A New Generation Mobile Communication Base Station Equipment Fiber Optic Link Convergence Monitoring System
By combining data acquisition, loss decomposition, and interval determination modules, the problem of loss detection lag caused by low-intensity disturbances in in-service optical fiber links is solved, enabling accurate detection of continuous loss and tracing of boundary formation.
Patent Information
- Application Number
- CN202611347214.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-02
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2046-09-02
AI Technical Summary
Existing technologies struggle to accurately distinguish between recoverable losses caused by low-intensity disturbances and persistent localized additional losses in in-service fiber optic links, resulting in detection delays and an inability to determine the loss formation boundary and its associated disturbance range.
The data acquisition module is used to obtain the dynamic disturbance time series and block-level longitudinal loss. The recoverable state is reconstructed and the recoverable loss is deducted by the loss decomposition module. The interval determination module is used to determine the continuous loss formation process and the associated disturbance interval.
It enables accurate detection of persistent loss and reverse tracing of loss formation boundaries, improving the reliability and timeliness of fiber optic link fault diagnosis.
Smart Images

Figure CN122870682B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of optical fiber monitoring technology, and in particular to a new generation of optical fiber link fusion monitoring system for mobile communication base station equipment. Background Technology
[0002] Current monitoring of in-service fiber optic links mainly employs longitudinal loss monitoring, dynamic disturbance monitoring, and multi-source telemetry fault location. The literature (Sasai T, nakamura M, Yamazaki E, et al. digital longitudinal monitoring of optical fiber communication link[J]. Journal of Lightwave Technology, 2022, 40: 2390-2408) discloses a method to recover the longitudinal loss distribution of the fiber optic link based on the receiver's communication signal and the fiber optic transmission model, in order to determine local abnormal losses and their locations along the link. The literature (Abdellik, Grießer H, Ehrle P, et al. Reflective fiber fault detection and characterization using longshort-term memory[J]. Journal of Optical Communications and networking, 2021, 13: 32-41) discloses the use of optical time-domain reflectometry data to detect and locate reflective faults at fiber optic connectors, mechanical splices, and microbends. While these techniques can identify existing local losses, they cannot directly determine which external disturbance caused the loss based on the loss results.
[0003] To improve the reliability of fault diagnosis, existing technologies also integrate monitoring data from different devices and link locations. The literature (Mayerks, Soares JA, Pinto RP, et al. Machine-learning-based soft-failure localization with partial software-defined networking telemetry[J]. Journal of Optical Communications and networking, 2021, 13: 122-131) discloses the collection of streaming telemetry data in optical networks and the use of machine learning methods to locate soft faults in optical fibers, optical amplifiers, and optical transceivers. The literature (Wangds, Zhang CY, Chen WB, et al. A review of machine-learning-based failure management in optical networks[J]. Science China Information Sciences, 2022, 65(11): 211302) describes the use of optical time-domain reflectometry data, receiver digital signals, optical module operating parameters, and alarm data for fault detection, fault identification, and fault location. The above-mentioned fusion method mainly determines the suspected fault location based on the temporal relationship, spatial relationship or statistical correlation between monitoring parameters, alarm events and performance degradation, and usually takes the abnormal loss or service performance degradation that has already occurred as the starting point of the analysis.
[0004] In the fronthaul fiber optic links of in-service base stations, multiple low-intensity disturbances at the same fiber location may occur consecutively before the previous transient additional loss has recovered, causing the immediate responses and recovery tails of different disturbances to overlap. Simultaneously, the reliability of small, persistent local additional losses is lower than that of a single short-time data block, requiring accumulation, integration, or averaging across multiple disturbance processes. When persistent local additional loss first forms within a loss data block, the baseline before loss formation, the immediate disturbance response, the overlapping recovery tail, and the persistent residual state after formation are compressed into a single block-level loss result, resulting in the first reliable detection time being later than the actual formation time. Existing fusion monitoring methods struggle to distinguish between recoverable and persistent residual components from this block-level loss result, and cannot determine whether the persistent loss formed after a single disturbance or was formed by the combined effects of multiple consecutive disturbances. Therefore, they cannot reversely determine the initial formation boundary of persistent local additional loss and its minimum state transition associated disturbance interval. Summary of the Invention
[0005] The technical problem this invention aims to solve is that existing technologies suffer from low-intensity disturbance intervals shorter than the recovery time of transient additional losses. This results in a high degree of overlap between the recoverable response tails of adjacent disturbances and the small persistent residual states in block-level loss measurements. The baseline, immediate response, overlapping recovery tails, and persistent residual states are compressed into the same data block and cannot be separated. This causes the initial detection time of persistent loss, which requires accumulation across disturbances for reliable confirmation, to lag behind the actual formation time. Furthermore, it is impossible to reverse-determine the formation boundary of persistent loss and its minimum state transformation associated disturbance interval from the aliasing results. To address this issue, we propose a new generation of fiber optic link fusion monitoring system for mobile communication base station equipment.
[0006] To achieve the above objectives, this application adopts the following technical solution: a new generation mobile communication base station equipment fiber optic link convergence monitoring system, including a data acquisition module, a loss decomposition module, and an interval determination module.
[0007] The data acquisition module is used to acquire the dynamic disturbance timing, block-level longitudinal loss, and time range and time weight of each longitudinal loss data block in the same local fiber segment of the in-service base station fiber link. The in-service base station fiber link is a base station fronthaul link, base station transmission link, or communication fiber link between base station equipment that continuously carries mobile communication services. The same local fiber segment is a fiber segment whose position difference is within a preset registration tolerance after the dynamic disturbance location and longitudinal loss location are registered along a unified fiber distance axis. The dynamic disturbance timing includes the fiber position of the dynamic disturbance, start time, end time, loading process, unloading process, recovery process after the disturbance stops, and data quality marker. The block-level longitudinal loss is the local loss obtained by accumulating multiple valid echoes, multiple frames of valid data, or multiple valid sampling results within the corresponding time range. The time weight is the actual contribution weight of each valid echo, valid frame, or valid sampling time to the block-level longitudinal loss. The time weight is a non-negative weight and is normalized within the corresponding longitudinal loss data block.
[0008] The loss decomposition module is used to reconstruct the recoverable state that evolves continuously between adjacent dynamic disturbances based on the continuous disturbance path and the continuous path state model calibrated and independently verified by recoverable disturbance test data. The continuous disturbance path is a state change path formed by continuously connecting the loading process, unloading process, and recovery process between adjacent dynamic disturbances according to the occurrence order of each dynamic disturbance within the same local fiber segment. The continuous path state model is a model that can continuously update the internal state according to the mechanical state input, retain the unfinished recovery state between adjacent dynamic disturbances, and represent the state decay process after the disturbance stops. The continuous path state model is only used within the calibration range that has been independently verified. When the model verification fails or the running input exceeds the calibration range, the reconstruction of the corresponding recoverable state stops. The recoverable state is an internal state quantity that characterizes the changes of the optical fiber under mechanical disturbance with the loading, unloading, and recovery processes, and can decay after the disturbance stops. The recoverable state is not directly used as an optical loss value.
[0009] The loss decomposition module converts the recoverable state into recoverable additional loss based on the mapping relationship between recoverable state and optical loss calibrated for the local fiber segment, and performs weighted integration according to the time range and time weight of the corresponding longitudinal loss data block to obtain the block-level recoverable loss. The mapping relationship between recoverable state and optical loss is obtained by pairing recoverable state with synchronous independent loss using fully recoverable disturbance test data. The mapping relationship satisfies that zero recoverable state corresponds to zero recoverable additional loss, and within the calibration range, the recoverable additional loss does not decrease when the recoverable state amplitude increases. The block-level recoverable loss is the block-level result obtained by converting the recoverable additional loss that changes with time according to the actual time range and time weight used when forming the block-level longitudinal loss.
[0010] The loss decomposition module is also used to set a persistent loss formation process for candidate disturbances, which monotonically transitions from an unformed state to a formed state. The proportion of persistent states within a block is determined based on the overlap between the persistent loss formation process and each longitudinal loss data block. The candidate disturbance is a dynamic disturbance located in the same local fiber segment as the target block-level longitudinal loss and may be related to the formation of newly added persistent residual loss. The persistent loss formation process is a candidate state change process set for candidate disturbances, whose value monotonically changes from zero (representing an unformed state) to a value (representing a fully formed state), and remains at a value after full formation. The persistent loss formation process represents the time change of the corresponding persistent residual loss from not yet formed to fully formed, without directly using the moment when persistent loss is first reliably detected as the formation moment. The proportion of persistent states within a block is a value obtained by normalizing and weighting the persistent loss formation process according to the time range and time weight of the corresponding longitudinal loss data block, used to represent the effective proportion of the corresponding persistent state in the actual observation process of that longitudinal loss data block.
[0011] The loss decomposition module subtracts the loss baseline and the block-level recoverable loss from the block-level longitudinal loss to obtain the block-level loss residual. Using the proportion of persistent states within the block as the observation coefficient, it inverts the persistent residual loss increment corresponding to each candidate disturbance under the constraint that the persistent residual loss increment is not less than zero, and constrains the sum of the persistent residual loss increments with the total persistent loss determined in the subsequent stable reference period. The loss baseline includes the local loss that existed before the candidate disturbance occurred and the common drift component generated by changes in the monitoring light source, receiver drift, temperature changes, and overall link fluctuations during the monitoring period. The loss baseline is converted according to the same time range and time weight as the corresponding longitudinal loss data block. The block-level loss residual is the result of subtracting the loss baseline and block-level recoverable loss from the block-level longitudinal loss. The persistent residual loss increment is the newly added local additional loss formed by the corresponding candidate disturbance or indistinguishable disturbance group that does not recover on its own in the current monitoring period. The total persistent loss is the total newly added stable persistent loss determined in the subsequent stable reference period relative to the baseline state before the candidate disturbance occurred, excluding the historical loss that existed at that location before the candidate disturbance occurred.
[0012] The subsequent stable reference period is the period during which recoverable additional losses have decayed to a preset error range, there are no new in-situ dynamic disturbances, no maintenance or reset operations, and the common drift reference data is stable. The subsequent stable reference period is used to determine the total amount of continuous losses, but not to determine the formation time of continuous residual losses. The formation time or formation range is determined in reverse based on the previously recorded dynamic disturbance path, the proportion of continuous states within the block, and multiple block-level loss residuals.
[0013] The interval determination module is used to determine the state transition associated disturbance interval that cannot be further reduced by excluding boundary disturbances from the continuous candidate disturbance range that meets the block-level loss residual interpretation condition, cross-data block persistence condition, persistence total closure condition, and intra-block persistence state proportion sequence distinguishable condition. The continuous candidate disturbance range is the range formed by one or more co-position candidate disturbances arranged continuously in chronological order. The starting point of the continuous candidate disturbance range is the first co-position candidate disturbance that appears after the last longitudinal loss data block that meets the baseline condition before the persistence loss is formed, and the ending point of the continuous candidate disturbance range is the last co-position candidate disturbance that appears before obtaining the subsequent stable reference period. When the recoverable state of the preceding dynamic disturbance continues beyond the starting point, the corresponding preceding dynamic disturbance is included forward into the continuous candidate disturbance range.
[0014] The block-level loss residual interpretation condition is that the difference between the block-level loss residual reconstructed based on the proportion of persistent states within the block corresponding to the candidate perturbation and the incremental residual loss within the block, and the measured block-level loss residual, is within a preset residual range. The cross-block persistence condition is that, after the persistent state is formed, the block-level loss residual of at least two valid longitudinal loss data blocks is not lower than the persistence loss detection limit, and the total persistence loss is not lower than the persistence loss detection limit. The total persistence loss closure condition is that the difference between the sum of each incremental residual loss and the total persistence loss does not exceed a preset total closure allowable error. The intra-block persistent state proportion sequence distinguishable condition is that, within the same continuous candidate perturbation range, the difference in the intra-block persistent state proportion sequence of the candidate perturbation or indistinguishable perturbation group to be solved separately in all valid longitudinal loss data blocks is not lower than a preset minimum distinguishable difference.
[0015] The state transition associated perturbation interval is a continuous candidate perturbation range that simultaneously satisfies the block-level loss residual interpretation condition, cross-block persistence condition, total persistence loss closure condition, and the distinguishable condition of the intra-block persistence state proportion sequence, and no longer satisfies all of the above conditions after excluding boundary perturbations from its start or end. When multiple continuous candidate perturbation ranges cannot be distinguished, the interval determination module outputs the minimum continuous range covering the multiple continuous candidate perturbation ranges; the multiple continuous candidate perturbation ranges cannot be distinguished means that two or more continuous candidate perturbation ranges all satisfy the corresponding determination conditions, and no range can be excluded based on the existing block-level loss residual, cross-block persistence result, total persistence loss closure result, and intra-block persistence state proportion sequence relationship; the minimum continuous range is the continuous time range formed from the earliest starting perturbation to the latest ending perturbation among the multiple continuous candidate perturbation ranges.
[0016] Specifically, the data acquisition module includes a dynamic disturbance acquisition channel, a longitudinal loss acquisition channel, a time synchronization component, a fiber-to-fiber distance calibration component, and a monitoring host. The dynamic disturbance acquisition channel is used to generate mechanical state data that retains the loading direction, unloading direction, and recovery process. The longitudinal loss acquisition channel is used to accumulate fiber-to-fiber echo data or other longitudinal loss measurement data and generate the block-level longitudinal loss. The time synchronization component is used to make the dynamic disturbance data, longitudinal loss data, and equipment state data adopt a unified time reference. The fiber-to-fiber distance calibration component is used to map the dynamic disturbance position and the longitudinal loss position to a unified fiber-to-fiber coordinate system. The monitoring host is used to perform data quality judgment, spatiotemporal registration, recoverable state reconstruction, continuous residual loss inversion, and state transformation associated disturbance interval output. The dynamic disturbance acquisition channel and the longitudinal loss acquisition channel can share the monitoring optical path and receiving channel, or they can use independent monitoring channels that are registered with a unified time reference and a unified fiber-to-fiber coordinate system.
[0017] Specifically, the dynamic disturbance acquisition channel includes a monitoring optical signal generation component, a monitoring optical access component, a fiber-backscattered signal receiving component, a sampling component, and a disturbance state processing component. The monitoring optical signal generation component generates a monitoring optical signal that is isolated from the service optical signal. The monitoring optical access component connects the monitoring optical signal to the fiber optic link of the in-service base station and isolates the service receiving end. The fiber-backscattered signal receiving component receives the backscattered signal returned from the fiber optic link of the in-service base station. The sampling component converts the backscattered signal into sampled data. The disturbance state processing component obtains the phase change, local strain, curvature, or equivalent mechanical state quantities that are calibrated and retain the direction of change corresponding to the fiber position from the sampled data.
[0018] Specifically, the longitudinal loss acquisition channel reuses the echo amplitude data from the fiber-backscattered signal receiving component, or adopts an independent longitudinal loss monitoring channel that shares a time reference and fiber-distance axis with the dynamic disturbance acquisition channel. The longitudinal loss acquisition channel accumulates valid echoes or valid frames and records the start time, end time, fiber-distance position, valid data marker, block-level longitudinal loss, and time weight corresponding to each valid echo, valid frame, or valid sampling time for each longitudinal loss data block. When the number of valid data is lower than a preset valid proportion, the corresponding longitudinal loss data block is marked as invalid, preventing the invalid longitudinal loss data block from participating in the continuous residual loss increment inversion.
[0019] Specifically, the data acquisition module performs spatiotemporal registration of dynamic disturbance timing and block-level longitudinal loss based on a unified timestamp and a unified fiber distance axis. When the difference between the dynamic disturbance location and the longitudinal loss location does not exceed the preset registration tolerance, the corresponding dynamic disturbance data and longitudinal loss data are grouped into the same local fiber segment; when the difference exceeds the preset registration tolerance, no candidate relationship is established between the corresponding dynamic disturbance and the block-level longitudinal loss.
[0020] Specifically, the loss decomposition module includes a recoverable contribution reconstruction unit and a persistent residual inversion unit; the recoverable contribution reconstruction unit is used to form continuous disturbance paths, continuously update recoverable states, convert recoverable states into recoverable additional losses, and form block-level recoverable losses; the persistent residual inversion unit is used to set the persistent loss formation process, calculate the proportion of persistent states within a block, form block-level loss residuals, invert persistent residual loss increments, and complete the closure of the total persistent loss; the interval determination module includes a candidate range verification unit and a state transition interval output unit; the candidate range verification unit is used to verify whether the range of continuous candidate disturbances meets the corresponding determination conditions, and the state transition interval output unit is used to shrink the range of continuous candidate disturbances and output the state transition associated disturbance interval.
[0021] Specifically, the data acquisition module sends the dynamic disturbance timing sequence, block-level longitudinal loss, data block time range, time weight, and loss baseline, which have been completed for data quality judgment and spatiotemporal registration, to the loss decomposition module; the loss decomposition module outputs the proportion of persistent states within the block, the increment of persistent residual loss, the block-level loss residual, and the closure result of the total persistent loss corresponding to each candidate disturbance to the interval determination module; the interval determination module checks and narrows the range of continuous candidate disturbances based on the proportion of persistent states within the block, the increment of persistent residual loss, the block-level loss residual, and the closure result of the total persistent loss, and outputs the state transformation associated disturbance interval or the corresponding indeterminate state.
[0022] Preferably, the loss decomposition module forms the continuous disturbance path according to the start and end times of each dynamic disturbance, continues the recovery process of the recoverable state between adjacent dynamic disturbances, and uses the recoverable state evolved from the end of the previous dynamic disturbance to the start of the next dynamic disturbance as the initial state of the next dynamic disturbance; the recovery process is the process of the recoverable state decaying towards a silent state over time after the end of the dynamic disturbance. If the recoverable state formed by the previous dynamic disturbance has not yet recovered to the silent range, the next dynamic disturbance continues to update from the current recoverable state without resetting the recoverable state to zero.
[0023] Preferably, the loss decomposition module determines the mechanical state sequence based on the dynamic disturbance time sequence, divides the mechanical state sequence into a loading segment, an unloading segment, and a recovery segment between adjacent dynamic disturbances, and uses the recoverable state at the previous sampling time as the initial state value at the next sampling time, continuously updating the recoverable state according to the time sequence of each segment; the mechanical state sequence is a sequence formed by local strain, curvature, distributed phase change, or calibrated equivalent mechanical state quantities in time sequence; the mechanical state sequence retains the direction of mechanical state change and can distinguish the loading process, the unloading process, and the recovery process after the disturbance stops.
[0024] Furthermore, the mapping relationship between the recoverable state and optical loss is calibrated through a recoverable perturbation test, satisfying the condition that zero recoverable state corresponds to zero recoverable additional loss, and within the calibration range, the recoverable additional loss does not decrease when the recoverable state amplitude increases; the time weight is the actual weight of each valid frame or each valid sampling moment in the corresponding longitudinal loss data block on the block-level longitudinal loss; the loss decomposition module performs a weighted integral of the recoverable additional loss according to the normalized time weight to obtain the block-level recoverable loss of the corresponding longitudinal loss data block; the recoverable perturbation test is a test in which both the mechanical state and the independently measured additional loss can return to the error range before the perturbation after the perturbation ends. When the operating state exceeds the calibration range of the mapping relationship, the formation of the corresponding block-level recoverable loss stops, and the mapping relationship is not extrapolated without constraints.
[0025] Specifically, the continuous path state model can also employ segmented path state relationships established through recoverable disturbance experiments; these segmented path state relationships include loading segment state update relationships, unloading segment state update relationships, and recovery segment state update relationships. When processing the current sampling time, the corresponding state update relationship is selected based on the direction of change in the mechanical state, and the recoverable state from the previous sampling time is used as the initial state value for the current sampling time. When the current mechanical state is located between adjacent calibration nodes, interpolation is performed based on the adjacent calibration nodes; when the current mechanical state or recovery time exceeds the calibration range, state updates are stopped, and the segmented path state relationships are not extrapolated without constraints.
[0026] Specifically, the mapping relationship between the recoverable state and optical loss adopts a piecewise linear mapping, a single-investigation table lookup, or an interpolation relationship that maintains non-negative monotonicity. All different mapping relationships are established using recoverable perturbation test data, satisfying the condition that zero recoverable state corresponds to zero recoverable additional loss, and that within the calibration range, the recoverable additional loss does not decrease when the recoverable state amplitude increases.
[0027] Preferably, the value of the continuous loss formation process is between zero and one and does not decrease over time. It is zero before the continuous loss forms, one after the continuous loss is fully formed, and a value between zero and one during the gradual formation of continuous loss. The candidate formation start point of the continuous loss formation process is no earlier than the start time of the corresponding candidate disturbance, and the candidate formation end point is no later than the creep observation cutoff time after the end of the corresponding candidate disturbance. The creep observation cutoff time is the sum of the end time of the corresponding candidate disturbance and the preset creep observation duration. The preset creep observation duration is determined based on the longest effective time required for the mechanical state and additional losses in the recoverable disturbance test to recover to the baseline error range.
[0028] Preferably, the loss decomposition module performs normalized weighted integration on the state values of each persistent loss formation process according to the time range and time weight of each longitudinal loss data block, and determines the integration result as the proportion of the persistent state within the corresponding longitudinal loss data block for each candidate disturbance; the proportion of the persistent state within the block is between zero and one. When the persistent state has been fully formed before the corresponding longitudinal loss data block, the proportion of the persistent state within the block is one; when the persistent state begins to form after the corresponding longitudinal loss data block, the proportion of the persistent state within the block is zero; when the persistent state is formed within the corresponding longitudinal loss data block, the proportion of the persistent state within the block is between zero and one.
[0029] Preferably, the loss decomposition module uses the proportion of each candidate disturbance's intra-block persistent state in each longitudinal loss data block as equation coefficients to establish a set of inversion equations between multiple block-level loss residuals and the persistent residual loss increments corresponding to each candidate disturbance, and solves the set of inversion equations under the constraint that the persistent residual loss increment does not exceed a preset residual loss calibration range; the set of inversion equations uses the block-level loss residuals of different longitudinal loss data blocks as observations, and the persistent residual loss increments corresponding to candidate disturbances or indistinguishable disturbance groups as unknowns. During the current monitoring period without maintenance, clamp release, manual reset, or optical cable re-laying, the persistent residual loss increment is not less than zero; when maintenance, clamp release, manual reset, or optical cable re-laying occurs, the current inversion is terminated, and the loss baseline is re-established after the corresponding operation is completed.
[0030] Furthermore, the subsequent stable reference period satisfies the following conditions: recoverable additional losses have decayed to a preset error range, there are no new in-situ dynamic disturbances, there are no maintenance reset operations, and the common drift reference data is stable. The loss decomposition module determines the new stable persistent loss relative to the baseline state before the candidate disturbance occurs within the subsequent stable reference period as the total persistent loss, and excludes candidate disturbance combinations whose difference between the sum of persistent residual loss increments and the total persistent loss exceeds a preset total closure allowable error. The subsequent stable reference period is not used to determine the formation time of persistent residual loss. The total closure allowable error is determined based on the subsequent stable loss error, the loss baseline correction error, and the persistent residual loss increment inversion error, and is fixed before processing data that may form persistent loss.
[0031] Preferably, the interval determination module merges candidate perturbations whose intra-block persistent state proportion sequence difference is lower than a preset minimum distinguishable difference into an indistinguishable perturbation group, and determines the range of continuous candidate perturbations that simultaneously meets the following conditions as the effective candidate perturbation range: the difference between the block-level loss residual reconstructed from the inversion result and the measured block-level loss residual does not exceed a preset residual threshold; the block-level loss residual of at least two effective longitudinal loss data blocks after the persistent state is formed is not lower than the persistent loss detection limit, and the total persistent loss is not lower than the persistent loss detection limit; the difference between the sum of each persistent residual loss increment and the total persistent loss does not exceed the total closure allowable error; the intra-block persistent state proportion sequence difference of each candidate perturbation or indistinguishable perturbation group after merging is not lower than the preset minimum distinguishable difference; the preset minimum distinguishable difference is the minimum difference determined by independent verification data used to distinguish the intra-block persistent state proportion sequence of two candidate perturbations.
[0032] Furthermore, the interval determination module successively eliminates boundary disturbances from the start and end of the effective candidate disturbance range. After each elimination, the determination conditions of the effective candidate disturbance range are re-examined. If the determination conditions are still met after elimination, the reduced range is retained and elimination continues until both eliminating boundary disturbances from the start and end will cause the reduced range to no longer meet the determination conditions. The range after stopping elimination is then determined as the state transition associated disturbance interval. The elimination of boundary disturbances means that the corresponding boundary disturbance is not used as the solution object for the continuous residual loss increment in the current inversion relationship, and the original dynamic disturbance data and longitudinal loss data corresponding to the boundary disturbance are not deleted.
[0033] Furthermore, when there are two or more valid candidate perturbation ranges that both satisfy the determination conditions of the valid candidate perturbation range and cannot be distinguished from each other, the interval determination module outputs the smallest continuous range from the earliest starting perturbation to the latest ending perturbation in the valid candidate perturbation range, and determines the smallest continuous range as the state transition associated perturbation interval; the two or more valid candidate perturbation ranges that cannot be distinguished from each other can all explain the block-level loss residual, satisfy the cross-data block persistence condition and the persistence total loss closure condition within the preset error range, and the existing valid longitudinal loss data blocks cannot exclude any of the valid candidate perturbation ranges.
[0034] Furthermore, the recoverable state is calculated using a Bouc-Wen type internal state model with a relaxation term. This internal state model takes the mechanical state sequence as input, and its state changes are determined by the mechanical state change term, the loading and unloading path term, and the relaxation term after the disturbance stops. The model parameters are calibrated using recoverable disturbance test data and validated using verification data independent of the calibration data. The persistent residual loss, as a state variable independent of the recoverable state, is accumulated according to the persistent residual loss increment corresponding to each candidate disturbance. The relaxation term is used to decay the recoverable state over time after the mechanical state stops loading; the persistent residual loss does not participate in the relaxation calculation of the recoverable state.
[0035] Specifically, when the positional difference between dynamic disturbance data and longitudinal loss data exceeds the registration tolerance, the input of the continuous path state model exceeds the calibration range, there is insufficient effective longitudinal loss data blocks, a subsequent stable reference period cannot be obtained, or the total continuous loss is lower than the continuous loss detection limit, the system stops outputting the determined state transition associated disturbance interval and outputs the operating states of position misregistration, model exceeding the applicable range, insufficient data, lack of stable reference, or continuous loss pending confirmation, respectively. When candidate disturbances cannot be further distinguished at the current observation scale, the system outputs the minimum continuous range containing the corresponding candidate disturbance.
[0036] The technical effects and advantages of this invention are as follows:
[0037] In this invention, the data acquisition module correlates the dynamic disturbance timing, block-level longitudinal loss, data block time range, and time weight of the same local fiber segment, establishing a unified observation basis for disturbance and loss data with different time resolutions. The loss decomposition module reconstructs the incompletely recovered state between adjacent disturbances according to the continuous disturbance path, and converts the recoverable state into block-level recoverable loss at the same observation scale as the block-level longitudinal loss. Thus, without shortening the longitudinal loss data block or reducing the cumulative sample size, the immediate response and recovery tail of the disturbance are deducted from the mixed loss result, avoiding the misidentification of recoverable loss as continuous loss. By considering the proportion of persistent states within a block, the non-negativity constraint of the incremental persistent residual loss, and the closure of the total amount of subsequent stable persistent loss, the system can recover the degree of persistent loss formation within the data block and eliminate non-damaging disturbances that occur after loss formation. The interval determination module further combines block-level loss residual interpretation, cross-block persistence, total closure, and the distinguishable condition of the proportion sequence to gradually eliminate boundary disturbances and output the state transition-related disturbance interval that cannot be further reduced under the current observation conditions. When multiple disturbances cannot be reliably distinguished, the system outputs the minimum continuous range, avoiding the forced specification of a single disturbance when evidence is insufficient. The system thus balances the reliability of detecting small persistent losses with the time traceability capability of the formation process, enabling the reverse recovery of the first boundary formation of persistent losses that have been compressed by block-level cumulative processing. Attached Figure Description
[0038] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts:
[0039] Figure 1 This is a schematic diagram of the overall system architecture of the present invention;
[0040] Figure 2 This is a flowchart of the recoverable state calibration and block-level recoverable loss calculation of the present invention;
[0041] Figure 3 This is a flowchart of the continuous residual loss inversion process of the present invention;
[0042] Figure 4 This is a flowchart for determining the state transition associated perturbation interval of the present invention. Detailed Implementation
[0043] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.
[0044] Example 1
[0045] The system described in this embodiment is used for an in-service optical fiber link that continuously carries mobile communication services; the optical fiber link can be a fronthaul optical fiber link between a distributed unit (DU) and a wireless unit (RU) or an active antenna unit (AAU), or it can be a communication optical fiber link between a centralized unit, a distributed unit, a base station transmission device, and a radio frequency device.
[0046] When optical cables pass through bends in cable trays, clamp locations, outdoor entry points, duct bends, near junction boxes, and equipment cabinet entry / exit points, they may be continuously subjected to vibration, tension, bending, localized compression, or displacement of the fixed structure. Some disturbances can gradually recover after the external force disappears, and the corresponding additional losses also decrease accordingly; other disturbances may cause the clamps to slip, or cause changes in the sheath, bending state, and fixed structure that cannot be immediately recovered, thus forming continuous localized additional losses.
[0047] like Figure 1 As shown, the system in this embodiment includes a monitoring data acquisition unit, a recoverable contribution reconstruction unit, a persistent residual inversion unit, and a state transition interval output unit. The monitoring data acquisition unit obtains dynamic disturbance data and block-level longitudinal loss data for the same local fiber segment. The recoverable contribution reconstruction unit calculates the block-level recoverable loss based on the continuous disturbance path. The persistent residual inversion unit calculates the persistent residual loss increment from the block-level loss after deducting the baseline and recoverable contribution. The state transition interval output unit determines the state transition associated disturbance interval accordingly. These units can be located within the same monitoring host, or they can be located separately in the base station-side edge computing device and the remote server. Data transmitted between units is marked with a time stamp, fiber-side location stamp, and parameter version stamp.
[0048] In one feasible configuration, the monitoring data acquisition unit is located on the base station transmission equipment side. Its optical components include a narrow-linewidth monitoring light source, a pulse modulator, an optical amplifier, a bandpass filter, an optical circulator, a wavelength division multiplexer, a polarization diversity coherent receiver, an analog-to-digital converter (ADC), and a field-programmable gate array (FPGA). The monitoring light source outputs monitoring light separated from the service optical band. The pulse modulator modulates the monitoring light into probe pulses, which, after passing through the optical amplifier and bandpass filter, are coupled to the in-service optical fiber by the optical circulator and wavelength division multiplexer. The backscattered light along the fiber returns along the original path and is sent to the polarization diversity coherent receiver by the optical circulator. After analog-to-digital conversion, it forms two in-phase, orthogonal sampling data with two polarizations. The FPGA completes pulse triggering, sampling timing, data block boundary recording, and preliminary quality judgment. The monitoring host completes subsequent phase processing, loss calculation, and state inversion.
[0049] When the service link uses the 1310nm or 1550nm band, 1625nm or other monitoring wavelengths with sufficient spacing from the service optical band can be selected. The passband and isolation of the wavelength division multiplexer and the peer filter are determined based on the receiving band of the service optical module, the allowable crosstalk, and the monitoring optical power. In this embodiment, the monitoring light is isolated by a cutoff filter before entering the service receiver. The service received power, bit error rate, and link alarm status are measured before and after the monitoring optical path is connected. After confirming that the wavelength division multiplexer insertion loss and monitoring optical crosstalk do not cause the service indicators to exceed the allowable range of the original link, continuous monitoring is then initiated.
[0050] For base station fronthaul links no longer than 10km, monitoring pulses of 20ns to 100ns can be used. The pulse repetition frequency should ensure that the effective echo of the previous detection pulse returns before transmitting the next pulse, allowing for margins required for equipment processing and fiber optic pigtail length; in a 10km link, a repetition frequency of 5kHz to 8kHz can be used. The analog-to-digital converter's sampling rate should ensure at least five sampling points within one pulse width; when using 20ns pulses, a sampling rate of at least 250MS / s can be selected. The bandwidth of the polarization diversity receiver should match the analog-to-digital converter's sampling rate, and the coherence length of the monitoring light source should cover the effective round-trip optical path of the monitoring link. The above parameters represent a feasible equipment configuration and are not intended to limit other link lengths, spatial resolutions, or equipment types.
[0051] The monitoring host can be an industrial computer or edge server with a high-speed data interface. A field-programmable gate array (FPGA) connects to the monitoring host via PCIe, Gigabit Ethernet, or other interfaces capable of continuously transmitting raw sampled data. Dynamic disturbance data, longitudinal loss data, and device status data are recorded using the same hardware-triggered counter. When the dynamic acquisition device and the loss acquisition device are independent, they are synchronized using the IEEE 1588 precision time protocol, a synchronization pulse, or an equivalent unified clock; the remaining time error should be less than the minimum duration of the dynamic disturbance event. Uniform calibration is performed along the fiber distance axis using known connectors, reflective ends, or calibrated fiber segments.
[0052] The monitoring data acquisition unit acquires dynamic disturbance data and longitudinal loss data for the same optical fiber link. The dynamic disturbance data is used to record the location along the fiber where the disturbance occurs, the start time, the end time, the loading process, the unloading process, and the recovery process after the disturbance stops.
[0053] When using the aforementioned polarization diversity coherent receiver, the monitoring host first calculates the phase based on the in-phase and quadrature components of each polarization channel. For each polarization channel, phase unwinding, abnormal jump repair, and amplitude quality assessment are performed. Then, phase difference is calculated for sampling positions along the fiber separated by one gauge length. After calibration with a disturbance-free zero point and loading direction, the phase difference results yield equivalent mechanical state quantities retaining both positive and negative directions. When all polarization channels are valid, they are weighted according to the variance of the silent noise of each channel. When a channel experiences coherent fading, receiver saturation, or a signal-to-noise ratio below the quality threshold, the weight of that channel at the corresponding time and location is set to zero. When all channels are invalid, the data segment is marked as missing, and the disturbance direction is not forcibly supplemented using adjacent data.
[0054] The monitoring host determines the silent fluctuation range of the equivalent mechanical state amplitude and rate of change based on the undisturbed data. A disturbance is recorded when either the amplitude or rate of change continuously exceeds the entry threshold and reaches the minimum duration; the disturbance is recorded when both continuously fall below the exit threshold and reach the exit duration. The entry threshold, exit threshold, and minimum duration are determined by the false trigger rate of the undisturbed data and recoverable disturbance verification data, and are fixed during a single monitoring task. Each disturbance event is recorded with at least the event number, start time, end time, fiber position, peak mechanical state value, loading direction, unloading direction, and data quality flag.
[0055] Dynamic disturbance data can also be generated from local strain, curvature, or other monitoring quantities that can reflect changes in mechanical state. When absolute strain cannot be directly obtained for the equipment, equivalent mechanical state quantities that have been calibrated and retain the direction of change can be used, but this substitute data should still be able to distinguish between loading, unloading, and the recovery process after the disturbance stops.
[0056] Longitudinal loss data is used to reflect local loss variations at different fiber locations along the link. Longitudinal loss acquisition can reuse the echo amplitude data from the aforementioned coherent receiving channel, or it can be obtained from an independent longitudinal loss channel that shares wavelength division access and a unified clock with the dynamic disturbance acquisition channel. When reusing the coherent receiving channel, the monitoring host calculates the echo power envelope of each polarization channel based on in-phase and quadrature sampling data, removes saturated frames, triggered abnormal frames, and frames with insufficient effective sampling ratios, and then accumulates the effective echoes or effective frames.
[0057] For the target fiber location, reference segments are first selected upstream and downstream, respectively, excluding connectors, fusion splices, fiber type changes, and known anomalies. A normal fiber-along attenuation trend is then fitted to these reference segments. The average power difference between the upstream and downstream reference segments, after trend correction, is converted into the local loss at the target location. When calculating from the backscattered signal, the two-way loss is converted into a one-way local loss. If a new reflection event, a sudden change in the backscattering coefficient, or a common drift anomaly occurs in the reference segment, the original reference segment is discontinued, and a new reference segment is selected, or a pre-calibrated event loss calculation relationship is used.
[0058] Longitudinal loss monitoring accumulates multiple echoes or frames of data to reduce the impact of scattered noise and equipment fluctuations on the assessment of minor losses. Each accumulation forms a longitudinal loss data block, which at least stores the start time, end time, corresponding fiber-along position, valid frame number, block-level loss value, and the actual weight of each valid frame or time point to the final loss value. During equal-weighted accumulation, each valid frame uses the same weight; after removing some abnormal frames, the remaining valid weights are renormalized. If the number of valid frames is lower than a predetermined minimum proportion, the data block is marked as invalid.
[0059] The length of the longitudinal loss data block is determined based on the minimum sustained loss level that needs to be reliably verified. Specifically, under conditions of no significant disturbance, multiple sets of blank data are acquired, and the loss fluctuations corresponding to different cumulative durations are compared. When a certain cumulative duration allows the blank data fluctuations to be below a preset error and can distinguish the minute sustained losses to be monitored, this cumulative duration is determined as the data block length for this monitoring task.
[0060] Once the data block length is determined, it remains unchanged during a single loss formation judgment period. The system does not improve the temporal resolution at the formation time by shortening the data block, nor does it reduce the number of echoes or frames involved in loss estimation. This maintains the reliability of detecting minor continuous losses and prevents the actual continuous loss from falling back into the device noise range after the data block is shortened.
[0061] The temporal resolution of dynamic disturbance data is higher than the output frequency of longitudinal loss data blocks. Therefore, within the time covered by a longitudinal loss data block, the system can record the start, end, loading, unloading, and recovery processes of multiple disturbances, while longitudinal loss monitoring only outputs a single block-level loss result after cumulative processing.
[0062] The monitoring data acquisition unit performs spatiotemporal registration of dynamic disturbance data and longitudinal loss data. Temporal registration is completed using a unified clock or a corrected timestamp. Spatial registration maps the dynamic disturbance location and loss location to the same fiber-along coordinates through distance axis calibration. When the fiber-along position difference between the two types of data does not exceed the registration range jointly determined by their respective positioning errors, they are grouped into the same local fiber segment.
[0063] The registration range is obtained before formal monitoring. Specifically, at least three known locations along the fiber are selected, and repeatable mechanical perturbations and known additional attenuations are applied to each location. The dynamic perturbation positioning results and longitudinal loss positioning results are recorded multiple times. The dynamic positioning error, loss positioning error, and distance axis calibration error are calculated separately, and the root mean square composite value of the three errors is multiplied by a predetermined position confidence coefficient as the registration tolerance. The position confidence coefficient is determined based on the coverage ratio of real co-location events in the calibration data. The registration tolerance remains unchanged during a single monitoring task; if the difference between the dynamic position and the loss position exceeds this tolerance, no co-location candidate relationship is established.
[0064] The system can also acquire data on the monitoring optical output power, receiver gain, equipment temperature, and loss changes in adjacent undisturbed fiber segments. This data is used to correct for changes in the monitoring light source, receiver drift, temperature changes, and overall link fluctuations, and is not used directly to determine that a particular disturbance has caused sustained loss.
[0065] In one feasible baseline correction method, undisturbed data is used, with changes in monitored optical output power, equipment temperature, and block loss changes in adjacent undisturbed fiber segments as inputs, and block loss changes at the target location as output, to establish a linear regression relationship. This regression relationship is applied to independent undisturbed data. If the corrected loss fluctuation at the target location is no higher than before the correction and no systematic deviation increases over time, the regression relationship is adopted. If the linear relationship fails verification, a piecewise linear relationship is established according to the operating range of temperature or monitored optical power. The regression coefficients, applicable range, and verification error are written into the parameter version and remain fixed during a single judgment period.
[0066] When multiple short-interval disturbances occur in succession, the recoverable additional loss caused by the previous disturbance has not yet fully attenuated before the next disturbance begins. If a disturbance or multiple consecutive disturbances cause the fiber to transition to a continuous loss state, and this state is formed within the longitudinal loss data block, then the data block will simultaneously contain the baseline before the disturbance, the immediate response of multiple disturbances, the recovery tail of the preceding disturbance, and the continuous loss formed afterward.
[0067] Therefore, the first data block exceeding the loss threshold only indicates that the persistent loss has reached a level that can be reliably detected; it does not indicate that the persistent loss occurred precisely at the beginning of that data block, nor can it directly determine which disturbance within the data block's coverage area is related to the formation of persistent loss. Subsequent processing, while maintaining the data block length, utilizes dynamic disturbance data to recover the compressed time process during block-level accumulation.
[0068] Example 2
[0069] Based on the in-situ dynamic disturbance data and block-level longitudinal loss data obtained in Example 1, this embodiment reversely recovers the formation range of persistent local additional loss.
[0070] Let the position along the fiber be The time is The local loss at this location is expressed as:
[0071] ,
[0072] in, This represents the local losses that existed before the candidate disturbance occurred and the common drift component during the monitoring period. This refers to the recoverable additional loss caused by mechanical disturbances that can decay after the disturbance stops. This represents the newly formed persistent residual loss within the scope of this candidate perturbation. Old losses existing before the candidate perturbation occurred are included in the baseline and will not be recalculated as persistent residual increments.
[0073] like Figure 2 As shown, the recoverable contribution reconstruction unit continuously reconstructs the loading, unloading, and relaxation paths of the same local fiber segment according to the actual occurrence sequence of each disturbance. The continuous path state model is first calibrated using recoverable disturbance test data, and then its interpretation error of the recoverable loss process is checked using independent verification data. After verification, the model parameters are fixed. During operation, the recoverable state is not cleared between adjacent disturbances; when a subsequent disturbance occurs, the recovery tail segment of the previous disturbance that has not yet attenuated is retained, and the subsequent disturbance continues to be calculated from the current state. Reinitialization is only allowed when the dynamic state continuously returns to the quiescent range and the local loss returns to the baseline error range.
[0074] The recoverable contribution reconstruction unit continuously reconstructs the loading, unloading, and relaxation paths of the same local fiber segment according to the actual occurrence sequence of each disturbance. The recoverable state is not reset between adjacent disturbances. When a subsequent disturbance occurs, the recovered tail segment that has not yet attenuated from the previous disturbance is retained, and the calculation for the subsequent disturbance continues from the current state. Reinitialization is only allowed when the dynamic state continuously returns to the quiescent range and the synchronous local loss returns to the baseline error range.
[0075] The reconstruction can employ a continuous path model with relaxation terms, or a segmented path model obtained through full recoverable perturbation experiments. To provide a directly executable approach, this embodiment uses a Bouc-Wen type internal state model with relaxation terms:
[0076] ,
[0077] in, This represents the equivalent mechanical state after zero-point and amplitude calibration. This indicates that the dimensionless property can be recovered to its internal state. Indicates the input conversion coefficients. and Indicates the load and unload path parameters. Indicates the order of nonlinearity of the path. This represents the relaxation coefficient after the disturbance stops. The parameters satisfy... , and .
[0078] Parameters were obtained from all recoverable perturbation experiments. Before each independent experiment, the initial internal state value was set to zero when both the mechanical state and independent loss returned to the baseline error range; the initial value was not reset to zero within the same short-interval repeated perturbation experiment. For a set of candidate parameters, the internal state was obtained by integrating point by point according to the dynamic sampling interval, and then the predicted block-level recoverable loss was calculated by mapping the mechanical state to optical loss and the observation weights of each data block. The weighted sum of squares of the differences between the measured block-level loss and the predicted block-level recoverable loss in each recoverable experiment was used as the fitting objective, and the parameters were obtained using the bounded nonlinear least squares method. The upper and lower limits of the search for each parameter were determined based on the maximum state amplitude, loading and unloading slopes, and recovery time after the perturbation stopped in a single perturbation experiment.
[0079] When multiple sets of parameters with similar fitting errors exist, the set with the smallest independent validation set error, whose internal state does not increase on its own after the perturbation stops, whose recovery direction does not reverse, and whose parameter amplitude does not increase excessively is selected sequentially. If no parameter set simultaneously meets the above conditions, or if the independent validation set error exceeds the predetermined recoverable contribution allowable error, the recoverable model calibration at that location is considered to have failed, and the continuous loss formation interval judgment is not enabled. During numerical calculation, the internal state at the previous sampling time is used as the initial value at the next sampling time; if the sampling interval is abnormal or the continuous missing time exceeds the allowable duration, the recursion of that segment of state is stopped.
[0080] The internal state output by the model cannot be directly regarded as optical loss. Instead, it should be converted into recoverable additional loss through the mapping relationship obtained from the target fiber segment or equivalent calibration segment. In one specific calibration method, the sampling times at each sampling point in all recoverable tests are... Paired with synchronous independent loss values, according to The system is segmented from smallest to largest, and the representative loss of each segment is calculated. When adjacent segments show a decrease in loss as the internal state increases, the corresponding segments are merged until a non-negative monotonic sequence is formed. Then, piecewise linear interpolation is used to obtain a mapping from the internal state to the recoverable loss. This mapping satisfies that zero internal state corresponds to zero recoverable additional loss, and within the calibration range, the recoverable additional loss does not decrease as the amplitude of the internal state increases. When the running data exceeds the calibrated internal state range, no unconstrained extrapolation of the mapping is performed.
[0081] The output of a longitudinal loss data block is the result of accumulating multiple echoes or frames of data over a period of time, while the dynamic disturbance path corresponds to a state that changes over time. Therefore, recoverable additional loss cannot be directly compared with the block-level loss using the instantaneous value at the center of the data block.
[0082] For the For each longitudinal loss data block, calculate the recoverable loss contribution within the block according to the actual time range and weighting relationship used in that block:
[0083] ,
[0084] in, Indicates the first The weights of observations at each time point within each data block. This represents the recoverable loss contribution within the data block that can be explained by the loading, unloading, and recovery processes of dynamic disturbances. The observation weights are non-negative normalized weights, taking the value of zero outside the corresponding data block, and the sum of all effective weights is 1. It has the same dimensions as the block-level local loss output by the longitudinal loss device.
[0085] When a uniform average is used for data blocks, the observation weight is equal to the reciprocal of the data block length. When the equipment uses different weights for different echoes or different times, the weights are converted according to the actual weights of the equipment. If all valid samples of a data block are missing, or if all samples are deemed invalid due to coherent fading, receiver saturation, or other reasons, then that data block will not participate in subsequent inversion.
[0086] This embodiment employs a processing method that first calculates the local loss for each valid echo or each valid frame, and then forms a block-level loss according to the observation weights. Therefore, the block-level loss, baseline correction, and recoverable loss contribution are all at the same dB observation scale. If the equipment first accumulates in the linear optical power domain and then converts it to dB, the equivalent observation relationship of the equipment should be calibrated using undisturbed data and known additional loss data, and the baseline, recoverable loss, and sustained loss should be processed according to the same relationship. The linear power averaging result should not be directly substituted into the dB additive relationship of this embodiment.
[0087] Subsequently, from the The block-level residual loss is obtained by subtracting the baseline correction and recoverable loss contribution corresponding to each data block from the measured local loss of that data block. The baseline correction is calculated based on the local loss before the candidate disturbance occurred and the common drift during the monitoring period, according to the [missing information]. The block-level results are calculated with the same observation weights for each data block. Therefore, the dimensions and observation time range of the measured block-level loss, baseline correction, recoverable loss contribution, and remaining block-level loss remain consistent.
[0088] Block-level residual loss includes any newly formed persistent residual loss, as well as longitudinal loss measurement errors, baseline correction errors, dynamic path reconstruction errors, and mechanical state-to-optical loss mapping errors. These errors are treated as a unified block-level composite error and are not calculated repeatedly in other formulas.
[0089] like Figure 3 As shown, persistent loss may form during the disturbance or gradually during clamp slippage, sheath creep, or bending retention after the disturbance ends. The system sets a possible time range for the formation of persistent states for each candidate disturbance, and calculates the proportion of persistent states within a block based on the overlap between this time range and the longitudinal loss data block. After deducting the baseline correction and block-level recoverable loss from the block-level longitudinal loss, the system forms the block-level residual loss used for inversion. Using the proportion of persistent states within a block as the observation coefficient, the system calculates the persistent residual loss increment corresponding to each candidate disturbance under the condition that the incremental persistent residual loss is not less than zero, and sums this increment with the newly added stable persistent loss verification increment obtained in subsequent stable reference periods.
[0090] Persistent losses may develop rapidly during the disturbance or gradually during clamp slippage, sheath creep, or bending retention after the disturbance ends. The system sets a time range for the possible formation of the persistent state for each candidate disturbance and uses... Indicates the first The process of forming a persistent state corresponding to each candidate perturbation.
[0091] The starting point of the candidate formation time range is no earlier than the start time of the corresponding disturbance, and the ending point is no later than the creep observation cutoff time after the end of the disturbance; the creep observation duration is predetermined based on the longest effective time required for the mechanical state and additional losses in all recoverable disturbance tests to recover to the baseline error range. The system enumerates the candidate formation start and end points according to the sampling time of the dynamic disturbance data, and does not arbitrarily set the formation time period outside the disturbance cluster.
[0092] The value of is between 0 and 1 and remains constant over time. It is 0 when the persistent state has not yet formed, 1 after it is fully formed, and a value between 0 and 1 during gradual formation. The formation process can adopt a step-like, linear, or monotonic variation form predetermined through independent displacement and loss tests. The formation form and its optional parameters are determined before processing the damage data and are not temporarily adjusted based on the desired disturbance range. When the candidate formation start and end points are at the same sampling time, a step-like formation form is used, and time difference division is not performed.
[0093] No. The candidate persistent state is in the... The effective proportion in each data block is represented as follows:
[0094] ,
[0095] Since the observation weights have been normalized and the persistent state formation process is between 0 and 1, therefore It also lies between 0 and 1. This occurs when the persistent state was fully formed before this data block. The value equals 1; when the persistent state is formed after this data block, Equals 0; when a persistent state is formed within a data block, It lies between 0 and 1.
[0096] Location The persistent residual loss at a given point is formed by the sequential superposition of persistent residual loss increments generated by candidate perturbations. Each increment transitions from an unformed state to a fully formed state according to the corresponding persistent state formation process. During monitoring periods without maintenance, clamp release, manual reset, or fiber optic cable re-laying, the newly added persistent residual loss increment is not less than zero. When the above operations occur, the system terminates the current inversion and re-establishes the loss baseline after the operation is completed, without using the method of canceling out positive and negative increments to continue the original judgment.
[0097] Let the range of continuous candidate perturbations currently being tested be denoted as the i-th The second disturbance to the first The relationship between each residual increment and the block-level residual loss within the candidate range after the second perturbation is expressed as follows:
[0098] ,
[0099] in, Indicates the first The incremental residual loss corresponding to the sub-disturbance or indistinguishable disturbance group. Indicates the first The block-level synthesis error of each data block. The incremental residual loss outside the candidate range is taken as zero in this solution.
[0100] The system utilizes multiple consecutive data blocks to jointly invert the persistent residual loss increment corresponding to each candidate perturbation. Specifically, a non-negative constrained least squares method can be used to minimize the weighted squared difference between the measured residual loss of each data block and the predicted residual loss calculated from the block's proportion and residual increment. The weights of each data block are determined based on the block-level comprehensive error obtained from unperturbed experiments and independent verification experiments. The solution results should also satisfy conditions such as residual increment not being less than zero, being persistent after formation, and not exceeding the calibration range.
[0101] Whether a candidate range can explain block-level observations is determined based on the standardized residuals of each valid data block. The standardized residual is the multiple of the difference between the measured and predicted block-level residual losses relative to the overall standard deviation of that data block. The residual threshold is determined by independent recoverable experiments and known persistent incremental validation experiments, ensuring that validation data without persistent residuals are not misclassified extensively, while data with known persistent residuals can pass the test. If the standardized residuals of all valid data blocks do not exceed this threshold, the candidate range is considered to meet the block-level residual interpretation criteria.
[0102] Once a persistent state is established, if the remaining block-level loss of at least two subsequent valid data blocks does not return to within the persistent loss detection limit, and the new loss obtained during the subsequent stable reference period is not lower than the detection limit, then the persistent state is considered to satisfy the cross-block persistence condition. If there are fewer than two valid data blocks, confirmation is not made solely based on cross-block persistence; instead, subsequent data is awaited, or only the persistent loss is output for confirmation.
[0103] When the number of valid data blocks is less than the number of residual increment groups to be solved separately, each increment is not solved individually. Instead, candidate perturbations that are temporally adjacent and have similar proportions are first merged. For two candidate perturbations, their proportion sequences in all valid data blocks are compared. When the difference between the two proportion sequences does not exceed the minimum distinguishable difference determined by independent verification experiments, they are considered indistinguishable at the current observation scale. The minimum distinguishable difference is obtained by repeatedly changing the formation time and adding block-level errors equivalent to actual measurements, so that the single perturbation identification result when the proportion difference is less than this value is no longer stable. At this time, instead of forcibly obtaining a single result by adding constraints, they are merged into an indistinguishable perturbation group, and only the total incremental residual loss of this perturbation group is solved.
[0104] After the last in-situ disturbance, the system selects a subsequent reference period where recoverable losses have sufficiently decayed, there are no new in-situ disturbances, no repair resets, and the common drift reference segment is stable. The data obtained from this reference period... This represents the additional stable and persistent loss relative to the baseline state before the candidate disturbance occurred, rather than the absolute value of all historical losses at that location.
[0105] The sum of the incremental residual losses obtained from the inversion satisfies:
[0106] ,
[0107] in, This represents the allowable error for total closure. Specifically, it is obtained by calculating the standard deviation of subsequent stable loss, the standard deviation of baseline correction, and the standard deviation of residual increment inversion based on subsequent stable reference periods, independent baseline validation data, and known continuous increment validation data. The square root of the sum of the squares of these three standard deviations is then multiplied by a predetermined closure confidence coefficient to obtain the allowable error for total closure. The closure confidence coefficient is determined based on the false pass rate without continuous residual data and the correct pass rate with known continuous residual data. The data is fixed before processing the data that could cause damage; it is not set to zero, nor is it relaxed afterward based on the final output.
[0108] when If the loss falls below the pre-determined persistent loss detection limit based on unperturbed data, the system cannot confirm that persistent loss has formed, nor will it continue to output a perturbation range. At this point, the unexplained block-level residual loss record is retained, awaiting further confirmation from subsequent data.
[0109] Subsequent stable reference periods are only used to confirm the total amount of persistent loss, and are not used as the time when persistent loss is formed. The formation period is still determined in reverse based on the previously recorded dynamic disturbance paths, the proportion of persistent states within the block, and the loss results of multiple blocks.
[0110] like Figure 4 As shown, the state transition interval output unit first checks a relatively wide range of continuous candidate disturbances. A candidate range must simultaneously satisfy the following conditions: block-level residual loss is explainable, the incremental residual loss is not less than zero, the formed state is maintained across data blocks, and the sum of residual increments and subsequent stable increments are closed. After a candidate range passes the check, disturbances at the beginning or end of the range are sequentially deleted and rechecked. If the above conditions are still met after deletion, the candidate range is further narrowed. If all conditions cannot be met after deleting any end disturbance, the current range is determined as the state transition associated disturbance interval. After a candidate range passes the check, the system sequentially attempts to delete disturbances at the beginning or end of the range. If the above conditions are still met after deletion, the candidate range is further narrowed. If all conditions cannot be met after deleting any end disturbance, the current range is determined as the minimum state transition associated disturbance interval corresponding to the first formation of persistent loss.
[0111] When multiple minimum candidate ranges cannot be distinguished under the existing error level, the output is the smallest continuous union covering these candidate ranges, without forcibly specifying a single perturbation based on the perturbation amplitude, the number of occurrences, or the distance from the first time the threshold is exceeded.
[0112] When the dynamic disturbance path exceeds the calibration range of the recoverable model, there is a lack of effective data blocks, subsequent stable loss increments cannot be obtained, or there is no identifiability among candidate disturbances, the system outputs the minimum identifiable continuous disturbance interval, respectively: model exceeds applicable range, data is insufficient, continuous loss has been confirmed but the range of formation cannot be determined, or the disturbance is expanded.
[0113] Through the above processing, the longitudinal loss data block retains its original cumulative length, and the reliability of detecting minute persistent losses does not decrease due to the increased temporal resolution. High temporal resolution dynamic perturbation data is used to reconstruct the time process of recoverable losses within the data block. The conversion to the same observation scale is used to subtract the recoverable contribution from the block-level loss. The proportion of persistent states within the block is used to represent the actual degree of formation of persistent losses within the data block. Subsequent stable losses are used to close the historical inversion total. Finally, the minimum state transformation associated perturbation interval is determined in reverse from the already averaged and mixed block-level losses.
[0114] Example 3
[0115] This embodiment illustrates the methods for obtaining parameters, freezing parameters, system operation, physical test verification, and handling insufficient data.
[0116] The test setup includes a single-mode communication optical fiber, a controllable bending or pulling mechanism, an optical fiber fixing clamp, a displacement or bending radius measuring device, a dynamic disturbance acquisition channel, a longitudinal loss acquisition channel, an independent optical loss measurement channel, and a unified clock.
[0117] In one test configuration, a single-mode optical fiber with a length of 1km to 2km is used, with a target test section of at least 2m reserved in the middle. The remaining fiber is coiled and fixed, free from external forces. One end of the target test section is fixed to a base fixture, and the other end is connected to a servo linear actuator. A cylindrical bending member with an interchangeable radius is passed around the middle of the target section to create a controllable bend. The actuator's displacement is recorded by an encoder. Positioning marks or displacement sensors are placed on both sides of the bending member to determine whether the fiber returns to its original position after unloading. An independent displacement sensor is placed near the fixture to detect fixture slippage. The actuator, encoder, displacement sensors, and monitoring equipment are all sampled starting from the same trigger signal.
[0118] The dynamic disturbance acquisition channel and longitudinal loss acquisition channel adopt the monitoring optical path described in Example 1, with the target test section located at a known distance along the fiber. An optical power detection branch is set up at the far end of the optical fiber in the monitoring wavelength channel to record the average pulse energy transmitted through the target test section. Since the optical fiber, connectors, and fixtures, except for the target test section, remain fixed during the test, the change in the far-end transmitted power relative to the pre-disturbance reference can be used as an independent measurement of the total additional loss of the target test section. A calibration OTDR is used before and after each test to verify whether new local loss events have occurred at the target location. Neither the optical power detection branch nor the calibration OTDR uses the recoverable contribution or state transition interval retrieved by this system as input.
[0119] Equipment selection should be based on its ability to cover the magnitude to be verified. The minimum step size of the displacement actuator should be less than the mechanical state difference between recoverable and persistent losses, and the resolution of the displacement sensor should not exceed half of this minimum step size. The repeatability error of the independent optical power detection channel should be less than one-third of the persistent loss to be verified. The standard deviation of the longitudinal loss data block under undisturbed conditions should be lower than the allowable error of the persistent loss to be verified. For the 0.120dB persistent loss calculation magnitude in this embodiment, an independent loss detection channel with repeatability better than 0.02dB can be selected, and the blank standard deviation of the longitudinal loss data block should be controlled within 0.03dB. The above values are used to illustrate that the equipment performance should match the target loss magnitude and are not limited to the only selection.
[0120] The dynamic disturbance acquisition channel records the mechanical state changes of the test section, which can employ distributed phase changes, local strain, curvature, or other mechanical state quantities that can preserve the loading and unloading directions. The longitudinal loss acquisition channel outputs the block-level local loss of the test section according to a preset data block length. The independent optical loss measurement channel can employ an optical power meter, a calibrated OTDR, or other loss measurement devices that do not rely on the inversion results of this system.
[0121] Before the formal test, dynamic data, longitudinal loss data, monitored optical power, equipment temperature, and data from an undisturbed reference fiber segment were continuously collected without any artificial disturbance. This data was used to determine the dynamic quiet range, longitudinal loss baseline, common drift correction relationship, loss detection limit, and various error ranges.
[0122] Subsequently, multiple sets of fully recoverable single disturbances were applied. Each set of disturbances covered different directions, amplitudes, durations, and unloading speeds. After each set of disturbances, the mechanical position and independent losses were allowed to return to their pre-disturbance error range. Tests meeting the full recovery criteria were used to obtain the mapping relationship between the mechanical state change path and recoverable additional losses.
[0123] Multiple sets of short-interval, repetitive perturbations are then applied. Each subsequent perturbation begins before the recovery process of the previous perturbation has finished. This set of experiments is used to determine the state inheritance relationship between adjacent perturbations and to verify that the recoverable state cannot be reset to zero at the beginning of each perturbation.
[0124] Calibration and validation data are separated according to the complete test number. Data from the same test cannot be used for both parameter acquisition and independent validation simultaneously. Once the recoverable model can control the loss recovery process after disturbance within a preset error range on independent validation data, it generates a fixed parameter version.
[0125] The fixed parameter version should at least record the dynamic state zero point, recoverable path parameters and their search range, mechanical state to optical loss mapping relationship, baseline correction parameters and applicable range, dynamic positioning error, loss positioning error, registration tolerance, longitudinal loss detection limit, block-level integrated error, standardized residual threshold, total closure allowable error, minimum resolvable percentage difference, allowable dynamic path range, creep observation duration, and candidate forms of continuous state formation process.
[0126] When processing data that may cause ongoing losses, do not use the damaging data to modify the recoverable model. When equipment is replaced, the fiber optic cable fixing method is changed, or the long-term operating status changes significantly, create a new version of the parameters and re-perform independent verification.
[0127] In the sustained loss formation test, several fully recoverable disturbances are first applied to the same location. Then, by increasing the bending displacement, maintaining the local bending state, or causing measurable slippage in the fixture, one or several consecutive disturbances are used to form sustained local loss. After sustained loss is formed, at least one non-damaging disturbance within the recoverable range is applied to verify whether the system can eliminate irrelevant disturbances after formation.
[0128] The true state transition range cannot be labeled by the system itself. Its determination simultaneously considers independent displacement, bending radius, fixture position, and independent loss results. When the mechanical position initially cannot be recovered, and the independent loss remains above the pre-disturbance baseline during the subsequent stable period, the time range from the initial maintenance of the mechanical state to the formation of stable loss is taken as the true state transition range.
[0129] If different independent measurement channels cannot determine a unique formation time, the smallest continuous interval covering the boundaries of each independent measurement is used as the true range, and the instantaneous time of any single sensor is not used as the absolute true value.
[0130] During system operation, the system first reads the verified fixed parameter version, and then continuously acquires dynamic disturbance data and longitudinal loss data. Based on a unified timestamp and fiber-along distance coordinates, the system groups the two types of data into the same local fiber segment.
[0131] For disturbances occurring consecutively at the same location, the system continuously reconstructs recoverable additional losses along the actual path, and then obtains the block-level recoverable loss contribution according to the observation process of each longitudinal loss data block. After subtracting the baseline and recoverable contribution from the measured block-level loss, the remaining block-level loss is formed.
[0132] The system sets up a persistent state formation process for continuous candidate disturbances, calculates the effective proportion of each candidate state in each data block, and inverts the persistent residual loss increment under the condition that the residual increment is not less than zero. After obtaining the subsequent stable loss increment, the inversion results are then closed by total closure.
[0133] By using block-level interpretation, cross-block persistence, and total closure checks to narrow down the candidate range from the front end and back end until end perturbations can no longer be removed, the minimum state transition associated perturbation range is obtained.
[0134] Each experiment generates at least one verifiable processing record. This record includes the start and end times of the disturbance event, the target position along the fiber, actuator displacement, fixture slippage, equivalent mechanical state, the number of effective frames and measured local loss for each longitudinal loss data block, recoverable loss contribution converted with observation weights, block-level residual loss, the block proportion of each candidate formation process, the persistent residual increment obtained from the inversion, subsequent stable loss increments, and the final output range. Calibration experiments, verification experiments, and experiments that may cause persistent loss use different experiment numbers. The processing program reads a fixed parameter version based on the experiment number and does not allow data from verification experiments and loss-causing experiments to be backfilled into the original calibration data.
[0135] To illustrate the calculation process of the proportion of persistent state within a block, a set of physical constraint calculation data is set. The total calculation time is 60s, and the length of the longitudinal loss data block is 10s. The third perturbation occurs from 23.0s to 25.0s, the persistent state begins to form from 24.0s and is fully formed by 27.0s, and the increment of persistent residual loss is set to 0.120dB.
[0136] For data blocks ranging from 20s to 30s, the persistent state forms linearly between 24.0s and 27.0s, and remains fully formed between 27.0s and 30.0s. Within this data block, the linear formation phase is equivalent to a 1.5s full formation time, and the full formation phase is 3.0s; therefore, the persistent state accounts for 0.45%. The block-level contribution from persistent residual loss in this data block is 0.054dB.
[0137] The above values are calculated based on the defined formation process and data block boundaries, and are used to illustrate the proportion conversion and inversion process, not as the accuracy of field measurements. If the persistent state within the 30s to 40s data block has been fully formed and the recoverable contribution has been deducted, then the persistent residual loss contribution corresponding to this data block is 0.120dB. Using the proportion relationship between the preceding and following data blocks and the subsequent stable total, it can be confirmed in reverse that the 0.120dB residual increment was formed within the time range corresponding to the third disturbance.
[0138] In one dataset, the total computation time was 60 seconds, with a sampling interval of 0.5 seconds, divided into six 10-second longitudinal loss data blocks. The first and second perturbations only generated recoverable losses. The third perturbation occurred between 23.0 and 25.0 seconds and triggered persistent residual losses. The fourth perturbation occurred after the persistent state was formed and did not increase persistent losses. The persistent state gradually formed between 24.0 and 27.0 seconds, and the actual block size of the third data block was 0.45%.
[0139] Using the output of the mechanical path to recoverable loss mapping as a known input, calculations are performed according to the non-negative inversion and subsequent total closure of this embodiment. The persistent residual increments of the 1st, 2nd, and 4th perturbations are all 0, and the calculated persistent residual increment of the 3rd perturbation is 0.120202dB, with an absolute error of 0.000202dB relative to the set true value of 0.120000dB. The inversion total is 0.120202dB, the minimum effective candidate range is the 3rd perturbation, and the output time range is 23.0s to 25.0s. The 4th non-damaging perturbation is excluded.
[0140] When the location difference between dynamic disturbance and longitudinal loss exceeds the registration tolerance, the system does not establish a candidate relationship between them. When the magnitude, direction, duration, or adjacent interval of the dynamic path exceeds the calibration range, unconstrained extrapolation of recoverable loss is not performed.
[0141] When new disturbances, monitoring optical path switching, receiver gain changes, or fixture adjustments occur continuously during the subsequent reference period, and reliable subsequent stable loss increments cannot be obtained, the system suspends the current interval judgment.
[0142] When multiple candidate perturbations form the same or indistinguishable proportions within different data blocks, the system outputs the minimum continuous range covering these perturbations, instead of outputting a single perturbation. Only when the newly added stable loss is not lower than the detection limit, the recoverable model is independently validated, and the candidate range satisfies the total closure requirement and is identifiable, will the minimum state transition associated perturbation interval be output.
[0143] Each test and field monitoring session saves the original dynamic data, original longitudinal loss data, data block boundaries, observation weights, fiber-along location, baseline data, independent displacement data, independent loss data, equipment status, parameter version, candidate range, exclusion reasons, and final output results. The original data and processing results are saved separately; parameter updates do not overwrite old parameter versions, nor do they alter judgments already made in the old versions.
[0144] Comparative Example 1
[0145] This comparative example uses the same dynamic perturbation sequence, longitudinal loss data block length, and persistent residual loss increment as in Example 3, but does not reconstruct or deduct recoverable loss contributions, nor does it calculate the effective proportion of persistent states within the data block.
[0146] The comparative example uses the data block that reliably exceeds the loss detection limit for the first time as the data block for the formation of continuous loss, and takes the disturbance closest to the detection time before the data block as the main candidate disturbance.
[0147] In the physical constraint calculation described in Example 3, the persistent state begins to form at 24.0s and is fully formed by 27.0s, with a persistent residual loss increment of 0.120dB. The persistent residual loss contribution within the data block from 20s to 30s is only 0.054dB.
[0148] When the loss is 0.054dB, which is below the reliable detection limit determined by the actual equipment for a single data block, the 20s to 30s data block cannot reliably confirm persistent loss. By the 30s to 40s data block, the persistent state has been fully formed, and the block-level persistent loss contribution reaches 0.120dB, at which point the comparative scale first confirms persistent loss.
[0149] Therefore, the first reliable detection time obtained from the comparative model is after 30 seconds, which is later than the actual formation range of the sustained state between 24.0 seconds and 27.0 seconds.
[0150] When another non-damaging disturbance occurs between 28.0s and 29.0s, this disturbance is located before the first reliable detection data block and is closer to the time of the first reliable detection. When using the most recent disturbance, disturbance amplitude sorting, or fixed backtracking window, it is easy to include this non-damaging disturbance in the main candidate range, or even mistakenly identify it as a disturbance that causes continuous loss.
[0151] Example 3 maintains the same 10s longitudinal loss data block without reducing the cumulative sample size. It reconstructs the data using dynamic perturbation paths and subtracts the recoverable loss caused by the perturbation from 28.0s to 29.0s. Then, it explains the 0.054dB block-level contribution based on the proportion of 0.45 in the data block of the continuous state from 20s to 30s, and completes the total closure with a subsequent stable continuous loss increment of 0.120dB.
[0152] Therefore, without shortening the longitudinal loss data block, Example 3 can shrink the formation range of persistent loss to the 23.0s to 25.0s range corresponding to the third disturbance, and eliminate non-damaging disturbances after formation. The comparative example can only obtain the first measurable data block and its nearby disturbances, and cannot recover the initial formation range of persistent loss from the already averaged and mixed block-level loss.
[0153] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.
Claims
1. A new generation of fiber optic link convergence monitoring system for mobile communication base station equipment, characterized in that, include: The data acquisition module is used to acquire the dynamic disturbance timing, block-level longitudinal loss, and time range and time weight of each longitudinal loss data block of the same local fiber segment of the fiber optic link of the in-service base station. The loss decomposition module is used to reconstruct the recoverable state of continuous evolution between adjacent disturbances based on the continuous disturbance path and the calibrated and independently verified continuous path state model. The recoverable state of the local fiber segment is converted into recoverable additional loss through the mapping relationship between the recoverable state and optical loss. The loss is then integrated according to the time range and time weight of the corresponding longitudinal loss data block to obtain the block-level recoverable loss. A persistent loss formation process is set for candidate disturbances. The proportion of persistent state within the block is determined according to the overlap relationship between the formation process and each longitudinal loss data block. The loss baseline and block-level recoverable loss are deducted from the block-level longitudinal loss to obtain the block-level loss residual. Using the proportion as the observation coefficient, the persistent residual loss increment corresponding to each candidate disturbance is inverted under the condition that the persistent residual loss increment is not less than zero. The sum of the increments is constrained by the total persistent loss determined by the subsequent stable reference period. The mapping relationship between the recoverable state and optical loss is calibrated through recoverable perturbation tests, satisfying that zero recoverable state corresponds to zero recoverable additional loss, and within the calibration range, the recoverable additional loss does not decrease when the recoverable state amplitude increases; the time weight is the actual weight of each effective frame or each effective sampling time in the corresponding longitudinal loss data block to the block-level longitudinal loss, and the loss decomposition module performs a weighted integral of the recoverable additional loss according to the normalized time weight to obtain the block-level recoverable loss of the corresponding longitudinal loss data block; The interval determination module is used to determine the state transition associated perturbation interval that cannot be further reduced by excluding boundary perturbations from the continuous candidate perturbation ranges that meet the conditions of residual interpretation, cross-data block continuity, total closure and distinguishable percentage sequence, and when multiple candidate ranges cannot be distinguished, it outputs the smallest continuous range covering the multiple candidate ranges. The interval determination module merges candidate perturbations whose intra-block persistent state proportion sequence difference is lower than a preset minimum distinguishable difference into an indistinguishable perturbation group, and determines the range of continuous candidate perturbations that simultaneously meets the following conditions as the effective candidate perturbation range: the difference between the block-level loss residual reconstructed from the inversion result and the measured block-level loss residual does not exceed a preset residual threshold; the block-level loss residual of at least two effective longitudinal loss data blocks after the persistent state is formed is not lower than the persistent loss detection limit, and the total persistent loss is not lower than the persistent loss detection limit; the difference between the sum of each persistent residual loss increment and the total persistent loss does not exceed a preset total closure allowable error; and the intra-block persistent state proportion sequence difference of each candidate perturbation or indistinguishable perturbation group after merging is not lower than the preset minimum distinguishable difference.
2. The fiber optic link fusion monitoring system for a new generation of mobile communication base station equipment according to claim 1, characterized in that, The loss decomposition module forms the continuous disturbance path according to the start and end times of each dynamic disturbance, continues the recovery process of the recoverable state between adjacent dynamic disturbances, and takes the recoverable state evolved from the end of the previous dynamic disturbance to the start of the next dynamic disturbance as the initial state of the next dynamic disturbance.
3. The fiber optic link fusion monitoring system for a new generation of mobile communication base station equipment according to claim 2, characterized in that, The loss decomposition module determines the mechanical state sequence based on the dynamic disturbance time sequence, divides the mechanical state sequence into a loading segment, an unloading segment, and a recovery segment between adjacent dynamic disturbances, and uses the recoverable state at the previous sampling time as the initial state value at the next sampling time, and continuously updates the recoverable state according to the time sequence of each segment.
4. The fiber optic link fusion monitoring system for a new generation of mobile communication base station equipment according to claim 1, characterized in that, The value of the continuous loss formation process is between zero and one and does not decrease over time. It is zero before the continuous loss is formed, one after the continuous loss is fully formed, and a value between zero and one during the gradual formation of continuous loss. The candidate formation start point of the continuous loss formation process is no earlier than the start time of the corresponding candidate disturbance, and the candidate formation end point is no later than the creep observation cutoff time after the end of the corresponding candidate disturbance.
5. The fiber optic link fusion monitoring system for a new generation of mobile communication base station equipment according to claim 4, characterized in that, The loss decomposition module performs normalized weighted integration on the state values of each persistent loss formation process according to the time range and time weight of each longitudinal loss data block, and determines the integration result as the proportion of the intra-block persistent state of each candidate disturbance in the corresponding longitudinal loss data block.
6. The fiber optic link fusion monitoring system for a new generation of mobile communication base station equipment according to claim 5, characterized in that, The loss decomposition module uses the proportion of the intra-block persistent state of each candidate disturbance in each longitudinal loss data block as equation coefficients to establish a set of inversion equations between multiple block-level loss residuals and the persistent residual loss increments corresponding to each candidate disturbance, and solves the set of inversion equations under the constraint that the persistent residual loss increments do not exceed the preset residual loss calibration range.
7. The fiber optic link fusion monitoring system for a new generation of mobile communication base station equipment according to claim 6, characterized in that, The subsequent stable reference period meets the following conditions: the recoverable additional loss has been attenuated to the preset error range, there are no new in-situ dynamic disturbances, there are no maintenance and reset operations, and the common drift reference data is stable. The loss decomposition module determines the total continuous loss as the newly added stable continuous loss relative to the baseline state before the candidate disturbance occurred within the subsequent stable reference period, and excludes candidate disturbance combinations whose difference between the sum of continuous residual loss increments and the total continuous loss exceeds the preset total closure allowable error. The subsequent stable reference period is not used to determine the formation time of continuous residual loss.
8. The fiber optic link fusion monitoring system for a new generation of mobile communication base station equipment according to claim 1, characterized in that, The interval determination module successively eliminates boundary disturbances from the start and end of the effective candidate disturbance range. After each elimination, the determination conditions of the effective candidate disturbance range are re-examined. If the determination conditions are still met after elimination, the reduced range is retained and elimination continues until both eliminating boundary disturbances from the start and the end will cause the reduced range to no longer meet the determination conditions. The range after stopping elimination is then determined as the state transition associated disturbance interval.
9. A new generation mobile communication base station equipment fiber optic link fusion monitoring system according to claim 8, characterized in that, When there are two or more valid candidate disturbance ranges that both meet the determination conditions of the valid candidate disturbance range and cannot be distinguished from each other, the interval determination module outputs the minimum continuous range from the earliest starting disturbance to the latest ending disturbance in the valid candidate disturbance range, and determines the minimum continuous range as the state transition associated disturbance interval.
10. A new generation mobile communication base station equipment fiber optic link fusion monitoring system according to claim 3, characterized in that, The recoverable state is calculated using a Bouc-Wen type internal state model with relaxation terms. The internal state model takes the mechanical state sequence as input, and its state changes are jointly determined by the mechanical state change term, the loading and unloading path term, and the relaxation term after the disturbance stops. The model parameters are calibrated using recoverable disturbance test data and verified by independent validation data. The persistent residual loss is a state variable independent of the recoverable state and is accumulated according to the persistent residual loss increment corresponding to each candidate disturbance.
Citation Information
Patent Citations
Optical fiber loss dynamic prediction method and system based on free curve form
CN120471186A
Optical signal attenuation analysis-based optical fiber connector loss automatic detection method
CN122293186A
