A Deep Learning-Based Method and System for Assessing the Status of Fiber Backup Channels

Abstract

This invention discloses a deep learning-based method and system for assessing the status of optical fiber backup channels, relating to the field of communication transmission status assessment technology. The invention obtains the current disturbance response sequence by applying an individualized small disturbance test sequence to the target optical fiber backup channel. Based on the current disturbance response sequence and the individual baseline response spectrum, a current status assessment sample is generated. This sample is then input into a status assessment model to output the status assessment result of the target optical fiber backup channel. This achieves transient carrying capacity assessment of long-term low-activity optical fiber backup channels under real switching conditions, alleviating the problem that existing methods can only reflect static health status and cannot identify the risk of transient transmission distortion during switching.

Description

Technical Field

[0001] This invention relates to the field of communication transmission status assessment technology, and in particular to a method and system for assessing the status of optical fiber backup channels based on deep learning. Background Technology

[0002] In fiber optic communication networks, to improve service continuity when the primary transmission path experiences anomalies, a backup fiber optic channel is typically configured for the primary transmission path. This backup channel remains in a low-activity state most of the time, only being activated when the primary transmission path experiences anomalies, maintenance switching, or drill switching. Maintenance personnel regularly conduct offline tests on the backup channel, such as sending test frames and loopback tests. The start and end times of these tests are also recorded in the logs. Current test activations are typically used to verify the basic connectivity of the backup channel, such as whether it can transmit light, receive test frames, or measure static parameters such as optical power and bit error rate. The testing methods are fixed, periodic, and standardized. The purpose is to determine whether the backup channel is currently operational and whether its static indicators meet the requirements.

[0003] Currently, most methods for determining the status of fiber optic backup channels rely on static monitoring. This static monitoring typically includes received optical power, transmitted optical power, bit error rate statistics, forward error correction statistics, alarm status, and some link quality parameters. While this method can reflect the current status of the fiber optic backup channel under static observation conditions, it cannot reflect the dynamic response capability of the backup channel during actual handover. In practical applications, fiber optic backup channels in a long-term low-activity state may still experience transient transmission distortion during actual handover, even if their static monitoring status is normal. This transient transmission distortion manifests as: short-term fluctuations in received signal after handover begins, a sudden increase in bit error rate statistics, an abnormally large increase in forward error correction, prolonged carrier recovery process, reduced equalization convergence speed, amplified phase error fluctuations, or increased stabilization time. These distortions are usually not apparent during static monitoring and only become apparent briefly after the actual handover is triggered. This presents the following problems: First, existing methods typically determine the availability of backup channels based on whether static monitoring values ​​exceed limits. This conclusion corresponds to a static health status, not the transient carrying capacity under actual handover conditions. Second, existing methods often use a uniform threshold across the entire network to judge the status of different backup channels. However, different backup channels differ in link length, device aging, amplification characteristics, historical activation frequencies, and tuning states at both ends. The response amplitude generated by the same disturbance on different backup channels will not be consistent, making a uniform threshold prone to misjudgment. Third, existing technologies usually rely on actual handover results to expose problems, failing to identify the transient distortion risks of the fiber optic backup channels in advance when they are not carrying real services for extended periods. Summary of the Invention

[0004] To address the technical problems existing in the background art, this invention proposes a method and system for evaluating the status of optical fiber backup channels based on deep learning.

[0005] The optical fiber backup channel status assessment method proposed in this invention includes the following steps: S1. Obtain historical monitoring data and historical switching records of the target optical fiber backup channel, and construct the original historical sample set; S2. Historical disturbance response records extracted from the original historical sample set are used to generate individual baseline response spectra corresponding to the target optical fiber backup channel; S3. Generate individualized small perturbation test sequences corresponding to the target optical fiber backup channel based on the original historical sample set; S4. Apply an individualized small perturbation test sequence to the target optical fiber backup channel to obtain the current perturbation response sequence; S5. Generate a current state assessment sample based on the current disturbance response sequence and individual baseline response spectrum; S6. A state assessment model is obtained by deep learning training. The state assessment model is used to output the state assessment results of the target fiber optic backup channel. The state assessment results include: direct switching is possible, reversible switching is possible, and switching is prohibited. Input the current state assessment sample into the state assessment model, and the state assessment model outputs the state assessment result corresponding to the current state assessment sample.

[0006] Preferably, in S1, the historical monitoring data includes: transmitted optical power, received optical power, optical signal-to-noise ratio, forward error correction statistics, bit error statistics, frequency offset estimate, phase error statistics, polarization-related state variables, and equalization tap changes. Historical handover records include: actual handover start time, actual handover completion time, actual handover stable establishment time, actual handover result record, test activation start time, test activation end time, test stable establishment time, and test handover result record; Both the actual handover result record and the test handover result record include: handover prohibited, handover resumable, and handover can be directly performed; When constructing the original historical sample set, the historical monitoring data and historical switching records are arranged in chronological order to obtain the original historical sample set.

[0007] Preferably, in S2, historical disturbance response records extracted from the original historical sample set are used to generate the individual baseline response spectrum corresponding to the target optical fiber backup channel, as follows: Candidate perturbation response records are extracted from the original historical sample set to form a candidate baseline response set. Historical perturbation response records are then identified from the candidate baseline response set to form a historical baseline response set. For each candidate perturbation response record, a corresponding response feature vector is generated; the response feature vector includes: response start delay, peak offset, and recovery slope; Obtain all response feature vectors to form a candidate response feature vector set; Extract the response feature vector corresponding to each historical disturbance response record from the candidate response feature vector set to form a response feature vector set; In the set of response feature vectors, obtain all response start delays and get the median, first quartile, third quartile, and interquartile range of the response start delay; Obtain all peak offsets to get the median, first quartile, third quartile, and interquartile range of the peak offsets; Obtain all recovery slopes and get the median, first quartile, third quartile, and interquartile range of the recovery slope; The median, first quartile, third quartile, and interquartile range of the response start delay, the median, first quartile, third quartile, and interquartile range of the peak offset, and the median, first quartile, third quartile, and interquartile range of the recovery slope form the individual baseline response spectrum.

[0008] Preferably, in S2, candidate perturbation response records are extracted from the original historical sample set to form a candidate baseline response set; historical perturbation response records are identified from the candidate baseline response set to form a historical baseline response set; a corresponding response feature vector is generated for each candidate perturbation response record; the response feature vector includes: response start delay, peak offset, and recovery slope, as follows: Historical monitoring data corresponding to the range from the start time of test activation to the end time of test activation are extracted from the original historical sample set and used as candidate disturbance response records; Obtain all candidate perturbation response records to form a candidate baseline response set; In the candidate baseline response set, if the test handover result record corresponding to the candidate disturbance response record in the original historical sample set is a recoverable handover or a direct handover, then the candidate disturbance response record is regarded as a historical disturbance response record. Obtain all historical disturbance response records to form a historical baseline response set; For each candidate perturbation response record, a corresponding response feature vector is generated; the response feature vector includes: response start delay, peak offset, and recovery slope; The response start delay, peak offset, and recovery slope are generated as follows: Historical monitoring data corresponding to the range from the start time of the actual handover to the completion time of the actual handover are extracted from the original historical sample set and used as the actual handover response record; In the original historical sample set, all candidate disturbance response records and actual switch response records are removed from the historical monitoring data to form inactive phase monitoring data; Extract all received optical power from the monitoring data of the inactive phase, arrange them in chronological order, and obtain the received optical power sequence of the inactive phase; The normal fluctuation range of received optical power in the received optical power sequence during the inactive phase was obtained by using a 1.5 times interquartile range. As an explanation, 1.5 times the interquartile range is a commonly used standard in existing statistics for identifying outliers in data. Its core principle is based on the extended calculation of the interquartile range. The upper boundary of the normal fluctuation range is the third quartile of the received optical power sequence in the inactive stage plus 1.5 times the interquartile range, and the lower boundary is the first quartile of the received optical power sequence in the inactive stage minus 1.5 times the interquartile range. The interquartile range refers to the difference between the third quartile and the first quartile. For each candidate perturbation response record, extract all received optical power in that candidate perturbation response record and arrange them in chronological order to obtain the received optical power sequence during the test activation phase; In the received optical power sequence during the test activation phase, the moment corresponding to the first received optical power exceeding the normal fluctuation range is taken as the response start moment. The time difference between the response start time and the test activation start time corresponding to the candidate perturbation response record is used as the response start delay; Obtain the average value of the received optical power in the received optical power sequence during the inactive phase as the steady-state value; In the received optical power sequence during the test activation phase, the maximum absolute difference between the received optical power and the steady-state value is used as the peak offset. In the received optical power sequence during the test activation phase, the moment when the peak offset occurs is taken as the peak moment. For the received optical power after the peak moment, when the value of the received optical power first drops to within the normal fluctuation range, the moment corresponding to the received optical power is recorded as the recovery moment. The recovery slope is obtained by subtracting the received optical power at the recovery time from the received optical power at the peak time, taking the absolute value, and then dividing it by the time difference between the peak time and the recovery time.

[0009] Preferably, in S3, based on the original historical sample set, an individualized small perturbation test sequence corresponding to the target optical fiber backup channel is generated, as follows: For each real handover response record, the transmitted optical power is extracted from each real handover response record and arranged in chronological order to obtain the transmitted optical power sequence; All transmitted optical power sequences are obtained, and the transmitted optical power in each transmitted optical power sequence is normalized to obtain the amplitude-normalized transmitted optical power as the amplitude-normalized transmitted optical power, thus forming an amplitude-normalized transmitted optical power sequence. All amplitude-normalized transmitted optical power sequences are subjected to time normalization to obtain time-normalized transmitted optical power sequences. Each normalized time position in the time-normalized transmitted optical power sequence corresponds to an amplitude-normalized transmitted optical power. Obtain all time-normalized emitted optical power sequences to form a set of historical variation trajectories; In the historical trajectory set, at each normalized time position, the median value of the amplitude normalized emitted optical power of all time-normalized emitted optical power sequences at that normalized time position is obtained as the trajectory median value; Obtain the median value of all trajectories, and connect the median values ​​of the trajectories at each normalized time position in chronological order to obtain the target change prototype trajectory. A stable fluctuation scale is generated based on monitoring data from the inactive phase. Obtain the steady-state value of the amplitude-normalized emitted optical power sequence; As an explanation, the steady-state value of the amplitude-normalized transmitted optical power sequence is 1 or 0, depending on the amplitude normalization method. If the existing amplitude normalization adopts the proportional rule, the steady-state value is 1; if the difference proportional rule is adopted, the steady-state value is 0. In the target change prototype trajectory, the absolute value of the difference between the position value and the steady-state value of the amplitude normalized emitted light power sequence in each trajectory is obtained as the deviation value, forming a deviation value set. The maximum value of the deviation value in the deviation value set is taken as the maximum deviation of the target change prototype trajectory. Using the ratio of the maximum deviation between the stable fluctuation scale and the target change prototype trajectory as a scaling factor, the median value of each trajectory in the target change prototype trajectory is multiplied by the scaling factor to obtain an individualized small perturbation test sequence.

[0010] Preferably, in S3, a stable fluctuation scale is generated based on the monitoring data from the inactive phase, as follows: Extract all emitted optical power from the monitoring data of the inactive phase, arrange them in chronological order, and obtain the inactive emitted optical power sequence; The amplitude of the emitted optical power in the inactive emitted optical power sequence is normalized to obtain the amplitude-normalized inactive emitted optical power sequence. The first and third quartiles of the amplitude-normalized unactivated emitted optical power sequence were obtained respectively; The difference between the third quartile and the first quartile of the amplitude-normalized unactivated emitted optical power sequence is used as the stable fluctuation scale.

[0011] Preferably, in S4, an individualized small perturbation test sequence is applied to the target optical fiber backup channel to obtain the current perturbation response sequence, as follows: An individualized small perturbation test sequence is applied to the target optical fiber backup channel; Starting from the moment the individualized small perturbation test sequence is applied to the backup channel of the target optical fiber, the target monitoring quantity is collected at a fixed sampling interval until the individualized small perturbation test sequence is applied. The collected target monitoring quantities are arranged in chronological order to form the current disturbance response sequence; Target monitoring quantities include: transmitted optical power, received optical power, optical signal-to-noise ratio, forward error correction statistics, bit error statistics, frequency offset estimates, phase error statistics, polarization-related state quantities, and equalization tap changes. For clarification, the data collected for both target monitoring and historical monitoring include transmitted optical power, received optical power, optical signal-to-noise ratio, forward error correction statistics, bit error statistics, frequency offset estimates, phase error statistics, polarization-related state variables, and equalization tap changes.

[0012] Preferably, in S5, a current state assessment sample is generated based on the current disturbance response sequence and the individual baseline response spectrum, as follows: Generate the current response feature vector based on the current disturbance response sequence; The current response feature vector includes the current response start delay, the current peak offset, and the current recovery slope; The current response start delay, current peak offset, and current recovery slope are generated in the following manner: Extract all received optical power from the current disturbance response sequence and arrange them in chronological order to obtain the current received optical power sequence; In the current received optical power sequence, the moment corresponding to the first received optical power exceeding the normal fluctuation range is taken as the current response start moment; The start time of applying an individualized small perturbation test sequence to the backup channel of the target optical fiber is taken as the start time of the current small perturbation test; The time difference between the current response start time and the current small disturbance test start time is used as the current response start delay; In the current received optical power sequence, the maximum absolute difference between the received optical power and the steady-state value is used as the current peak offset; In the current received optical power sequence, the moment when the current peak offset occurs is denoted as the current peak moment; For the received optical power after the current peak moment, the moment when the received optical power first drops to the normal fluctuation range is taken as the current recovery moment; The current recovery slope is obtained by subtracting the received optical power at the current peak time from the received optical power at the current recovery time, taking the absolute value, and then dividing it by the time difference between the current peak time and the current recovery time. Based on the individual baseline response spectrum, the difference between the current response start delay and the median value of the response start delay in the individual baseline response spectrum is used as the current response start delay offset; the difference between the current peak offset and the median value of the peak offset in the individual baseline response spectrum is used as the current peak offset deviation; and the difference between the current recovery slope and the median value of the recovery slope in the individual baseline response spectrum is used as the current recovery slope offset. The current response start delay offset, the current peak offset deviation, and the current recovery slope offset form the current feature offset vector; The current disturbance response sequence, the current response feature vector, and the current feature offset vector are concatenated to generate the current state evaluation sample.

[0013] Preferably, in S6, a state evaluation model is obtained by training with deep learning, as follows: Obtain the test switch result record corresponding to each candidate disturbance response record, and form a test switch result record set; For each response feature vector in the candidate response feature vector set, a corresponding historical feature offset vector is generated to obtain the historical feature offset vector set; For each response feature vector in the candidate response feature vector set, based on the individual baseline response spectrum, the historical response start delay offset is the difference between the response start delay and the median value of the response start delay in the individual baseline response spectrum; the historical peak offset offset is the difference between the peak offset and the median value of the peak offset in the individual baseline response spectrum; and the historical recovery slope offset is the difference between the recovery slope and the median value of the recovery slope in the individual baseline response spectrum. The historical response start delay offset, historical peak offset deviation, and historical recovery slope offset form a historical feature offset vector; Obtain all historical feature offset vectors to form a set of historical feature offset vectors; Based on the candidate baseline response set, the candidate response feature vector set, and the historical feature offset vector set, test the set of switching results records; For each candidate disturbance response record, the response feature vector and historical feature offset vector corresponding to the candidate disturbance response record are concatenated with the candidate disturbance response record to generate a historical state evaluation sample. Obtain all historical state assessment samples to form a historical state assessment sample set; Based on the historical state assessment sample set and the test handover result record set, the state assessment model is trained using deep learning. The historical state assessment sample is used as input, and the corresponding test handover result record in the test handover result record set of the candidate disturbance response record of the historical state assessment sample is used as output. The state assessment model is used to output the state assessment result of the target fiber backup channel. The state assessment result includes: direct handover is possible, handover is recoverable, and handover is prohibited.

[0014] A deep learning-based optical fiber backup channel status assessment system includes: Data acquisition module: acquires historical monitoring data and historical switching records of the target fiber optic backup channel to construct the original historical sample set; Individual baseline response spectrum generation module: Extracts historical disturbance response records from the original historical sample set and generates the individual baseline response spectrum corresponding to the target optical fiber backup channel; Disturbance test sequence generation module: Generates individualized small disturbance test sequences corresponding to the target fiber backup channel based on the original historical sample set; Current Disturbance Response Sequence Generation Module: Apply an individualized small disturbance test sequence to the backup channel of the target optical fiber to obtain the current disturbance response sequence; Current state assessment sample generation module: Generates current state assessment samples based on the current disturbance response sequence and individual baseline response spectrum; Channel status assessment module: The status assessment model is trained using deep learning. The status assessment model is used to output the status assessment results of the target fiber backup channel. The status assessment results include: direct switching is possible, switchable is recoverable, and switching is prohibited. Input the current state assessment sample into the state assessment model, and the state assessment model outputs the state assessment result corresponding to the current state assessment sample.

[0015] The optical fiber backup channel status assessment method and system proposed in this invention have the following beneficial technical effects: 1. This invention obtains the current disturbance response sequence by applying an individualized small disturbance test sequence to the target optical fiber backup channel, and generates a current state assessment sample based on the current disturbance response sequence and the individual baseline response spectrum. The current state assessment sample is then input into the state assessment model to output the state assessment result of the target optical fiber backup channel. This enables the assessment of the transient carrying capacity of long-term low-activity optical fiber backup channels under real switching conditions, alleviating the problem that existing methods can only reflect static health status and cannot identify the risk of transient transmission distortion during switching.

[0016] 2. This invention generates an individual baseline response spectrum corresponding to the target optical fiber backup channel by extracting historical disturbance response records from the original historical sample set. Based on the individual baseline response spectrum, the difference between the current response start delay and the median value of the response start delay in the individual baseline response spectrum is used as the current response start delay offset; the difference between the current peak offset and the median value of the peak offset in the individual baseline response spectrum is used as the current peak offset deviation; and the difference between the current recovery slope and the median value of the recovery slope in the individual baseline response spectrum is used as the current recovery slope offset. This allows the state assessment model to make judgments based on the individual baseline response spectrum of each optical fiber backup channel, alleviating the problem of misjudgment caused by individual differences between different channels when using a uniform threshold across the entire network.

[0017] 3. This invention generates individualized small-disturbance test sequences corresponding to the target optical fiber backup channel based on the original historical sample set, and applies the individualized small-disturbance test sequences to the target optical fiber backup channel. Under the condition of not carrying real services for a long time, the dynamic response process is obtained. By combining the current disturbance response sequence, the current response feature vector, and the current feature offset vector, a current state evaluation sample is generated and input into the state evaluation model. This enables the early identification of transient distortion risks under low-risk testing methods, alleviating the problem of existing methods that rely on real switching results to expose problems and cannot evaluate them in advance. Attached Figure Description

[0018] Figure 1 This is a flowchart of the deep learning-based optical fiber backup channel status assessment method of the present invention; Figure 2 This is a block diagram illustrating the principle of the deep learning-based optical fiber backup channel status assessment system of the present invention. Detailed Implementation

[0019] Embodiments of the present invention are described in detail below. Examples of these embodiments are illustrated in the accompanying drawings, wherein the same or similar symbols denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0020] like Figure 1 The deep learning-based optical fiber backup channel status assessment method shown includes the following steps: S1. Obtain historical monitoring data and historical switching records of the target optical fiber backup channel, and construct the original historical sample set; In an optional embodiment, in S1, the historical monitoring data includes: transmitted optical power, received optical power, optical signal-to-noise ratio, forward error correction statistics, bit error statistics, frequency offset estimation, phase error statistics, polarization-related state variables, and equalization tap changes. Historical handover records include: actual handover start time, actual handover completion time, actual handover stable establishment time, actual handover result record, test activation start time, test activation end time, test stable establishment time, and test handover result record; Both the actual handover result record and the test handover result record include: handover prohibited, handover resumable, and handover can be directly performed; When constructing the original historical sample set, the historical monitoring data and historical switching records are arranged in chronological order to obtain the original historical sample set; S2. Historical disturbance response records extracted from the original historical sample set are used to generate individual baseline response spectra corresponding to the target optical fiber backup channel; In an optional embodiment, in S2, historical disturbance response records extracted from the original historical sample set are used to generate the individual baseline response spectrum corresponding to the target fiber backup channel, as follows: Candidate perturbation response records are extracted from the original historical sample set to form a candidate baseline response set. Historical perturbation response records are then identified from the candidate baseline response set to form a historical baseline response set. For each candidate perturbation response record, a corresponding response feature vector is generated; the response feature vector includes: response start delay, peak offset, and recovery slope; Obtain all response feature vectors to form a candidate response feature vector set; Extract the response feature vector corresponding to each historical disturbance response record from the candidate response feature vector set to form a response feature vector set; In the set of response feature vectors, obtain all response start delays and get the median, first quartile, third quartile, and interquartile range of the response start delay; Obtain all peak offsets to get the median, first quartile, third quartile, and interquartile range of the peak offsets; Obtain all recovery slopes and get the median, first quartile, third quartile, and interquartile range of the recovery slope; The median, first quartile, third quartile, and interquartile range of the response start delay, the median, first quartile, third quartile, and interquartile range of the peak offset, and the median, first quartile, third quartile, and interquartile range of the recovery slope form the individual baseline response spectrum. In an optional embodiment, in S2, candidate perturbation response records are extracted from the original historical sample set to form a candidate baseline response set; historical perturbation response records are identified from the candidate baseline response set to form a historical baseline response set; a corresponding response feature vector is generated for each candidate perturbation response record; the response feature vector includes: response start delay, peak offset, and recovery slope, as follows: Historical monitoring data corresponding to the range from the start time of test activation to the end time of test activation are extracted from the original historical sample set and used as candidate disturbance response records; Obtain all candidate perturbation response records to form a candidate baseline response set; In the candidate baseline response set, if the test handover result record corresponding to the candidate disturbance response record in the original historical sample set is a recoverable handover or a direct handover, then the candidate disturbance response record is regarded as a historical disturbance response record. Obtain all historical disturbance response records to form a historical baseline response set; For each candidate perturbation response record, a corresponding response feature vector is generated; the response feature vector includes: response start delay, peak offset, and recovery slope; The response start delay, peak offset, and recovery slope are generated as follows: Historical monitoring data corresponding to the range from the start time of the actual handover to the completion time of the actual handover are extracted from the original historical sample set and used as the actual handover response record; In the original historical sample set, all candidate disturbance response records and actual switch response records are removed from the historical monitoring data to form inactive phase monitoring data; Extract all received optical power from the monitoring data of the inactive phase, arrange them in chronological order, and obtain the received optical power sequence of the inactive phase; The normal fluctuation range of received optical power in the received optical power sequence during the inactive phase was obtained by using a 1.5 times interquartile range. As an explanation, 1.5 times the interquartile range is a commonly used standard in existing statistics for identifying outliers in data. Its core principle is based on the extended calculation of the interquartile range. The upper boundary of the normal fluctuation range is the third quartile of the received optical power sequence in the inactive stage plus 1.5 times the interquartile range, and the lower boundary is the first quartile of the received optical power sequence in the inactive stage minus 1.5 times the interquartile range. The interquartile range refers to the difference between the third quartile and the first quartile. For each candidate perturbation response record, extract all received optical power in that candidate perturbation response record and arrange them in chronological order to obtain the received optical power sequence during the test activation phase; In the received optical power sequence during the test activation phase, the moment corresponding to the first received optical power exceeding the normal fluctuation range is taken as the response start moment. The time difference between the response start time and the test activation start time corresponding to the candidate perturbation response record is used as the response start delay; Obtain the average value of the received optical power in the received optical power sequence during the inactive phase as the steady-state value; In the received optical power sequence during the test activation phase, the maximum absolute difference between the received optical power and the steady-state value is used as the peak offset. In the received optical power sequence during the test activation phase, the moment when the peak offset occurs is taken as the peak moment. For the received optical power after the peak moment, when the value of the received optical power first drops to within the normal fluctuation range, the moment corresponding to the received optical power is recorded as the recovery moment. The recovery slope is obtained by subtracting the received optical power at the recovery time from the received optical power at the peak time, taking the absolute value, and then dividing it by the time difference between the peak time and the recovery time. S3. Generate individualized small perturbation test sequences corresponding to the target optical fiber backup channel based on the original historical sample set; In an optional embodiment, in S3, based on the original historical sample set, an individualized small perturbation test sequence corresponding to the target fiber backup channel is generated as follows: For each real handover response record, the transmitted optical power is extracted from each real handover response record and arranged in chronological order to obtain the transmitted optical power sequence; All transmitted optical power sequences are obtained, and the transmitted optical power in each transmitted optical power sequence is normalized to obtain the amplitude-normalized transmitted optical power as the amplitude-normalized transmitted optical power, thus forming an amplitude-normalized transmitted optical power sequence. All amplitude-normalized transmitted optical power sequences are subjected to time normalization to obtain time-normalized transmitted optical power sequences. Each normalized time position in the time-normalized transmitted optical power sequence corresponds to an amplitude-normalized transmitted optical power. Obtain all time-normalized emitted optical power sequences to form a set of historical variation trajectories; In the historical trajectory set, at each normalized time position, the median value of the amplitude normalized emitted optical power of all time-normalized emitted optical power sequences at that normalized time position is obtained as the trajectory median value; Obtain the median value of all trajectories, and connect the median values ​​of the trajectories at each normalized time position in chronological order to obtain the target change prototype trajectory. A stable fluctuation scale is generated based on monitoring data from the inactive phase. Obtain the steady-state value of the amplitude-normalized emitted optical power sequence; As an explanation, the steady-state value of the amplitude-normalized transmitted optical power sequence is 1 or 0, depending on the amplitude normalization method. If the existing amplitude normalization adopts the proportional rule, the steady-state value is 1; if the difference proportional rule is adopted, the steady-state value is 0. In the target change prototype trajectory, the absolute value of the difference between the position value and the steady-state value of the amplitude normalized emitted light power sequence in each trajectory is obtained as the deviation value, forming a deviation value set. The maximum value of the deviation value in the deviation value set is taken as the maximum deviation of the target change prototype trajectory. Using the ratio of the stable fluctuation scale to the maximum deviation of the target change prototype trajectory as a scaling factor, the median value of each trajectory in the target change prototype trajectory is multiplied by the scaling factor to obtain an individualized small perturbation test sequence. In an optional embodiment, in S3, a stable fluctuation scale is generated based on the monitoring data from the inactive phase, as follows: Extract all emitted optical power from the monitoring data of the inactive phase, arrange them in chronological order, and obtain the inactive emitted optical power sequence; The amplitude of the emitted optical power in the inactive emitted optical power sequence is normalized to obtain the amplitude-normalized inactive emitted optical power sequence. The first and third quartiles of the amplitude-normalized unactivated emitted optical power sequence were obtained respectively; The difference between the third quartile and the first quartile of the amplitude-normalized unactivated emission power sequence is used as the stable fluctuation scale. S4. Apply an individualized small perturbation test sequence to the target optical fiber backup channel to obtain the current perturbation response sequence; In an optional embodiment, in S4, an individualized small perturbation test sequence is applied to the target fiber backup channel to obtain the current perturbation response sequence, as follows: an individualized small perturbation test sequence is applied to the target fiber backup channel; Starting from the moment the individualized small perturbation test sequence is applied to the backup channel of the target optical fiber, the target monitoring quantity is collected at a fixed sampling interval until the individualized small perturbation test sequence is applied. The collected target monitoring quantities are arranged in chronological order to form the current disturbance response sequence; Target monitoring quantities include: transmitted optical power, received optical power, optical signal-to-noise ratio, forward error correction statistics, bit error statistics, frequency offset estimates, phase error statistics, polarization-related state quantities, and equalization tap changes. As an explanation, the data collected for both target monitoring and historical monitoring include transmitted optical power, received optical power, optical signal-to-noise ratio, forward error correction statistics, bit error statistics, frequency offset estimates, phase error statistics, polarization-related state quantities, and equalization tap changes. S5. Generate a current state assessment sample based on the current disturbance response sequence and individual baseline response spectrum; In an optional embodiment, in S5, a current state assessment sample is generated based on the current perturbation response sequence and the individual baseline response spectrum, as follows: Generate the current response feature vector based on the current disturbance response sequence; The current response feature vector includes the current response start delay, the current peak offset, and the current recovery slope; The current response start delay, current peak offset, and current recovery slope are generated in the following manner: Extract all received optical power from the current disturbance response sequence and arrange them in chronological order to obtain the current received optical power sequence; In the current received optical power sequence, the moment corresponding to the first received optical power exceeding the normal fluctuation range is taken as the current response start moment; The start time of applying an individualized small perturbation test sequence to the backup channel of the target optical fiber is taken as the start time of the current small perturbation test; The time difference between the current response start time and the current small disturbance test start time is used as the current response start delay; In the current received optical power sequence, the maximum absolute difference between the received optical power and the steady-state value is used as the current peak offset; In the current received optical power sequence, the moment when the current peak offset occurs is denoted as the current peak moment; For the received optical power after the current peak moment, the moment when the received optical power first drops to the normal fluctuation range is taken as the current recovery moment; The current recovery slope is obtained by subtracting the received optical power at the current peak time from the received optical power at the current recovery time, taking the absolute value, and then dividing it by the time difference between the current peak time and the current recovery time. Based on the individual baseline response spectrum, the difference between the current response start delay and the median value of the response start delay in the individual baseline response spectrum is used as the current response start delay offset; the difference between the current peak offset and the median value of the peak offset in the individual baseline response spectrum is used as the current peak offset deviation; and the difference between the current recovery slope and the median value of the recovery slope in the individual baseline response spectrum is used as the current recovery slope offset. The current response start delay offset, the current peak offset deviation, and the current recovery slope offset form the current feature offset vector; The current disturbance response sequence, the current response feature vector, and the current feature offset vector are concatenated to generate the current state evaluation sample.

[0021] S6. A state assessment model is obtained by deep learning training. The state assessment model is used to output the state assessment results of the target fiber optic backup channel. The state assessment results include: direct switching is possible, reversible switching is possible, and switching is prohibited. Input the current state assessment sample into the state assessment model, and the state assessment model outputs the state assessment result corresponding to the current state assessment sample.

[0022] This invention generates an individual baseline response spectrum for the target fiber backup channel by extracting historical disturbance response records from the original historical sample set. Based on the individual baseline response spectrum, the difference between the current response start delay and the median value of the response start delay in the individual baseline response spectrum is used as the current response start delay offset; the difference between the current peak offset and the median value of the peak offset in the individual baseline response spectrum is used as the current peak offset deviation; and the difference between the current recovery slope and the median value of the recovery slope in the individual baseline response spectrum is used as the current recovery slope offset. This allows the state assessment model to make judgments based on the individual baseline response spectrum of each fiber backup channel, alleviating the problem of misjudgment caused by individual differences between different channels when using a uniform threshold across the entire network.

[0023] In an optional embodiment, in S6, a state evaluation model is trained using deep learning, as follows: Obtain the test switch result record corresponding to each candidate disturbance response record, and form a test switch result record set; For each response feature vector in the candidate response feature vector set, a corresponding historical feature offset vector is generated to obtain the historical feature offset vector set; For each response feature vector in the candidate response feature vector set, based on the individual baseline response spectrum, the historical response start delay offset is the difference between the response start delay and the median value of the response start delay in the individual baseline response spectrum; the historical peak offset offset is the difference between the peak offset and the median value of the peak offset in the individual baseline response spectrum; and the historical recovery slope offset is the difference between the recovery slope and the median value of the recovery slope in the individual baseline response spectrum. The historical response start delay offset, historical peak offset deviation, and historical recovery slope offset form a historical feature offset vector; Obtain all historical feature offset vectors to form a set of historical feature offset vectors; Based on the candidate baseline response set, the candidate response feature vector set, and the historical feature offset vector set, test the set of switching results records; For each candidate disturbance response record, the response feature vector and historical feature offset vector corresponding to the candidate disturbance response record are concatenated with the candidate disturbance response record to generate a historical state evaluation sample. Obtain all historical state assessment samples to form a historical state assessment sample set; Based on the historical state assessment sample set and the test handover result record set, the state assessment model is trained using deep learning. The historical state assessment sample is used as input, and the corresponding test handover result record in the test handover result record set of the candidate disturbance response record of the historical state assessment sample is used as output. The state assessment model is used to output the state assessment result of the target fiber backup channel. The state assessment result includes: direct handover is possible, handover is recoverable, and handover is prohibited.

[0024] This invention obtains the current disturbance response sequence by applying an individualized small disturbance test sequence to the target optical fiber backup channel, and generates a current state assessment sample based on the current disturbance response sequence and the individual baseline response spectrum. The current state assessment sample is then input into the state assessment model to output the state assessment result of the target optical fiber backup channel. This enables the assessment of the transient carrying capacity of long-term low-activity optical fiber backup channels under real switching conditions, alleviating the problem that existing methods can only reflect static health status and cannot identify the risk of transient transmission distortion during switching.

[0025] This invention generates individualized small-disturbance test sequences corresponding to the target optical fiber backup channel based on the original historical sample set, and applies these individualized small-disturbance test sequences to the target optical fiber backup channel. Under the condition of not carrying real services for a long time, the dynamic response process is obtained. By combining the current disturbance response sequence, the current response feature vector, and the current feature offset vector, a current state evaluation sample is generated and input into the state evaluation model. This enables the early identification of transient distortion risks under low-risk testing methods, alleviating the problem of existing methods that rely on real switching results to expose problems and cannot evaluate them in advance.

[0026] As an explanation, this invention assesses the risk of transient distortion by outputting state evaluation results of direct switchability, recoverable switchability, and prohibited switchability. Direct switchability corresponds to almost no transient distortion, stability after switchability, and no abnormal fluctuations or distortions. Recoverable switchability corresponds to short-term transient distortion, but it can recover on its own. After switchability, there are brief signal fluctuations and a sudden increase in bit errors, but it quickly returns to normal. Prohibited switchability corresponds to transient distortion, continuous distortion after switchability, inability to establish a stable connection, or direct rollback.

[0027] like Figure 2 The deep learning-based fiber optic backup channel status assessment system shown includes: Data acquisition module: acquires historical monitoring data and historical switching records of the target fiber optic backup channel to construct the original historical sample set; Individual baseline response spectrum generation module: Extracts historical disturbance response records from the original historical sample set and generates the individual baseline response spectrum corresponding to the target optical fiber backup channel; Disturbance test sequence generation module: Generates individualized small disturbance test sequences corresponding to the target fiber backup channel based on the original historical sample set; Current Disturbance Response Sequence Generation Module: Apply an individualized small disturbance test sequence to the backup channel of the target optical fiber to obtain the current disturbance response sequence; Current state assessment sample generation module: Generates current state assessment samples based on the current disturbance response sequence and individual baseline response spectrum; Channel status assessment module: The status assessment model is trained using deep learning. The status assessment model is used to output the status assessment results of the target fiber backup channel. The status assessment results include: direct switching is possible, switchable is recoverable, and switching is prohibited. Input the current state assessment sample into the state assessment model, and the state assessment model outputs the state assessment result corresponding to the current state assessment sample.

[0028] For clarification, "acquisition" in this application refers to obtaining the required content or data using existing technical means.

[0029] Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0030] In the embodiments provided by this invention, it should be understood that the disclosed system or method can be implemented in other ways. For example, the embodiments of the invention described above are merely illustrative; for instance, the division of modules is only a logical functional division, and there may be other division methods in actual implementation.

[0031] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0032] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or in the form of hardware plus software functional modules.

[0033] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the basic characteristics of the present invention.

[0034] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A deep learning-based method for assessing the status of optical fiber backup channels, characterized in that, Includes the following steps: S1. Obtain historical monitoring data and historical switching records of the target optical fiber backup channel, and construct the original historical sample set; S2. Historical disturbance response records extracted from the original historical sample set are used to generate individual baseline response spectra corresponding to the target optical fiber backup channel; S3. Generate individualized small perturbation test sequences corresponding to the target optical fiber backup channel based on the original historical sample set; S4. Apply an individualized small perturbation test sequence to the target optical fiber backup channel to obtain the current perturbation response sequence; S5. Generate a current state assessment sample based on the current disturbance response sequence and individual baseline response spectrum; S6. A state assessment model is obtained by deep learning training. The state assessment model is used to output the state assessment results of the target fiber optic backup channel. The state assessment results include: direct switching is possible, reversible switching is possible, and switching is prohibited. Input the current state assessment sample into the state assessment model, and the state assessment model outputs the state assessment result corresponding to the current state assessment sample.

2. The deep learning-based optical fiber backup channel status assessment method according to claim 1, characterized in that, In S1, historical monitoring data includes: transmitted optical power, received optical power, optical signal-to-noise ratio, forward error correction statistics, bit error statistics, frequency offset estimates, phase error statistics, polarization-related state variables, and equalization tap changes. Historical handover records include: actual handover start time, actual handover completion time, actual handover stable establishment time, actual handover result record, test activation start time, test activation end time, test stable establishment time, and test handover result record; Both the actual handover result record and the test handover result record include: handover prohibited, handover resumable, and handover can be directly performed; When constructing the original historical sample set, the historical monitoring data and historical switching records are arranged in chronological order to obtain the original historical sample set.

3. The deep learning-based optical fiber backup channel status assessment method according to claim 2, characterized in that, In S2, historical disturbance response records extracted from the original historical sample set are used to generate the individual baseline response spectrum corresponding to the target fiber backup channel, as follows: Candidate perturbation response records are extracted from the original historical sample set to form a candidate baseline response set. Historical perturbation response records are then identified from the candidate baseline response set to form a historical baseline response set. Generate a corresponding response feature vector for each candidate perturbation response record; The response feature vector includes: response start delay, peak offset, and recovery slope; Obtain all response feature vectors to form a candidate response feature vector set; Extract the response feature vector corresponding to each historical disturbance response record from the candidate response feature vector set to form a response feature vector set; In the set of response feature vectors, obtain all response start delays and get the median, first quartile, third quartile, and interquartile range of the response start delay; Obtain all peak offsets to get the median, first quartile, third quartile, and interquartile range of the peak offsets; Obtain all recovery slopes and get the median, first quartile, third quartile, and interquartile range of the recovery slope; The median, first quartile, third quartile, and interquartile range of the response start delay, the median, first quartile, third quartile, and interquartile range of the peak offset, and the median, first quartile, third quartile, and interquartile range of the recovery slope form the individual baseline response spectrum.

4. The deep learning-based optical fiber backup channel status assessment method according to claim 3, characterized in that, In S2, candidate disturbance response records are extracted from the original historical sample set to form a candidate baseline response set, and historical disturbance response records are identified from the candidate baseline response set to form a historical baseline response set. Generate a corresponding response feature vector for each candidate perturbation response record; The response feature vector includes: response start delay, peak offset, and recovery slope, as follows: Historical monitoring data corresponding to the range from the start time of test activation to the end time of test activation are extracted from the original historical sample set and used as candidate disturbance response records; Obtain all candidate perturbation response records to form a candidate baseline response set; In the candidate baseline response set, if the test handover result record corresponding to the candidate disturbance response record in the original historical sample set is a recoverable handover or a direct handover, then the candidate disturbance response record is regarded as a historical disturbance response record. Obtain all historical disturbance response records to form a historical baseline response set; For each candidate perturbation response record, a corresponding response feature vector is generated; the response feature vector includes: response start delay, peak offset, and recovery slope; The response start delay, peak offset, and recovery slope are generated as follows: Historical monitoring data corresponding to the range from the start time of the actual handover to the completion time of the actual handover are extracted from the original historical sample set and used as the actual handover response record; In the original historical sample set, all candidate disturbance response records and actual switch response records are removed from the historical monitoring data to form inactive phase monitoring data; Extract all received optical power from the monitoring data of the inactive phase, arrange them in chronological order, and obtain the received optical power sequence of the inactive phase; The normal fluctuation range of received optical power in the received optical power sequence during the inactive phase was obtained by using a 1.5 times interquartile range. For each candidate perturbation response record, extract all received optical power in that candidate perturbation response record and arrange them in chronological order to obtain the received optical power sequence during the test activation phase; In the received optical power sequence during the test activation phase, the moment corresponding to the first received optical power exceeding the normal fluctuation range is taken as the response start moment. The time difference between the response start time and the test activation start time corresponding to the candidate perturbation response record is used as the response start delay; Obtain the average value of the received optical power in the received optical power sequence during the inactive phase as the steady-state value; In the received optical power sequence during the test activation phase, the maximum absolute difference between the received optical power and the steady-state value is used as the peak offset. In the received optical power sequence during the test activation phase, the moment when the peak offset occurs is taken as the peak moment. For the received optical power after the peak moment, when the value of the received optical power first drops to within the normal fluctuation range, the moment corresponding to the received optical power is recorded as the recovery moment. The recovery slope is obtained by subtracting the received optical power at the recovery time from the received optical power at the peak time, taking the absolute value, and then dividing it by the time difference between the peak time and the recovery time.

5. The deep learning-based optical fiber backup channel status assessment method according to claim 3, characterized in that, In S3, based on the original historical sample set, an individualized small perturbation test sequence corresponding to the target fiber backup channel is generated, as follows: For each real handover response record, the transmitted optical power is extracted from each real handover response record and arranged in chronological order to obtain the transmitted optical power sequence; All transmitted optical power sequences are obtained, and the transmitted optical power in each transmitted optical power sequence is normalized to obtain the amplitude-normalized transmitted optical power as the amplitude-normalized transmitted optical power, thus forming an amplitude-normalized transmitted optical power sequence. All amplitude-normalized transmitted optical power sequences are subjected to time normalization to obtain time-normalized transmitted optical power sequences. Each normalized time position in the time-normalized transmitted optical power sequence corresponds to an amplitude-normalized transmitted optical power. Obtain all time-normalized emitted optical power sequences to form a set of historical variation trajectories; In the historical trajectory set, at each normalized time position, the median value of the amplitude normalized emitted optical power of all time-normalized emitted optical power sequences at that normalized time position is obtained as the trajectory median value; Obtain the median value of all trajectories, and connect the median values ​​of the trajectories at each normalized time position in chronological order to obtain the target change prototype trajectory. A stable fluctuation scale is generated based on monitoring data from the inactive phase. Obtain the steady-state value of the amplitude-normalized emitted optical power sequence; In the target change prototype trajectory, the absolute value of the difference between the position value and the steady-state value of the amplitude normalized emitted light power sequence in each trajectory is obtained as the deviation value, forming a deviation value set. The maximum value of the deviation value in the deviation value set is taken as the maximum deviation of the target change prototype trajectory. Using the ratio of the maximum deviation between the stable fluctuation scale and the target change prototype trajectory as a scaling factor, the median value of each trajectory in the target change prototype trajectory is multiplied by the scaling factor to obtain an individualized small perturbation test sequence.

6. The deep learning-based optical fiber backup channel status assessment method according to claim 5, characterized in that, In S3, a stable fluctuation scale is generated based on the monitoring data from the inactive phase, as follows: Extract all emitted optical power from the monitoring data of the inactive phase, arrange them in chronological order, and obtain the inactive emitted optical power sequence; The amplitude of the emitted optical power in the inactive emitted optical power sequence is normalized to obtain the amplitude-normalized inactive emitted optical power sequence. The first and third quartiles of the amplitude-normalized unactivated emitted optical power sequence were obtained respectively; The difference between the third quartile and the first quartile of the amplitude-normalized unactivated emitted optical power sequence is used as the stable fluctuation scale.

7. The deep learning-based optical fiber backup channel status assessment method according to claim 1, characterized in that, In S4, an individualized small perturbation test sequence is applied to the target fiber backup channel to obtain the current perturbation response sequence, as follows: An individualized small perturbation test sequence is applied to the target fiber backup channel; Starting from the moment the individualized small perturbation test sequence is applied to the backup channel of the target optical fiber, the target monitoring quantity is collected at a fixed sampling interval until the individualized small perturbation test sequence is applied. The collected target monitoring quantities are arranged in chronological order to form the current disturbance response sequence; The target monitoring quantities include: transmitted optical power, received optical power, optical signal-to-noise ratio, forward error correction statistics, bit error statistics, frequency offset estimates, phase error statistics, polarization-related state quantities, and equalization tap changes.

8. The deep learning-based optical fiber backup channel status assessment method according to claim 5, characterized in that, In S5, based on the current disturbance response sequence and the individual baseline response spectrum, a current state assessment sample is generated as follows: Generate the current response feature vector based on the current disturbance response sequence; The current response feature vector includes the current response start delay, the current peak offset, and the current recovery slope; The current response start delay, current peak offset, and current recovery slope are generated in the following manner: Extract all received optical power from the current disturbance response sequence and arrange them in chronological order to obtain the current received optical power sequence; In the current received optical power sequence, the moment corresponding to the first received optical power exceeding the normal fluctuation range is taken as the start time of the current response; The start time of applying an individualized small perturbation test sequence to the backup channel of the target optical fiber is taken as the start time of the current small perturbation test; The time difference between the current response start time and the current small disturbance test start time is used as the current response start delay; In the current received optical power sequence, the maximum absolute difference between the received optical power and the steady-state value is used as the current peak offset; In the current received optical power sequence, the moment when the current peak offset occurs is denoted as the current peak moment; For the received optical power after the current peak moment, the moment when the received optical power first drops to the normal fluctuation range is taken as the current recovery moment; The current recovery slope is obtained by subtracting the received optical power at the current peak time from the received optical power at the current recovery time, taking the absolute value, and then dividing it by the time difference between the current peak time and the current recovery time. Based on the individual baseline response spectrum, the difference between the current response start delay and the median value of the response start delay in the individual baseline response spectrum is used as the current response start delay offset; the difference between the current peak offset and the median value of the peak offset in the individual baseline response spectrum is used as the current peak offset deviation; and the difference between the current recovery slope and the median value of the recovery slope in the individual baseline response spectrum is used as the current recovery slope offset. The current response start delay offset, the current peak offset deviation, and the current recovery slope offset form the current feature offset vector; The current disturbance response sequence, the current response feature vector, and the current feature offset vector are concatenated to generate the current state evaluation sample.

9. The method for assessing the status of optical fiber backup channels based on deep learning according to claim 8, characterized in that, In S6, a state evaluation model is obtained by training with deep learning, as follows: Obtain the test switch result record corresponding to each candidate disturbance response record, and form a test switch result record set; For each response feature vector in the candidate response feature vector set, a corresponding historical feature offset vector is generated to obtain the historical feature offset vector set; For each response feature vector in the candidate response feature vector set, based on the individual baseline response spectrum, the historical response start delay offset is the difference between the response start delay and the median value of the response start delay in the individual baseline response spectrum; the historical peak offset offset is the difference between the peak offset and the median value of the peak offset in the individual baseline response spectrum; and the historical recovery slope offset is the difference between the recovery slope and the median value of the recovery slope in the individual baseline response spectrum. The historical response start delay offset, historical peak offset deviation, and historical recovery slope offset form a historical feature offset vector; Obtain all historical feature offset vectors to form a set of historical feature offset vectors; Based on the candidate baseline response set, the candidate response feature vector set, and the historical feature offset vector set, test the set of switching results records; For each candidate disturbance response record, the response feature vector and historical feature offset vector corresponding to the candidate disturbance response record are concatenated with the candidate disturbance response record to generate a historical state evaluation sample. Obtain all historical state assessment samples to form a historical state assessment sample set; Based on the historical state assessment sample set and the test handover result record set, the state assessment model is trained using deep learning. The historical state assessment sample is used as input, and the corresponding test handover result record in the test handover result record set of the candidate disturbance response record of the historical state assessment sample is used as output. The state assessment model is used to output the state assessment result of the target fiber backup channel. The state assessment result includes: direct handover is possible, handover is recoverable, and handover is prohibited.

10. A deep learning-based optical fiber backup channel status assessment system, used to apply the deep learning-based optical fiber backup channel status assessment method according to any one of claims 1 to 9, characterized in that, include: Data acquisition module: acquires historical monitoring data and historical switching records of the target fiber optic backup channel to construct the original historical sample set; Individual baseline response spectrum generation module: Extracts historical disturbance response records from the original historical sample set and generates the individual baseline response spectrum corresponding to the target optical fiber backup channel; Disturbance test sequence generation module: Generates individualized small disturbance test sequences corresponding to the target fiber backup channel based on the original historical sample set; Current Disturbance Response Sequence Generation Module: Apply an individualized small disturbance test sequence to the backup channel of the target optical fiber to obtain the current disturbance response sequence; Current state assessment sample generation module: Generates current state assessment samples based on the current disturbance response sequence and individual baseline response spectrum; Channel status assessment module: The status assessment model is trained using deep learning. The status assessment model is used to output the status assessment results of the target fiber backup channel. The status assessment results include: direct switching is possible, switchable is recoverable, and switching is prohibited. Input the current state assessment sample into the state assessment model, and the state assessment model outputs the state assessment result corresponding to the current state assessment sample.

Images