Optical fiber transmission signal monitoring system for long-distance communication

Abstract

The invention discloses an optical fiber transmission signal monitoring system for long-distance communication, which relates to the technical field of optical fiber transmission signal monitoring and can comprehensively acquire key parameters in an optical fiber network by integrating various high-precision sensors, including multi-dimensional data such as optical fiber signal attenuation rate Af, signal time delay Tf, temperature Te, humidity He and optical fiber vibration Vf. The collection of the multi-source data can accurately reflect the state of the optical fiber in long-distance transmission, the health condition of the optical fiber is comprehensively known, and high-quality basic data is provided for subsequent fault prediction and repair. And the data preprocessing unit automatically eliminates abnormal values through cleaning and normalization processing, and standardizes the data, so that the reliability and stability of the data in the subsequent analysis and model training process are ensured. And particularly, abnormal value identification and elimination are carried out by using a median absolute deviation method, so that the fault tolerance of the system in a complex network environment is effectively improved, and the interference of noise on a prediction result is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of optical fiber transmission signal monitoring, in particular to an optical fiber transmission signal monitoring system for long-distance communication. BACKGROUND

[0002] Communication engineering is one of the core fields of current technology development, covering various technologies and systems for information transmission. With the deepening of global informatization, long-distance communication has become a crucial component of modern communication infrastructure. In this field, optical fiber communication has become the mainstream technology for long-distance information transmission due to its high speed, high capacity, and low loss. In the specific management of optical fiber networks, the monitoring and fault diagnosis of optical fibers are the core links, ensuring the continuous and stable operation of the optical fiber network and avoiding communication interruption.

[0003] The existing optical fiber network monitoring and fault diagnosis technology has certain limitations, especially in the context of long-distance communication. Traditional optical fiber monitoring methods mostly rely on periodic inspection and manual inspection, which cannot grasp the network status in real time. This approach not only makes it difficult to quickly identify potential optical fiber faults, but also cannot efficiently locate the specific position of the fault, thereby prolonging the fault recovery time. In addition, existing systems have difficulty accurately predicting the time and location of fault occurrence when faced with complex network environments and variable transmission conditions, leading to frequent communication service interruptions or signal quality degradation, affecting network reliability and user experience.

[0004] These problems are usually caused by the inefficiency and lag of optical fiber monitoring technology. Traditional optical fiber fault diagnosis methods are mostly based on conventional sensor monitoring data, which cannot meet the real-time requirements of long-distance communication networks. When a fault occurs, traditional systems can only be handled through post-analysis or manual inspection, resulting in the failure to discover the fault in a timely manner and the lack of accurate positioning means after the fault occurs. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides an optical fiber transmission signal monitoring system for long-distance communication, which solves the problems mentioned in the background art.

[0006] To achieve the above purpose, the present application is implemented by the following technical scheme: an optical fiber transmission signal monitoring system for long-distance communication, comprising a data acquisition and processing module, a feature extraction and signal analysis module, a fault prediction model module, a fault location and alarm module, a fault repair and recovery strategy module, and a maintenance feedback and optimization module;

[0007] The data acquisition and processing module collects signal data transmitted by the optical fiber through sensors, fits the original data set GX, and performs preprocessing to obtain the optical fiber data set GW.

[0008] The feature extraction and signal analysis module analyzes the optical fiber dataset GW and extracts features, fitting into a feature set FG for diagnosing the optical fiber;

[0009] The fault prediction model module establishes a fault diagnosis model by using a support vector machine algorithm, predicts faults of the optical fiber according to the feature set FG, obtains a fault risk value Frisk, and judges the probability of the optical fiber occurring a fault through the fault risk value Frisk, to obtain a fault probability value Pva;

[0010] The fault location and alarm module locates the area where the fault occurs according to the obtained fault risk value Frisk and fault probability value Pva, combines the optical fiber network topology structure and the feature set FG, obtains a fault location Lf and a joint trigger factor Φ(Lf) of the fault location, and issues an alarm notification;

[0011] The fault repair and recovery strategy module calculates a repair response strength RS according to the obtained fault risk value Frisk, fault location Lf and joint trigger factor Φ(Lf) of the fault location, and formulates and executes a repair strategy for the optical fiber fault;

[0012] The maintenance feedback and optimization module collects data after the repair strategy is executed, obtains a repair effect score SPre, and adjusts the fault diagnosis model according to the repair effect score SPre.

[0013] Preferably, the data acquisition and processing module comprises a multi-source signal acquisition unit and a data preprocessing unit;

[0014] The multi-source signal acquisition unit acquires signal data through a sensor, including an optical fiber signal attenuation rate Af, a signal time delay Tf, a temperature Te, a humidity He and an optical fiber vibration Vf, and fits into an original data set GX;

[0015] The optical fiber signal attenuation rate Af is acquired through an optoelectronic sensor;

[0016] The optical fiber signal attenuation rate Af is obtained by the following formula:

[0017] ;

[0018] In the formula, Pin represents the input power of the optical signal, Pout represents the output power of the optical signal, and Lgx represents the transmission length of the optical fiber;

[0019] The signal time delay Tf is acquired through high-precision clock and optical pulse timing acquisition;

[0020] The temperature Te is acquired through temperature sensors distributed around the optical fiber;

[0021] Humidity He is acquired by humidity sensitive sensors distributed in the walls of the communication tunnels surrounding the optical fiber;

[0022] Optical fiber vibration Vf is acquired by fiber Bragg grating vibration sensors fixed to the outer sheath of the optical cable;

[0023] The data preprocessing unit cleans and normalizes the acquired raw data set GX to obtain the optical fiber data set GW;

[0024] Cleaning includes outlier rejection, which identifies and rejects outliers in the raw data set GX by using the median absolute deviation method;

[0025] Normalization processing is performed on the raw data set GX by using the Min-Max normalization method to obtain the optical fiber data set GW;

[0026] The optical fiber data set GW is obtained by the following formula:

[0027] ;

[0028] In the formula, GWb represents the b-th data in the optical fiber data set GW, GXb represents the b-th data in the raw data set GX, minGXb represents the valley value of the b-th data in the raw data set GX, and maxGXb represents the peak value of the b-th data in the raw data set GX.

[0029] Preferably, the feature extraction and signal analysis module includes a time series fluctuation modeling unit and an environmental dynamic fitting unit;

[0030] The time series fluctuation modeling unit extracts features from the optical fiber data set GW by using a sliding time window technique, including signal intensity fluctuation ΔAf and time delay fluctuation ΔTf;

[0031] Signal intensity fluctuation ΔAf is obtained by the following formula:

[0032] ;

[0033] In the formula, L represents the length of the sliding window, represents the observation range of L unit time from the current time, and is used to construct local features, represents the weight coefficient, Af(t-i) represents the optical fiber signal attenuation rate at time t-i, and pAf(t) represents the average attenuation rate at time t;

[0034] Time delay fluctuation ΔTf is obtained by the following formula:

[0035] ;

[0036] In the formula, Tf (t-i) represents signal time delay at time t-i, maxTf (t-i) represents maximum time delay value, and minTf (t-i) represents minimum time delay value;

[0037] The environmental dynamic fitting unit analyzes the influence of temperature Te, humidity He and fiber vibration Vf on signal transmission in the optical fiber data set GW, and calculates to obtain the temperature and humidity change rate ΔHT and the vibration change rate ΔVf;

[0038] The temperature and humidity change rate ΔHT is obtained by the following formula:

[0039] ;

[0040] In the formula, Te (t) represents temperature at time t, Te (t-1) represents temperature at time t-1, He (t) represents humidity at time t, and He (t-1) represents humidity at time t-1;

[0041] The vibration change rate ΔVf is obtained by the following formula:

[0042] ;

[0043] In the formula, Δt represents time interval, Vf (t) represents fiber vibration at time t, and Vf (t-1) represents fiber vibration at time t-1;

[0044] The obtained signal intensity fluctuation ΔAf, time delay fluctuation ΔTf, temperature and humidity change rate ΔHT and vibration change rate ΔVf are integrated to obtain the feature set FG.

[0045] Preferably, the fault prediction model module comprises a feature mapping and classification decision unit and a probability mapping and risk normalization unit;

[0046] The feature mapping and classification decision unit performs high-dimensional mapping on the feature set FG, establishes a classification hyperplane through a support vector machine, constructs a fault diagnosis model, realizes binary classification decision of the fault state, calculates to obtain a fault risk value Frisk, and judges the fault state of the optical fiber through the fault risk value Frisk;

[0047] The fault risk value Frisk is obtained by the following formula:

[0048] ;

[0049] In the formula, FG (t) represents a feature vector at time t, FG (tj) represents the jth feature sample in the training set at time t, M represents the total number of feature samples, represents a parameter of the fault diagnosis model, yj represents a fault label of the jth feature sample, and ranges from -1 to +1, B represents a bias term, and K (, ) represents a kernel function;

[0050] The fault state of the optical fiber is obtained by matching in the following way:

[0051] When the fault risk value Frisk> 0, it indicates that the fault state of the optical fiber is in a risk state;

[0052] When the fault risk value Frisk≤ 0, it indicates that the fault state of the optical fiber is in a normal state.

[0053] Preferably, the probability mapping and risk normalization unit performs probability mapping on the obtained fault risk value Frisk, and calculates an obtained fault probability value Pva;

[0054] The fault probability value Pva is obtained by the following formula:

[0055] ;

[0056] In the formula, e represents a constant, and c represents a steepness control parameter of a Sigmoid function;

[0057] By analyzing the fault probability value Pva, an uncertainty entropy rHF is calculated and compared with a preset judgment threshold Trh to judge the accuracy of the fault probability occurrence;

[0058] The uncertainty entropy rHF is obtained by the following formula:

[0059] ;

[0060] In the formula, represents a logarithm function with base 2;

[0061] The accuracy of the fault occurrence is obtained by matching in the following way:

[0062] When the uncertainty entropy rHF≥ the judgment threshold Trh, it indicates that the fault probability judgment is inaccurate;

[0063] When the uncertainty entropy rHF< the judgment threshold Trh, it indicates that the fault probability judgment is accurate.

[0064] Preferably, the fault positioning and alarm module includes a spatial weight inversion positioning unit and an alarm grading decision and linkage unit;

[0065] The spatial weight inversion positioning unit constructs a node space model based on the optical fiber network topology structure, maps the obtained fault risk value Frisk and fault probability value Pva to the topology path, and deduces a fault position Lf of the fault occurrence probability through a weighted inversion algorithm;

[0066] The fault position Lf is obtained by the following formula:

[0067] ;

[0068] In the formula, Na represents the total number of topological nodes, Pva(ai) represents the failure probability value of the ai-th topological node, Frisk(ai) represents the failure risk value of the ai-th topological node, S(ai) represents the sensor sensitivity factor of the ai-th topological node, and X(ai) represents the position value of the ai-th topological node, which is the kilometer number or geographic coordinates from the starting point.

[0069] Preferably, the alarm grading decision and linkage unit obtains the joint triggering factor Φ(Lf) of the failure position according to the obtained failure position Lf, in combination with the failure risk value Frisk, the uncertainty entropy rHF and the failure probability value Pva, and triggers a three-level alarm response mechanism, including early warning, confirmation alarm and forced intervention.

[0070] The joint triggering factor Φ(Lf) of the failure position is obtained by the following formula:

[0071] ;

[0072] In the formula, Frisk, Pva and rHF represent the preset weight values of the failure risk value Frisk, the failure probability value Pva and the uncertainty entropy rHF, respectively, and ;

[0073] The three-level alarm response mechanism is matched and obtained by the following method:

[0074] When 0 < the joint triggering factor Φ(Lf) of the failure position < 0.4, it indicates the first risk level, low-risk reminder, early warning is performed on the failure position Lf, and the record is waited for the trend development.

[0075] When 0.4 ≤ the joint triggering factor Φ(Lf) of the failure position ≤ 0.7, it indicates the second risk level, medium-risk alarm, re-inspection and encrypted sampling are performed on the failure position Lf.

[0076] When 0.7 < the joint triggering factor Φ(Lf) of the failure position < 1.0, it indicates the third risk level, high-risk alarm, forced signal switching and cable re-inspection are performed on the failure position Lf.

[0077] Preferably, the failure repair and recovery strategy module includes a dynamic repair strategy decision unit and a strategy execution and link reconstruction unit.

[0078] The dynamic repair strategy decision unit comprehensively analyzes the failure state, obtains the repair response strength RS in combination with the failure risk value Frisk, the failure position Lf and the joint triggering factor Φ(Lf) of the failure position, and determines the response level and the strategy content of the repair strategy. ​

[0079] The repair response strength RS is obtained by the following formula:

[0080] ;

[0081] Where A(Lf) represents the influence function of the fault location, which is defined as the influence weight coefficient of the fault point on the backbone path. The backbone node is assigned a high value, and the edge path is assigned a low value. represents the adjustable coefficient, and ;

[0082] The response level and policy content of the repair policy are obtained in the following ways:

[0083] When 0 < repair response strength RS < 0.3, it indicates the first response level; strategy content: dynamic signal gain adjustment, adjustment of optical power amplifier, automatic control of optical attenuator to compensate for local attenuation fluctuations;

[0084] When 0.3 ≤ Repair Response Strength RS ≤ 0.6, it indicates the second response level; Strategy content: Active / standby Fibre Channel switchover. In the event of an incomplete physical link interruption, switch to the parallel / peer link path to maintain connection integrity;

[0085] When 0.6 < Repair Response Strength RS < 0.8, it indicates the third response level. Strategy content: Network segment-level rerouting, by reconstructing the routing table and automatically detouring known fault segments, is applicable to ring networks or redundant network structures.

[0086] When 0.8≤Repair Response Strength RS<1.0, it indicates the fourth response level; policy content: activation of backup links and alarm push, activation of backup optical fibers including wavelength division redundant channels, and issuance of mandatory alarms and manual intervention notifications.

[0087] Preferably, after the repair strategy is executed, the policy execution and link reconstruction unit automatically performs network resource management and link reconstruction operations, including calling the control channel and allocating the optical switch, routing equipment and amplifier.

[0088] Preferably, the data after the execution of the repair strategy of the maintenance feedback and optimization module is collected, including the signal quality recovery evaluation nQs, the delay recovery evaluation QTf and the fault recovery time evaluation nRt;

[0089] The signal quality recovery evaluation nQs is obtained by the following formula:

[0090] ;

[0091] Where nAf represents the attenuation rate of the repaired optical fiber signal;

[0092] The acquired signal quality recovery evaluation nQs is compared with a preset recovery threshold TQs to determine the repair effect;

[0093] The repair effect is acquired by matching in the following manner:

[0094] When the signal quality recovery evaluation nQs is greater than the recovery threshold TQs, it indicates that the repair effect is normal and the signal quality recovery is significant.

[0095] When the signal quality recovery evaluation nQs is less than or equal to the recovery threshold TQs, it indicates that the repair effect is abnormal and the signal quality is not recovered.

[0096] The time delay recovery evaluation QTf is acquired by the following formula:

[0097] ;

[0098] In the formula, nTf represents the time delay of the repaired signal; if the time delay recovery ratio is high, it indicates that the system response speed is significantly improved after repair.

[0099] The fault recovery time evaluation nRt is acquired by the following formula: the shorter the fault recovery time, the faster the repair strategy response and the stronger the system recovery ability.

[0100] ;

[0101] In the formula, Trq represents the time when the fault repair is completed, and Tfa represents the time when the fault occurs.

[0102] The acquired signal quality recovery evaluation nQs, time delay recovery evaluation QTf, and fault recovery time evaluation nRt are weighted and averaged to obtain a repair effect score SPre.

[0103] ;

[0104] In the formula, respectively represent the preset weight values of the signal quality recovery evaluation nQs, the time delay recovery evaluation QTf, and the fault recovery time evaluation nRt, and ;

[0105] The acquired repair effect score SPre is normalized and combined with the parameters in the fault diagnosis model in the same order of magnitude to obtain a new parameter , and the fault diagnosis model is readjusted.

[0106] The new parameter is acquired by the following formula:

[0107] ;

[0108] In the formula, SPre(t) represents the repair effect score at time t, SPre(t-1) represents the repair effect score at time t-1, maxSPre represents the maximum value of the repair effect score, represents the learning rate.

[0109] The application provides an optical fiber transmission signal monitoring system for long-distance communication, which has the following advantages:

[0110] (1) When the system is running, the fault diagnosis model integrated with the support vector machine algorithm can perform fault prediction based on the real-time collected data, and issue a fault warning in advance. This function can significantly reduce the response time after the fault occurs, avoid the long communication interruption or signal quality degradation caused by the lag in traditional methods, and thus improve the stability and availability of the optical fiber network. Combined with the topology of the optical fiber network and the real-time fault prediction data, the system can accurately locate the area where the fault occurs, and quickly determine the fault point through intelligent positioning algorithm. This accurate fault positioning capability compared with the manual inspection or periodic monitoring of the traditional system not only improves the fault response efficiency, but also optimizes the allocation of maintenance resources, avoiding the delay in repair caused by manual inspection.

[0111] (2) By integrating multiple high-precision sensors, the system can comprehensively collect key parameters in the optical fiber network, including optical fiber signal attenuation rate Af, signal time delay Tf, temperature Te, humidity He, and optical fiber vibration Vf, and other multi-dimensional data. The collection of such multi-source data can accurately reflect the state of the optical fiber in long-distance transmission, and comprehensively understand the health status of the optical fiber, providing high-quality basic data for subsequent fault prediction and repair. The data preprocessing unit automatically removes outliers through cleaning and normalization processing, and standardizes the data, ensuring the reliability and stability of the data in the subsequent analysis and model training process. Especially using the median absolute deviation method for outlier identification and removal, effectively improving the fault tolerance of the system in complex network environment, and reducing the interference of noise on the prediction result.

[0112] (3) Through the spatial weight inversion positioning unit, the system can combine the topology of the optical fiber network with the fault risk value Frisk and the fault probability value Pva to accurately locate the specific position of the fault. This capability enables the system to quickly narrow down the fault range from the overall network, avoiding the inefficiency of traditional methods relying on manual inspection or step-by-step troubleshooting. Through the weighted inversion algorithm, the system can dynamically derive the fault area and take targeted measures in time, thereby greatly improving the efficiency and accuracy of fault response.

[0113] In combination with the fault risk value Frisk, the fault probability value Pva and the uncertainty entropy rHF, the alarm grading decision and linkage unit can generate the joint trigger factor of the fault location, and respond according to the risk level. Through the three-level alarm mechanism, the system can take different levels of response measures such as early warning, re-inspection, encrypted sampling, and forced switching signal and re-inspection according to the severity and uncertainty of the fault. This intelligent grading response greatly improves the processing capacity of the system for different fault conditions, so that the system can more flexibly cope with the diversified fault conditions in the network environment.

[0114] (4) Through the dynamic repair strategy decision unit, the repair response intensity RS is automatically determined according to the fault risk value Frisk, the fault location Lf and the joint trigger factor Φ(Lf) of the fault location. According to the different repair response intensity RS, the response level and content of the repair strategy are determined. This intelligent decision system can automatically select the most suitable repair strategy according to the severity and location of the fault, such as signal gain adjustment, fiber channel switching, network segment level rerouting, etc. Compared with traditional manual operation or fixed strategy, this system can flexibly cope with different fault situations and provide more accurate and efficient repair solutions.

[0115] Through the strategy execution and link reconstruction unit, the system can automatically manage network resources and dynamically allocate optical switches, routing devices and amplifiers after the execution of the repair strategy. This process greatly reduces the need for manual intervention and improves the efficiency and accuracy of network management. Especially in long-distance communication networks, link reconstruction operations usually need to be executed quickly and accurately, and the automation of this module can ensure that network resources can be restored to the best state in time after fault repair, ensuring the continuity and stability of communication services. BRIEF DESCRIPTION OF DRAWINGS

[0116] Figure 1 A flowchart of the block diagram of the optical fiber transmission signal monitoring system for long-distance communication of the present application;

[0117] Figure 2 A flowchart of the optical fiber repair strategy execution of the present application;

[0118] Figure 3 A line graph of the joint trigger factor of the fault location of the present application;

[0119] Figure 4 A line graph of the matching of the joint trigger factor of the fault location and the risk level of the present application. DETAILED DESCRIPTION

[0120] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the protection scope of the present application.

[0121] Embodiment 1

[0122] The present application provides an optical fiber transmission signal monitoring system for long-distance communication, please refer to Figures 1 to 4 , comprising a data acquisition and processing module, a feature extraction and signal analysis module, a fault prediction model module, a fault location and alarm module, a fault repair and recovery strategy module and a maintenance feedback and optimization module.

[0123] The data acquisition and processing module collects the signal data transmitted by the optical fiber through the sensor, fits the original data set GX, and performs preprocessing to obtain the optical fiber data set GW.

[0124] The feature extraction and signal analysis module analyzes the optical fiber data set GW and performs feature extraction, fitting the feature set FG for diagnosing the optical fiber.

[0125] The fault prediction model module establishes a fault diagnosis model by using a support vector machine algorithm, performs fault prediction on the optical fiber according to the feature set FG, obtains a fault risk value Frisk, and judges the probability of the fault of the optical fiber through the fault risk value Frisk, and obtains a fault probability value Pva.

[0126] The fault location and alarm module locates the area where the fault occurs according to the obtained fault risk value Frisk and fault probability value Pva, combines the optical fiber network topology structure and feature set FG, obtains the fault location Lf and the joint trigger factor Φ (Lf) of the fault location, and sends an alarm notification.

[0127] The fault repair and recovery strategy module calculates the repair response strength RS according to the obtained fault risk value Frisk, fault location Lf and joint trigger factor Φ (Lf) of the fault location, and formulates and executes the repair strategy of the optical fiber fault.

[0128] The maintenance feedback and optimization module collects the data after the execution of the repair strategy, obtains a repair effect score SPre, and adjusts the fault diagnosis model according to the repair effect score SPre.

[0129] In this embodiment, through the integration of the fault diagnosis model supporting the support vector machine algorithm, the system can perform fault prediction based on real-time collected data and issue early fault warnings. This function can significantly reduce the response time after a fault occurs, avoid long communication interruptions or signal quality degradation caused by the lag in traditional methods, and thus improve the stability and availability of the optical fiber network. Combined with the topology of the optical fiber network and real-time fault prediction data, the system can accurately locate the area where the fault occurs and quickly determine the fault point through intelligent positioning algorithms. This precise fault positioning capability, compared to manual inspection or regular monitoring in traditional systems, not only improves the fault response efficiency but also optimizes the allocation of maintenance resources, avoiding delays in repairs due to manual inspection.

[0130] The system automatically selects and executes the optimal repair strategy based on the fault diagnosis results and positioning information, including switching to a backup path, adjusting signal gain, or performing channel switching. This automated repair mechanism ensures that the optical fiber transmission network quickly recovers communication after a fault occurs, greatly reducing the system recovery time caused by manual intervention, thereby ensuring the continuity of communication. After the execution of the repair strategy, the system evaluates the repair effect in real time and adjusts the parameters of the fault diagnosis model through the feedback and optimization module. Through continuous learning and optimization, the system can continuously improve the diagnostic accuracy and the execution efficiency of the repair strategy based on historical fault and repair data. This mechanism can adapt to changes in the network environment, improve the prediction accuracy and adaptability of future faults, and further enhance the intelligence and adaptability of the system.

[0131] The fault positioning and alarm module implements a three-level alarm mechanism based on the fault risk value and probability, combined with the joint trigger factor of the fault location, to ensure that the system responds reasonably under different risk levels. This intelligent alarm system can respond more flexibly to various fault situations and take appropriate measures according to the severity of the fault, effectively reducing potential losses caused by delayed or inaccurate alarm responses.

[0132] Embodiment 2

[0133] This embodiment is an explanation and description in Embodiment 1, please refer to Figure 1 and Figure 2 , specifically: the data acquisition and processing module includes a multi-source signal acquisition unit and a data preprocessing unit;

[0134] The multi-source signal acquisition unit collects signal data through sensors, including the optical fiber signal attenuation rate Af, the signal time delay Tf, the temperature Te, the humidity He, and the optical fiber vibration Vf, and fits them into the original data set GX;

[0135] Among them, the optical fiber signal attenuation rate Af is obtained by an optoelectronic sensor;

[0136] The optical fiber signal attenuation rate Af is obtained by the following formula:

[0137] ;

[0138] In the formula, Pin represents the optical signal input power, Pout represents the optical signal output power, and Lgx represents the optical fiber transmission length.

[0139] The signal time delay Tf is obtained by high-precision clock and optical pulse timing acquisition;

[0140] The temperature Te is obtained by temperature sensors distributed around the optical fiber;

[0141] The humidity He is obtained by humidity sensors in the communication tunnel wall distributed around the optical fiber;

[0142] The optical fiber vibration Vf is obtained by optical fiber Bragg grating vibration sensors fixed to the optical cable outer sheath;

[0143] The data preprocessing unit cleans and normalizes the obtained raw data set GX to obtain the optical fiber data set GW;

[0144] The cleaning includes outlier rejection, and the outliers in the raw data set GX are identified and rejected by using the median absolute deviation method;

[0145] The normalization processing is performed on the raw data set GX by using the Min-Max normalization method to obtain the optical fiber data set GW;

[0146] The optical fiber data set GW is obtained by the following formula:

[0147] ;

[0148] In the formula, GWb represents the bth data in the optical fiber data set GW, GXb represents the bth data in the raw data set GX, minGXb represents the valley value of the bth data in the raw data set GX, and maxGXb represents the peak value of the bth data in the raw data set GX.

[0149] The feature extraction and signal analysis module includes a time series fluctuation modeling unit and an environmental dynamic fitting unit;

[0150] The time series fluctuation modeling unit extracts features from the optical fiber data set GW by using a sliding time window technique, including signal intensity fluctuation ΔAf and time delay fluctuation ΔTf;

[0151] The signal intensity fluctuation ΔAf is obtained by the following formula:

[0152] ;

[0153] In the formula, L represents a sliding window length, represents a weight coefficient, Af(t-i) represents an optical fiber signal attenuation rate at time t-i, and pAf(t) represents an average attenuation rate at time t;

[0154] The time delay fluctuation ATf is obtained by the following formula:

[0155]

[0156] In the formula, Tf(t-i) represents a signal time delay at time t-i, maxTf(t-i) represents a maximum time delay value, and minTf(t-i) represents a minimum time delay value;

[0157] The environmental dynamic fitting unit analyzes the influence of temperature Te, humidity He, and optical fiber vibration Vf on signal transmission in the optical fiber data set GW, and calculates to obtain a temperature and humidity change rate AHt and a vibration change rate AVf;

[0158] The temperature and humidity change rate AHt is obtained by the following formula:

[0159]

[0160] In the formula, Te(t) represents a temperature at time t, Te(t-1) represents a temperature at time t-1, He(t) represents a humidity at time t, and He(t-1) represents a humidity at time t-1;

[0161] The vibration change rate AVf is obtained by the following formula:

[0162]

[0163] In the formula, At represents a time interval, Vf(t) represents an optical fiber vibration at time t, and Vf(t-1) represents an optical fiber vibration at time t-1;

[0164] The obtained signal intensity fluctuation DAf, time delay fluctuation ATf, temperature and humidity change rate AHt, and vibration change rate AVf are integrated to obtain a feature set FG.

[0165] ​​​In this embodiment, by integrating multiple high-precision sensors, the system can comprehensively collect key parameters in the optical fiber network, including optical fiber signal attenuation rate Af, signal time delay Tf, temperature Te, humidity He, and optical fiber vibration Vf, and other multi-dimensional data. This multi-source data collection can accurately reflect the state of the optical fiber in long-distance transmission, and fully understand the health status of the optical fiber, providing high-quality basic data for subsequent fault prediction and repair. The data preprocessing unit automatically removes outliers through cleaning and normalization processing, and standardizes the data, ensuring the reliability and stability of the data in the subsequent analysis and model training process. Especially using the median absolute deviation method for outlier identification and removal, effectively improving the fault tolerance of the system in complex network environment, and reducing the interference of noise on the prediction results.

[0166] The feature extraction and signal analysis module intelligently extracts the fluctuation characteristics of the optical fiber signal through time series fluctuation modeling and environmental dynamic fitting, and analyzes the influence of external environmental factors on optical fiber transmission. Through this multi-dimensional and dynamic analysis, the system can timely identify the possible fault area of the optical fiber, and enhance the sensitivity and accuracy of fault detection. The system can dynamically adjust the analysis model according to the real-time state of the optical fiber, especially under the influence of environmental factors such as temperature and humidity and vibration changes, to provide more accurate environmental perception for fault prediction. This intelligent dynamic adjustment mechanism can significantly improve the adaptability of the optical fiber network to environmental changes, reduce false positives caused by environmental changes, and enhance the stability of the system.

[0167] Through the integration of signal intensity fluctuation ΔAf, time delay fluctuation ΔTf, temperature and humidity change rate ΔHT, and vibration change rate ΔVf, the system can generate a highly representative feature set, which provides more comprehensive and accurate input data for the subsequent fault prediction model. Combined with machine learning algorithms, these feature sets can effectively identify potential faults in the optical fiber network, thereby providing early warnings before problems occur, improving the timeliness of fault response, and reducing the risk of communication interruption.

[0168] The modules in this embodiment improve the system's monitoring ability of the optical fiber network state through accurate data collection, real-time analysis, and dynamic optimization, and reduce the need for manual intervention, making the fault detection and repair process more automated and intelligent. This efficient automated processing method improves the reliability and self-healing ability of the network, while reducing maintenance costs and the risk of human error.

[0169] Embodiment 3

[0170] This embodiment is an explanation and description in Embodiment 2, please refer to Figure 1 , specifically: the fault prediction model module includes a feature mapping and classification decision unit and a probability mapping and risk normalization unit;

[0171] The feature mapping and classification decision unit performs high-dimensional mapping on the feature set FG, establishes a classification hyperplane through a support vector machine, constructs a fault diagnosis model, realizes binary classification decision of the fault state, calculates and obtains a fault risk value Frisk, and judges the fault state of the optical fiber through the fault risk value Frisk;

[0172] The fault risk value Frisk is obtained through the following formula:

[0173] ;

[0174] In the formula, FG(t) represents a feature vector at time t, FG(tj) represents the jth feature sample in the training set at time t, M represents the total number of feature samples, represents a parameter of the fault diagnosis model, yj represents a fault label of the jth feature sample, B represents a bias term, and K(, ) represents a kernel function;

[0175] The fault state of the optical fiber is matched and obtained through the following manner:

[0176] When the fault risk value Frisk> 0, it indicates that the fault state of the optical fiber is in a risk state;

[0177] When the fault risk value Frisk≤ 0, it indicates that the fault state of the optical fiber is in a normal state.

[0178] The probability mapping and risk normalization unit performs probability mapping on the obtained fault risk value Frisk, and calculates and obtains a fault probability value Pva;

[0179] The fault probability value Pva is obtained through the following formula:

[0180] ;

[0181] In the formula, e represents a constant, and c represents a steepness control parameter of a Sigmoid function;

[0182] Through analysis on the fault probability value Pva, an uncertainty entropy rHF is calculated and obtained, and compared with a preset judgment threshold value Trh to judge the accuracy of the fault probability occurrence;

[0183] The uncertainty entropy rHF is obtained through the following formula:

[0184] ;

[0185] In the formula, represents a logarithm function with 2 as the base;

[0186] The accuracy of the fault occurrence is matched and obtained through the following manner:

[0187] When the uncertainty entropy rHF≥ judgment threshold Trh, it indicates that the fault probability judgment is inaccurate.

[0188] When the uncertainty entropy rHF< judgment threshold Trh, it indicates that the fault probability judgment is accurate.

[0189] In this embodiment, through the feature mapping and classification decision unit, the system can map the multi-dimensional features of the optical fiber transmission signal to a high-dimensional space, and establish a fault diagnosis model through support vectors. This process not only enhances the recognition ability of optical fiber faults, but also can accurately divide the fault state of the optical fiber, so as to realize early fault prediction. Compared with the traditional method, this high-precision prediction method greatly shortens the fault detection time, improves the response speed to potential faults, and avoids long-time communication interruption or signal quality degradation. The fault risk value Frisk is calculated through the fault diagnosis model, and the value is used to judge whether the optical fiber is in a risk state. The system adopts a binary classification decision mechanism. Once the fault risk value is greater than zero, the system will enter the fault alarm state, prompting the network management personnel to check and repair in time. This intelligent judgment method based on data driving avoids the lag and uncertainty of manual experience judgment, effectively reduces the false judgment and delayed response caused by human factors.

[0190] The probability mapping and risk normalization unit converts the fault risk value into a fault probability value Pva, so that the system can not only judge whether a fault occurs, but also quantify the probability of fault occurrence, providing more accurate decision support. This probability-based fault evaluation method makes the management of optical fiber networks more intelligent, and can adjust network resources and maintenance plans according to the level of fault probability, thereby optimizing the overall operation efficiency of the network.

[0191] The system measures the accuracy of the fault prediction result by calculating the uncertainty entropy rHF. Combined with the preset judgment threshold Trh, the system can automatically evaluate the reliability of fault prediction. If the entropy value is high, it indicates that there is a large uncertainty in fault judgment, and the system can choose to delay decision or request manual review. This adaptive judgment mechanism can effectively reduce the false alarm rate and improve the accuracy of fault prediction of the system, avoiding the premature or incorrect fault diagnosis that may occur in traditional methods.

[0192] Embodiment 4

[0193] This embodiment is an explanation and description in embodiment 3. Please refer to Figure 3 and Figure 4 , in detail: the fault positioning and alarm module includes a spatial weight inversion positioning unit and an alarm grading decision and linkage unit;

[0194] The spatial weighted inversion positioning unit builds a node space model based on the optical fiber network topology, maps the acquired fault risk value Frisk and fault probability value Pva to the topological path, and derives the fault location Lf of the fault probability through a weighted inversion algorithm;

[0195] The fault location Lf is obtained by the following formula:

[0196] ;

[0197] Where Na represents the total number of topological nodes, Pva(ai) represents the failure probability value of the ai-th topological node, Frisk(ai) represents the failure risk value of the ai-th topological node, S(ai) represents the sensor sensitivity factor of the ai-th topological node, and X(ai) represents the position value of the ai-th topological node.

[0198] The alarm classification decision and linkage unit obtains the joint trigger factor Φ(Lf) of the fault location based on the acquired fault location Lf, combined with the fault risk value Frisk, uncertainty entropy rHF and fault probability value Pva, and triggers the three-level alarm response mechanism, including early warning, confirmed alarm and forced intervention;

[0199] The joint trigger factor Φ (Lf) of the fault location is obtained by the following formula, as shown in Table 1:

[0200] ;

[0201] In the formula, in the formula, represent the preset weight values ​​of the fault risk value Frisk, the fault probability value Pva and the uncertainty entropy rHF, respectively, and ;

[0202] The three-level alarm response mechanism is matched and obtained in the following ways:

[0203] When 0 < the combined trigger factor Φ (Lf) of the fault location < 0.4, it indicates the first risk level, and an early warning is issued for the fault location Lf, and the warning is recorded, waiting for the trend to develop;

[0204] When 0.4≤the joint trigger factor Φ(Lf) of the fault location≤0.7, it indicates the second risk level and the fault location Lf is re-inspected and sampling is increased;

[0205] When 0.7 < the joint trigger factor Φ (Lf) of the fault location < 1.0, it indicates the third risk level, and forced signal switching and cable re-inspection are performed on the fault location Lf.

[0206] Specific examples:

[0207] Preset Settings ;

[0208] The fault risk value Frisk = 0.2 is obtained;

[0209] The uncertainty entropy rHF = 0.05 is obtained;

[0210] The fault probability value Pva = 0.1 is obtained;

[0211] The joint trigger factor Φ (Lf) for fault location is calculated:

[0212] ;

[0213] The joint trigger factor Φ (Lf) for fault location = 0.125 is obtained, which belongs to the first risk level, and a pre-warning is given for the fault location Lf;

[0214] Table 1 Fault location trigger factor table:

[0215] Group number Failure risk value Frisk Failure probability value (Pva) Uncertainty entropy (rHF) Joint trigger factor for failure location (Φ(Lf)) Risk level A group 0.2 0.1 0.05 0.125 First risk level B group 0.5 0.4 0.3 0.41 Second risk level C group 0.8 0.7 0.6 0.71 Third risk level D group 0.3 0.2 0.2 0.24 First risk level E group 0.6 0.5 0.4 0.51 Second risk level

[0216] In this embodiment, through the spatial weight inversion positioning unit, the system can combine the topological structure of the optical fiber network with the fault risk value Frisk and the fault probability value Pva to accurately locate the specific location of the fault occurrence. This capability enables the system to quickly narrow down the fault range from the overall network, avoiding the inefficiency of relying on manual inspection or step-by-step investigation in traditional methods. Through the weighted inversion algorithm, the system can dynamically derive the fault area and take targeted measures in a timely manner, thereby greatly improving the efficiency and accuracy of fault response.

[0217] Combined with the fault risk value Frisk, the fault probability value Pva, and the uncertainty entropy rHF, the alarm grading decision and linkage unit can generate the joint trigger factor for fault location and respond according to the risk level. Through a three-level alarm mechanism, the system can take different levels of response measures such as pre-warning, re-inspection, encrypted sampling, and forced signal switching and re-inspection according to the severity and uncertainty of the fault. This intelligent grading response greatly improves the system's processing capability for different fault conditions, enabling the system to more flexibly cope with diverse fault conditions in the network environment.

[0218] Through the linkage fault risk and probability evaluation mechanism, combined with the comprehensive information of environmental changes and network topology, the system can quickly make decisions and start the corresponding repair process when potential faults occur. When the system determines that the fault risk is high through fault location and joint trigger factor, it can automatically start repair strategies such as encrypted sampling and signal switching, avoiding the lag in traditional methods and the delay caused by manual intervention, greatly improving the real-time performance and efficiency of fault response.

[0219] The module ensures flexible and multi-level response of the system when faults occur through a three-level alarm mechanism. When the risk of failure is low, the system will only record and warn, waiting for the trend to develop; when the risk is high, it will intervene forcibly, such as switching signals or rechecking optical fibers. This response mechanism effectively balances the sensitivity to faults and prevents excessive intervention, ensuring the stable operation of the network and reducing unnecessary maintenance operations.

[0220] Embodiment 5

[0221] This embodiment is an explanation and illustration in Embodiment 4, please refer to Figure 1 , in particular: the fault repair and recovery strategy module includes a dynamic repair strategy decision unit and a strategy execution and link reconstruction unit;

[0222] The dynamic repair strategy decision unit comprehensively analyzes the fault state, combines the fault risk value Frisk, the fault location Lf, and the joint trigger factor Φ(Lf) of the fault location, obtains the repair response strength RS, and determines the response level and strategy content of the repair strategy;

[0223] The repair response strength RS is obtained by the following formula:

[0224] ;

[0225] In the formula, A(Lf) represents the influence function of the fault location, represents an adjustable coefficient, and ;

[0226] The response level and strategy content of the repair strategy are obtained by the following method:

[0227] When 0 < repair response strength RS < 0.3, it represents the first response level; the strategy content is signal dynamic gain adjustment;

[0228] When 0.3 ≤ repair response strength RS ≤ 0.6, it represents the second response level; the strategy content is main and backup optical fiber channel switching;

[0229] When 0.6 < repair response strength RS < 0.8, it represents the third response level; the strategy content is network segment level rerouting;

[0230] When 0.8 ≤ repair response strength RS < 1.0, it represents the fourth response level; the strategy content is backup link activation and alarm pushing.

[0231] The strategy execution and link reconstruction unit automatically performs network resource management and link reconstruction operations after the execution of the repair strategy, including calling the control channel, and adjusting the optical switch, routing device, and amplifier.

[0232] The maintenance feedback and optimization module collects data after the repair strategy is executed, including signal quality recovery evaluation nQs, time delay recovery evaluation QTf, and fault recovery time evaluation nRt;

[0233] The signal quality recovery evaluation nQs is obtained by the following formula:

[0234]

[0235] In the formula, nAf represents the signal attenuation rate of the repaired optical fiber;

[0236] The obtained signal quality recovery evaluation nQs is compared with a preset recovery threshold TQs to determine the repair effect;

[0237] The repair effect is obtained by matching in the following manner:

[0238] When the signal quality recovery evaluation nQs is greater than the recovery threshold TQs, it indicates that the repair effect is normal, and the signal quality is significantly recovered;

[0239] When the signal quality recovery evaluation nQs is less than or equal to the recovery threshold TQs, it indicates that the repair effect is abnormal, and the signal quality is not recovered;

[0240] The time delay recovery evaluation QTf is obtained by the following formula:

[0241]

[0242] In the formula, nTf represents the signal time delay after repair;

[0243] The fault recovery time evaluation nRt is obtained by the following formula:

[0244]

[0245] In the formula, Trq represents the time when the fault repair is completed, and Tfa represents the time when the fault occurs;

[0246] The obtained signal quality recovery evaluation nQs, time delay recovery evaluation QTf, and fault recovery time evaluation nRt are weighted and averaged to obtain a repair effect score SPre;

[0247]

[0248] In the formula, respectively represent preset weight values of the signal quality recovery evaluation nQs, the time delay recovery evaluation QTf, and the fault recovery time evaluation nRt, and

[0249] The obtained repair effect score SPre is normalized and compared with parameters in a fault diagnosis model in the same order of magnitude​​​​​ acquire new parameters re-adjust the fault diagnosis model;

[0250] new parameters are acquired by the following formula:

[0251] ;

[0252] wherein SPre(t) represents the repair effect score at time t, SPre(t-1) represents the repair effect score at time t-1, maxSPre represents the maximum value of the repair effect score, denotes the learning rate.

[0253] In this embodiment, the dynamic repair strategy decision unit automatically determines the repair response intensity RS according to the fault risk value Frisk, the fault location Lf, and the joint trigger factor Φ(Lf) of the fault location, and determines the response level and content of the repair strategy according to the different repair response intensities RS. This intelligent decision system can automatically select the most appropriate repair strategy according to the severity and location of the fault, such as signal gain adjustment, fiber channel switching, network segment-level rerouting, etc. Compared with traditional manual operation or fixed strategy, this system can flexibly cope with different fault situations and provide more accurate and efficient repair solutions.

[0254] Through the strategy execution and link reconstruction unit, the system can automatically manage network resources and dynamically allocate optical switches, routing devices, and amplifiers after the execution of the repair strategy. This process greatly reduces the need for manual intervention and improves the efficiency and accuracy of network management. In particular, in long-distance communication networks, link reconstruction operations often need to be executed quickly and accurately, and the automated execution of this module can ensure that network resources can be restored to the best state in a timely manner after fault repair, ensuring the continuity and stability of communication services.

[0255] The system collects and evaluates the repaired data through the maintenance feedback and optimization module, including signal quality recovery, time delay recovery, and fault recovery time, etc. This real-time feedback mechanism can provide data support for the optimization of repair strategies, ensuring that the effect of each repair operation can be accurately evaluated, and dynamically adjusting the fault diagnosis model according to the evaluation results, further improving the accuracy of subsequent fault prediction and the efficiency of repair response. This adaptive optimization mechanism can achieve continuous system improvement, thereby improving the reliability of the overall network.

[0256] The system comprehensively considers multiple indexes such as signal quality, time delay recovery and fault recovery time in the repair effect evaluation, obtains a comprehensive score through weighted average, and compares with a preset threshold, so as to ensure that the system can accurately judge whether the repair is successful or not on the basis of different repair effects. This refined evaluation can not only detect the repair effect in real time, but also provide feedback for the fault diagnosis model, optimize the model parameters, and thus improve the accuracy of fault prediction.

[0257] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A fiber optic transmission signal monitoring system for long-distance communication, characterized by: It includes data acquisition and processing module, feature extraction and signal analysis module, fault prediction model module, fault location and alarm module, fault repair and recovery strategy module and maintenance feedback and optimization module; The data acquisition and processing module collects the signal data transmitted by the optical fiber through the sensor, fits it into the original data set GX, and performs preprocessing to obtain the optical fiber data set GW; The feature extraction and signal analysis module analyzes the optical fiber data set GW, extracts features, and fits them into a feature set FG for optical fiber diagnosis; The fault prediction model module uses the support vector machine algorithm to establish a fault diagnosis model, predicts optical fiber faults based on the feature set FG, obtains the fault risk value Frisk, and uses the fault risk value Frisk to determine the probability of optical fiber failure and obtain the fault probability value Pva; The fault location and alarm module locates the fault area based on the acquired fault risk value Frisk and fault probability value Pva, combined with the optical fiber network topology and feature set FG, obtains the fault location Lf and the joint trigger factor Φ(Lf) of the fault location, and issues an alarm notification; The fault repair and recovery strategy module calculates the repair response strength RS based on the acquired fault risk value Frisk, the fault location Lf, and the joint trigger factor Φ(Lf) of the fault location, and formulates and implements the repair strategy for the optical fiber fault. The maintenance feedback and optimization module collects data after the repair strategy is executed, obtains the repair effect score SPre, and adjusts the fault diagnosis model according to the repair effect score SPre.

2. The optical fiber transmission signal monitoring system for long-distance communication according to claim 1, characterized in that: The data acquisition and processing module includes a multi-source signal acquisition unit and a data pre-processing unit; The multi-source signal acquisition unit collects signal data through sensors, including optical fiber signal attenuation rate Af, signal delay Tf, temperature Te, humidity He and optical fiber vibration Vf, and fits them into the original data set GX; Among them, the optical fiber signal attenuation rate Af is acquired through photoelectric sensor; The optical fiber signal attenuation rate Af is obtained by the following formula: ； Where Pin represents the optical signal input power, Pout represents the optical signal output power, and Lgx represents the optical fiber transmission length; The signal delay Tf is acquired through high-precision clock and optical pulse timing acquisition; The temperature Te is acquired through temperature sensors distributed around the optical fiber; The humidity He is collected by humidity sensors distributed in the communication tunnel wall around the optical fiber; The optical fiber vibration Vf is acquired by a fiber Bragg grating vibration sensor fixed to the outer sheath of the optical cable; The data preprocessing unit cleans and normalizes the acquired original data set GX to obtain the optical fiber data set GW; Cleaning includes outlier removal, which is done by identifying and removing outliers from the original dataset GX using the median absolute deviation method; Normalization processing: The original data set GX is processed by using the Min-Max normalization method to obtain the fiber data set GW; The optical fiber dataset GW is obtained using the following formula: ； Where GWb represents the b-th data item in the optical fiber dataset GW, GXb represents the b-th data item in the original dataset GX, minGXb represents the valley value of the b-th data item in the original dataset GX, and maxGXb represents the peak value of the b-th data item in the original dataset GX.

3. The optical fiber transmission signal monitoring system for long-distance communication according to claim 1, characterized in that: The feature extraction and signal analysis module includes a time series fluctuation modeling unit and an environmental dynamic fitting unit; The time series fluctuation modeling unit extracts features from the optical fiber dataset GW using the sliding time window technology, including signal intensity fluctuation ΔAf and delay fluctuation ΔTf; The signal intensity fluctuation ΔAf is obtained by the following formula: ； Where L represents the sliding window length, represents the weight coefficient, Af(ti) represents the optical fiber signal attenuation rate at time ti, and pAf(t) represents the average attenuation rate at time t; The delay fluctuation ΔTf is obtained by the following formula: ； Where Tf(ti) represents the signal delay at time ti, maxTf(ti) represents the maximum delay value, and minTf(ti) represents the minimum delay value; The environmental dynamic fitting unit analyzes the effects of temperature Te, humidity He, and fiber vibration Vf in the fiber data set GW on signal transmission, and calculates the temperature and humidity change rate ΔHT and the vibration change rate ΔVf; The temperature and humidity change rate ΔHT is obtained by the following formula: ； Where Te(t) represents the temperature at time t, Te(t-1) represents the temperature at time t-1, He(t) represents the humidity at time t, and He(t-1) represents the humidity at time t-1; The vibration change rate ΔVf is obtained by the following formula: ； Where Δt represents the time interval, Vf(t) represents the optical fiber vibration at time t, and Vf(t-1) represents the optical fiber vibration at time t-1; The acquired signal strength fluctuation ΔAf, delay fluctuation ΔTf, temperature and humidity change rate ΔHT, and vibration change rate ΔVf are integrated to obtain the feature set FG.

4. The optical fiber transmission signal monitoring system for long-distance communication according to claim 1, characterized in that: The fault prediction model module includes a feature mapping and classification decision unit and a probability mapping and risk normalization unit; The feature mapping and classification decision unit performs high-dimensional mapping on the feature set FG, establishes a classification hyperplane through a support vector machine, builds a fault diagnosis model, implements a binary classification decision on the fault status, calculates the fault risk value Frisk, and uses the fault risk value Frisk to determine the fault status of the optical fiber; The failure risk value Frisk is obtained by the following formula: ； Where FG(t) represents the feature vector at time t, FG(tj) represents the jth feature sample in the training set at time t, M represents the total number of feature samples, represents the parameters of the fault diagnosis model, yj represents the fault label of the jth feature sample, B represents the bias term, and K(,) represents the kernel function; The fault status of the optical fiber is obtained by matching the following methods: When the fault risk value Frisk>0, it means that the fault status of the optical fiber is in a risky state; When the fault risk value Frisk≤0, it indicates that the fault state of the optical fiber is normal.

5. The optical fiber transmission signal monitoring system for long-distance communication according to claim 4, characterized in that: The probability mapping and risk normalization unit performs probability mapping on the acquired fault risk value Frisk and calculates the fault probability value Pva; The failure probability value Pva is obtained by the following formula: ； In the formula, e represents a constant, c represents the steepness control parameter of the Sigmoid function; By analyzing the fault probability value Pva, the uncertainty entropy rHF is calculated and compared with the preset judgment threshold Trh to judge the accuracy of the fault probability; The uncertainty entropy rHF is obtained by the following formula: ； Where, represents the logarithmic function with base 2; The accuracy of the fault occurrence is obtained by matching the following methods: When the uncertainty entropy rHF ≥ the judgment threshold Trh, it means that the fault probability judgment is inaccurate; When the uncertainty entropy rHF is less than the judgment threshold Trh, it means that the fault probability judgment is accurate.

6. The optical fiber transmission signal monitoring system for long-distance communication according to claim 5, characterized in that: The fault location and alarm module includes a spatial weight inversion location unit and an alarm classification decision and linkage unit; The spatial weighted inversion positioning unit builds a node space model based on the optical fiber network topology, maps the acquired fault risk value Frisk and fault probability value Pva to the topological path, and derives the fault location Lf of the fault probability through a weighted inversion algorithm; The fault location Lf is obtained by the following formula: ； Where Na represents the total number of topological nodes, Pva(ai) represents the failure probability value of the ai-th topological node, Frisk(ai) represents the failure risk value of the ai-th topological node, S(ai) represents the sensor sensitivity factor of the ai-th topological node, and X(ai) represents the position value of the ai-th topological node.

7. The optical fiber transmission signal monitoring system for long-distance communication according to claim 6, characterized in that: The alarm classification decision and linkage unit obtains the joint trigger factor Φ(Lf) of the fault location based on the acquired fault location Lf, combined with the fault risk value Frisk, uncertainty entropy rHF and fault probability value Pva, and triggers the three-level alarm response mechanism, including early warning, confirmed alarm and forced intervention; The joint trigger factor Φ (Lf) of the fault location is obtained by the following formula: ； In the formula, in the formula, represent the preset weight values ​​of the fault risk value Frisk, the fault probability value Pva and the uncertainty entropy rHF, respectively, and ; The three-level alarm response mechanism is matched and obtained in the following ways: When 0 < the combined trigger factor Φ (Lf) of the fault location < 0.4, it indicates the first risk level, and an early warning is issued for the fault location Lf, and the warning is recorded, waiting for the trend to develop; When 0.4≤the joint trigger factor Φ(Lf) of the fault location≤0.7, it indicates the second risk level and the fault location Lf is re-inspected and sampling is increased; When 0.7 < the joint trigger factor Φ (Lf) of the fault location < 1.0, it indicates the third risk level, and forced signal switching and cable re-inspection are performed on the fault location Lf.

8. The optical fiber transmission signal monitoring system for long-distance communication according to claim 7, characterized in that: The fault repair and recovery strategy module includes a dynamic repair strategy decision unit and a strategy execution and link reconstruction unit; The dynamic repair strategy decision unit performs a comprehensive analysis of the fault status, combines the fault risk value Frisk, the fault location Lf, and the joint trigger factor Φ(Lf) of the fault location, obtains the repair response strength RS, and determines the response level and strategy content of the repair strategy; The repair response strength RS is obtained by the following formula: ； Where A(Lf) represents the influence function of the fault location, represents the adjustable coefficient, and ; The response level and policy content of the repair policy are obtained in the following ways: When 0<repair response strength RS<0.3, it indicates the first response level; strategy content: signal dynamic gain adjustment; When 0.3≤Repair Response Strength RS≤0.6, it indicates the second response level; Policy content: Active / standby Fibre Channel switchover; When 0.6<repair response strength RS<0.8, it indicates the third response level; Policy content: network segment-level rerouting; When 0.8≤repair response strength RS<1.0, it indicates the fourth response level; policy content: backup link activation and alarm push.

9. The optical fiber transmission signal monitoring system for long-distance communication according to claim 8, characterized in that: After the repair strategy is executed, the policy execution and link reconstruction unit automatically performs network resource management and link reconstruction operations, including calling control channels and deploying optical switches, routing equipment and amplifiers.

10. The optical fiber transmission signal monitoring system for long-distance communication according to claim 1, characterized in that: The maintenance feedback and optimization module collects data after the repair strategy is executed, including signal quality recovery evaluation nQs, delay recovery evaluation QTf and fault recovery time evaluation nRt; The signal quality recovery evaluation nQs is obtained by the following formula: ； Where nAf represents the attenuation rate of the repaired optical fiber signal; Compare the obtained signal quality recovery evaluation nQs with the preset recovery threshold TQs to determine the repair effect; The repair effect is obtained by matching the following methods: When the signal quality recovery evaluation nQs> the recovery threshold TQs, it means that the repair effect is normal and the signal quality has been significantly restored; When the signal quality recovery evaluation nQs ≤ the recovery threshold TQs, it means that the repair effect is abnormal and the signal quality has not been restored; The time delay recovery evaluation QTf is obtained by the following formula: ； Where nTf represents the signal delay after repair; The fault recovery time evaluation nRt is obtained by the following formula: ； Where Trq represents the time when the fault is repaired, and Tfa represents the time when the fault occurs; Take a weighted average of the obtained signal quality recovery evaluation nQs, delay recovery evaluation QTf, and fault recovery time evaluation nRt to obtain the repair effect score SPre; ； Where, They represent the preset weight values ​​of signal quality recovery evaluation nQs, delay recovery evaluation QTf and fault recovery time evaluation nRt, respectively, and ; The repair effect score SPre is normalized and compared with the parameters in the fault diagnosis model at the same level. Combined to obtain new parameters , re-adjust the fault diagnosis model; New parameters Obtained by the following formula: ； In the formula, SPre(t) represents the repair effect score at time t, SPre(t-1) represents the repair effect score at time t-1, and maxSPre represents the maximum value of the repair effect score. Represents the learning rate.

Images