste of human and material resources. It is difficult to find a balance between operation and maintenance cost control and service assurance requirements. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent operation and maintenance platform for fiber optic communication lines for operators, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent operation and maintenance platform for optical fiber communication lines for operators, wherein the data semantic fusion module is specifically as follows:

[0008] First, three semantic libraries are established: the device semantic library transforms OTDR curve features and alarm codes into descriptions that maintenance personnel can directly understand; the environment semantic library associates soil subsidence, temperature and humidity changes with concerns including optical cable compression risks and accelerated sheath aging; and the business semantic library maps bandwidth utilization and business interruption records to load pressure levels and business impact levels. Based on historical maintenance data, a weight model is trained, and the semantic association coefficients are adjusted in conjunction with real-time business status to ensure that the semantic associations align with actual business needs.

[0009] The design incorporates an edge-cloud collaborative semantic processing architecture. At the edge, a semantic extraction unit is deployed to locally perform semantic conversion of OTDR curves and temperature and humidity data. Only high-value semantic data, including those related to fiber optic cable compression risks and high business load pressure, along with scenario parameters (such as laying type, real-time weather, and construction status), are uploaded to the cloud. The cloud then uses a cross-domain semantic conflict arbitration mechanism to correct semantic discrepancies by combining the scenario parameters from the edge with the global business data from the cloud. This reduces the computing power consumption at the edge and improves the fusion accuracy.

[0010] Add an SLA compliance risk level field to the business semantic library to associate bandwidth utilization and downtime with the business SLA agreement, so that semantic data matches business performance assessment.

[0011] Furthermore, the digital twin module is specifically as follows:

[0012] Based on the semantic data of the data semantic fusion module, a digital twin of the optical fiber line is established. In the service and resource coupling mapping layer of the twin, the service load heat map function is embedded. Through the red, yellow and green color gradients, the service load density of different sections of the optical cable is presented, and hot spots with load exceeding the threshold are marked.

[0013] Develop a multi-scenario parallel pre-simulation engine with the ability to simulate both fault and business growth scenarios simultaneously. During the pre-simulation process, the semantic data of the data semantic fusion module and the optical cable health data (such as real-time attenuation value and aging coefficient) of the health prediction module are called. Finally, the SLA loss rate of the affected business and the priority ranking of resource expansion are output to provide a basis for operation and maintenance decisions.

[0014] A two-way calibration mechanism between the digital twin and the physical entity is established. Real-time data of the physical optical cable is collected by edge sensors and compared point by point with the simulated data of the digital twin. When the deviation exceeds the threshold, the scene loss factor and other parameters of the digital twin are automatically adjusted. The calibrated parameters are synchronously fed back to the data semantic fusion module and the health prediction module.

[0015] Furthermore, the fault migration diagnosis module is specifically as follows:

[0016] Based on the semantic data of the data semantic fusion module and the scene parameters of the digital twin in the digital twin module, a scene feature distillation algorithm is designed to extract core features from the original scene parameters. The new scene only needs to input 10-15 sets of scene samples to complete the transfer adaptation.

[0017] The system distinguishes between false alarms and genuine alarms by considering three dimensions: environment, business, and equipment. The environmental dimension refers to parameters such as rainstorm intensity and wind level. The business dimension monitors whether bandwidth fluctuations exceed the threshold. The equipment dimension analyzes whether the OTDR curve shows continuous attenuation. The alarm type is determined by a weighted score: a score less than 40 is considered a false alarm, a score greater than 60 is considered a genuine alarm, and scores in between are manually reviewed.

[0018] An embedded unknown fault feature capture module is used. When a new fault that does not match the historical fault database is diagnosed, the module automatically records the scene features, equipment data and handling solutions of the fault, and updates the scene and fault association feature database through federated learning.

[0019] Furthermore, the scheduling module is specifically as follows:

[0020] Combining the fault diagnosis results of the fault migration diagnosis module with the health prediction data of the health prediction module, a resource capability assessment system is first established in three dimensions: personnel, equipment, and tools. The personnel dimension assesses whether underwater welding certification is available and the continuous operation time is sufficient; the equipment dimension monitors the welding machine accuracy and tests the equipment's battery life; and the tool dimension verifies the integrity rate of underwater operation equipment and outputs a resource adaptability score.

[0021] Develop a service loss prediction and resource scheduling linkage algorithm. When the health prediction module predicts that the health of high-priority service-bearing optical cables is declining, it calculates the potential service loss rate in advance and performs resource pre-scheduling.

[0022] A regional resource pool and collaborative dispatch center architecture is established. When special repair equipment is lacking in remote mountainous areas, the collaborative dispatch center queries the idle resources in the nearby areas, calculates the balance between resource allocation costs and business losses, and generates the optimal transportation route by combining real-time road conditions and equipment transportation requirements. After the dispatch is completed, the relevant resource usage will be fed back to the closed-loop module.

[0023] Furthermore, the health prediction module is specifically as follows:

[0024] Based on the business semantic data of the data semantic fusion module and the scene parameters of the digital twin module, a coupling coefficient is introduced to quantify the synergistic effect. The load factor of high-frequency data transmission services is higher than that of ordinary services, the loss factor of high temperature and high humidity environment is higher than that of normal temperature environment, and the longer the service life of optical cable, the higher the correction coefficient. The actual loss of optical cable is reflected by the health calculation formula.

[0025] When the predicted health of the optical cable is below the threshold, adjustment suggestions are automatically pushed to the business management system. At the same time, the business distribution of the digital twin in the digital twin module is updated synchronously, so that the digital twin can reflect the resource status after the business adjustment in real time. It has prediction functions for three periods: short-term, medium-term and long-term. The short-term prediction uses the real-time attenuation value and bandwidth data of the data semantic fusion module; the medium-term prediction combines the weekly business fluctuation pattern; and the long-term prediction incorporates seasonal scene changes.

[0026] Furthermore, the closed-loop module is specifically as follows:

[0027] Based on the scene parameters of the digital twin module and the resource scheduling results of the scheduling module, exclusive task templates are designed for different laying scenarios, and the operational requirements of each step are clarified.

[0028] During the task execution phase, key indicator monitoring fields are set in the template. Data is collected in real time through edge terminals and the indicator compliance rate is displayed on the cloud dashboard. If the indicator is abnormal, an alert is automatically triggered and adjustment suggestions are pushed. After the task is completed, the template adaptability is scored from three dimensions: task time, fault recurrence rate, and business recovery quality. When the score is below 80 points, optimization is automatically started and the tools are identified.

[0029] In addition, a knowledge graph of scenarios and tasks is established to associate task experience in different scenarios. When creating a new template, reusable steps are automatically recommended to reduce the template development cycle. The optimized template is synchronized to the decision inference module.

[0030] Furthermore, the decision deduction module is specifically as follows:

[0031] By integrating the business semantic data from the data semantic fusion module, the twin simulation data from the digital twin module, and the health prediction results from the health prediction module, an evaluation system is established that considers five dimensions: cost, business assurance, long-term benefits, resource utilization, and carbon emissions, and a multi-objective decision-making scheme is generated.

[0032] When business parameters change, the dynamic simulation engine automatically recalculates and outputs suggestions for adjusting the plan. After the plan is implemented, actual effect data is collected periodically, compared with the simulation expectations, and an effect difference analysis report is generated. The analysis results are fed back to each module to correct the health prediction factors of the health prediction module and optimize the task template steps of the closed-loop module.

[0033] The beneficial effects of this invention are as follows:

[0034] 1. This invention constructs three semantic libraries—device, environment, and service—through a data semantic fusion module. The device semantic library transforms OTDR curve features and alarm codes into descriptions directly understandable to maintenance personnel, such as attenuation anomalies and connector failures. The environment semantic library associates soil subsidence, temperature and humidity changes with optical cable compression risks and accelerated sheath aging. The service semantic library maps bandwidth utilization and service interruption records as load pressure levels and service impact degrees. Simultaneously, it trains a weight model using historical maintenance data, dynamically adjusting semantic association coefficients in real-time service scenarios such as peak hours on government dedicated lines. This, combined with an edge-cloud collaborative semantic processing architecture, reduces edge computing power consumption and adds an SLA compliance risk level field to align with service fulfillment requirements. Furthermore, a digital twin module embeds a service load heatmap, using red, yellow, and green color gradients to represent the load density of different optical cable segments, quickly locating hotspot segments and transforming scattered data into effective information supporting maintenance decisions.

[0035] 2. This invention employs a scene feature distillation algorithm to extract core features from original scene parameters such as soil hardness, underwater pressure, and temperature and humidity fluctuation range. New scenes only require 10-15 sets of samples to complete migration adaptation. Simultaneously, it determines true and false alarms through three dimensions: environment, business, and equipment. The environment dimension considers the intensity of heavy rain and the level of strong winds; the business dimension monitors whether bandwidth fluctuations exceed thresholds; and the equipment dimension analyzes whether the OTDR curve exhibits continuous attenuation. A weighted score is then used to determine the alarm type: less than 40 points indicates a false alarm, and greater than 60 points indicates a true alarm. The intermediate range is manually verified. An unknown fault feature capture module is embedded, and federated learning is used to update the scene and fault association feature library, improving fault diagnosis accuracy in complex scenarios and reducing ineffective maintenance costs.

[0036] 3. This invention introduces a coupling coefficient to quantify the synergistic impact of service load, scene loss, and optical cable aging on optical cable health, enabling health prediction in the short, medium, and long term. When the optical cable health falls below a threshold, adjustment suggestions are automatically pushed to the business management system, and the digital twin business distribution is updated synchronously. The scheduling module establishes a three-dimensional resource capability assessment system for personnel, equipment, and tools, and performs resource pre-scheduling in conjunction with a business loss prediction algorithm. At the same time, it constructs a regional resource pool and a collaborative scheduling center to fill resource gaps in remote mountainous areas. The decision-making and deduction module establishes a five-dimensional assessment system for cost, business assurance, long-term benefits, resource utilization, and carbon emissions, dynamically optimizes the operation and maintenance plan, and compares the actual effects to effectively balance operation and maintenance cost control and business assurance requirements. Attached Figure Description

[0037] Figure 1 This is a flowchart of the intelligent operation and maintenance platform for fiber optic communication lines for operators according to the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] like Figure 1 As shown in the figure, this embodiment of the invention provides an intelligent operation and maintenance platform for optical fiber communication lines for operators. The data semantic fusion module is specifically as follows:

[0040] First, three semantic libraries are established. The device semantic library transforms OTDR curve features and alarm codes (such as LOS signal loss) into descriptions that maintenance personnel can directly understand, such as attenuation anomalies and connector failures. The environmental semantic library associates soil subsidence, temperature and humidity changes with concerns including optical cable compression risks and accelerated sheath aging. The business semantic library maps bandwidth utilization and business interruption records to load pressure levels and business impact levels. To adapt to business fluctuation scenarios, a weight model is trained based on historical maintenance data, and the semantic association coefficient is adjusted in combination with real-time business status (such as morning peak hours for government dedicated lines and busy hours for 5G base station backhaul). For example, when the business bandwidth exceeds the threshold, the association weight between "device alarm - LOS" and "loose optical cable connector" will increase by 30%, making the semantic association fit the actual business needs.

[0041] The specific implementation of semantic correlation coefficient adjustment: The weight model presets a basic correlation coefficient matrix for three core business scenarios (government dedicated lines, 5G base station backhaul, and home broadband access). The real-time business status trigger adjustment conditions are "bandwidth utilization rate exceeds 80%", "alarm frequency ≥ 2 times within 5 minutes" and "scenario parameter change (such as construction start-up, sudden weather change)". The adjustment range is calculated as "basic coefficient × (1 + business priority coefficient)", where the priority coefficient for government dedicated lines is 0.3, 5G base station backhaul is 0.2, and home broadband is 0.1. For example, the basic correlation coefficient for "equipment LOS alarm - loose optical cable connector" is 0.6. After the adjustment is triggered during the morning peak of government dedicated lines, the correlation coefficient is updated to 0.6 × (1 + 0.3) = 0.78.

[0042] The specific implementation of the edge-cloud collaborative architecture is as follows: The semantic extraction unit at the edge adopts a lightweight CNN model, with local processing latency ≤100ms. It only uploads combined semantic data of "risk level ≥ medium + business load pressure ≥ high" (data volume ≤15% of the original data) and 3 types of scenario parameters (laying type: pipeline / direct burial / underwater; real-time weather: sunny / rainy / strong wind; construction status: no construction / road occupation construction / near construction). The cloud arbitration mechanism goes through a three-step process of "scenario parameter matching → global business conflict detection → semantic deviation correction". The correction threshold is set to 0.05 (correction is triggered if the semantic similarity is lower than this value). The corrected data is synchronized back to the edge to update the semantic library.

[0043] The SLA compliance risk level field has been expanded and implemented as follows: The field is divided into four levels: "Low / Medium / High / Extremely High". The mapping rules are as follows: Bandwidth utilization ≤ 60% and downtime ≤ 5 minutes → Low risk; 60% < bandwidth utilization ≤ 80% or 5 minutes < downtime ≤ 15 minutes → Medium risk; 80% < bandwidth utilization ≤ 95% or 15 minutes < downtime ≤ 30 minutes → High risk; Bandwidth utilization > 95% or downtime > 30 minutes → Extremely high risk. The field is directly associated with the "Availability Commitment" in the business SLA agreement and supports automatic lookup of the corresponding SLA clause number.

[0044] Considering the limitations of edge computing power and the global scheduling requirements of the cloud, a collaborative semantic processing architecture between the edge and the cloud is designed. Semantic extraction units are deployed at the edge to locally complete the semantic conversion of OTDR curves and temperature and humidity data. For example, 5mm soil settlement is directly converted into optical cable compression risk. Only high-value combined semantic data, including optical cable compression risk and high business carrying pressure, and scene parameters (such as laying type, real-time weather, and construction status) are uploaded to the cloud. The cloud then uses a cross-domain arbitration mechanism for semantic conflicts to combine the scene parameters of the edge (such as whether there is construction and real-time weather) with the global business data (cross-regional bandwidth distribution and backup link status) held by the cloud to correct semantic deviations, thereby reducing the computing power consumption of the edge and improving the fusion accuracy.

[0045] To directly align with business performance requirements, an SLA compliance risk level field has been added to the business semantic library, linking bandwidth utilization and downtime with the business SLA agreement, thus matching semantic data with business performance assessment.

[0046] The digital twin module is specifically as follows:

[0047] Based on the semantic data of the aforementioned data semantic fusion module, a digital twin of the optical fiber line is established to realize the dynamic coupling and visualization of services and resources. In the service and resource coupling mapping layer of the twin, a service load heat map function is embedded. Through the color gradient of red, yellow, and green, red represents high load, yellow represents medium load, and green represents low load, presenting the service load density of different sections of the optical cable. For example, the government dedicated line section is often displayed in red due to high bandwidth demand, while the home broadband access section is mostly green. At the same time, hot spots with load exceeding the threshold are marked, allowing operation and maintenance personnel to quickly locate areas with concentrated resource pressure.

[0048] To balance short-term fault handling with long-term operation and maintenance planning, a multi-scenario parallel simulation engine was developed, capable of simulating both fault and business growth scenarios simultaneously. For example, when assessing the impact of fiber breakage in directly buried optical cables in mountainous areas, the business forecast of a 20% increase in bandwidth demand for 5G base stations in the area three months later can be incorporated. During the simulation, the semantic data (such as the business bearing semantics of the fiber breakage section) from the data semantic fusion module and the optical cable health data (such as real-time attenuation value and aging coefficient) from the health prediction module are called. Finally, the SLA loss rate of the affected services and the priority ranking of resource expansion are output, such as prioritizing fiber breakage repair over new bandwidth expansion, providing a basis for operation and maintenance decisions at different time dimensions.

[0049] To ensure consistency between the digital twin and the physical network, a two-way calibration mechanism is established. Real-time data of the physical optical cable, such as fiber attenuation value and connector temperature, is collected by edge sensors and compared point by point with the simulated data of the digital twin. When the deviation exceeds the threshold, such as the deviation between the simulated attenuation value and the actual value > 0.1dB / km, the scene loss factor and other parameters of the digital twin are automatically adjusted. The calibrated parameters are synchronously fed back to the data semantic fusion module and the health prediction module to optimize the semantic fusion accuracy and the health prediction model.

[0050] The fault migration diagnosis module is specifically as follows:

[0051] Based on the semantic data of the data semantic fusion module and the scene parameters of the digital twin in the digital twin module, in order to solve the fault diagnosis adaptation problem under different laying scenarios and distinguish between real and false alarms to avoid ineffective operation and maintenance, a scene feature distillation algorithm is designed to extract core features from the original scene parameters (soil hardness, underwater pressure, temperature and humidity fluctuation range). For example, soil hardness >20MPa in mountainous areas will be refined into rock compression risk features. New scenarios only need to input 10-15 sets of scene samples to complete the migration adaptation, reducing the sample requirements.

[0052] To avoid misjudgments caused by environmental interference, the system determines the authenticity of alarms based on three dimensions: environment, business, and equipment. The environmental dimension considers parameters such as rainstorm intensity (>50mm / 24h) and wind level. The business dimension monitors whether bandwidth fluctuations exceed the ±10% threshold. The equipment dimension analyzes whether there is continuous attenuation in the OTDR curve. The alarm type is determined by a weighted score (environment 30% + business 40% + equipment 30%). A score less than 40 is considered a false alarm (such as temporary attenuation caused by rainstorms), a score greater than 60 is considered a true alarm (such as fiber optic cable breakage), and scores in between are manually verified.

[0053] To address new types of faults, an unknown fault feature capture module is embedded. When a new fault that does not match the historical fault database is diagnosed, such as intermittent attenuation caused by aging of the optical cable sheath, the system automatically records the scene characteristics, equipment data, and handling solutions for the fault. The system updates the scene and fault association feature database through federated learning, and the updated feature database will synchronously support the health prediction module's health prediction.

[0054] Weighted scoring formula for true and false alarms:

[0055] ;

[0056] In the formula: This represents the total alarm score, ranging from 0 to 100. A value less than 40 is considered a false alarm. A value greater than 60 indicates a true alarm, while a value less than or equal to 40 indicates a false alarm. ≤60 requires manual review;

[0057] This represents the environmental dimension score, calculated based on environmental parameters, with a value ranging from 0 to 100. It references parameters such as rainfall intensity (>50mm / 24h) and wind speed (>6). The more severe the environmental interference, such as rainfall intensity >100mm / 24h or wind speed >8, the higher the score. The higher the level, the more severe the interference. ≥80 points, without interference ≤30 points;

[0058] This represents the environmental dimension weight, which is fixed at 0.3, reflecting the degree of influence of environmental interference on alarm judgment.

[0059] This represents a business-level score, calculated based on the business operation status, with a value ranging from 0 to 100. It monitors whether bandwidth fluctuations exceed a ±10% threshold; the higher the synchronization between business fluctuations and alarms, the better. "The higher";

[0060] This represents the weight of the business dimension, which is fixed at 0.4, reflecting the core role of business impact in alarm determination;

[0061] This represents the device-level score, calculated based on device data, with a value ranging from 0 to 100. It analyzes whether the OTDR curve shows continuous decay; the more obvious the device's abnormal characteristics, the better. The higher;

[0062] This represents the weight of the device dimension, which is fixed at 0.3, reflecting the fundamental role of device status in alarm determination.

[0063] The scheduling module is specifically as follows:

[0064] By combining the fault diagnosis results of the fault migration diagnosis module with the health prediction data of the health prediction module, operation and maintenance resources are dynamically scheduled to ensure fault recovery and resource adaptation for high-priority services. First, a resource capability assessment system is established in three dimensions: personnel, equipment, and tools. The personnel dimension assesses whether they have underwater welding certification and continuous operation time; the equipment dimension monitors the accuracy of the welding machine (e.g., whether the welding loss is ≤0.05dB) and tests the equipment's battery life; the tool dimension verifies the integrity rate of underwater operation equipment, etc., and outputs a resource adaptation score. For example, a resource combination with underwater welding certified personnel, high-precision welding machines, and 100% integrity rate of underwater operation equipment can achieve an adaptation rate of 90%, avoiding resource mismatch.

[0065] To proactively address potential failures, a business loss prediction and resource scheduling linkage algorithm was developed. When the health prediction module predicts a decline in the health of high-priority service-bearing optical cables, it calculates the potential business loss rate in advance. For example, if the health level is less than 60%, the loss rate is 50%. Resources are pre-scheduled, such as adjusting nearby splicing teams with the corresponding skills to stand by along the optical cable line, thus shortening the response time after a failure occurs.

[0066] To address the issue of insufficient regional resources, a regional resource pool and collaborative dispatch center architecture are established. When special repair equipment is lacking in remote mountainous areas, the collaborative dispatch center queries the idle resources in neighboring areas, calculates the balance between resource allocation costs and business losses, and triggers dispatch if the allocation cost is less than the hourly business loss. At the same time, it generates the optimal transportation route by combining real-time road conditions and equipment transportation requirements. After the dispatch is completed, the relevant resource usage will be fed back to the closed-loop module to provide a reference for the resource configuration of scenario-based task templates.

[0067] The specific implementation of the business loss prediction and resource scheduling linkage algorithm is as follows: The algorithm input parameters include optical cable health, business priority, and historical fault loss data. The loss rate calculation formula is "potential loss rate = (1 - optical cable health / threshold) × business priority coefficient × historical average loss of similar faults", where the priority coefficient for core business is 1.5 and that for ordinary business is 1.0. The pre-schedule trigger condition is "potential loss rate ≥ 30%". The scheduling process is "algorithm calculates loss rate → marks high-risk optical cable sections → queries regional resource pool for suitable resources → generates pre-schedule instruction → notifies the operation and maintenance team to stand by". The pre-schedule response delay is ≤ 1 hour.

[0068] The specific implementation of the regional resource pool and collaborative dispatch center architecture is as follows: The regional resource pool is stored in categories of "personnel / equipment / tools". Personnel resources are labeled with certification qualifications, location, and idle status; equipment resources are labeled with model, accuracy, battery life, and idle status; tool resources are labeled with type, integrity rate, and storage location. The collaborative dispatch center uses the Dijkstra algorithm to plan transportation routes. The cost and loss balance value is calculated as "total allocation cost (transportation fee + idle loss) ≤ hourly business loss × estimated repair time". If this condition is met, cross-regional dispatch is triggered. During the dispatch process, "resource location → transportation progress → arrival time" are synchronized to the cloud dashboard in real time. After the resources arrive, the resource pool status is automatically updated to "occupied". After the repair is completed, it is updated to "idle" and the usage time is recorded.

[0069] The health prediction module is specifically as follows:

[0070] Based on the business semantic data of the data semantic fusion module and the scene parameters of the digital twin module, the synergistic impact of business load and scene loss on optical cable health is quantified, multi-cycle health prediction is achieved, and business adjustments are linked. A coupling coefficient K is introduced to quantify the synergistic impact. The load factor of high-frequency data transmission services (such as 5G enterprise leased lines) is higher than that of ordinary services, the loss factor of high temperature and high humidity environments is higher than that of normal temperature environments, and the longer the service life of the optical cable, the higher the correction coefficient. The actual loss of the optical cable is reflected through the health calculation formula.

[0071] When the predicted health of the optical cable is below a threshold, adjustment suggestions are automatically pushed to the business management system. For example, if the health of the government dedicated line segment is less than 50%, 30% of non-core business is temporarily diverted to the backup optical cable. The business distribution of the digital twin module is updated simultaneously, so that the digital twin can reflect the resource status after the business adjustment in real time. It has three-week prediction functions: short-term (1 week), medium-term (1 month), and long-term (3 months). The short-term prediction uses the real-time attenuation value and bandwidth data of the data semantic fusion module. The medium-term prediction combines the weekly business fluctuation pattern, such as the difference between weekdays and weekends. The long-term prediction incorporates seasonal scene changes, such as rainy season corrosion and winter low temperature, to support the fault prevention of the fault migration diagnosis module, the resource scheduling plan of the scheduling module, and the spare parts procurement decision of the decision inference module, respectively.

[0072] The formula for calculating the coupling coefficient K is as follows:

[0073] ;

[0074] In the formula: This represents the coupling coefficient, quantifying the synergistic impact of service load and scenario loss on optical cable health. Its value range varies depending on the scenario, such as high-frequency services. High temperature environment 4-year fiber optic cable =1.8; Regular business normal temperature environment New optical cable, =0.6;

[0075] This represents the service load factor, a quantitative factor set according to the service type and bandwidth characteristics. The load factor for high-frequency data transmission services (such as 5G enterprise leased lines) is higher than that for ordinary services (such as home broadband). The value is dynamically set in combination with the service attributes.

[0076] This represents the scene loss factor, a quantitative factor set according to the characteristics of the laying environment. The loss factor in high temperature and high humidity environments is higher than that in normal temperature environments, and the loss factor in rainy season and winter low temperature environments is higher than that in normal environments. The value needs to be combined with environmental parameters, such as temperature, humidity, soil characteristics, etc.

[0077] This represents the correction factor, a compensation factor adjusted according to the service life of the optical cable. The longer the service life of the optical cable, the higher the correction factor. The correction factor is 1.0 for 1-3 years, 1.1 for 3-5 years, and 1.2 for more than 5 years. It is used to compensate for the cumulative effect of optical cable aging on loss.

[0078] Formula for calculating the health of optical cables:

[0079] ;

[0080] In the formula: This indicates the health status of the optical cable, used to quantify its current health condition. The health threshold for the core optical cable is 60, and the threshold for the access optical cable is 50. A health warning is triggered when the health status falls below the threshold.

[0081] This represents the basic health value, which is the initial health benchmark value set according to the optical cable model. The basic health value of single-mode G.652D optical cable is 100, and the basic health value of special underwater optical cable is 120. It is only related to the type of optical cable and is a fixed initial value.

[0082] This represents the coupling coefficient, and is calculated using the same formula as the coupling coefficient K. Quantify the impact of collaborative losses between business and scenarios to ensure that health calculations are relevant to actual operation and maintenance scenarios;

[0083] This indicates the runtime, the cumulative operating time of the optical cable after it has been put into use, expressed in months, and is used to calculate the impact of long-term losses on the health status.

[0084] The closed-loop module is specifically as follows:

[0085] Based on the scene parameters of the digital twin module and the resource scheduling results of the scheduling module, exclusive task templates are designed for different laying scenarios such as underwater, direct burial in mountainous areas, and permafrost in plateau regions. For example, the underwater optical cable repair template includes steps such as diving equipment inspection → water temperature monitoring and control (ensuring 5-25℃) → underwater fault location (using sonar detector) → low-pressure welding → waterproof sealing test (pressure 0.8MPa for 30 minutes). The operational requirements of each step are clearly defined, such as the requirement to control the water temperature to 5-25℃ for underwater welding.

[0086] During task execution, key indicator monitoring fields are set in the template, such as the welding loss value and pressure test value for underwater repair. Data is collected in real time through edge terminals, and the indicator compliance rate is displayed on the cloud dashboard. For example, welding loss ≤0.05dB is considered compliant, and the current compliance rate is 90%. If the indicator is abnormal (such as loss exceeding 0.1dB), an early warning is automatically triggered, and adjustment suggestions are pushed, such as calibrating the welding machine parameters. After the task is completed, the template adaptability is scored from three dimensions: task time, fault recurrence rate, and business recovery quality. When the score is below 80 points, optimization is automatically initiated. For example, if the time taken for the direct burial template in mountainous areas exceeds expectations, rock cleaning pretreatment steps are added and the tools are specified.

[0087] In addition, a knowledge graph of scenarios and tasks is established to link task experience in different scenarios. For example, underwater waterproofing technology can be transferred to pipeline repair during the rainy season, and reusable steps can be automatically recommended when creating a new template for desert optical cable maintenance, thereby reducing the template development cycle. The optimized template is synchronized to the decision-making module to provide task execution parameters for decision-making.

[0088] The three-dimensional scoring system is implemented as follows: the scoring dimensions are "task time (weight 40%), failure recurrence rate (weight 30%), and business recovery quality (weight 30%)".

[0089] The scoring criteria are as follows: Task time ≤ preset time → 40 points; exceeding the preset time by less than 10% → 30 points; exceeding by 10%-30% → 20 points; exceeding by more than 30% → 10 points.

[0090] Fault recurrence rate ≤1% → 30 points, 1% < recurrence rate ≤3% → 20 points, 3% < recurrence rate ≤5% → 10 points, recurrence rate >5% → 0 points;

[0091] Business recovery quality ≥ 99.9% → 30 points, 99.5% ≤ recovery quality < 99.9% → 20 points, 99% ≤ recovery quality < 99.5% → 10 points, recovery quality < 99% → 0 points;

[0092] The total score is the sum of the scores of each dimension. A score of 80 or above indicates a good fit, 60-79 indicates a basic fit (requires local optimization), and below 60 indicates an unsuitable fit (requires template reconstruction).

[0093] Specific implementation of knowledge graph construction and cross-scenario reuse: The knowledge graph is constructed according to a six-layer structure of "scenario type → fault type → task template → operation steps → tool requirements → experience parameters". The associated data includes the scenario parameters, fault characteristics and task execution data of each module.

[0094] The cross-scenario reuse trigger condition is "the similarity between the new scenario and the historical scenario is ≥70%". The similarity calculation is based on four core elements: "layout type, environmental parameters, business type, and fault characteristics". The reuse process is "new scenario trigger graph query → matching high similarity historical scenarios → recommending reusable steps and tool parameters → embedding the new template after manual confirmation", which can reduce the template development cycle and the reuse record is automatically synchronized to the closed-loop module knowledge base.

[0095] The decision deduction module is specifically as follows:

[0096] By integrating the business semantic data from the data semantic fusion module, the twin simulation data from the digital twin module, and the health prediction results from the health prediction module, an evaluation system is established that considers five dimensions: cost, business assurance, long-term benefits, resource utilization, and carbon emissions. Taking an aging optical cable as an example, solutions such as early replacement, enhanced inspection, and on-demand capacity expansion can be generated. The evaluation results of the early replacement solution include an increase of 200,000 yuan in maintenance costs, a 90% reduction in business interruption rate, a 40% increase in resource utilization over 3 years, and a reduction of 15 tons in carbon emissions, supporting multi-objective decision-making such as cost reduction and carbon neutrality.

[0097] When business parameters change, such as the addition of 5G base stations exceeding expectations by 50%, the dynamic simulation engine automatically recalculates and outputs suggestions for adjusting the plan. For example, the original on-demand expansion is changed to expansion three months in advance. At the same time, the decision deviation rate is calculated, and the difference in business assurance rate before and after the adjustment is less than 5% to be acceptable. After the plan is implemented, actual effect data is collected regularly, such as the bandwidth utilization rate after expansion and the failure rate after replacing optical cables. These are compared with the simulation expectations to generate an effect difference analysis report. For example, if the actual failure rate exceeds expectations by 10%, the reason is that the impact of low winter temperatures was not considered. The analysis results are fed back to each module to correct the health prediction factors of the health prediction module and optimize the task template steps of the closed-loop module. This enables multi-plan simulation, dynamic optimization, and effect tracking, providing operators with optimal operation and maintenance decisions.

[0098] Formula for calculating decision bias rate:

[0099] ;

[0100] In the formula: This represents the decision deviation rate, used to quantify the change in business assurance capabilities before and after adjustments to the operation and maintenance plan. Its value ranges from 0% to 100%. A deviation of less than 5% is acceptable and does not require re-analysis. ≥5% requires further optimization of the plan;

[0101] This represents the service availability rate after the plan adjustment. It's the predicted or actual service availability rate achieved after adjusting the operation and maintenance plan, ranging from 0-100%. For example, replacing the fiber optic cable in advance reduced the service interruption rate by 90%. =99.99%;

[0102] This represents the service availability rate before the plan was adjusted, the initial service availability rate before the operation and maintenance plan was adjusted, and its value ranges from 0 to 100%. They are from the same source and are both inverse indicators of SLA compliance rate or business interruption rate, ensuring that the deviation calculation benchmark is consistent.

[0103] This represents the absolute difference in the business assurance rate before and after the adjustment.

[0104] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0105] Although embodiments of the invention have been shown and described, it will be understood by those skilled 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 invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent operation and maintenance platform for fiber optic communication lines for telecom operators, characterized in that: The platform includes: The data semantic fusion module first establishes a device semantic library, an environment semantic library, and a business semantic library. The device semantic library transforms OTDR curve features and alarm codes into descriptions that maintenance personnel can directly understand. The environment semantic library associates soil subsidence, temperature and humidity changes with concerns such as optical cable compression risks and accelerated sheath aging. The business semantic library maps bandwidth utilization and service interruption records to load pressure levels and service impact levels. By adjusting the semantic association coefficients to ensure that semantic associations align with actual business needs, an edge-cloud collaborative semantic processing architecture is designed. Semantic extraction units are deployed at the edge to locally complete the semantic conversion of OTDR curves and temperature and humidity data, uploading only high-value combined semantic data and scenario parameters, including optical cable compression risks and high service load pressure, to the cloud. The cloud then corrects semantic deviations by combining edge scenario parameters with global business data from the cloud through a cross-domain semantic conflict arbitration mechanism. Finally, an SLA compliance risk level field is added to the business semantic library. Digital twin module: Establish a digital twin of the optical fiber line, embed a service load heat map, develop a multi-scenario pre-drilling engine, establish a two-way calibration mechanism between the twin and the physical optical cable, collect real-time data of the physical optical cable through edge sensors, and compare it point by point with the simulated data of the digital twin to ensure consistency between the twin and the physical optical cable; Fault migration diagnosis module: It adopts the scene feature distillation algorithm to reduce the sample requirements of new scenes, establishes a three-dimensional model of environment, business and equipment to judge true and false alarms, sets up an unknown fault feature capture module, and updates the scene and fault association feature library through federated learning; Scheduling module: First, establish a resource capability assessment system for personnel, equipment and tools; then, develop a business loss prediction and resource scheduling linkage algorithm to deal with potential failures; and finally, establish a regional resource pool and collaborative scheduling center architecture to fill resource gaps. Health prediction module: It uses coupling coefficients to quantify the impact of business and scenarios on the health of optical cables, pushes adjustment suggestions to the business management system, and updates the digital twin. It has short-term, medium-term and long-term prediction functions. Closed-loop module: Real-time monitoring of task metrics, scoring template adaptability through three dimensions: task time, fault recurrence rate, and business recovery quality, and establishing a knowledge graph for cross-scenario knowledge reuse; Decision simulation module: Establish an evaluation system with five dimensions: cost, business assurance, long-term benefits, resource utilization, and carbon emissions; generate operation and maintenance plans; dynamically simulate and optimize the plans; and generate a performance difference analysis report by comparing the effect data with the simulation expectations.

2. The intelligent operation and maintenance platform for fiber optic communication lines for operators according to claim 1, characterized in that: The data semantic fusion module trains a weight model based on historical operation and maintenance data, and adjusts the semantic association coefficients in combination with real-time business status to make the semantic association fit the actual business needs. Add an SLA compliance risk level field to the business semantic library to associate bandwidth utilization and downtime with the business SLA agreement, so that semantic data matches business performance assessment.

3. The intelligent operation and maintenance platform for fiber optic communication lines for operators according to claim 2, characterized in that: The digital twin module establishes a digital twin of the optical fiber line based on the semantic data of the data semantic fusion module. In the service and resource coupling mapping layer of the twin, a service load heat map function is embedded. Through the red, yellow and green color gradients, the service load density of different sections of the optical cable is presented, and hot spots with load exceeding the threshold are marked. A multi-scenario parallel pre-simulation engine was developed, which has the function of simulating both fault and business growth scenarios simultaneously. During the pre-simulation process, the semantic data of the data semantic fusion module and the optical cable health data of the health prediction module are called. Finally, the SLA loss rate of the affected business and the priority ranking of resource expansion are output to provide a basis for operation and maintenance decisions. A bidirectional calibration mechanism is established between the digital twin and the physical optical cable. Real-time data of the physical optical cable is collected by an edge sensor and compared point by point with the simulated data of the digital twin. When the deviation exceeds a threshold, the parameters of the digital twin are automatically adjusted. The calibrated parameters are synchronously fed back to the data semantic fusion module and the health prediction module.

4. The intelligent operation and maintenance platform for fiber optic communication lines for operators according to claim 3, characterized in that: The fault migration diagnosis module designs a scene feature distillation algorithm based on the semantic data of the data semantic fusion module and the scene parameters of the digital twin in the digital twin module. It extracts core features from the original scene parameters. The new scene only needs to input 10-15 sets of scene samples to complete the migration adaptation. The system distinguishes between false alarms and genuine alarms by considering three dimensions: environment, business, and equipment. The environmental dimension refers to parameters such as rainstorm intensity and wind level. The business dimension monitors whether bandwidth fluctuations exceed the threshold. The equipment dimension analyzes whether the OTDR curve shows continuous attenuation. The alarm type is determined by a weighted score: a score less than 40 is considered a false alarm, a score greater than 60 is considered a genuine alarm, and scores in between are manually reviewed. An embedded unknown fault feature capture module is used. When a new fault that does not match the historical fault database is diagnosed, the module automatically records the scene features, equipment data and handling solutions of the fault, and updates the scene and fault association feature database through federated learning.

5. The intelligent operation and maintenance platform for fiber optic communication lines for operators according to claim 4, characterized in that: The scheduling module combines the fault diagnosis results of the fault migration diagnosis module with the health prediction data of the health prediction module to first establish a resource capability assessment system in three dimensions: personnel, equipment, and tools. The personnel dimension assesses whether underwater welding certification is available and the continuous operation time; the equipment dimension monitors the welding machine accuracy and tests the equipment's battery life; and the tool dimension verifies the integrity rate of underwater operation equipment and outputs a resource adaptability score. Develop a service loss prediction and resource scheduling linkage algorithm. When the health prediction module predicts that the health of high-priority service-bearing optical cables is declining, it calculates the potential service loss rate in advance and performs resource pre-scheduling. A regional resource pool and collaborative dispatch center architecture is established. When special repair equipment is lacking in remote mountainous areas, the collaborative dispatch center queries the idle resources in the nearby areas, calculates the balance between resource allocation costs and business losses, and generates the optimal transportation route by combining real-time road conditions and equipment transportation requirements. After the dispatch is completed, the relevant resource usage will be fed back to the closed-loop module.

6. The intelligent operation and maintenance platform for fiber optic communication lines for operators according to claim 5, characterized in that: The health prediction module is based on the business semantic data of the data semantic fusion module and the scene parameters of the digital twin module. It introduces a coupling coefficient to quantify the synergistic effect. The load factor of high-frequency data transmission services is higher than that of ordinary services, the loss factor of high temperature and high humidity environment is higher than that of normal temperature environment, and the longer the service life of the optical cable, the higher the correction coefficient. The actual loss of the optical cable is reflected through the health calculation formula. When the predicted optical cable health is below the threshold, adjustment suggestions are automatically pushed to the business management system. At the same time, the business distribution of the digital twin in the digital twin module is updated synchronously, so that the digital twin can reflect the resource status after the business adjustment in real time. It has prediction functions for three periods: short-term, medium-term and long-term. The short-term prediction uses the real-time attenuation value and bandwidth data of the data semantic fusion module; the medium-term prediction combines the weekly business fluctuation pattern; and the long-term prediction incorporates seasonal scene changes.

7. The intelligent operation and maintenance platform for fiber optic communication lines for operators according to claim 6, characterized in that: Based on the scenario parameters of the digital twin module and the resource scheduling results of the scheduling module, the closed-loop module designs exclusive task templates for different laying scenarios and clarifies the operational requirements of each step. During the task execution phase, key indicator monitoring fields are set in the template. Data is collected in real time through edge terminals and the indicator compliance rate is displayed on the cloud dashboard. If the indicator is abnormal, an alert is automatically triggered and adjustment suggestions are pushed. After the task is completed, the template adaptability is scored from three dimensions: task time, fault recurrence rate, and business recovery quality. When the score is below 80 points, optimization is automatically started and the tools are identified. In addition, a knowledge graph of scenarios and tasks is established to associate task experience in different scenarios. When creating a new template, reusable steps are automatically recommended to reduce the template development cycle. The optimized template is synchronized to the decision inference module.

8. The intelligent operation and maintenance platform for fiber optic communication lines for operators according to claim 7, characterized in that: The decision-making module integrates the business semantic data from the data semantic fusion module, the twin simulation data from the digital twin module, and the health prediction results from the health prediction module to establish an evaluation system with five dimensions: cost, business assurance, long-term benefits, resource utilization, and carbon emissions, and generates multi-objective decision-making schemes. When business parameters change, the dynamic simulation engine automatically recalculates and outputs suggestions for adjusting the plan. After the plan is implemented, actual effect data is collected periodically, compared with the simulation expectations, and an effect difference analysis report is generated. The analysis results are fed back to each module to correct the health prediction factors of the health prediction module and optimize the task template steps of the closed-loop module.

Images