Ai-based self-improving integrated control system for tunnel optical cable anomaly response

Abstract

The present invention analyzes attenuation data, which includes at least one of the optical signal loss value per unit length (dB / km) installed in each of a plurality of tunnels, reflected signal strength (dB) at a specific point, change in attenuation value over time, average attenuation value per section, and cumulative change in attenuation value, as well as environmental data obtained by measuring the environment of each tunnel and including at least one of internal temperature, vibration, and water level, to calculate location coordinates and a risk index. When the risk index of a specific optical cable exceeds a reference value, the optical path to which the optical cable belongs is switched, and the location of the damage and the route change of the optical cable are notified to the operator, thereby enabling the repair of the optical cable without interruption of communication.

Description

Technology Field

[0001] The present invention relates to an AI-based self-improving integrated control system for responding to tunnel optical cable anomalies, which monitors the condition of optical communication lines installed in tunnels in real time to predict anomalies at a pre-damage stage and enables the maintenance of uninterrupted communication in a multi-tunnel environment. Background Technology

[0002] Condition detection devices, environmental sensors, and switch control devices are installed inside the tunnel to form a communication network. As tunnels are formed with a sealed structure, they are continuously exposed to various environmental factors such as fire, flooding, vehicle vibration, structural deformation, and temperature rise. These environmental factors can cause sheath degradation, localized crimping, increased bending, or microcracks in optical cables, which can ultimately lead to a deterioration in communication quality or disconnection.

[0003] Conventional tunnel optical cable management technologies mostly rely on detecting failures only after the cable is completely severed or a communication failure occurs. This reactive approach leads to a time delay between the occurrence and detection of a failure, resulting in the interruption of communication services. In particular, since in-tunnel CCTVs and security communication networks require real-time performance, a post-severance response method presents a critical limitation.

[0004] Furthermore, conventional technologies often target only rapid attenuation changes exceeding a certain threshold, which presents a problem in that they cannot precisely analyze the rate of minute attenuation changes accumulated over the long term. Although optical cable damage often manifests as an increase in minute attenuation prior to disconnection, the inability to systematically analyze these early signs makes it difficult to detect the risk immediately preceding the disconnection in advance.

[0005] Furthermore, conventional systems have structures that indicate the location of damage solely by distance values ​​or require manual verification procedures. Since the interior of a tunnel has a complex structure, it is difficult to pinpoint the exact physical location using only simple distance information. This necessitates additional on-site search by maintenance personnel, leading to delays in recovery time.

[0006] Meanwhile, in conventional technology, optical path switching is often performed manually or switching decisions are made in a decentralized manner by individual devices in the event of a failure. In such cases, the absence of an integrated risk assessment structure leads to a lack of consistency in switching criteria and causes problems such as switching delays or the need for human intervention. As a result, it is difficult to completely prevent communication interruptions.

[0007] Furthermore, although multiple tunnels are often connected to the same network in wide-area communication network environments, conventional technology lacks a structure for clearly identifying and integrally managing tunnels. This can lead to problems such as data interference between tunnels, location identification errors, and unclear management history.

[0008] Furthermore, conventional systems typically determine risk based on fixed reference values ​​or preset thresholds. However, tunnel environments are constantly changing due to seasonal variations, traffic volume, structural conditions, and external environmental shifts. Fixed threshold methods fail to adapt to these environmental changes, leading to increased false positives or missed detections, and suffer from limitations such as degraded prediction accuracy during long-term operation.

[0009] Furthermore, conventional technology has not sufficiently incorporated a closed-loop structure to reflect maintenance results in the analysis model. In other words, judgment criteria are not improved even after fault response is completed, and there is a possibility that the same type of false positives will be repeated.

[0010] Therefore, a system with a central integrated processing structure is required to analyze the rate of change in minute attenuation prior to the disconnection stage, accurately calculate the location of damage, automatically switch optical paths based on a risk index, and perform self-improvement after relearning by reflecting maintenance results. Prior art literature

[0012] (Patent Document 0001) KR 10-2899844 B1(2025.12.09) The problem to be solved

[0013] The present invention aims to provide an AI-based self-improving integrated control system for responding to tunnel optical cable abnormalities, which detects the condition of an optical cable installed in a tunnel in real time and analyzes the measurement data at a central integrated control server to calculate the rate of change in minute attenuation, the coordinates of the predicted location of damage, and a risk index, thereby determining the possibility of damage in advance prior to the optical cable disconnection stage.

[0014] In addition, the present invention aims to provide a centralized optical path switching control structure that can minimize or prevent communication interruptions by automatically transmitting an optical path switching control command to activate a backup optical path when the risk index exceeds a reference value.

[0015] In addition, the present invention aims to continuously improve prediction accuracy by providing a self-improving structure that enables rapid on-site response by displaying damage prediction location information to the operator, stores maintenance results as training data, performs retraining, and corrects reference values ​​or weights based on the retraining results.

[0016] In addition, the present invention aims to improve the stability and management efficiency of a wide-area tunnel communication network by integrally managing multiple tunnels and accurately identifying the location of damage through a tunnel identification and location matching structure.

[0017] Furthermore, the present invention aims to implement an intelligent tunnel optical cable management system by forming a closed-loop structure in which risk assessment, automatic switching, operator notification, reflection of maintenance results, and self-improvement after relearning are organically linked and performed sequentially. means of solving the problem

[0018] An embodiment of the present invention includes an AI-based self-improving tunnel optical cable anomaly response integrated control system comprising: a state detection device that outputs attenuation data including at least one of an optical signal loss value per unit length (dB / km) installed in each of a plurality of tunnels, a reflected signal strength (dB) at a specific point, a change in attenuation value over time, an average attenuation value per section, and a cumulative change in attenuation value as time-series data collected along with time information; an environment sensor installed in each of a plurality of tunnels that measures the environment of each tunnel and outputs environment data including at least one of internal temperature, vibration, and water level; an integrated control server that receives and analyzes attenuation data and environment information from the state detection device and environment sensor installed in a plurality of tunnels to calculate a fine attenuation change rate, location coordinates, and a risk index; and a switch control device that switches the optical path to which the optical cable belongs under the control of the integrated control server when the risk index of a specific optical cable exceeds a reference value in the integrated control server.

[0019] In the above embodiment, the integrated control server is characterized by forming a central closed-loop structure that provides the operator with information on the predicted location of damage and performs relearning and self-improvement after storing maintenance results.

[0020] In addition, the integrated control server includes: a network interface unit that communicates with a state detection device, an environment sensor, and a switch control unit; a database that stores attenuation data received through the network interface unit and environment data from the environment sensor; an analysis control unit that analyzes the attenuation data stored in the database to calculate the attenuation change rate, calculates the distance value of the expected damage location based on the reflection signal delay information of the attenuation data, and transmits an optical path switching control command to the switch control unit if the risk index calculated by synthesizing the attenuation change rate and the environment data from the environment sensor exceeds a reference value; a tunnel management unit that aligns the distance value calculated by the analysis control unit with tunnel section information to convert and output it as physical location coordinates; and an AI self-improvement unit that stores risk judgment results and maintenance results as training data, and retrains a risk index calculation model using the stored training data to correct weights or reference values ​​used in calculating the risk index.

[0021] Here, the analysis control unit comprises: a fine attenuation analysis module that arranges attenuation data in chronological order, calculates the average of N measurements over a certain time interval to determine the difference from the previous reference average, and calculates a quantitative analysis value indicating the possibility of an anomaly in at least one of the attenuation increase rate per unit time and the cumulative increase amount over a certain time; a damage location calculation module that calculates a distance value using reflection delay time information of an optical signal and calculates physical location coordinates by matching the calculated distance value with the interval information provided by the tunnel management unit; a risk analysis module that determines the presence of risk by calculating a risk index by applying a weight set for vibration intensity to any one of the attenuation change rate, cumulative increase amount, and deviation from the reference line analyzed by the fine attenuation analysis module, the temperature change rate calculated by analyzing environmental data including temperature and vibration received from an environmental sensor, and a risk level analysis module that determines the presence of risk by comparing the calculated risk index with a set reference value; and a bypass control module that receives a judgment result from the analysis result of the risk analysis module indicating that the risk index has exceeded the reference value, generates an optical path switching control command to bypass the optical cable damage risk section, and transmits it to a switch control device. The system includes an operator notification module that outputs display information including at least one of tunnel identification information, damage location coordinates, risk index value, and attenuation change trend after transmitting a light path switching control command, or transmits it to a set operator terminal; wherein the weight of the risk analysis module is at least one of a preset value, a statistical analysis-based setting value, a past failure history-based setting value, and a dynamic correction value by an AI self-improvement unit.

[0022] In addition, the bypass control module is characterized by, if the optical cable network structure is a ring-type network structure where multiple tunnels or nodes are interconnected in a circular or closed-loop form and each node is connected to adjacent nodes in both directions, checking the direction of the section where the risk of damage is determined, controlling data bypass in the opposite direction of the currently used path, or checking whether the backup path in that direction is in a normal state, and then executing optical path switching.

[0023] Alternatively, if the optical cable network structure is a mesh network structure where multiple tunnels or nodes are interconnected via multiple paths and each node is directly connected to two or more nodes, the bypass control module generates a list of possible paths excluding sections at risk of damage, evaluates the communication quality, latency, and load status of each path, determines priority according to established optimization criteria to select a path, and transmits an optical path switching control command to the selected optical path. The optimization criteria include at least one of minimizing communication delay, maximizing communication quality, traffic load balancing, complete avoidance of risk sections, and minimizing energy consumption. The criterion for minimizing communication delay is a criterion for prioritizing the path with the lowest transmission delay time calculated based on path length, the number of relay nodes, or network status among selectable reserve optical paths. The criterion for maximizing communication quality is a criterion for selecting the path with the most stable signal-to-noise ratio (SNR), packet loss rate, or current attenuation state. The criterion for traffic load balancing is a criterion for prioritizing a path with a low current load to prevent excessive traffic concentration on a specific optical path. The criterion for complete avoidance of risk sections is a criterion for selecting a path including a relatively stable section while excluding a path containing a section where the risk index is close to a reference value. The criterion for minimizing energy consumption is a network It is characterized by being a criterion for selecting a path that minimizes the power consumption of the equipment.

[0024] In addition, the bypass control module is characterized by outputting a light path switching control command if the risk index remains in a state exceeding a standard for a certain period of time or longer.

[0025] In addition, the bypass control module is characterized by continuously monitoring the attenuation change rate and risk index trends of the relevant section after the optical path switch, and controlling the return to the original optical path when the risk is resolved.

[0026] Additionally, the AI ​​self-improvement unit includes: a learning data storage module that stores at least one of the following: a rate of change value measured by a fine damping analysis module, environmental data from an environmental sensor, whether a risk index calculated by a risk analysis module and a threshold value are exceeded, actual maintenance results input from an operator terminal or maintenance management system, and whether there is a false positive or a false negative when comparing actual risk judgment results with actual maintenance results; an AI learning module that generates updated parameters by retraining weights, threshold values, or judgment model coefficients applied within the risk analysis module stored in the learning data storage module, as any one of a supervised learning-based model, a regression model, a neural network model, or a time series prediction model; and a self-improvement control module that applies the updated parameters, reflecting the retraining results of the AI ​​learning module, to the risk analysis module to correct at least one of weights, threshold values, anomaly judgment duration conditions, and path selection optimization criteria.

[0027] Additionally, the tunnel management unit includes: a tunnel identification management module that generates and registers identification information, including tunnel names, management codes, or network address information, to uniquely identify each tunnel; a tunnel unique identifier storage module that stores the unique identifier of each tunnel and the network information, physical location information, and operational history information corresponding to the tunnel; a tunnel section mapping module that divides the interior of the tunnel into multiple management sections and defines the start and end points for each section; and a location matching module that matches the distance or coordinate values ​​of the damage location calculated by the damage location calculation module with the actual tunnel section information defined in the tunnel section mapping module and outputs them to the operator notification module. Effects of the invention

[0029] The present invention analyzes measurement data received from a condition detection device to calculate the rate of change in minute attenuation, the coordinates of the predicted damage location, and a risk index, thereby enabling the determination of the possibility of damage in advance prior to the disconnection of the optical cable, and allowing for a preemptive response before a failure occurs.

[0030] In addition, the present invention can improve the stability of CCTV and safety communication networks within tunnels by automatically transmitting an optical path switching control command when the risk index exceeds a reference value, thereby enabling the rapid activation of a backup optical path and minimizing or preventing communication interruptions.

[0031] In addition, by displaying the location coordinates of a damaged optical cable, the present invention enables maintenance personnel to quickly and accurately identify the expected location of damage, thereby reducing on-site search time and improving recovery efficiency.

[0032] In addition, the present invention stores maintenance results as training data, performs retraining, and corrects reference values ​​or weights based on the retraining results, thereby allowing risk judgment criteria to be adaptively improved in response to environmental changes, which can reduce false positives and false negatives and continuously improve prediction accuracy during long-term operation.

[0033] Furthermore, since the present invention integrates the management of multiple tunnels and can accurately pinpoint the location of damage in each tunnel through a tunnel identification and location matching structure, stable management without data interference is possible even in a wide-area tunnel communication network environment.

[0034] In addition, the present invention can improve the consistency of judgment and management efficiency of the entire system by forming a central closed-loop structure in which measurement data reception, analysis, risk judgment, automatic optical path switching, operator notification, storage of maintenance results, and self-improvement after relearning are performed sequentially. Brief explanation of the drawing

[0035] FIG. 1 is a block diagram illustrating an AI-based self-improving tunnel optical cable anomaly response integrated control system according to the present invention. Figure 2 is a block diagram illustrating an integrated control server. Figure 3 is a block diagram illustrating an analysis control unit. Figure 4 is a block diagram illustrating the AI ​​self-improvement section. Figure 5 is a block diagram illustrating the tunnel management department. FIG. 6 is a flowchart illustrating a tunnel environment adaptive optical cable condition detection and AI self-improvement-based optical path switching control method according to the present invention. Specific details for implementing the invention

[0036] Although the present invention may be subject to various modifications and may have various embodiments, specific embodiments are to be described in detail by way of illustration in the drawings. This is not intended to limit the present invention to specific embodiments, and it should be understood that any modification, equivalent, or substitution falls within the spirit and scope of the present invention for connecting and / or fixing structures extending in different directions.

[0037] The terms used in this specification are used merely to describe specific embodiments and are not intended to limit the invention. The singular expression includes the plural expression unless the context clearly indicates otherwise.

[0038] In this specification, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should not be understood as precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0039] Hereinafter, a preferred embodiment of an AI-based self-improving tunnel optical cable anomaly response integrated control system according to the present invention will be described in detail with reference to the accompanying drawings.

[0040] FIG. 1 is a schematic diagram illustrating the overall configuration of an AI-based self-improving tunnel optical cable anomaly response integrated control system according to the present invention.

[0041] Referring to FIG. 1, a plurality of tunnels (e.g., tunnel 1, tunnel 2, tunnel 3, etc.) are each equipped with a state detection device (20) for detecting the state of an optical cable, an environment sensor (30), and a switch control device (40).

[0042] The state detection device (20) measures the attenuation state of the optical cable installed in each tunnel and generates measurement data. The measurement data may include optical signal reflection information or attenuation value information.

[0043] The environmental sensor (30) measures temperature, vibration, water level, or other environmental information inside the tunnel and generates auxiliary data related to the possibility of damage to the optical cable. That is, the environmental sensor (30) outputs environmental data measured for temperature, vibration, water level, or other environmental information to the integrated control server (10).

[0044] The switch control device (40) is configured to switch the optical path of the optical cable installed in the tunnel according to a control command received from the integrated control server (10). The switch control device (40) can perform the function of activating a backup optical path or blocking an existing optical path.

[0045] According to one embodiment, the switch control device (40) includes an input port and a plurality of output ports, and is configured to connect an optical signal connected to the input port to a selected output port based on switching target optical path identification information received from an integrated control server.

[0046] Additionally, the switch control device (40) may include a status detection circuit that detects the current active optical path status, the availability status of the backup optical path, and whether there is a fault, and may be configured to send the result of the switching command execution to the integrated control server.

[0047] Additionally, the switch control device (40) can be implemented as at least one of a mechanical optical switch, a MEMS-based optical switch, a semiconductor optical switch, or a software-defined network (SDN)-based logical path control method.

[0048] Additionally, the switch control device (40) can set the switching target path differently depending on the ring network structure or mesh network structure, and a hitless switching or delay-minimizing switching method may be applied to minimize traffic loss at the time of switching.

[0049] The integrated control server (10) is configured to receive and analyze measurement data received from multiple tunnels to calculate the fine attenuation change rate, location coordinates, and risk index, and to transmit an optical path switching control command to the switch control device (40) when the risk index exceeds a reference value.

[0050] In addition, the integrated control server (10) provides the operator with information on the predicted location of damage and forms a central closed-loop structure that performs relearning and self-improvement after storing maintenance results.

[0051] The detailed configuration of the integrated control server (10) is described below.

[0052] Figure 2 is a block diagram illustrating an integrated control server.

[0053] Referring to FIG. 2, the integrated control server (10) includes a network interface unit (100), a database (200), an analysis control unit (300), an AI self-improvement unit (400), and a tunnel management unit (500).

[0054] The network interface unit (100) performs a communication function of receiving measurement data from a state detection device (20) and an environment sensor (30) installed in each tunnel using any one of Ethernet, an optical communication network, LTE, 5G, and a dedicated network, and transmitting control commands to a switch control device (40). The network interface unit (100) can perform data transmission and reception through a wired or wireless communication network.

[0055] The database (200) stores measurement data, analysis results, risk index, location coordinate information, maintenance results, and learning data. The information stored in the database (200) is referenced by the analysis control unit (300) and the AI ​​self-improvement unit (400).

[0056] The analysis control unit (300) is configured to calculate the fine attenuation change rate, damage location coordinates, and risk index based on the received measurement data. Additionally, if the risk index exceeds a reference value, the analysis control unit (300) generates an optical path switching control command and transmits it to the switch control device (40) through the network interface unit (100). Furthermore, the analysis control unit (300) may be configured to provide damage prediction location information to an operator terminal (not shown).

[0057] Here, the operator terminal (not shown) may be either a mobile terminal or a desktop, or a server installed at a remote control center.

[0058] The AI ​​self-improvement unit (400) is configured to retrain a model based on training data and to correct risk judgment criteria according to the retraining results. The AI ​​self-improvement unit (400) sequentially performs a learning function that performs retraining and a self-improvement function that corrects reference values ​​or weights by reflecting the retraining results.

[0059] The tunnel management unit (500) is configured to identify and manage multiple tunnels, stores unique tunnel identifiers and section information, and aligns damaged location coordinates with specific tunnel sections. This enables accurate location identification even in a multiple tunnel environment.

[0060] In this way, the integrated control server (10) forms a structure that centrally performs measurement data reception, data analysis, risk judgment, optical path switching control, operator notification, maintenance result storage, and self-improvement after relearning.

[0061] A detailed description of each component of the integrated control server (10) is given with reference to FIGS. 3 to 5.

[0062] FIG. 3 is a block diagram illustrating an analysis control unit (300).

[0063] Referring to FIG. 3, the analysis control unit (300) includes a fine attenuation analysis module (310), a damage location calculation module (320), a risk analysis module (330), a bypass control module (340), and an operator notification module (350) to analyze the condition of the optical cable based on measurement data received through the network interface unit (100), predict the risk of damage, and perform optical path switching control and operator notification to bypass the predicted damage location according to the predicted judgment result.

[0064] The fine damping analysis module (310) arranges the damping data received from the state detection device (20) in chronological order, calculates the moving average value of a certain time interval, and then calculates the deviation from the normal baseline.

[0065] The normal baseline is a reference value established based on attenuation data collected during the period when the optical cable was determined to be in a normal state, and it refers to a standard value used by the fine attenuation analysis module to determine whether there is an abnormality in attenuation changes.

[0066] For example, the normal baseline is set in at least one of the following methods: the average value of attenuation data collected over a certain period immediately after initial installation, the moving average value of N measurements recently determined to be in a normal state, a confidence interval including the average value and standard deviation range calculated by a statistical method, or a reference value dynamically corrected according to the retraining results of the AI ​​self-improvement unit.

[0067] According to one embodiment, the normal baseline may be set as the average attenuation value of a certain time interval recently determined to be in a normal state, and said average value may be updated over time. Additionally, the normal baseline may be individually set by tunnel, by section, or by type of optical cable, and may be periodically recalculated to reflect long-term environmental changes such as environmental conditions (temperature, vibration, humidity, etc.), seasonal changes, increased traffic volume, and structural aging.

[0068] The fine damping analysis module (310) quantifies the possibility of anomalies by calculating the deviation between the currently measured damping data and the normal baseline, the rate of change per unit time, or the cumulative increase.

[0069] Here, the normal baseline is not limited to a fixed value and can be corrected by the AI ​​self-improvement unit (400) based on the history of false detection and non-detection. Accordingly, a dynamic reference system that adapts to changes in the long-term operating environment can be formed.

[0070] Attenuation data is data indicating the change in intensity or loss characteristics of an optical signal transmitted through an optical cable, and may include information on the change in optical signal intensity over time, the loss value per unit length, or the reflected signal intensity at a specific point by the state detection device (20).

[0071] Attenuation data can be obtained by an OTDR (Optical Time Domain Reflectometer)-based measurement method in a state detection device (20), and in this case, it may include information on the arrival time of the reflected signal and the intensity of the reflected signal. Additionally, if a distributed optical fiber sensor is used, it may include a change in signal intensity due to a change in the scattering characteristics of the optical signal.

[0072] Additionally, attenuation data may consist of continuously collected time-series data and may include not only a single measurement at a specific point in time but also cumulative values ​​or average values ​​over a certain period of time.

[0073] Therefore, attenuation data includes at least one of the optical signal loss value per unit length (dB / km), reflected signal intensity at a specific point (dB), change in attenuation value over time, average attenuation value per section, and cumulative change in attenuation value.

[0074] Then, the fine damping analysis module (310) arranges the above damping data in chronological order, calculates the moving average value of a certain time interval, and calculates the deviation from the normal baseline.

[0075] According to one embodiment, the fine damping analysis module (310) calculates a quantitative analysis value indicating the possibility of at least one of the following: calculating the average of the most recent N measurements, calculating the difference value from the previous reference average, calculating the damping increase rate per unit time, and the cumulative increase amount over a certain period of time.

[0076] The damage location calculation module (320) calculates a distance value using the reflection delay time information of the optical signal, and calculates physical location coordinates by matching the calculated distance value with the section information provided by the tunnel management unit (500).

[0077] According to one embodiment, the damage location calculation module (320) measures the time t of arrival of the reflected signal and calculates the distance d = v × t / 2 by applying the optical fiber propagation speed v, and determines the actual location within the tunnel by comparing the calculated distance d with tunnel section mapping information.

[0078] Through this, specific location information, such as “1.85 km point of Tunnel 2,” is generated instead of simple distance information.

[0079] The risk analysis module (330) calculates a risk index that quantifies the degree of damage risk of the optical cable by combining the attenuation change analysis result transmitted from the fine attenuation analysis module (310) and the environmental data received from the environmental sensor (30).

[0080] The risk analysis module (330) does not directly interpret the damping data itself, but uses analysis results such as the damping change rate, cumulative increase, or deviation from the baseline calculated by the fine damping analysis module (310) as input values. In addition, the risk analysis module (330) can also incorporate information on temperature, vibration, water level, or other external environmental factors received from the environmental sensor (30).

[0081] According to one embodiment, the risk analysis module (330) calculates a risk index by substituting at least one of the damping change rate, damping increase duration, cumulative damping increase amount, tunnel internal temperature change, vibration intensity, and water level rise information into the following risk index calculation formula.

[0083] Risk Index = w1 × Rate of change in damping + w2 × Change in temperature + w3 × Vibration intensity

[0084] Here, w1, w2, and w3 are weights.

[0086] For example, the risk index can be calculated by applying a first weight (w1) to the damping change rate, a second weight (w2) to the temperature change, and a third weight (w3) to the vibration intensity using a weighted summation method. At this time, the weights (w1, w2, w3) can be corrected by the AI ​​self-improvement unit (400).

[0087] Here, the first weight (w1) reflects the importance of the rate of change in attenuation, which is directly related to the possibility of physical damage to the optical cable itself. For example, if the change in attenuation is determined to be the most important factor in damage prediction, the first weight (w1) can be set relatively large.

[0088] The second weight (w2) reflects the effect of temperature change among external environmental factors on the occurrence of damage. For example, the second weight (w2) can be increased in tunnel sections where the likelihood of fire is high.

[0089] The third weight (w3) reflects the impact of structural vibration or external shock on optical cable damage. In sections where vibration is frequent, the third weight (w3) can be set relatively high.

[0090] That is, a weight refers to a parameter that adjusts the relative contribution of the corresponding input element to the final risk index, and is set as at least one of a preset value, a setting value based on statistical analysis, a setting value based on past failure history, and a setting value based on the analysis results of data collected during the test operation period.

[0091] Additionally, each weight (w1, w2, w3) is set in one of the above ways during the initial setup process and then automatically corrected according to the retraining results of the AI ​​self-improvement unit (400).

[0092] The risk analysis module (330) determines whether there is a risk by comparing the calculated risk index with a reference value. If the risk index exceeds the set reference value, the section is determined to be in a state of damage risk and a control signal is transmitted to the bypass control module (340). Conversely, if the risk index is below the reference value, it is determined to be in a normal state and continuous monitoring is maintained.

[0093] Here, the reference value is a judgment threshold value for determining whether there is a risk of damage by comparing it with the risk index calculated by the risk analysis module (330), and is not limited to a fixed single value. It is set as one of an initial threshold value set by the manager during the initial operation phase of the system, a statistically based threshold value derived by statistically analyzing the history of past failures, a segment-specific threshold value that is differentially set according to tunnel-specific characteristics, segment-specific environmental conditions, or network structure, or an adaptive threshold value that is dynamically corrected according to the relearning results of the AI ​​self-improvement unit (400).

[0094] According to one embodiment, the reference value may be set differently for each tunnel or for each section within the tunnel, and, for example, a relatively high reference value may be applied in sections with high traffic volume or sections with frequent vibration.

[0095] In addition, the threshold value may be set as a threshold range rather than a single numerical value, and the system may be configured to determine a risk state if the risk index exceeds the said range or continuously approaches it for a certain period of time.

[0096] The threshold value can be configured to be adjusted upward when a false positive occurs and downward when a false negative occurs, based on the retraining results of the AI ​​self-improvement unit.

[0097] In addition, the risk analysis module (330) can analyze not only the risk index at a single point in time but also the trend of the risk index over a certain period of time. For example, it can distinguish between cases where the risk index rises rapidly in a short period and cases where it increases gradually over a long period.

[0098] The risk analysis module (330) may be configured to store judgment results and risk index values ​​in the database (200), and the data is subsequently used for retraining in the AI ​​self-improvement unit (400).

[0099] In this way, the risk analysis module (330) functions as a decision-making entity that ultimately determines the risk of damage by synthesizing the micro-attenuation analysis results and environmental information, and is linked with the bypass control module (340) and the AI ​​self-improvement unit (400) to perform the core decision-making function of the central integrated control structure.

[0100] The bypass control module (340) is configured to generate an optical path switching control command to bypass the optical cable damage risk section and transmit it to the switch control device (40) when it receives a determination result from the risk analysis module (330) that the risk index has exceeded a reference value.

[0101] The bypass control module (340) is not limited to simply outputting a switching signal, but generates a control command by comprehensively determining the necessity of switching, the target path for switching, and the timing of switching according to the network structure (e.g., ring network structure, mesh network structure).

[0102] In this specification, a ring network structure refers to a network structure in which a plurality of tunnels or nodes are interconnected in a circular or closed-loop form, and each node is connected to an adjacent node in both directions.

[0103] In a ring network structure, each node is connected to two adjacent nodes, and data transmission is possible via the opposite path even if a failure occurs in a specific section. In other words, it features a closed-loop structure that enables reverse bypass transmission in the event of a single path failure.

[0104] Furthermore, in this specification, a mesh network structure refers to a network structure in which multiple tunnels or nodes are interconnected via multiple paths, and each node is directly connected to two or more nodes. Therefore, in a mesh network structure, even if a failure occurs in a specific section, data can be transmitted by selecting one of multiple alternative paths. In other words, it is characterized by a multipath structure in which the entire network is not disconnected even if a single section is severed.

[0105] Therefore, if the tunnel sections are connected in a ring network structure and each tunnel is connected to an adjacent tunnel in both directions, the bypass control module (340) checks the direction of the section where damage is predicted or the section where the risk of damage is determined, controls the data bypass in the opposite direction of the currently used path, checks whether the backup path in that direction is in a normal state, and then sequentially executes the optical path switching.

[0106] For example, if a risk occurs in the path from Tunnel 1 to Tunnel 2, the optical path can be switched to divert data from Tunnel 1 to Tunnel 3. In this way, in a ring network structure, communication can be maintained by using the opposite direction of the disconnection point.

[0107] Additionally, the bypass control module (340) selects the optimal path among multiple reserve paths when each tunnel is in a mesh network structure connected to multiple tunnels and multiple paths. Specifically, the bypass control module (340) generates a list of possible paths excluding sections at risk of damage, evaluates the communication quality, latency, and load status of each path, determines the priority according to the set optimization criteria, and sequentially executes the process of transmitting optical path switching control commands to the selected path.

[0108] Here, the optimization criteria include at least one of minimizing communication delay, maximizing communication quality, distributing traffic load, completely avoiding risk zones, and minimizing energy consumption.

[0109] The communication delay minimization criterion is a standard for prioritizing the path with the lowest transmission delay among available reserve optical paths, and the delay time can be calculated based on the path length, the number of relay nodes, or network conditions.

[0110] The criterion for maximizing communication quality is to select the path with the most stable signal-to-noise ratio (SNR), packet loss rate, or current attenuation state.

[0111] Traffic load balancing criteria are standards for prioritizing paths with low current loads to prevent excessive traffic concentration on specific routes.

[0112] The criterion for complete avoidance of risk sections is a standard for selecting a path that includes relatively stable sections while excluding paths that include sections where the risk index is close to a reference value.

[0113] The energy consumption minimization standard is a criterion for selecting a path that minimizes the power consumption of network equipment.

[0114] That is, the bypass control module (340) can select a bypass path based on optimization criteria (minimizing communication delay, balancing traffic load, maximizing signal quality, and completely avoiding the failure section). At this time, the selection criteria may be set as a fixed value and may be dynamically adjusted by the AI ​​self-improvement unit (400).

[0115] Additionally, the bypass control module (340) may be configured to perform a light path switch only when the risk index exceeds a reference value, and may be set to switch only when the risk index shows an upward trend or when the state of exceeding the reference value persists for a certain period of time.

[0116] In addition, the bypass control module (340) can continuously monitor the attenuation change rate and risk index trend of the section even after the optical path switch, and control it to return to the original optical path when the risk is resolved.

[0117] In this way, the bypass control module (340) automatically activates a backup optical path based on the risk (prediction) result and prevents communication interruption.

[0118] The operator notification module (350) displays damage prediction location information on the control screen after transmitting the optical path switching control command. The displayed information is at least one of tunnel identification information, damage location coordinates, risk index value, and attenuation change trend.

[0119] Here, the operator notification module (350) transmits display information to an operator terminal configured as a mobile or desktop and / or a remote control center via the network interface unit (100) or a separate wireless device.

[0120] The AI ​​self-improvement unit (400) sequentially performs the processes of storing learning data, performing re-learning by the AI ​​learning module (410), and correcting reference values ​​and weights by the self-improvement control module (420). Additionally, the AI ​​self-improvement unit (400) performs re-learning to improve the judgment accuracy of the analysis control unit (300). A detailed description of the AI ​​self-improvement unit (400) is provided with reference to FIG. 4.

[0121] Figure 4 is a block diagram illustrating the AI ​​self-improvement section.

[0122] Referring to FIG. 4, the AI ​​self-improvement unit (400) includes an AI learning module (410), a self-improvement control module (420), and a learning data storage module (430).

[0123] The learning data storage module (430) is configured to store risk analysis results, optical path switching history, and maintenance results. Specifically, the learning data storage module (430) stores the rate of change measured by the fine attenuation analysis module (310), environmental sensor (30) data such as temperature, vibration, and water level of the environmental sensor (30), the risk index calculated by the risk analysis module (330) and whether the threshold value is exceeded, actual maintenance results input from the operator terminal and / or maintenance management system, and whether there is a false positive or a false negative when comparing the actual risk judgment result and the actual maintenance result in the AI ​​self-improvement unit (400).

[0124] For example, if a path switching is performed because the risk index exceeds a threshold value but the actual inspection reveals no abnormalities, the corresponding data may be stored as a false positive case. Conversely, if the risk index is below the threshold value but actual damage is detected, the data may be stored as a undetected case.

[0125] The AI ​​learning module (410) is configured to perform retraining using data stored in the learning data storage module (430). At this time, the AI ​​learning module (410) performs retraining as any one of a supervised learning-based model, a regression model, a neural network model, or a time series prediction model.

[0126] According to one embodiment, the AI ​​learning module (410) can reconstruct a risk index calculation model using past attenuation change rates and environmental data and actual damage status as learning data.

[0127] For example, if vibration intensity has a higher correlation with actual damage occurrence than temperature change in a specific tunnel section, the learning results can be reflected in a direction that increases the importance of the vibration factor.

[0128] The self-improvement control module (420) is configured to correct the judgment criteria of the risk analysis module (330) by reflecting the retraining results of the AI ​​learning module (410). The self-improvement control module (420) performs at least one of correcting the weights of the risk index calculation formula, adjusting the reference value, changing the abnormality judgment duration condition, and adjusting the path selection optimization criteria.

[0129] For example, if the self-improvement control module (420) analyzes that the rate of change in attenuation resulting from relearning has a high correlation with actual damage, it can increase the first weight corresponding to the rate of change in attenuation and decrease the weight corresponding to other factors.

[0130] Additionally, the self-improvement control module (420) can increase the risk judgment threshold value when false detections occur frequently, and decrease the threshold value when no detection occurs.

[0131] Additionally, the self-improvement control module (420) can change the duration condition in which the risk index exceeds a reference value, for example, the condition of exceeding 10 seconds or more can be adjusted to the condition of exceeding 30 seconds or more.

[0132] Additionally, the self-improvement control module (420) can adjust the weights of the optimization criteria applied when selecting a path based on the relearning results, for example, by reducing the weight corresponding to communication delay and increasing the weight corresponding to risk.

[0133] According to another embodiment, the AI ​​self-improvement unit (400) is not limited to a supervised learning-based retraining method and can be implemented as a reinforcement learning-based structure.

[0134] In a reinforcement learning-based structure, the AI ​​self-improvement unit (400) operates as an agent, recognizes the tunnel environment as a state, performs risk judgment criteria or weight adjustments as an action, and has a structure that receives a reward according to the result.

[0135] In reinforcement learning, the state includes at least one of the following: a damping change rate value, environmental sensor (30) data (temperature, vibration, etc.), a calculated risk index, a recent optical path switching history, and whether a false positive or a non-detection occurred. That is, the current integrated operational status of the tunnel section can be defined as a single state vector.

[0136] Additionally, the actions performed by the AI ​​self-improvement unit (400) in the reinforcement learning structure include at least one of adjusting the weight of the risk index calculation formula, changing the risk judgment criterion value, changing the abnormal duration condition, adjusting the path selection optimization criterion, and deciding whether to perform an immediate light path switch.

[0137] For example, policy changes such as increasing the first weight by +0.05, raising the threshold value from 0.75 to 0.80, and adjusting the bypass transition delay time from 10 seconds to 20 seconds can be one action.

[0138] The reward reflects the effect of the self-improvement policy on actual operational results. According to one embodiment, the reward can be defined as a positive reward if no communication interruption occurs, a negative reward if a false positive occurs, a large negative reward if no detection occurs, a positive reward if the number of switching is minimized, and a positive reward if recovery time is shortened. For example, a reward system can be established such as +10 if communication is maintained for 24 hours without interruption, -5 if unnecessary optical path switching occurs, and +20 if actual damage is not detected.

[0139] In reinforcement learning structures, the policy is updated by accumulating rewards as described above. Therefore, as the positive reward value increases, the probability of the corresponding action increases (the policy is used more), and as negative rewards accumulate, the probability of the corresponding action decreases (the policy is used less).

[0140] That is, the reward is a quantitative evaluation value of the policy adjustment results performed by the AI ​​self-improvement unit (400). A positive value is assigned when no communication interruption occurs or when unnecessary optical path switching is reduced, and a negative value is assigned when false detection or non-detection occurs. The reward value is used in policy updates to determine future risk judgment criteria or weight adjustment directions.

[0141] The tunnel management unit (500) identifies multiple tunnels, manages section information and location information for each tunnel, and is configured to align distance information calculated by the damage location calculation module (320) with the actual physical location. This is explained in more detail with reference to FIG. 5.

[0142] Figure 5 is a block diagram illustrating the tunnel management department.

[0143] Referring to FIG. 5, the tunnel management unit (500) includes a tunnel identification management module (510), a tunnel unique identifier storage module (520), a tunnel section mapping module (530), and a location matching module (540).

[0144] The tunnel identification management module (510) distinguishes and manages multiple tunnels. Specifically, the tunnel identification management module (510) generates or registers identification information to uniquely identify each tunnel.

[0145] Identification information may include tunnel name, management code, or network address information. For example, a unique identifier may be assigned to each tunnel, such as Tunnel 1 → ID: TNL-001, Tunnel 2 → ID: TNL-002, Tunnel 3 → ID: TNL-003.

[0146] Through this, the integrated control server (10) can distinguish and process measurement data received from multiple tunnels.

[0147] The tunnel unique identifier storage module (520) stores the unique identifier of each tunnel and network information, physical location information, and operational history information corresponding to the tunnel.

[0148] For example, the tunnel unique identifier storage module (520) stores unique identifiers for tunnel 2, such as a unique identifier (e.g., TL-002), length (e.g., 3.2 km), connection structure (e.g., ring type or mesh type), and the number of installed optical cable sections (e.g., 5). The identifier information can be referenced when calculating the location of damage.

[0149] The tunnel section mapping module (530) divides the interior of the tunnel into multiple management sections and defines a start point and an end point for each section.

[0150] For example, a tunnel with a length of 3.2 km can be divided into sections as follows.

[0151] Section A: 0 ~ 0.8 km

[0152] Section B: 0.8 ~ 1.6 km

[0153] Section C: 1.6 ~ 2.4 km

[0154] Section D: 2.4 ~ 3.2 km

[0155] This section information is used to convert the distance value calculated by the damage location calculation module (320) into an actual management section.

[0156] The position matching module (540) is configured to match the distance value or coordinate value calculated by the damage location calculation module (320) with the actual tunnel section information.

[0157] For example, if the damage location calculation module (320) detects an abnormal signal at the 1.85 km point, the location matching module (540) matches this to the location corresponding to section C of Tunnel 2. Accordingly, the operator notification module (350) can display matched location information such as "Tunnel 2, section C, 1.85 km point, danger occurred" rather than a simple distance value.

[0158] Alternatively, if the attenuation change rate at the 2.15 km point in Tunnel 3 increases and the risk index exceeds the reference value, the damage location calculation module (320) calculates the 2.15 km distance value, the location matching module (540) matches it to the location corresponding to section B of Tunnel 3, the tunnel identification management module (510) confirms that the data corresponds to TNL-003, and the operator notification module (350) transmits or displays notification information such as "Danger occurred at Tunnel 3, section B, 2.15 km point" to the operator terminal.

[0159] Accordingly, the tunnel management department (500) prevents data confusion among multiple tunnels, identifies the exact location of damage, improves the efficiency of maintenance work, enables statistical analysis per tunnel, and enables the selection of a path linked to the network structure.

[0160] That is, the tunnel management unit (500) performs not only simple data storage functions, but also tunnel identification, section division, alignment of distances and sections, and location specification.

[0161] As another embodiment, the present invention includes a tunnel environment adaptive optical cable condition detection and AI self-improvement-based optical path switching control method according to the configurations of the above-described embodiment. This is explained with reference to FIG. 6 below.

[0162] FIG. 6 is a flowchart illustrating a tunnel environment adaptive optical cable condition detection and AI self-improvement-based optical path switching control method according to the present invention.

[0163] Referring to FIG. 6, the present invention includes a step S100 of receiving measurement data from an integrated control server (10), a step S200 of analyzing the measurement data received from the integrated control server (10), a step S300 of comparing a risk index with a reference value, a step S400 of transmitting an optical path switching control command, a step S500 of displaying location information as an operator notification, a step S600 of storing maintenance results, and a step S700 of self-improvement after relearning.

[0164] Step S100 is a step of collecting measurement data from the network interface unit (100) of the integrated control server (10). The network interface unit (100) receives attenuation data from the condition detection device (20) installed in each tunnel and receives temperature, vibration, or water level data from the environment sensor (30). The received data is stored in the database (200) or transmitted to the analysis control unit (300).

[0165] Step S200 is a step in which the analysis control unit (300) of the integrated control server (10) analyzes measurement data. The analysis control unit (300) analyzes attenuation data to calculate the attenuation change rate or cumulative increase, calculates a distance value using reflected signal delay information, and calculates the actual tunnel section coordinates in conjunction with the position matching module (540) of the tunnel management unit (500). Then, the analysis control unit (300) calculates a risk index by combining the attenuation change rate and environmental data.

[0166] More specifically, the fine damping analysis module (310) sorts the damping data by time interval and calculates the deviation from the normal baseline, the damping change rate, or the cumulative increase. For example, if the damping value gradually increases over a certain period of time in a specific interval, the fine damping analysis module (310) calculates the damping change rate and the cumulative increase to generate an analysis result indicating the damping change characteristics of the interval.

[0167] Next, the damage location calculation module (320) calculates a distance value of the predicted damage location using optical signal reflection delay information or distance calculation results. The distance value calculated by the damage location calculation module (320) is transmitted to the location matching module (540) of the tunnel management unit (500), and the location matching module (540) converts the distance value into physical coordinates of the actual tunnel section based on the section information stored in the tunnel section mapping module (530). Accordingly, the coordinates of the predicted damage location are calculated.

[0168] Next, the risk analysis module (330) calculates a risk index by combining the analysis results of the fine damping analysis module (310) and environmental data. The risk index can be calculated by applying a first weight corresponding to the damping change rate, a second weight corresponding to the temperature change, and a third weight corresponding to the vibration intensity, and the weights can be set to values ​​that can be corrected by the AI ​​self-improvement unit (400).

[0169] The analysis control unit (300) can record the calculated risk index and the predicted damage location coordinates in the database (200).

[0170] Step S300 is a step in which the risk analysis module (330) compares the calculated risk index with a risk judgment threshold value to determine whether it exceeds the threshold. The threshold value may be set differently for each tunnel and may be corrected by the AI ​​self-improvement unit (400).

[0171] And if the risk index is below the threshold value, the risk analysis module (330) determines the state to be normal and returns to step S100, and the result of the determination of the normal state can be recorded in the database (200).

[0172] On the other hand, if the risk index exceeds the threshold value, the risk analysis module (330) determines the condition as a risk of damage and transmits a control trigger signal to the bypass control module (340) so that step S400 is performed.

[0173] Step S400 is a step in which the bypass control module (340) generates an optical path switching control command to bypass a section determined to be in a state of risk of damage. The bypass control module (340) can select a switching target path by referring to the status of the optical path currently in use, the availability of the backup optical path, and the network structure.

[0174] For example, if the network structure is ring-shaped, the bypass control module (340) can control to switch to the opposite direction path, and if the network structure is mesh-shaped, it can control to switch to a path selected according to optimization criteria that consider at least one of communication delay, communication quality, traffic load, and risk index among a plurality of backup paths.

[0175] The bypass control module (340) transmits an optical path switching control command, including switching target optical path identification information and switching target section information, to the network interface unit (100), and the network interface unit (100) transmits the control command to the switch control device (40). The optical path switching result can be recorded in the database (200).

[0176] Step S500 is a step in which the operator notification module (350) displays location information on the control screen using the predicted damage location coordinates and risk index information. The operator notification module (350) provides guidance on the damage risk status after the optical path switching control command is transmitted.

[0177] The display information may include at least one of tunnel identification information, coordinates of the expected location of damage, a risk index value, and a trend of attenuation change. For example, it may be displayed on the control screen in the form of “Tunnel identifier TNL-002, 1.85 km point, risk index 0.87, performing a switchover”.

[0178] Step S600 is a step in which the integrated control server (10) receives inspection results from field maintenance personnel. The network interface unit (100) stores the input inspection results in the database (200) and the learning data storage module (430). The maintenance results stored here may include at least one of the presence or absence of actual damage, type of damage, location of damage, recovery time, or details of action taken.

[0179] Additionally, the AI ​​self-improvement unit (400) can generate comparison data to determine whether there is a false positive or a false negative by linking the risk judgment result, the optical path switching history, and the maintenance result.

[0180] The AI ​​learning module (410) of the AI ​​self-improvement unit (400) analyzes comparison data to determine whether there is a match between the risk judgment result and the actual maintenance result.

[0181] Specifically, if the risk index exceeds the threshold value and the optical path is switched, but the actual inspection results show no damage, the AI ​​learning module (410) classifies this as a false positive case.

[0182] If the risk index is determined to be below the threshold value but actual damage is confirmed, the AI ​​learning module (410) classifies this as an undetected case.

[0183] The classification result is stored in the learning data storage module (430), and the self-improvement control module (420) corrects the weights or reference values ​​used to calculate the risk index by reflecting the frequency of false positives or non-detections.

[0184] Step S700 is a step in which the AI ​​self-improvement unit (400) performs retraining using data stored in the learning data storage module (430) and corrects the risk judgment parameters of the analysis control unit (300) based on the retraining results. Specifically, the AI ​​learning module (410) retrains the model using the attenuation change rate, environmental data, risk index, optical path switching history, and maintenance results as learning data.

[0185] Subsequently, the self-improvement control module (420) corrects at least one of the weights, reference values, abnormality judgment duration conditions, or path selection optimization criteria used in calculating the risk index based on the output of the retrained model. The corrected parameters are then applied to the risk index calculation in step S200 and the reference value comparison in step S300 to improve the judgment accuracy of the next cycle.

[0186] That is, the present invention forms a closed-loop structure in which measurement data reception, analysis, risk determination, optical path switching control, location information display, maintenance result storage, and self-improvement after relearning are performed sequentially.

[0187] Accordingly, the integrated control server (10) can gradually improve the accuracy of risk judgment by reflecting result data accumulated during the operation process.

[0189] The present invention is not limited to the specific preferred embodiments described above, and it is obvious that anyone with ordinary knowledge in the technical field to which the invention pertains can make various modifications without departing from the essence of the invention as claimed in the claims, and that such modifications fall within the scope of the claims. Explanation of the symbols

[0191] 10: Integrated control server 20: Status detection device 30: Environmental sensor 40: Switch control unit 100: Network Interface Section 200: Database 300: Analysis Control Unit 310: Fine Attenuation Analysis Module 320: Damage Location Calculation Module 330: Risk Analysis Module 340: Bypass control module 350: Operator notification module 400: AI Self-Improvement Unit 410: AI Learning Module 420: Self-improvement control module 430: Training data storage module 500: Tunnel Management Department 510: Tunnel Identification Management Module 520: Tunnel Unique Identifier Storage Module 530: Tunnel Section Mapping Module 540: Position alignment module

Claims

Claim 1 A state detection device (20) that outputs attenuation data including at least one of a light signal loss value per unit length (dB / km) installed in each of a plurality of tunnels, a reflected signal strength (dB) at a specific point, a change in attenuation value over time, an average attenuation value per section, and a cumulative change in attenuation value, as time-series data collected along with time information; an environment sensor (30) that is installed in each of a plurality of tunnels to measure the environment of each tunnel and outputs environment data including at least one of internal temperature, vibration, and water level; a network interface unit (100) that receives attenuation data and environment data from the state detection device (20) and environment sensor (30) installed in a plurality of tunnels; a database (200) that stores attenuation data, environment data, risk judgment results, and maintenance results; and an integrated control server (10) including an analysis control unit (300), an AI self-improvement unit (400), and a tunnel management unit (500). The integrated control server (10) includes: a network interface unit (100) that communicates with a state detection device (20), an environment sensor (30), and a switch control device (40); a database (200) that stores attenuation data received through the network interface unit (100) and environment data of the environment sensor (30); an analysis control unit (300) that analyzes the attenuation data stored in the database (200) to calculate the attenuation change rate, calculates the distance value of the expected damage location based on the reflection signal delay information of the attenuation data, and transmits an optical path switching control command to the switch control device (40) when the risk index calculated by combining the attenuation change rate and the environment data of the environment sensor (30) exceeds a reference value; a tunnel management unit (500) that aligns the distance value calculated by the analysis control unit (300) with tunnel section information to convert and output it as physical location coordinates; and an AI self-improvement unit (400) that stores risk judgment results and maintenance results as learning data, and retrains a risk index calculation model using the stored learning data to correct weights or reference values ​​used in calculating the risk index. Includes,The analysis control unit (300) includes: a fine attenuation analysis module (310) that arranges attenuation data in chronological order, calculates the average of N measurements over a certain time interval to calculate the difference from the previous reference average set as the normal reference line, and calculates at least one quantitative analysis value among the attenuation increase rate per unit time and the cumulative increase amount over a certain time; a damage location calculation module (320) that calculates a distance value using reflection delay time information of an optical signal based on an OTDR (Optical Time Domain Reflectometer) measurement method, and calculates physical location coordinates by matching the calculated distance value with the tunnel section information of the tunnel management unit (500); and a risk analysis module (330) that determines whether there is a risk by calculating a risk index by applying a weight set for vibration intensity to any one of the attenuation change rate, cumulative increase amount, and deviation from the reference line analyzed by the fine attenuation analysis module (310), the temperature change rate calculated by analyzing environmental data including temperature and vibration received from the environmental sensor (30), and the weight set for the vibration intensity, respectively, and comparing it with a set reference value. The method includes: a bypass control module (340) that receives a judgment result from the analysis result of the risk analysis module (330) that the risk index has exceeded a reference value, generates an optical path switching control command to bypass the optical cable damage risk section, and transmits it to a switch control device (40); and an operator notification module (350) that outputs display information including at least one of tunnel identification information, damage location coordinates, risk index value, and attenuation change trend after transmitting the optical path switching control command, or transmits it to a set operator terminal. The risk analysis module (330) calculates a risk index by applying weights set to the temperature change rate and vibration intensity included in the environmental data to at least one of the attenuation change rate, cumulative increase amount, and deviation from the baseline, and analyzes the trend of the calculated risk index over time to distinguish between cases where it rises rapidly in the short term and cases where it rises gradually over the long term, thereby determining whether there is a risk.If it is determined that the risk index exceeds a reference value, the determination result is transmitted to the bypass control module (340). The normal baseline is set by at least one of the following: a moving average value of attenuation data recently determined to be in a normal state, a confidence interval including an average value and a standard deviation range calculated by a statistical method, or a reference value corrected according to the relearning result of the AI ​​self-improvement unit (400), and is individually set for each tunnel or section. If the optical cable network structure is a ring-type network structure where multiple tunnels or nodes are interconnected in a circular or closed-loop form and each node is connected to an adjacent node in both directions, the bypass control module (340) checks the direction of the section where the damage risk is determined, and executes optical path switching after checking whether the data is bypassed in the opposite direction of the currently used path or whether the backup path in that direction is in a normal state. If the optical cable network structure is a mesh-type network structure where multiple tunnels or nodes are interconnected in a multipath form and each node is directly connected to two or more nodes, it generates a list of possible paths excluding the damage risk section, evaluates the communication quality, latency, and load status of each path, and selects a path by determining priority according to the set optimization criteria. It is characterized by transmitting a control command for switching to an optical path to a selected optical path, wherein the optimization criteria include at least one of minimizing communication delay, maximizing communication quality, traffic load balancing, complete avoidance of critical sections, and minimizing energy consumption, the criterion for minimizing communication delay is a criterion for prioritizing the path with the lowest transmission delay time calculated based on path length, number of relay nodes, or network status among selectable reserve optical paths, the criterion for maximizing communication quality is a criterion for selecting the path with the most stable signal-to-noise ratio (SNR), packet loss rate, or current attenuation state, and the criterion for traffic load balancing is a criterion for prioritizing the path with the lowest current load so that traffic is not excessively concentrated on a specific optical path.The criterion for complete avoidance of risk sections is a criterion for selecting a path that includes relatively stable sections while excluding paths that include sections where the risk index is close to a reference value, and the criterion for minimizing energy consumption is a criterion for selecting a path that minimizes the power consumption of network equipment; The operator notification module (350) outputs display information including at least one of tunnel identification information, damage location coordinates, risk index value, and attenuation change trend after transmitting the optical path switching control command, and the AI ​​self-improvement unit (400) stores at least one of the following: a change rate value measured by the fine attenuation analysis module (310), environmental data from the environmental sensor (30), whether the risk index calculated by the risk analysis module (330) exceeds a reference value, actual maintenance results input from the operator terminal and / or maintenance management system, and whether there is a false detection or a false detection when comparing the actual risk judgment result with the actual maintenance result; the learning data storage module (430); and the AI ​​learning module (410) which generates updated parameters by re-learning the weights, reference values, or judgment model coefficients applied within the risk analysis module (330) stored in the learning data storage module (430) as any one of a supervised learning-based model, a regression model, a neural network model, or a time series prediction model. An AI-based self-improving tunnel optical cable anomaly response integrated control system comprising: a self-improving control module (420) that applies updated parameters, reflecting the retraining results of the AI ​​learning module (410), to the risk analysis module (330) to correct at least one of weights, reference values, anomaly detection duration conditions, and path selection optimization criteria. Claim 2 delete Claim 3 The AI-based self-improving tunnel optical cable anomaly response integrated control system according to claim 1, wherein the weight of the risk analysis module (330) is at least one of a preset value, a statistical analysis-based setting value, a past failure history-based setting value, and a dynamic correction value by the AI ​​self-improvement unit (400), and the bypass control module (340) continuously monitors the attenuation change rate and risk index trend of the corresponding section after the optical path switch, and controls to return to the original optical path when the risk is resolved. Claim 4 delete Claim 5 delete Claim 6 The AI-based self-improving tunnel optical cable anomaly response integrated control system according to claim 3, wherein the bypass control module (340) outputs an optical path switching control command when the risk index remains in a state exceeding a standard for a certain period of time or longer. Claim 7 delete Claim 8 In claim 3, the tunnel management unit (500) comprises: a tunnel identification management module (510) that generates and registers identification information including a tunnel name, management code, or network address information to uniquely identify each tunnel; a tunnel unique identifier storage module (520) that stores a unique identifier of each tunnel and network information, physical location information, and operational history information corresponding to the tunnel; a tunnel section mapping module (530) that divides the interior of the tunnel into a plurality of management sections and defines a start point and an end point for each section; and a location matching module (540) that matches the distance value or coordinate value of the damage location calculated by the damage location calculation module (320) with the actual tunnel section information defined in the tunnel section mapping module (530) and outputs it to an operator notification module (350); an AI-based self-improving tunnel optical cable anomaly response integrated control system.

Images