A method for locating fiber transient fault based on FEC error burst

Abstract

The application discloses a kind of based on FEC error code burst optical fiber transient fault positioning method, comprising the following steps: step 1: obtaining FEC decoding telemetry;Step 2: identify FEC error code burst event at each endpoint;Step 3: generate burst fingerprint for each FEC error code burst event;Step 4: get the correction arrival time of FEC error code burst event to corresponding endpoint, endpoint time synchronization residual and the uncertainty of the correction arrival time;Step 5: get the same source burst event pair, corresponding matching score and the difference between the first endpoint correction arrival time and the second endpoint correction arrival time;Step 6: calculate the distance of transient fault point relative to the first endpoint;Step 7: output the confidence interval or the reliable level of the distance.The application has the characteristics of on-line detection without service interruption, FEC code word level burst identification, multi-delay joint correction, double-end same source matching, non-fiber disturbance rejection and positioning uncertainty quantification.

Description

Technical Field

[0001] This invention belongs to the field of optical communication network operation and maintenance and optical fiber link fault location technology, specifically involving an optical fiber transient fault location method based on FEC bit error bursts. Background Technology

[0002] As the scale of data center interconnects, metropolitan area transport networks, backbone networks, OTN / WDM systems, and coherent optical communication systems continues to expand, the data rate carried on a single optical fiber link is constantly increasing. High-speed Ethernet links, coherent DWDM links, and interconnect links for AI clusters typically rely on forward error correction (FEC) to improve transmission reliability. FEC decoders can correct errors in the bitstream at the receiving end, which are affected by noise, dispersion, nonlinearity, polarization disturbances, connector reflections, and link micro-bending, and output monitoring data such as pre-FEC BER, post-FEC BER, number of corrected codewords, and number of uncorrectable codewords.

[0003] Current fiber optic fault location primarily relies on testing technologies such as OTDR, coherent OTDR, or OFDR. These technologies obtain information on reflection, scattering, or phase changes by injecting test pulses or sweeping signals, thereby determining the location of fiber breaks, strong reflection points, or abnormal attenuation. However, in online links carrying high-capacity services, dedicated test signals may require test windows, additional hardware, bypass links, or service scheduling, making it difficult to capture transient faults such as short-lived micro-bends, vibrations, loose connectors, transient contamination, or localized reflections in a timely manner. For millisecond-level or even microsecond-level disturbances, traditional periodic inspections and offline testing often only detect error alarms after the fault disappears, making it difficult to provide an accurate location for dispatching repairs.

[0004] On the other hand, modern optical modules, coherent transceivers, switching chips, and OTN equipment are capable of outputting a large amount of physical layer telemetry data, such as received optical power, SNR, Q factor, pre-FEC BER, post-FEC BER, FEC error correction statistics, module temperature, power supply voltage, and DSP lockout status. Existing network management systems typically use this data for link health monitoring, threshold alarms, service protection, or rerouting. For example, when the pre-FEC BER continues to rise, the system can determine that the link quality has deteriorated and trigger protection switching. However, this approach usually only indicates an anomaly in the end-to-end link and cannot convert the anomaly into a specific distance along the optical fiber.

[0005] Existing literature has proposed using the time history of optical transceiver synchronized physical layer parameters for fault location, or monitoring burst errors in FEC frames to determine link problems. However, FEC error bursts have unique characteristics such as discrete codeword indexes, FEC frame boundaries, interleaving / deinterleaving delays, DSP fixed processing delays, and sampling window phases. If the FEC bit error rate is treated as a normal physical layer time series without correcting for the above factors, the time estimation of the arrival of the same transient disturbance at both ends will show significant deviations, thus affecting the location accuracy. At the same time, DSP unlocking on the equipment side, module power supply disturbances, temperature anomalies, protection switching, and service-side congestion may also cause bit error spikes. Without a source matching and non-fiber disturbance rejection mechanism, mislocation or mis-assignment can easily occur. For example, US Patent Application Publication No. US20120224846A1 discloses an optical transceiver fault location scheme that synchronizes the clocks of the optical transceivers at both ends of the link, records the time history of physical layer parameters such as BER, OSNR, optical power, polarization state, and propagation delay, and estimates the fault location based on the event records at both ends. This scheme is based on changes in general physical layer parameters and time synchronization at both ends. It does not disclose the construction of burst events for FEC codeword-level error correction strength, nor does it perform joint correction of FEC frame boundaries, interleaving / deinterleaving delays, DSP fixed processing delays and codeword phases. Furthermore, it does not utilize FEC burst fingerprints, physical propagation intervals and device-side common-mode states for source matching and false alarm elimination. Therefore, under discrete FEC telemetry conditions, arrival time deviations and mislocations may still exist.

[0006] Therefore, a method is needed that can operate during normal service flow, without injecting dedicated detection sequences, without relying on interrupted service OTDR testing, and capable of online location of error bursts via telemetry naturally output by the FEC decoder. This method should specifically address FEC codeword discretization, codeword phase, interleaving delay, and transceiver fixed delay, and combine this with two- or multi-end source matching, physical propagation interval constraints, and confidence interval output to improve the location accuracy and engineering availability of fiber optic transient faults. Summary of the Invention

[0007] To overcome the shortcomings of the existing technology, the present invention aims to provide a fiber optic transient fault location method based on FEC error bursts. This method features uninterrupted online detection, FEC codeword-level burst identification, multi-delay joint correction, dual-end homogeneous matching, non-fiber disturbance elimination, and quantification of location uncertainty.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for locating transient optical fiber faults based on FEC (Fault-Error-Based Capture) bursts is applied to a bidirectional optical fiber link carrying normal service flow. The bidirectional optical fiber link has a first endpoint and a second endpoint, which are optical communication devices located at both ends of the bidirectional optical fiber link and capable of outputting FEC decoding telemetry. The bidirectional optical fiber link includes a first service direction from the first endpoint to the second endpoint and a second service direction from the second endpoint to the first endpoint. The method specifically includes the following steps: Step 1: Acquire FEC decoding telemetry in the first receiving direction of the first endpoint and the second receiving direction of the second endpoint, respectively. The first receiving direction and the second receiving direction correspond to opposite service directions in the bidirectional optical fiber link. Step 2: Based on the spatial clustering of continuous high error correction strength codewords, uncorrectable codewords, or error symbols in the FEC decoding telemetry, identify FEC error burst events at each endpoint; Step 3: Generate a burst fingerprint for each of the aforementioned FEC error burst events; Step 4: Based on the endpoint time base, FEC codeword period, FEC frame boundary, interleaving / deinterleaving delay, transceiver fixed processing delay and the codeword phase, correct the codeword time index of the FEC error burst event to obtain the corrected arrival time of the FEC error burst event to the corresponding endpoint, the endpoint time synchronization residual and the uncertainty of the corrected arrival time. Step 5: Based on the envelope similarity of the burst fingerprint and the physical propagation interval constraint limited by the link length and group velocity, perform source matching on the FEC error burst events of the first endpoint and the second endpoint to obtain the source burst event pairs, the corresponding matching scores, and the difference between the correction arrival time of the first endpoint and the correction arrival time of the second endpoint. Step 6: Using the first endpoint as the reference endpoint, when the corrected arrival time difference of the same-origin burst event pair satisfies the physical propagation interval constraint, calculate the distance of the transient fault point relative to the first endpoint based on the corrected arrival time difference, the link length of the optical fiber link, and the group velocity of light in the optical fiber link. Step 7: Based on the burst fingerprint matching score output in Step 5, the endpoint time synchronization residual output in Step 4, and the corrected arrival time uncertainty, output the confidence interval or confidence level of the distance.

[0009] In step 1, the FEC decoding telemetry includes error correction strength information corresponding to the FEC codeword time index; the error correction strength information includes at least one of the following: number of error correction symbols, number of error correction bits, number of error correction codewords, uncorrectable codeword flag, error symbol histogram, and pre-FEC BER estimate. The FEC decoding telemetry is either codeword-by-codeword data or short-window aggregated data. When the FEC decoding telemetry is short-window aggregated data, an equivalent FEC codeword time index is generated based on the window start codeword and window end codeword, and the window width is included in the uncertainty of the correction arrival time.

[0010] In step 2, for each endpoint, service direction, and wavelength or channel FEC codeword sequence obtained in step 1, the background mean and background scale of the error correction intensity are calculated using a sliding window, and a dynamic threshold is formed. When a number of consecutive codewords reach the dynamic threshold, an uncorrectable codeword appears, or the spatial clustering of error symbols reaches the clustering threshold, adjacent abnormal codewords are merged into candidate FEC error burst events, and the starting codeword index, peak codeword index, and ending codeword index are output. Here, the subscripts e, d, and c represent the endpoint, service direction, and wavelength or channel, respectively, γ is the threshold sensitivity coefficient set according to the historical false alarm rate, and m is the minimum number of consecutive abnormal codewords.

[0011] In the formula, k represents the FEC codeword time index, C represents the error correction strength of the k-th FEC codeword at the corresponding endpoint, service direction, and wavelength or channel, U represents the uncorrectable flag of the corresponding codeword, μ represents the background mean of the error correction strength within the sliding window, σ represents the corresponding background scale, θ represents the dynamic threshold determined by the background mean, background scale, and threshold sensitivity coefficient, B represents the codeword index interval corresponding to the candidate FEC error burst event, the subscripts s and e of k represent the starting and ending codeword indices of the interval, respectively, and m is the minimum number of consecutive abnormal codewords; the number of error correction symbols is denoted as... The uncorrectable error mark is .

[0012] In step 3, for each candidate FEC error burst event output in step 2, the error correction intensity sequence from the start codeword index to the end codeword index is extracted, the background mean is subtracted, and the burst envelope is obtained by normalizing according to the burst peak value; the codeword phase is determined according to the offset of the peak codeword relative to the FEC frame boundary, interleaving block boundary, or sampling window boundary; the spatial dispersion is calculated according to the error symbol position distribution, and the normalized envelope, duration, peak intensity, error symbol dispersion, codeword phase, and correlation coefficient with changes in received optical power or SNR are combined into a burst fingerprint; a small positive number is set in the normalization denominator to prevent division by zero; In the formula, C represents the FEC codeword error correction intensity, μ represents the background mean of the error correction intensity within the background window where the candidate burst is located, the subscript B of A represents the maximum amplitude of the candidate burst after deducting the background mean, ε represents a small positive number used to prevent the normalization denominator from being zero, the subscript s of k represents the starting codeword index of the candidate burst, q represents the relative codeword index within the candidate burst, N represents the number of codewords contained in the candidate burst, s represents the normalized burst envelope value, S represents the envelope sequence composed of each normalized burst envelope value, F represents the burst fingerprint vector, W represents the half-width at half-maximum (WHM) of the burst envelope, A represents the burst peak intensity, κ represents the error symbol dispersion, φ represents the codeword phase, and ρ represents the correlation coefficient between the burst envelope and the received optical power or SNR variation.

[0013] The burst fingerprint includes at least two of the following: burst envelope, codeword phase, error symbol distribution information, burst start time, peak time, end time, half width at half maximum (WHM), peak intensity, envelope shape, error symbol dispersion, and correlation coefficient with changes in received optical power or SNR. The codeword phase is determined based on the FEC frame boundary, interleaving depth, deinterleaving delay, DSP fixed processing delay, endpoint clock offset, or link round-trip measurement results.

[0014] In step 4, the local time of the peak codeword is first obtained based on the endpoint timestamp and peak codeword index recorded in step 1. Then, the endpoint time offset and synchronization residual are estimated using the PTP hardware timestamp, SyncE frequency synchronization, link round-trip delay, or controller time synchronization. Based on the codeword phase compensation codeword quantization offset obtained in step 3, the interleaving / deinterleaving delay and DSP fixed processing delay obtained through loopback testing, known position disturbance, or equipment calibration are deducted. The envelope shape offset is estimated based on the cross-correlation peak or centroid of the normalized burst envelope to obtain the corrected arrival time. The synchronization residual, codeword phase residual, fixed delay calibration error, envelope offset error, and sampling window quantization error are combined into the corrected arrival time uncertainty. Where e represents the endpoint index, t with a cap represents the corrected arrival time of the FEC error burst event reaching endpoint e, p of k represents the peak codeword index, T represents the local time of the peak codeword recorded at the endpoint, φ represents the normalized codeword phase compensation amount, cw of T represents the FEC codeword period, int of τ represents the interleaving or deinterleaving fixed delay, DSP of τ represents the DSP fixed processing delay, shape of δ represents the peak time offset caused by the burst envelope shape, t of σ represents the standard uncertainty of the corrected arrival time, sync, phase, delay, and shape of σ represent the uncertainties corresponding to the time synchronization residual, codeword phase compensation residual, fixed delay calibration error, and envelope shape offset estimation error, respectively, and win of T represents the telemetry sampling window width; a in the local codeword time model represents the endpoint time offset, and ε represents the frequency drift or time jitter term; endpoint The local typing time is: in For endpoint time offset, For frequency drift or jitter, the FEC interleaving / deinterleaving and DSP fixed processing delay are: The codeword phase offset is The peak shift caused by the sudden change in envelope shape is .

[0015] The endpoint time reference is calibrated by at least one of the following methods: PTP hardware timestamp, SyncE frequency synchronization, link round-trip delay measurement, controller time synchronization, or historical homogeneous burst cross-correlation.

[0016] In step 5, within the physical propagation interval defined by link length and group velocity, candidate event pairs are constructed for the first endpoint candidate event and the second endpoint candidate event. Normalized burst envelope correlation, duration similarity, peak intensity similarity, and error symbol distribution similarity are calculated respectively, and a matching score is obtained by combining the physical propagation constraint satisfaction term and the non-fiber disturbance penalty term formed by the device status in step 1. Event pairs are selected according to the principle of maximum matching score and one-to-one event correspondence. When the matching score reaches the matching threshold and the physical propagation constraint is satisfied, the same-source burst event pair, matching score, and corrected arrival time difference are output; otherwise, matching is rejected. Each matching weight is a non-negative value and is determined by the validation set of historically confirmed events. The propagation time difference tolerance in the formula is determined by the synchronization residuals at both ends and the time estimation uncertainty. Where i and j represent the indices of the first and second endpoint candidate events, respectively; M represents the matching score of the candidate event pair; subscripts 1 to 6 of w represent the non-negative weights of each matching component; subscript S of ρ represents the normalized burst envelope correlation; subscripts W, A, and E of S represent the duration similarity, peak intensity similarity, and error symbol distribution similarity, respectively; subscript phys of I represents the physical propagation constraint satisfaction term; subscript device of P represents the device-side non-fiber disturbance penalty term; Δt represents the corrected arrival time difference between the two candidate events; capped t represents the corrected arrival time of the candidate events to the first and second endpoints, respectively; L represents the fiber link length; v represents the group velocity of light in the fiber link; and subscript t of η represents the propagation time difference tolerance. Same-source matching includes calculating the envelope correlation, duration similarity, peak intensity similarity, error symbol distribution similarity, and physical propagation constraint satisfaction of two FEC error burst events. When the corrected arrival time difference is not within the physical range defined by the link length and group velocity, the corresponding event is rejected as a homogeneous burst event pair.

[0017] In step 6, the distance between the transient fault point and the first endpoint... satisfy: in For link length, For group velocity, This is the difference between the arrival time of the first endpoint and the arrival time of the second endpoint; The confidence interval is calculated based on the uncertainties of link length error, group velocity error, and corrected time of arrival difference.

[0018] In step 7, distance estimates are obtained based on multiple wavelengths, multiple directions, multiple ports, or multiple homogeneous burst events. For the q-th distance estimate, a fusion weight is constructed using the matching score output in step 5, the reliability formed by the synchronization residual and codeword phase stability in step 4, and the distance uncertainty in step 6. The distance estimates are then normalized and weighted to obtain the final distance. Among these factors, higher matching scores, smaller time synchronization residuals, more stable FEC codeword phases, and more reliable device status result in higher normalized reliability; smaller distance uncertainties lead to greater fusion weights. The fusion variance is calculated based on the weighted distance dispersion and the inherent uncertainty of each distance estimate, with the confidence interval limited to the fiber link length. A small positive number is used in the denominator of the formula to prevent division by zero, and the confidence coefficient is set to 1.96 at a 95% confidence level. Where q represents the distance estimation index, Q represents the total number of distance estimates participating in the fusion, x represents the distance estimate obtained for the corresponding homogeneous burst event pair, M represents the matching score of the homogeneous burst event pair, R represents the normalized reliability formed by time synchronization residual, codeword phase stability and device status, σ represents the standard uncertainty of the corresponding distance estimate, ε represents a small positive number used to prevent the denominator of the fusion weight from being zero, ω represents the fusion weight of the corresponding distance estimate, the capped x represents the final distance obtained by normalized weighting, the σ squared corresponding to the capped x represents the variance of the fused distance, CI represents the distance confidence interval limited to the range of zero to the link length, L represents the fiber link length, and z represents the confidence coefficient corresponding to the target confidence level, which is 1.96 when the confidence level is 95%. Before performing source matching, step 5 also includes forming a non-fiber disturbance criterion based on the device temperature, power supply voltage, DSP lock status, module restart log, protection switching record and synchronization anomalies of other ports of the same device obtained synchronously in step 1; events that meet the common mode anomaly conditions on the device side are eliminated and do not enter source matching or fiber distance calculation. When multiple non-co-fiber routing ports synchronously experience a burst of bit errors on the same device and there is no homogeneous burst event at the remote end that satisfies the physical propagation interval constraint, the event is marked as a device-side anomaly and the fiber distance is refused to be output.

[0019] A burst fiber optic fault location device for FEC (Fiber Optic Error Detection) includes a telemetry acquisition module, a time calibration module, a burst detection module, a burst fingerprint module, a homogeneous matching module, a location calculation module, and a confidence assessment module. The telemetry acquisition module is used to acquire the FEC decoding telemetry; The time calibration module is used to calibrate the endpoint time base, FEC frame boundary, interleaving / deinterleaving delay, and transceiver fixed processing delay. The burst detection module and burst fingerprint module are used to identify FEC error burst events and generate burst fingerprints; The same-origin matching module is used to obtain pairs of same-origin burst events based on burst fingerprints and physical propagation interval constraints. The location calculation module and confidence assessment module are used to calculate the distance to the transient fault point and output the confidence interval or confidence level.

[0020] An electronic device includes a processor and a memory, the memory storing a computer program that, when executed by the processor, implements the FEC error burst fiber optic fault location method. Alternatively, a computer-readable storage medium storing a computer program that, when executed by a processor, implements the FEC error burst fiber optic fault location method.

[0021] The beneficial effects of this invention are: By utilizing FEC error bursts in normal service flow for location, no dedicated test signal needs to be injected, making it suitable for online operation. Specifically, step 1 directly reads the FEC decoding telemetry during normal service reception, and steps 2 to 7 complete burst detection, end-to-end matching, and distance calculation without interrupting the service or sending additional test sequences, thus enabling continuous online operation.

[0022] The accuracy of arrival time estimation is improved by correcting FEC codeword phase, FEC frame boundary, interleaving / deinterleaving delay, and transceiver fixed processing delay. Specifically, step 3 gives the codeword phase of the burst relative to the FEC frame boundary, and step 4 converts the discrete codeword index to a unified time base and compensates for interleaving / deinterleaving delay, DSP fixed processing delay, and envelope shape offset, thereby reducing the arrival time deviation caused by codeword quantization and differences in device processing chains.

[0023] By using burst envelope, error symbol distribution, and physical propagation interval constraints for homogeneous matching, mislocalization caused by random noise and device-side common-mode anomalies is reduced. Specifically, step 2 first forms candidate bursts with clear boundaries, step 3 generates burst fingerprints containing envelope and error symbol distribution, and step 5 performs one-to-one matching only within the physical propagation interval, and excludes common-mode anomalies by using device state penalty terms, making it difficult for random noise or device-side spikes to simultaneously satisfy fingerprint similarity and propagation time difference conditions.

[0024] By outputting distance confidence intervals and confidence levels, the automatic dispatching, on-site inspection, and protection strategy decisions can be based on quantifiable data. By supporting codeword-by-codeword data and short-window aggregated data, the adaptability to different optical modules, coherent transceivers, switching chips, and OTN devices is improved. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the overall process of the FEC error burst fiber optic fault location method provided in an embodiment of the present invention.

[0026] Figure 2 This is a schematic diagram of the codeword-level envelope of an FEC error burst event in an embodiment of the present invention.

[0027] Figure 3 This is a schematic diagram of the time difference distance calculation model for arrival at both ends in an embodiment of the present invention.

[0028] Figure 4 This is a schematic diagram of the codeword phase and fixed delay correction process in an embodiment of the present invention.

[0029] Figure 5 This is a schematic diagram of the non-fiber disturbance rejection process in an embodiment of the present invention. Detailed Implementation

[0030] The present invention will now be described in further detail with reference to the accompanying drawings.

[0031] It should be noted that the first endpoint and the second endpoint in this application can be optical modules, coherent transceivers, OTN line boards, Ethernet switching ports, router ports, or other optical communication devices with FEC decoding capabilities at both ends of an optical fiber link. The bidirectional optical fiber link can be a single-fiber bidirectional link or a link composed of pairs of optical fibers with opposite service directions. The first receiving direction and the second receiving direction correspond to opposite service directions; that is, the first endpoint receives the service stream propagated from the direction of the second endpoint, and the second endpoint receives the service stream propagated from the direction of the first endpoint.

[0032] Example 1: Overall Flowchart. See also Figure 1 The FEC error burst fiber optic fault location method provided in this embodiment includes telemetry acquisition, time calibration, burst detection, burst fingerprint generation, source matching, location calculation, confidence assessment, and result output. The telemetry acquisition step obtains data related to the FEC codeword time index from the FEC decoder, DSP, switching chip, module management interface, or network controller at both ends of the link. The time calibration step converts the local time axes of different endpoints, directions, wavelengths, or ports to a unified time reference. The burst detection step identifies short-term anomalies in the error correction intensity sequence. The burst fingerprint generation step expresses the anomalous event as a comparable event vector. The source matching step determines whether the events at both ends originate from the same fiber transient disturbance. The location calculation step calculates the fault distance based on the time difference of arrival. The confidence assessment step outputs the reliability of the location result.

[0033] In this embodiment, FEC decoding telemetry includes, but is not limited to, the number of error-correcting symbols, the number of error-correcting bits, the number of error-correcting codewords, the number of uncorrectable codewords, the uncorrectable codeword flag, the error symbol histogram, the pre-FEC BER estimate, the post-FEC BER, the FEC frame count, the FEC codeword sequence number, the PCS / PMA timestamp, the received optical power, the SNR, the Q factor, the DSP lockout status, the module temperature, the power supply voltage, the module restart flag, the protection switching record, and the port timestamp. The above data can originate from the optical module's DSP internal registers, the coherent DSP diagnostic interface, the switching chip or MAC / PCS FEC statistics, the CMIS extension page, the vendor's proprietary I2C / MDIO interface, the OTN device management interface, or the network controller sampling interface.

[0034] In some embodiments, the device can output codeword-by-codeword FEC statistics, in which case the FEC codeword sequence number or FEC frame count can be directly used as the codeword time index. In other embodiments, the device can only output short-window aggregation statistics, such as outputting the total number of error-corrected codewords, the total number of error-corrected symbols, the number of uncorrectable codewords, or the pre-FECBER estimate every 20 microseconds, 100 microseconds, or 1 millisecond. For short-window aggregation statistics, an equivalent FEC codeword time index can be generated based on the window start codeword, the window end codeword, or the window center time, and the window width can be included as part of the arrival time uncertainty.

[0035] Example 2: FEC error burst detection. See also Figure 2 At the endpoint ,direction wavelength or channel FEC codeword sequence on In this context, the number of error correction symbols is denoted as . The uncorrectable error mark is The coding cycle is The system estimates the background mean within a sliding window. and scale When continuous More than one code word to meet: If the total number of error-correcting symbols within the window exceeds the historical quantile threshold, or if an uncorrectable codeword appears accompanied by an increase in error correction intensity for several preceding and following codewords, or if error symbols are spatially clustered within the FEC frame and their dispersion is lower than the random noise baseline, the system will identify the corresponding interval as a candidate FEC error burst event. This embodiment specifically illustrates the FEC error burst detection process in step 2 of the method described in Embodiment 1.

[0036] In this embodiment, FEC error bursts differ from long-term link degradation. Long-term link degradation typically manifests as a continuous increase in pre-FEC BER or a slow decrease in optical power, while transient faults typically manifest as short-duration, distinct peaks in error correction intensity with clear boundaries. The system can set a transient upper limit, for example, marking events with a duration exceeding a preset time threshold as slow degradation, and marking events with a duration within a preset range and a peak intensity significantly higher than the background noise as transient candidate events. This threshold can be adaptively adjusted based on link rate, FEC type, sampling granularity, and historical noise level.

[0037] Example 3: Burst Fingerprint Generation. For each candidate FEC error burst event, the system extracts the burst start time. Peak time End time Half height and width Peak intensity Envelope shape Error symbol dispersion Code word phase and the correlation coefficient with changes in received optical power or SNR. This forms a sudden fingerprint: Among them, envelope shape It can be represented by a normalized error correction intensity sequence, or by several statistical characteristics; error symbol dispersion Used to describe the spatial clustering of error symbols in an FEC frame or codeword; codeword phase Used to describe the offset of burst peaks or start points relative to FEC frame boundaries, interleaving block boundaries, or sampling window boundaries.

[0038] By using burst fingerprinting, the system no longer relies solely on average BER or a single peak for judgment. Instead, it utilizes the shape, intensity, duration, spatial distribution, and phase information of FEC error bursts for subsequent matching. This enables the differentiation between real fiber disturbances and random noise, and provides a more stable characteristic basis for end-to-end matching. For example, a real fiber microbend may form bursts with similar durations, envelope shapes, and error symbol distributions at both ends, while power supply disturbances on the device side may simultaneously produce spikes at multiple non-co-fiber ports, but no matching event occurs at the far end.

[0039] Example 4: Time of Arrival Estimation and Codeword Phase Correction. Discrete Peak Codewords for Candidate FEC Error Bursts. This does not equal the actual time when the disturbance arrives at the receiver. Let the endpoint be... The local typing time is: in For endpoint time offset, This represents the frequency drift or jitter term. Assume the FEC interleaving / deinterleaving and DSP fixed processing delays are... The codeword phase offset is The peak shift caused by the sudden change in envelope shape is Then the estimated arrival time can be expressed as: See Figure 4 The correction process includes, in sequence, unifying the endpoint time base, converting the local time from the codeword index, compensating for the codeword phase, deducting the interleaving / deinterleaving and DSP fixed processing delays, correcting the envelope shape offset, and synthesizing the time uncertainty.

[0040] In this embodiment, Frequency drift can be estimated by PTP hardware timestamps, SyncE frequency synchronization, link round-trip delay measurements, controller time synchronization, or historical homogeneous burst cross-correlation. It can be obtained by aligning the FEC frame boundary with the PCS / PMA timestamp, using known perturbations during training, estimating the peak value of the two-end cross-correlation, or estimating the sampling window boundary. Calibration can be performed at the device factory, during link activation, loopback testing, or during training. Through the above corrections, the system converts discrete codeword indices into continuous time-of-arrival estimates and outputs the time-of-arrival uncertainty. .

[0041] In an engineering calculation example, if the FEC statistical window width is 20 microseconds, without correction, burst peaks can only fall in the center of the window, resulting in a single-end arrival time quantization error of ±10 microseconds, which translates to a distance of approximately ±1.02 kilometers. After applying FEC frame boundary, codeword phase, interleaving / deinterleaving delay, and DSP fixed delay corrections, if the residual time error is reduced to ±1.2 microseconds, the equivalent distance error is approximately ±122 meters. This calculation example is used to illustrate the technical effectiveness of the correction mechanism. If real test data is available during the formal application, it can be replaced with real logs and test results.

[0042] Example 5: Homologous burst matching. For candidate events at the first endpoint. Candidate events of the second endpoint The system calculates the matching score. The matching score can include burst envelope correlation, duration similarity, peak intensity similarity, error symbol distribution similarity, correlation with changes in optical power or SNR, physical propagation constraint satisfaction, and internal device anomaly penalty. For example, it can be represented as: in Indicates burst envelope correlation. Indicates normalized similarity. This indicates terms related to received optical power or SNR synchronization disturbances. This indicates whether the time difference of arrival falls within the physical propagation range defined by the link length and group velocity. This indicates the penalty items for internal equipment anomalies. If... If the value exceeds a preset threshold and meets the exclusive matching constraint, the system will trigger the event. and events The pair is identified as a single-origin burst event. If the corrected arrival time difference exceeds... If the scope is similar, then even if the envelopes are similar, corresponding events will not be considered as a pair of common-origin sudden events.

[0043] Example 6: Distance Calculation. See also Figure 3 Let the total length of the fiber optic link be... Group speed The distance from the fault point to the first endpoint is The distance from the second endpoint is In a bidirectional service link, the first endpoint receives services from the second endpoint to the first endpoint, and the second endpoint receives services from the first endpoint to the second endpoint. When a transient bend, vibration, contamination, or connector disturbance occurs at the same physical location, the service streams arriving at each receiver from that location in the opposite direction are affected, resulting in a common-source FEC (Fault-Error-Enhanced) burst at both receivers.

[0044] If the corrected arrival times of the same transient disturbance at the first and second endpoints are respectively and arrival time difference satisfy Therefore, the distance from the fault point to the first endpoint For multi-wavelength, multi-directional, multi-port, or multiple burst events, the system can construct a weighted least squares or Bayesian estimation model to fuse multiple distance estimates. The weights are related to the event matching score, time synchronization residual, FEC codeword phase stability, and equipment status reliability.

[0045] In a simulation example, the link length 40 km, group refractive index The group velocity is 1.468. Approximately meters per second. If the actual fault point is 11.8 kilometers from the first endpoint, then the theoretical time difference of arrival is approximately... 80.4 microseconds. If the corrected time of arrival uncertainty is 1.2 microseconds, the link length data error is 20 meters, and the group velocity error converted to a position error of less than 5 meters, then the standard position uncertainty is approximately 123 meters. The system can output a positioning result of 11.8 kilometers from the first endpoint with a 95% confidence interval of approximately ±250 meters.

[0046] Example 7: Non-fiber disturbance rejection. See also Figure 5 When transient power supply disturbances on a device board cause multiple non-co-fiber routing ports on the same device to simultaneously experience pre-FEC BER spikes, a standard BER thresholding scheme might mark multiple links as fiber anomalies. This embodiment checks the port common-mode behavior, DSP lockout status, module temperature, power supply voltage, module restart logs, protection switching records, whether the received optical power decreases synchronously, whether the two-end bursts meet physical propagation interval constraints, and whether the error symbol distribution is similar to historical fiber disturbance templates. If the event only occurs synchronously on multiple ports of this device, and there is no matching burst at the remote end, the system marks it as a device-side anomaly and refuses to output fiber distance.

[0047] When a connector micro-motion or localized micro-bending causes a real fiber optic transient fault, it typically manifests as a burst of light occurring at both ends of the relevant link in opposite receiving directions, satisfying the propagation time difference range, and the received optical power or SNR may exhibit synchronous short-term changes. The system can retain this event and output the distance. Therefore, this embodiment not only detects bit error spikes but also reduces false alarms through end-to-end source matching, physical propagation interval constraints, and non-fiber disturbance penalty terms.

[0048] Example 8: Multi-wavelength and multi-node fusion. In a coherent DWDM link, the same physical fiber can carry multiple wavelengths. Each wavelength's coherent receiver can output pre-FEC BER, soft-decision FEC error correction strength, Q factor, SNR, dispersion estimation, polarization-related parameters, and DSP lock-in state. When external force vibration causes local polarization disturbances or micro-bending, multiple wavelengths may experience FEC error bursts at similar correction arrival times. The system can fuse the distance estimates of each wavelength according to confidence weights and mark events occurring only at a single wavelength and accompanied by DSP reconvergence as possible transceiver internal anomalies.

[0049] In metropolitan area rings or backbone links, multiple relay sites can collect FEC telemetry data. The system can map candidate events for each link segment to a topology map and use the time difference of arrival (TDOA) of multiple segments satisfying physical propagation constraints to jointly solve for the fault segment and the distance within the segment. When the same event simultaneously affects two adjacent segments but not other segments, the system prioritizes locating the fault near the common physical route. When multiple ports on the same device simultaneously experience in-phase bursts without optical power changes, the system determines it as a common-mode anomaly on the device side and suppresses fiber optic positioning output.

[0050] Example 9: Apparatus and Electronic Equipment. The FEC (Fault-Error-Induced) burst fiber optic fault location device includes a telemetry acquisition module, a time calibration module, a burst detection module, a burst fingerprint module, a source matching module, a location calculation module, a confidence assessment module, and an output interface module. The telemetry acquisition module reads FEC statistics, BER, Q factor, SNR, received optical power, and device status from optical modules, coherent transceivers, switching chips, OTN equipment, DSP internal diagnostic interfaces, or network controllers. The time calibration module performs a unified reference conversion between timestamps and codeword clocks at each end. The burst detection module identifies candidate FEC bursts. The burst fingerprint module generates event vectors. The source matching module filters for pairs of events from the same source. The location calculation module calculates the fault distance. The confidence assessment module outputs the confidence interval and confidence level.

[0051] The aforementioned devices can be deployed in network controllers, NMS systems, SDN controllers, servers, line cards, optical module management units, or embedded processors. Electronic devices may include processors, memory, communication interfaces, and buses. The memory stores a computer program, which the processor executes to implement the method of the present invention. Computer-readable storage media may be read-only memory, random access memory, flash memory, magnetic disks, optical disks, or other media capable of storing program code.

[0052] To further illustrate the non-obviousness of this invention, it should be noted that this invention does not simply use the FEC bit error rate as a common physical layer parameter. FEC bit error bursts have proprietary structures such as discrete codeword sampling, FEC frame boundaries, interleaving / deinterleaving delays, fixed DSP processing delays, and spatial distribution of error symbols. This invention establishes arrival time correction and burst fingerprint matching mechanisms for these structures, and further combines physical propagation interval constraints, non-fiber disturbance elimination, and confidence interval output. Therefore, this invention addresses the problem of how to transform service FEC telemetry in high-speed optical communication links into transient fault distance location, rather than simply superimposing ordinary alarm thresholds or ordinary two-way correlations.

[0053] It should be noted that, in this invention, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0054] Embodiments of the present invention have been shown and described above. It will be understood by those skilled in the art that various changes, modifications, substitutions, and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0055] In some embodiments, the telemetry acquisition module establishes a field mapping table for different types of devices. For optical modules that support internal diagnostic register reading, the field mapping table may include an FEC frame counter, codeword sequence number, total number of error-correcting symbols, uncorrectable codeword flag, and error symbol histogram; for coherent DSPs, the field mapping table may include soft-decision FEC error correction strength, Q factor, SNR, dispersion estimation, polarization correlation estimation, and DSP lock-in state; for switching chips or MAC / PCS modules, the field mapping table may include RS-FEC corrected codewords, uncorrected codewords, lane-level error correction count, and port hardware timestamp. The field mapping table does not require all devices to have all fields, but rather allows the system to select the minimum feasible combination of fields based on device capabilities.

[0056] The minimum feasible field combination includes codeword time index, error correction strength information, and endpoint timestamp. Enhanced field combinations further include error symbol distribution, FEC frame boundary, interleaving / deinterleaving delay, received optical power, SNR, Q factor, and DSP status. With only the minimum field combination, the system can output a wider confidence interval; with enhanced field combinations, the system can perform more accurate codeword phase correction and non-fiber disturbance rejection. This ensures feasibility across different devices while avoiding limiting the protection scope to a specific register name from a particular vendor.

[0057] The system performs baseline learning during the link activation phase. Baseline learning includes recording the error correction intensity distribution under normal service load, the pre-FEC BER fluctuation range under different sampling windows, endpoint timestamp deviation, the correspondence between FEC frame boundaries and local hardware timestamps, transceiver fixed processing latency, and the normal variation range of device temperature and voltage. The baseline learning results are used for subsequent adaptive threshold setting and anomaly penalty term calculation.

[0058] During the link commissioning phase, the system can calibrate the fixed delay through loopback testing, known-length fiber optic disc testing, or manually controlled micro-bending testing. During calibration, a short-duration repeatable perturbation is applied at a known location, and FEC error bursts are collected at both ends. The deviation between the theoretical and observed time difference of arrival is calculated, and this deviation is decomposed into endpoint time offset, FEC interleaving / deinterleaving delay, DSP fixed processing delay, and codeword phase offset. Calibration parameters can be stored separately by port, wavelength, FEC type, and rate.

[0059] FEC (Fault-Oriented Encoder) burst detection employs a multi-level thresholding system. The first-level threshold is used to quickly filter suspected events from the continuous telemetry stream. The second-level threshold is used to confirm candidate events based on burst width, peak intensity, and error symbol clustering. The third-level threshold is used to determine whether an event should enter the same-source matching process by combining received optical power, SNR, and equipment status. This multi-level thresholding avoids performing complex matching calculations for every minute fluctuation, thereby reducing the controller load.

[0060] The spatial clustering of error symbols can be represented by the variance of error symbol locations in the FEC frame, the length of the maximum consecutive error symbol, the number of error symbol clusters, or the proportion of symbols within a cluster. Random noise typically causes error symbols to be relatively dispersed in the FEC frame, while transient fiber disturbances may form consecutive or near-consecutive error symbol clusters within a short period of time. By incorporating the spatial clustering of error symbols into the burst fingerprint, the ability to identify transient disturbances can be improved.

[0061] Burst envelope correlation can be calculated using normalized cross-correlation, dynamic time warping, or eigenvector distance. Considering that the FEC processing links at both ends of the receiving direction may have different filtering and sampling windows, the system allows for a certain degree of envelope widening or narrowing. During matching, the maximum correlation peak can be searched within the physical propagation interval, and the position of the correlation peak can be used as a candidate value for correcting the time difference of arrival.

[0062] Physical propagation range constraints not only include This also includes the symbolic relationship determined by the link topology and service direction. For example, when the fault point is close to the first endpoint, the arrival time difference between the first and second endpoints should show a corresponding directional shift; when the fault point is near the midpoint of the link, the arrival time difference should be close to zero. If the calculated distance is less than zero or greater than the link length, the system rejects the event pair or marks it as a low-confidence match.

[0063] The confidence interval is composed of multiple error components. These error components include link length data error, group refractive index error, endpoint time synchronization residual, codeword phase residual, sampling window quantization error, burst envelope peak estimation error, and fixed delay calibration error. The system can calculate the confidence interval using the error propagation formula, Bootstrap resampling, or particle filtering methods. If the localization results of multiple events are consistent, the confidence interval can be narrowed; if the results of multiple events are discrete, the system's confidence level can be reduced.

[0064] In some embodiments, the fault type labels output by the system include microbending, connector loosening, external force vibration, transient contamination, transient reflection, slow attenuation, and device-side anomaly. Label generation can be based on rules or machine learning models. For example, an event accompanied by a short-term drop in received optical power and occurring synchronously across multiple wavelengths can be labeled as microbending or external force vibration; an event occurring only at a single wavelength and accompanied by DSP reconvergence can be labeled as an internal transceiver anomaly; and an event with a long-term increase in pre-FEC BER without a clear burst boundary can be labeled as slow attenuation.

[0065] The system can be integrated with the work order system. Location results include reference endpoint, distance, confidence interval, matching score, candidate fault type, relevant port, relevant wavelength, event occurrence time, and key telemetry evidence. The work order system can map the location of fiber optic cable segments, manholes, ODF racks, junction boxes, or equipment rooms in the resource system based on the distance and send on-site inspection suggestions to maintenance personnel. If the confidence level is low, the system can generate only observation alarms without triggering repair dispatch.

[0066] The system can be integrated with protection strategies. When the location results indicate that a physical route has recurring transient disturbances, the controller can reduce the scheduling priority of that route or increase the weight of the backup path in the service protection strategy. When the non-fiber disturbance elimination module determines that the event is an equipment-side anomaly, the system can direct the alarm to the equipment maintenance process instead of triggering fiber optic cable inspection.

[0067] Verification testing can be conducted using a fiber optic cable of known length and a controllable perturbation device. The tester applies a slight bend or vibration at a known location on the fiber optic cable, recording FEC decoding telemetry at both ends, endpoint timestamps, received optical power, and SNR. The system calculates the position in three modes: without codeword phase correction, using only BER peak value, and using the complete correction model of this invention, and the positioning errors are compared. This test demonstrates the contribution of codeword phase and fixed delay correction to positioning accuracy.

[0068] False alarm elimination testing can be performed by simulating device-side anomalies. For example, record FEC statistics during DSP unlocking, module restart, power supply voltage fluctuations, or protection switching, and observe whether such events form a co-origin burst at the remote end that satisfies the physical propagation range constraint. If multiple ports on the local end experience anomalies simultaneously, but there are no matching events at the remote end, the system should output a device-side anomaly label. This test demonstrates that co-origin matching and device status penalty terms can reduce false alarms.

[0069] Multi-wavelength fusion testing can be performed on DWDM links. If the same physical disturbance produces similar location results on multiple wavelengths, the system can increase the confidence level and output the fused distance; if only one wavelength shows an anomaly, and there are no corresponding bursts on other wavelengths and ports along the same route, the system lowers the confidence level or marks it as a single-channel anomaly. This mechanism avoids misjudging a problem in a single transceiver channel as a fault along the fiber optic cable.

[0070] The applicability test for short-window aggregation allows for setting different sampling windows. Using a 20-microsecond window, the system obtains a narrower confidence interval; using a 100-microsecond window, the confidence interval expands accordingly; using a 1-millisecond window, the system can output coarse-segment positioning or low-confidence positioning. The system labels the sampling window granularity and the resulting uncertainty in the output results, thus avoiding commitments to positioning accuracy that do not conform to the data granularity.

[0071] The system can cluster historical events. If multiple transient events from different times are located near the same distance and have similar fault type labels, the system can mark that distance as a repeating anomaly. Repeating anomalies may correspond to loose connectors, pipes affected by external forces, temperature-sensitive joints, or cabinets affected by mechanical vibration. This historical clustering result can be used as a basis for prioritizing maintenance.

[0072] The system can perform lightweight detection on edge devices and complex matching on the controller. Edge devices only identify candidate FEC (Fault-Oriented Error) bursts and report compressed burst fingerprints, while the controller collects events from multiple endpoints, wavelengths, and links, and then performs source matching and distance calculation. This deployment method can reduce telemetry bandwidth pressure while retaining cross-link correlation capabilities.

[0073] To protect business privacy and reduce data volume, the system does not need to upload business payload content, but only uploads FEC statistics, physical layer parameters, timestamps, and device status. Therefore, this invention is applicable to data center interconnection and carrier transmission network scenarios with high requirements for business data security.

[0074] The system can be configured with different codeword periods and phase interpretation methods for different FEC types. For example, the codeword length, interleaving depth, and error correction symbol statistics may differ for Ethernet RS-FEC, KP4 FEC, OTN hard-decision FEC, or coherent system soft-decision FEC. The system can record the FEC type, codeword length, symbol width, interleaving block size, FEC frame boundary acquisition method, and the meaning of the error correction strength field in the configuration file, and call the corresponding parameters during burst detection and time-of-arrival correction. Therefore, this invention is not limited to a single type of FEC encoding, but is applicable to various optical communication systems capable of outputting FEC error correction telemetry that can be mapped to a time index.

[0075] If the sampling periods at both ends of the link are different, the system first converts the telemetry data at each end to a unified time grid. During the conversion, the original sampling points and the interpolated sampling point markers can be retained to avoid mistaking the interpolated results for actual observations. For codeword-by-codeword data, the unified time grid can be the codeword period or an integer multiple of the codeword period; for short-window aggregated data, the unified time grid can be the window center time. The system uses the sampling period difference as an uncertainty input when calculating the matching score.

[0076] The system can assign event numbers to emergencies and save the event lifecycle. The event lifecycle includes candidate generation, fingerprint completion, endpoint time correction, same-source matching, distance calculation, confidence assessment, output confirmation, and historical archiving. Each stage can save key intermediate quantities, such as candidate window, background noise level, threshold, matching candidate list, rejection reason, and final confidence level. By saving the event lifecycle, subsequent auditing and maintenance personnel can review why the system outputs a certain distance or why it rejects location tracking.

[0077] Non-fiber disturbance penalty term It can consist of multiple sub-items, including DSP unlock sub-items, module restart sub-items, voltage anomaly sub-items, temperature surge sub-items, multi-port common-mode sub-items on the same board, and protection switching sub-items. Each sub-item can use a normalized value between 0 and 1. When When the preset threshold is exceeded, even if there is a sudden weak correlation between the two ends, the system can lower the confidence level or require more evidence for confirmation. This design avoids misjudgment caused by a single similarity index.

[0078] The system can also incorporate manual confirmation feedback. After completing on-site inspections, maintenance personnel can backfill the system with the actual fault location, fault type, and handling results. The system updates the background noise model, burst templates, equipment anomaly rules, and risk levels for different link segments based on the backfill results. This feedback process does not change the basic localization process of this invention, but it can improve the adaptability of thresholds and weights during long-term operation.

[0079] The controller can adjust the processing strategy according to the service level. For links carrying critical services, the system can use a lower candidate threshold and a higher sampling frequency to capture transient disturbances as early as possible; for ordinary service links, the system can use a higher candidate threshold and a lower sampling frequency to reduce telemetry bandwidth and computational overhead. Regardless of the strategy used, the system records the sampling strategy and the resulting uncertainty in the output.

[0080] The system can combine the location results with a geographic information system or a fiber optic cable resource system. If the calculated distance falls near a junction box, manhole, cabinet, or ODF rack, the system can output the corresponding resource name. If the confidence interval covers multiple resource points, the system can generate inspection priorities based on distance center, historical fault frequency, junction type, and environmental risk. This function enables FEC error burst location results to be directly converted into executable maintenance actions.

[0081] The system can perform lateral correlation on multiple service links along the same route. If multiple service links share the same physical optical cable segment and experience FEC (Fiber Optic Error) bursts that can be located in the same segment around the same time, the system increases the fault confidence level of that physical segment. If only one service port experiences an anomaly while other services along the same route are normal, the system decreases the fiber segment fault confidence level and investigates equipment-side factors. This lateral correlation further enhances the ability to distinguish between actual fiber optic faults and equipment-side faults.

Claims

1. A method for locating transient fiber optic faults based on FEC (Fault-Error Burst) errors, characterized in that, A bidirectional optical fiber link for carrying normal service flow is provided, wherein the bidirectional optical fiber link has a first endpoint and a second endpoint, and the first endpoint and the second endpoint are respectively optical communication devices disposed at both ends of the bidirectional optical fiber link and capable of outputting FEC decoding telemetry. The bidirectional optical fiber link includes a first service direction from the first endpoint to the second endpoint and a second service direction from the second endpoint to the first endpoint; Specifically, the following steps are included: Step 1: Acquire FEC decoding telemetry in the first receiving direction of the first endpoint and the second receiving direction of the second endpoint, respectively. The first receiving direction and the second receiving direction correspond to opposite service directions in the bidirectional optical fiber link. Step 2: Based on the spatial clustering of continuous high error correction strength codewords, uncorrectable codewords, or error symbols in the FEC decoding telemetry, identify FEC error burst events at each endpoint; Step 3: Generate a burst fingerprint for each of the aforementioned FEC error burst events; Step 4: Correct the codeword time index of the FEC error burst event to obtain the correction arrival time of the FEC error burst event to the corresponding endpoint, the endpoint time synchronization residual, and the uncertainty of the correction arrival time. Step 5: Perform source matching on the FEC error burst events of the first and second endpoints to obtain the source burst event pairs, the corresponding matching scores, and the difference between the correction arrival time of the first endpoint and the correction arrival time of the second endpoint. Step 6: Using the first endpoint as the reference endpoint, when the corrected arrival time difference of the same-origin burst event pair satisfies the physical propagation interval constraint, calculate the distance of the transient fault point relative to the first endpoint; Step 7: Based on the matching score output in Step 5, the endpoint time synchronization residual output in Step 4, and the corrected arrival time uncertainty, obtain the confidence interval or confidence level of the distance.

2. The fiber optic transient fault location method based on FEC bit error bursts according to claim 1, characterized in that, In step 1, the FEC decoding telemetry includes error correction strength information corresponding to the FEC codeword time index; the error correction strength information includes at least one of the following: number of error correction symbols, number of error correction bits, number of error correction codewords, uncorrectable codeword flag, error symbol histogram, and pre-FEC BER estimate. The FEC decoding telemetry is either codeword-by-codeword data or short-window aggregated data. When the FEC decoding telemetry is short-window aggregated data, an equivalent FEC codeword time index is generated based on the window start codeword and window end codeword, and the window width is included in the uncertainty of the correction arrival time.

3. The fiber optic transient fault location method based on FEC bit error bursts according to claim 2, characterized in that, In step 2, for each endpoint, service direction, and wavelength or channel FEC codeword sequence obtained in step 1, the background mean and background scale of the error correction intensity are calculated using a sliding window, and a dynamic threshold is formed. When a number of consecutive codewords reach a dynamic threshold, an uncorrectable codeword appears, or the clustering degree of the error symbol space reaches a clustering threshold, adjacent abnormal codewords are merged into a candidate FEC error burst event, and the starting codeword index, peak codeword index, and ending codeword index are output. Here, subscripts e, d, and c represent the endpoint, service direction, and wavelength or channel, respectively; γ is the threshold sensitivity coefficient set based on the historical false alarm rate; and m is the minimum number of consecutive abnormal codewords. In the formula, k represents the FEC codeword time index; C represents the error correction strength of the k-th FEC codeword on the corresponding endpoint, service direction, and wavelength or channel; U represents the uncorrectable flag of the corresponding codeword; μ represents the background mean of the error correction strength within the sliding window; σ represents the corresponding background scale; θ represents the dynamic threshold determined by the background mean, background scale, and threshold sensitivity coefficient; B represents the codeword index interval corresponding to the candidate FEC error burst event; the subscripts s and e of k represent the starting and ending codeword indices of this interval, respectively; and m is the minimum number of consecutive abnormal codewords. Let the number of error correction symbols be . The uncorrectable error mark is .

4. The fiber optic transient fault location method based on FEC bit error bursts according to claim 3, characterized in that, In step 3, for each candidate FEC error burst event, the error correction intensity sequence from the start codeword index to the end codeword index is extracted, the background mean is subtracted, and the burst envelope is obtained by normalizing according to the burst peak value; the codeword phase is determined according to the offset of the peak codeword relative to the FEC frame boundary, interleaving block boundary, or sampling window boundary; the spatial dispersion is calculated according to the error symbol position distribution, and the normalized envelope, duration, peak intensity, error symbol dispersion, codeword phase, and correlation coefficient with changes in received optical power or SNR are combined into a burst fingerprint; a small positive number is set in the normalization denominator to prevent division by zero; In the formula, C represents the FEC codeword error correction intensity, μ represents the background mean of the error correction intensity within the background window where the candidate burst is located, the subscript B of A represents the maximum amplitude of the candidate burst after deducting the background mean, ε represents a small positive number used to prevent the normalization denominator from being zero, the subscript s of k represents the starting codeword index of the candidate burst, q represents the relative codeword index within the candidate burst, N represents the number of codewords contained in the candidate burst, s represents the normalized burst envelope value, S represents the envelope sequence composed of each normalized burst envelope value, F represents the burst fingerprint vector, W represents the half-width at half-maximum (WHM) of the burst envelope, A represents the burst peak intensity, κ represents the error symbol dispersion, φ represents the codeword phase, and ρ represents the correlation coefficient between the burst envelope and the received optical power or SNR variation.

5. The fiber optic transient fault location method based on FEC bit error bursts according to claim 4, characterized in that, The burst fingerprint includes at least two of the following: burst envelope, codeword phase, error symbol distribution information, burst start time, peak time, end time, half width at half maximum (WHM), peak intensity, envelope shape, error symbol dispersion, and correlation coefficient with changes in received optical power or SNR. The codeword phase is determined based on the FEC frame boundary, interleaving depth, deinterleaving delay, DSP fixed processing delay, endpoint clock offset, or link round-trip measurement results.

6. The fiber optic transient fault location method based on FEC bit error bursts according to claim 5, characterized in that, In step 4, based on the endpoint time base, FEC codeword period, FEC frame boundary, interleaving / deinterleaving delay, transceiver fixed processing delay, and the codeword phase, the local time of the peak codeword is first obtained according to the endpoint timestamp and peak codeword index recorded in step 1. Then, the endpoint time offset and synchronization residual are estimated using the PTP hardware timestamp, SyncE frequency synchronization, link round-trip time delay, or controller time synchronization. The codeword quantization offset is compensated according to the codeword phase obtained in step 3. The interleaving / deinterleaving delay and DSP fixed processing delay obtained through loopback testing, known position disturbance, or equipment calibration are deducted. The envelope shape offset is estimated according to the cross-correlation peak or centroid of the normalized burst envelope to obtain the corrected arrival time. The synchronization residual, codeword phase residual, fixed delay calibration error, envelope offset error, and sampling window quantization error are combined into the corrected arrival time uncertainty. Where e represents the endpoint index, t with a cap represents the corrected arrival time of the FEC error burst event to endpoint e, p of k represents the peak codeword index, T represents the local time of the peak codeword recorded at the endpoint, φ represents the normalized codeword phase compensation amount, cw of T represents the FEC codeword period, int of τ represents the interleaving or deinterleaving fixed delay, DSP of τ represents the DSP fixed processing delay, shape of δ represents the peak time offset caused by the burst envelope shape, t of σ represents the standard uncertainty of the corrected arrival time, sync, phase, delay and shape of σ represent the uncertainties corresponding to the time synchronization residual, codeword phase compensation residual, fixed delay calibration error and envelope shape offset estimation error, respectively, and win of T represents the telemetry sampling window width; a in the local codeword time model represents the endpoint time offset, and ε represents the frequency drift or time jitter term; endpoints The local typing time is: in For endpoint time offset, For frequency drift or jitter, the FEC interleaving / deinterleaving and DSP fixed processing delay are: The codeword phase offset is The peak shift caused by the sudden change in envelope shape is ; The endpoint time reference is calibrated by at least one of the following methods: PTP hardware timestamp, SyncE frequency synchronization, link round-trip delay measurement, controller time synchronization, or historical homogeneous burst cross-correlation.

7. The fiber optic transient fault location method based on FEC bit error bursts according to claim 6, characterized in that, In step 5, based on the envelope similarity of the burst fingerprint and the physical propagation interval constraint limited by the link length and group velocity, the FEC error burst events of the first endpoint and the second endpoint are matched for origin. Within the physical propagation range defined by the link length and group velocity, candidate event pairs are constructed for the first endpoint candidate event and the second endpoint candidate event; Normalized burst envelope correlation, duration similarity, peak intensity similarity, and error symbol distribution similarity are calculated separately, and a matching score is obtained by combining the physical propagation constraint satisfaction term and the non-fiber disturbance penalty term formed by the equipment status in step 1. Event pairs are selected according to the principle of maximum matching score and one-to-one event correspondence. When the matching score reaches the matching threshold and the physical propagation constraint is satisfied, the same-source burst event pair, matching score, and corrected arrival time difference are output; otherwise, matching is rejected. Each matching weight is a non-negative value and is determined by the validation set of historically confirmed events. The propagation time difference tolerance in the formula is determined by the synchronization residuals at both ends and the time estimation uncertainty. Where i and j represent the indices of the first endpoint candidate event and the second endpoint candidate event, respectively; M represents the matching score of the candidate event pair; subscripts 1 to 6 of w represent the non-negative weights of each matching component; subscript S of ρ represents the normalized burst envelope correlation; subscripts W, A, and E of S represent the duration similarity, peak intensity similarity, and error symbol distribution similarity, respectively; subscript phys of I represents the physical propagation constraint satisfaction term; subscript device of P represents the device-side non-fiber disturbance penalty term; Δt represents the corrected arrival time difference between the two candidate events; t with a cap represents the corrected arrival time of the candidate events to the first endpoint and the second endpoint, respectively; L represents the fiber link length; v represents the group velocity of light in the fiber link; and subscript t of η represents the propagation time difference tolerance. Source matching includes calculating the envelope correlation, duration similarity, peak intensity similarity, error symbol distribution similarity, and physical propagation constraint satisfaction of two FEC error burst events; When the corrected arrival time difference is not within the physical range defined by the link length and group velocity, the corresponding event is rejected as a homogeneous burst event pair.

8. The fiber optic transient fault location method based on FEC bit error bursts according to claim 7, characterized in that, In step 6, the time difference of arrival is corrected, the length of the optical fiber link is calculated, and the group velocity of light in the optical fiber link is calculated. Distance of the transient fault point relative to the first endpoint satisfy ,in For link length, For group velocity, This is the difference between the arrival time of the first endpoint and the arrival time of the second endpoint; The confidence interval is calculated based on the uncertainties of link length error, group velocity error, and corrected time of arrival difference.

9. The fiber optic transient fault location method based on FEC bit error bursts according to claim 8, characterized in that, In step 7, distance estimates are obtained based on multiple wavelengths, multiple directions, multiple ports, or multiple homogeneous burst events. For the q-th distance estimate, a fusion weight is constructed using the matching score output in step 5, the reliability formed by the synchronization residual and codeword phase stability in step 4, and the distance uncertainty in step 6. The distance estimates are then normalized and weighted to obtain the final distance. Among them, the higher the matching score, the smaller the time synchronization residual, the more stable the FEC codeword phase and the more reliable the equipment status, the higher the normalized reliability; the smaller the distance uncertainty, the greater the fusion weight; the fusion variance is calculated based on the weighted distance dispersion and the uncertainty of each distance estimate, and the confidence interval is limited to the fiber link length range; a small positive number is set in the denominator of the formula to prevent division by zero; Where q represents the distance estimation index, Q represents the total number of distance estimates participating in the fusion, x represents the distance estimate obtained for the corresponding homogeneous burst event pair, M represents the matching score of the homogeneous burst event pair, R represents the normalized reliability formed by time synchronization residual, codeword phase stability and device status, σ represents the standard uncertainty of the corresponding distance estimate, ε represents a small positive number used to prevent the denominator of the fusion weight from being zero, ω represents the fusion weight of the corresponding distance estimate, the capped x represents the final distance obtained by normalized weighting, the σ squared corresponding to the capped x represents the variance of the fused distance, CI represents the distance confidence interval limited to the range of zero to the link length, L represents the fiber link length, and z represents the confidence coefficient corresponding to the target confidence level; Before performing source matching, step 5 also includes forming a non-fiber disturbance criterion based on the device temperature, power supply voltage, DSP lock status, module restart log, protection switching record, and synchronization anomalies of other ports of the same device obtained synchronously in step 1; events that meet the common-mode anomaly conditions on the device side are eliminated and do not enter source matching or fiber distance calculation.

10. An FEC error burst fiber optic fault location device for implementing the method of any one of claims 1-9, characterized in that, It includes a telemetry acquisition module, a time calibration module, a burst detection module, a burst fingerprint module, a homogeneous matching module, a location calculation module, and a confidence assessment module; The telemetry acquisition module is used to acquire the FEC decoding telemetry; The time calibration module is used to calibrate the endpoint time base, FEC frame boundary, interleaving / deinterleaving delay, and transceiver fixed processing delay. The burst detection module and burst fingerprint module are used to identify FEC error burst events and generate burst fingerprints; The same-origin matching module is used to obtain pairs of same-origin burst events based on burst fingerprints and physical propagation interval constraints. The location calculation module and confidence assessment module are used to calculate the distance to the transient fault point and output the confidence interval or confidence level.

Images