Fiber connector mating position deviation detection method and system

Abstract

The application relates to the technical field of optical element testing, and discloses a fiber connector butt joint position deviation detection method and system. The method first performs physical butt joint on a standard reference jumper fiber, collects a static reference insertion loss value and a transient Fresnel reflection dithering waveform of a butt joint full cycle, extracts an end face structure scattering entropy, and constructs a multi-dimensional optical transmission characteristic reference library; then real-time transient waveforms of on-site butt joint are collected and real-time scattering entropy is calculated, feature matching is performed, decoupling branch determination of end face physical defects and mechanical geometric position deviation is completed, and a to-be-corrected optical link data packet is generated according to the position deviation; finally, a connection point characteristic signal is obtained through iterative deconvolution signal reconstruction, three-dimensional position deviation is analyzed, and a correction instruction is generated to perform closed-loop verification; the application effectively avoids end face damage caused by end face defect misjudgment, and realizes high-precision detection and adaptive correction of butt joint position deviation.

Description

Technical Field

[0001] This invention relates to the field of optical component testing technology, and more specifically, to a method and system for detecting misalignment of fiber optic connector mating positions. Background Technology

[0002] With the continuous expansion of power communication networks, fiber optic connectors, as core components of optical physical layer connections, directly determine the transmission quality of the link through their mating accuracy. In remote automated operation and maintenance and unattended optical distribution scenarios, real-time monitoring and precise optimization of connector mating status have become crucial for ensuring highly reliable transmission of power grid dispatch data.

[0003] Currently, numerous studies have been conducted in the industry regarding the inspection of fiber optic connectors. For example, Chinese patent application CN101852587A discloses a method and apparatus for inspecting the end face of fiber optic connectors. This technology utilizes the principle of interferometric imaging, reconstructing a three-dimensional topographic image of the fiber core region at the end face through coherent interference of reference light and object light combined with holographic denoising technology, thereby achieving precise measurement of static geometric parameters such as fiber core height and concentricity. Furthermore, Chinese patent application CN120894276A discloses a deep learning-based method for detecting defects in fiber optic connectors. This scheme introduces deep learning algorithms and a multimodal data fusion architecture, utilizing an improved YOLOv8 and GAN network to enhance the features of the end face image, achieving high-precision identification of defects such as scratches and dirt on the end face.

[0004] However, existing detection technologies mainly focus on static end-face topography mapping or defect classification of two-dimensional images. During the dynamic closure process of fiber optic connectors, they face accuracy bottlenecks caused by microscopic medium scattering interference and multidimensional deviation coupling. Since the insertion loss (IL) of a fiber optic connection is a scalar physical quantity affected by lateral misalignment, longitudinal gap, and end-face cleanliness, in the absence of transient evolution information, a single loss index constitutes a typical complex set of underdetermined equations. The system cannot effectively isolate the scattering noise caused by micro-dust contamination and the mode field mismatch caused by geometric position deviations from the physical layer solely through scalar feedback. This deep confusion of physical characteristics often leads to certain decision-making errors in automated operation and maintenance systems: when there is micron-level hard dust on the end face, if the system misjudges it as a positional deviation and drives the mechanical mechanism to perform compensatory fine-tuning, the dust particles are likely to embed into the ceramic ferrule at the moment of mechanical contact, causing irreversible Hertzian contact stress damage; at the same time, due to the lack of transient entropy analysis of the air gap disappearance process in the entire docking cycle, the system is unable to decouple the three-dimensional spatial deviation vector in real time at the submicron scale, which may cause the calibration algorithm to fall into the "pseudo-peak trap" of local optima, making it impossible to achieve accurate docking and closed-loop verification of long-distance, high-reliability optical links. Summary of the Invention

[0005] To overcome the aforementioned shortcomings of existing technologies, this invention provides a method and system for detecting optical fiber connector mating position deviations. By extracting the transient Fresnel reflection jitter waveform throughout the entire cycle of the optical fiber mating operation, the physical characteristics of end-face defects and mechanical deviations are decoupled using the end-face structure scattering entropy. Based on iterative deconvolution signal reconstruction technology, this invention can extract a precise three-dimensional position deviation vector from aliased signals. While effectively avoiding frictional damage, it achieves high-precision, automated closed-loop calibration of optical fiber connectors, significantly improving the reliability of automated operation and maintenance systems and the service life of connectors.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Methods for detecting misalignment of fiber optic connector mating positions include:

[0008] Physically connect the standard reference jumper fiber, collect the static reference insertion loss value at the connection port and the transient Fresnel reflection jitter waveform throughout the entire connection operation cycle, extract the end face structure scattering entropy from the transient Fresnel reflection jitter waveform, and construct a multi-dimensional optical transmission feature reference library;

[0009] The system collects real-time transient Fresnel reflection jitter waveforms during the on-site docking process and calculates the real-time end-face structure scattering entropy. It then performs feature matching between the real-time end-face structure scattering entropy and a multi-dimensional optical transmission feature benchmark library, and performs decoupling branch determination of anomaly types, including end-face physical defects and mechanical geometric position deviations. For mechanical geometric position deviations, it generates optical link data packets to be corrected.

[0010] Based on the optical link data packets to be corrected, iterative deconvolution signal reconstruction is performed to obtain the connection point feature signal. The longitudinal gap deviation value and the lateral misalignment deviation vector are analyzed from the connection point feature signal. The longitudinal gap deviation value and the lateral misalignment deviation vector are summarized to generate a three-dimensional position correction command and perform closed-loop verification.

[0011] The method for extracting the scattering entropy of the end face structure includes:

[0012] The transient Fresnel reflection jitter waveform is decomposed into low-frequency trend components and high-frequency detail components by performing wavelet transform, and the information entropy of the high-frequency detail components is calculated to obtain the end face structure scattering entropy.

[0013] The method for constructing the multidimensional optical transmission feature reference library includes:

[0014] The static reference insertion loss value, transient Fresnel reflection jitter waveform, and end-face structure scattering entropy corresponding to each type of prefabricated standard sample in multiple types of prefabricated standard samples are combined into the feature vector of the corresponding category. The discrimination boundary threshold between each category is determined, and the feature vector of each category and its discrimination boundary threshold are stored as a multi-dimensional optical transmission feature reference library. The discrimination boundary threshold includes the scattering entropy cleanliness threshold, the scattering entropy damage threshold, and the insertion loss deviation threshold.

[0015] The various prefabricated standard samples include four connector types: standard connectors representing an ideal, unbiased state; connectors representing a purely mechanical positional deviation state; connectors representing a contaminated end face state; and connectors representing a damaged end face state.

[0016] The method for matching the real-time end-face structure scattering entropy with the multi-dimensional optical transmission feature reference library includes: reading the scattering entropy cleanliness threshold, scattering entropy damage threshold, and insertion loss deviation threshold from the multi-dimensional optical transmission feature reference library;

[0017] The first layer of discrimination is performed, including: if the real-time end-face structure scattering entropy is less than or equal to the scattering entropy cleanliness threshold, the end face is determined to be in a clean state and the second layer of cleanliness branch discrimination is entered; if the real-time end-face structure scattering entropy is greater than the scattering entropy cleanliness threshold, the end face is determined to have surface anomalies and the second layer of anomaly branch discrimination is entered.

[0018] The execution method for the second-layer clean branch discrimination is as follows: if the real-time insertion loss value is less than or equal to the insertion loss deviation threshold, the current docking state is determined to be an ideal no-deviation state; if the real-time insertion loss value is greater than the insertion loss deviation threshold, the current docking state is determined to be a pure mechanical position deviation state.

[0019] The execution method for the second-level abnormal branch discrimination is as follows: if the real-time end-face structure scattering entropy is less than or equal to the scattering entropy damage threshold, the current docking state is determined to be an end-face contamination state; if the real-time end-face structure scattering entropy is greater than the scattering entropy damage threshold, the current docking state is determined to be an end-face damage state; the scattering entropy cleanliness threshold is lower than the scattering entropy damage threshold.

[0020] The method for determining the decoupled branch of the execution exception type includes:

[0021] Collect the real-time attenuation curve of the current full optical path and the real-time insertion loss value at the current docking port;

[0022] If the current docking status is an ideal, deviation-free state, the docking is considered successful, and the docking inspection process ends. If the current docking status is an end-face contamination or end-face damage state, the current anomaly type is determined to be an end-face physical defect, an end-face cleaning or maintenance alarm is generated, and the subsequent position deviation correction process is terminated. If the current docking status is a purely mechanical position deviation state, the current anomaly type is determined to be a mechanical geometric position deviation, and the real-time insertion loss value and real-time attenuation curve are packaged to generate an optical link data packet to be corrected.

[0023] The method for obtaining the connection point feature signal includes:

[0024] Extract the real-time attenuation curve from the data packets of the optical link to be calibrated, locate the original waveform data of the reflection peak at the docking port in the real-time attenuation curve, and establish the measurement impulse response function;

[0025] The real-time insertion loss value is extracted from the data packets of the optical link to be calibrated. The robotic arm is controlled to perform micro-motion detection and record the change of insertion loss value at the docking port to calculate the optical power displacement sensitivity. Based on the measured impulse response function and constrained by the optical power displacement sensitivity, iterative deconvolution operation is performed on the original waveform data of the reflection peak at the connection point to obtain the connection point characteristic signal.

[0026] The method for resolving longitudinal gap deviation values ​​and lateral misalignment deviation vectors from connection point feature signals includes:

[0027] Perform spectral analysis on the characteristic signal of the connection point, detect the periodic interference modulation component, and calculate the longitudinal gap deviation value;

[0028] The theoretical additional loss corresponding to the longitudinal gap is calculated based on the longitudinal gap deviation value. The difference between the real-time insertion loss value and the static reference insertion loss value is calculated to obtain the excess loss. The theoretical additional loss is subtracted from the excess loss to obtain the remaining loss component. The lateral misalignment is calculated based on the remaining loss component. The lateral misalignment is decomposed into components in two orthogonal directions to obtain the lateral misalignment deviation vector.

[0029] A fiber optic connector mating position deviation detection system, used to implement the aforementioned fiber optic connector mating position deviation detection method, the system comprising:

[0030] Feature reference construction module: Perform physical connection on standard reference jumper, collect static reference insertion loss value at the connection port and transient Fresnel reflection jitter waveform throughout the connection action cycle, extract end face structure scattering entropy from transient Fresnel reflection jitter waveform, and construct a multi-dimensional optical transmission feature reference library;

[0031] Anomaly decoupling determination module: used to acquire real-time transient Fresnel reflection jitter waveforms during the on-site docking process and calculate the real-time end-face structure scattering entropy, match the real-time end-face structure scattering entropy with the multi-dimensional optical transmission feature reference library, and perform decoupling branch determination of anomaly types, including end-face physical defects and mechanical geometric position deviations, and generate optical link data packets to be corrected for mechanical geometric position deviations;

[0032] Deviation quantization correction module: Based on the optical link data packets to be corrected, iterative deconvolution signal reconstruction is performed to obtain the connection point feature signal. The longitudinal gap deviation value and the lateral misalignment deviation vector are analyzed from the connection point feature signal. The longitudinal gap deviation value and the lateral misalignment deviation vector are summarized to generate a three-dimensional position correction command and perform closed-loop verification.

[0033] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0034] This invention introduces the transient Fresnel reflection jitter waveform and end-face structure scattering entropy characteristics throughout the entire docking cycle to construct a multi-dimensional optical transmission feature benchmark library that includes static optical transmission characteristics and dynamic time-domain response. This overcomes the limitations of traditional fiber optic docking detection, which relies solely on static optical power thresholds. Based on multi-dimensional feature matching and comparison, it achieves accurate decoupling and judgment of two types of anomalies: end-face physical defects and mechanical geometric position deviations. This fundamentally avoids mechanical miscalibration caused by feature confusion between end-face contamination and position deviations, effectively preventing irreversible scratches on connector end faces and ineffective wear of mechanical components due to invalid calibration. Simultaneously, it eliminates the potential for intermittent bit errors during link operation despite passing static testing. Furthermore, through iterative deconvolution signal reconstruction of the optical link data packets to be calibrated, it achieves the identification of connection point characteristics... The super-resolution restoration of the characteristic signal breaks through the physical limitations of the spatial resolution of traditional optical time-domain reflectometers. It achieves independent analysis and precise quantization of longitudinal gap deviation value and lateral misalignment deviation vector, solving the problem of underdetermined solution of three-dimensional spatial deviation vector that cannot be solved by a single scalar loss value. This provides a clear deviation direction and quantitative basis for mechanical correction, avoiding the inefficiency and end-face wear problems caused by traditional blind trial-and-error adjustments. Finally, through the execution and closed-loop verification of three-dimensional position correction commands, a closed-loop feedback control of the entire process of detection, quantification, correction, and verification is formed, ensuring the convergence and accuracy of correction actions. This significantly improves the reliability, stability, and operational efficiency of automated connection and maintenance of fiber optic connectors, providing stable and reliable technical support for remote automated maintenance of power communication optical cables. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 A flowchart of the fiber optic connector mating position deviation detection method provided in an embodiment of the present invention;

[0037] Figure 2 This is a schematic diagram illustrating the influence of the end face state on the transient waveform provided in an embodiment of the present invention;

[0038] Figure 3 This is a schematic diagram of the end face state of four types of prefabricated standard samples provided in the embodiments of the present invention;

[0039] Figure 4 A flowchart illustrating the hierarchical discrimination logic provided in this embodiment of the invention;

[0040] Figure 5 A schematic diagram of longitudinal gap deviation provided in an embodiment of the present invention;

[0041] Figure 6 This is a functional block diagram of the fiber optic connector mating position deviation detection system provided in an embodiment of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Example 1:

[0044] Please see Figure 1 As shown, this embodiment provides a method for detecting misalignment of fiber optic connector mating positions, including:

[0045] Step S10: Perform physical connection on the standard reference jumper, collect the static reference insertion loss value at the connection port and the transient Fresnel reflection jitter waveform of the entire connection operation cycle, extract the end face structure scattering entropy from the transient Fresnel reflection jitter waveform, and construct a multi-dimensional optical transmission feature reference library.

[0046] Further, step S10 includes:

[0047] Step S11: Perform physical connection on the standard reference jumper, collect the static reference insertion loss value at the connection port, and collect the transient Fresnel reflection jitter waveform throughout the entire connection operation cycle;

[0048] A standard reference patch cord refers to a standard fiber optic patch cord connector that has undergone rigorous cleaning and is free of contaminants and scratches. Its end-face grinding precision meets industry standards, and the geometric tolerances of its ceramic ferrule are controlled within the micrometer range. It represents the optical transmission characteristics of a fiber optic connector under ideal physical conditions. Standard reference patch cords are used as the benchmark because traditional fiber optic splice quality testing methods typically rely solely on a single static optical power threshold. When insertion loss exceeds a preset range, the system cannot distinguish whether the increased loss is due to dirt or physical damage on the end face, or mechanical misalignment between the two fibers. These two anomalies exhibit similar loss increases in static optical power measurements. If the system misinterprets end-face dirt as a misalignment and triggers a robotic arm to perform position correction, it not only fails to eliminate the loss caused by dirt but also causes ineffective wear on mechanical components and may even exacerbate end-face damage due to repeated adjustments. The introduction of standard reference jumpers provides the system with a clear optical performance benchmark, enabling the subsequent acquisition of static reference insertion loss values ​​and transient Fresnel reflection jitter waveforms to accurately reflect the optical transmission characteristics under defect-free and bias-free conditions, thereby establishing a comparable behavioral standard in the time dimension.

[0049] The static reference insertion loss value is acquired by sending test commands to the intelligent optical distribution terminal (IPC) through the power optical cable network management platform. The adaptive fiber optic docking robotic arm in the IPC responds to the command and physically docks its built-in standard light source module with the standard reference jumper. After the docking is complete and the optical signal transmission enters a steady state, the optical power meter in the multi-functional test module measures the optical power at the docking port. The difference between this value and the transmitted optical power is the static reference insertion loss value. The static reference insertion loss value represents the optical energy attenuation level of the optical link under ideal physical conditions. Its magnitude is directly related to the contact state of the connector end face and the fiber core alignment accuracy. In subsequent real-time detection, it serves as a numerical reference zero point for measuring the degree of positional deviation. The purpose of acquiring the static reference insertion loss value is to establish a quantitative benchmark at the optical power level for the system. This allows the insertion loss value acquired during real-time detection to be calculated by subtracting from this benchmark, thereby determining the degree of deviation of the current docking state from the ideal state. This enables the system to transform abstract optical power measurement results into concrete quantitative deviation indicators, avoiding ambiguity caused by the lack of a reference benchmark and providing a starting point for subsequent positional deviation quantitative correction.

[0050] Acquiring the transient Fresnel reflection jitter waveform requires controlling an adaptive fiber optic docking robotic arm to perform a complete docking sequence on a standard reference jumper. This sequence sequentially includes four continuous processes: separation, approach, contact, and clamping stabilization. The start time of the acquisition time window is set to the point just before the ceramic ferrule of the fiber optic connector makes contact, and the end time is set to the point when the signal reaches a stable state after contact. Within this complete time window, the optical time domain reflectometer in the multi-functional test module continuously acquires return loss variation data. For example, the acquisition time window can be set from 50 milliseconds before contact to 500 milliseconds after contact, and the sampling frequency can be set to 10 kHz. The resulting return loss time sequence is the transient Fresnel reflection jitter waveform. The physical formation mechanism of the transient Fresnel reflection jitter waveform involves optical interference during the fiber end-face approach process: when the end faces of two fibers gradually approach but have not yet fully contacted, the air gap between the end faces forms a weak reflection cavity structure. The incident light signal undergoes Fresnel reflection at the two boundary surfaces of this cavity. The reflected light undergoes multiple round trips within the cavity and superimposes with each other, producing an interference effect. As the robotic arm drives the connector closer, the thickness of the air gap decreases, causing the optical path length of the cavity to change. This results in a periodic shift in the phase of the interference signal, manifesting as characteristic signal jitter in return loss measurements. See also Figure 2 This is a schematic diagram illustrating the influence of the end face state on the transient waveform provided in an embodiment of this application. Figure 2 The diagram illustrates two typical fiber optic connector end-face states: a clean end-face and a defective end-face, and the corresponding transient Fresnel reflection jitter waveforms for each state. The microscopic surface morphology of the end-face significantly affects the morphology of this jitter waveform: if the end-face is... Figure 2 The smooth, clean end face shown in the diagram reflects light primarily as specular reflection, resulting in a regular, periodic interference signal with a relatively smooth waveform, corresponding to the formation of... Figure 2 The smooth waveform shown; if the end face is Figure 2 The defective end face shown has dust or scratches. Incident light undergoes diffuse reflection and scattering on the microscopic irregular structure. These additional scattered components are superimposed on the interference signal, resulting in a waveform exhibiting disordered high-frequency spikes and sawtooth characteristics, corresponding to the formation of... Figure 2The high-frequency glitch waveform shown is illustrated. Acquiring transient Fresnel reflection jitter waveforms extends the docking process from static single-point measurement to dynamic, full-process monitoring. This allows the system to capture the optical response characteristics of the end-face microstructure at the moment of contact, characteristics that are unavailable in steady-state measurements after docking. This provides the system with information to distinguish between the physical and mechanical positional states of the end face. The high-frequency details in the transient waveform are directly related to the end-face microstructure, while the low-frequency trend characteristics primarily reflect energy changes caused by mechanical displacement. Furthermore, the introduction of this time-domain information enables subsequent anomaly type decoupling and discrimination. The system no longer relies on a single loss threshold for judgment but can identify the specific type of anomaly through differences in transient optical characteristics, avoiding the risk of misjudging end-face contamination as positional deviation and performing ineffective correction actions.

[0051] Step S12: Perform wavelet transform on the transient Fresnel reflection jitter waveform to decompose it into low-frequency trend components and high-frequency detail components, and calculate the information entropy of the high-frequency detail components to obtain the end face structure scattering entropy.

[0052] Wavelet transform is a signal analysis method with time-frequency localization characteristics, capable of extracting component features at different frequency scales while preserving signal temporal information. After inputting the transient Fresnel reflection jitter waveform into the wavelet transform processing module, the original signal is decomposed into two parts: a low-frequency trend component and a high-frequency detail component. The low-frequency trend component corresponds to the smooth envelope curve of the signal, reflecting the overall change trend of optical signal energy with mechanical displacement during the docking process. This trend is mainly determined by the macroscopic change in the fiber endface spacing and has a weak correlation with the irregularity of the endface microstructure. The high-frequency detail component corresponds to the rapidly changing wave components in the signal, reflecting the light scattering noise caused by the endface microstructure. When there are dust particles or scratches on the endface, the diffuse reflection and scattering of incident light on these irregular structures will superimpose high-frequency disturbances on the transient waveform, causing significant irregularities in the amplitude and distribution of the high-frequency detail component. The reason for using wavelet transform instead of Fourier transform is that Fourier transform completely converts the signal to the frequency domain, losing time information and failing to reflect the frequency characteristic changes at different moments during the docking process; while wavelet transform preserves the signal's time positioning capability through multi-scale decomposition, and can distinguish the frequency characteristic differences between the approach phase, the moment of contact, and the stable phase, so that subsequent entropy calculation can more accurately quantify the irregularity of the end face structure.

[0053] The calculation of the scattering entropy of the end-face structure is based on the information entropy value of the high-frequency detail components over the entire time window. Information entropy is a mathematical indicator that measures the randomness and uncertainty of a signal. Its calculation process is as follows: the amplitude distribution of the high-frequency detail components is divided into multiple discrete intervals, and the probability distribution of the signal falling into each interval is statistically analyzed. The information entropy is equal to the negative of the sum of the products of the probabilities of each interval and their logarithms. The magnitude of information entropy directly reflects the randomness of high-frequency detail components: if the end face is smooth and flat, the high-frequency detail components during the docking process mainly originate from the background noise of the measurement system, with amplitude distribution concentrated in a narrow band near zero, and a highly concentrated probability distribution, resulting in a low information entropy value. If there are dust particles attached to the end face, light is locally scattered on the particle surface, causing discrete spike disturbances in the high-frequency detail components, expanding the amplitude distribution to a wider range, and the probability distribution tends to be more dispersed, resulting in a higher information entropy value. If there are scratches or damage on the end face, the continuous irregular structure of the scratches leads to more intense and chaotic light scattering, causing continuous and disordered fluctuations in the high-frequency detail components, further expanding the amplitude distribution range, and making the probability of each interval tend to be more uniform, resulting in a higher information entropy value. This information entropy is defined as the end face structure scattering entropy, used to quantify the irregularity of the end face's microstructure. A higher value indicates a more irregular end face structure and a greater possibility of dirt or damage. Information entropy is used as the quantitative indicator of end-face structure instead of directly using the amplitude or variance of high-frequency components because: information entropy comprehensively considers the morphological characteristics of amplitude distribution rather than just focusing on numerical magnitude, enabling more accurate differentiation between random noise and structural scattering, and exhibiting stronger robustness to changes in measurement system gain, avoiding misjudgments caused by fluctuations in light source power or drift in detector sensitivity. End-face structure scattering entropy transforms the complex optical response of the end-face microstructure into a single numerical indicator, allowing subsequent threshold comparison and discrimination to be performed within a concise numerical space, reducing the computational complexity of real-time discrimination. The combined use of end-face structure scattering entropy and static reference insertion loss value allows the system to simultaneously discriminate on two independent feature dimensions: the insertion loss value reflects the degree of light energy attenuation, and the end-face structure scattering entropy reflects the irregularity of the end-face structure. Their synergistic discrimination significantly improves the accuracy of anomaly type differentiation.

[0054] Step S13: Perform the same data acquisition and processing procedures as steps S11 to S12 on multiple types of prefabricated standard samples. Combine the static reference insertion loss value, transient Fresnel reflection jitter waveform, and end-face structure scattering entropy corresponding to each type of prefabricated standard sample into a feature vector of the corresponding category. Determine the discrimination boundary threshold between each category. Store the feature vector of each category and its discrimination boundary threshold as a multi-dimensional optical transmission feature reference library. The discrimination boundary threshold includes the scattering entropy cleanliness threshold, the scattering entropy damage threshold, and the insertion loss deviation threshold.

[0055] See Figure 3This is a schematic diagram of the end face state of four types of prefabricated standard samples provided in the embodiments of this application. The multiple types of prefabricated standard samples include four connector types, namely... Figure 3 The diagram shows four connectors: a clean and well-aligned standard connector representing an ideal, unbiased state; a clean connector with a known lateral offset representing a purely mechanically misaligned state; a connector with end-face contamination (end-face covered with microparticles of known size); and a connector with end-face damage (end-face scratches). The clean and well-aligned standard connector representing an ideal, unbiased state has a rigorously cleaned end-face and is perfectly aligned with the fiber core. This type of sample exhibits the lowest static reference insertion loss, a smooth transient Fresnel reflection jitter waveform, and the lowest end-face structure scattering entropy. The clean connector representing a purely mechanically misaligned state with a known lateral offset also has a clean end-face but a known lateral misalignment with the fiber core. The static reference insertion loss of this type of sample increases due to the decreased coupling efficiency caused by the core misalignment, but because the end-face itself is clean, the transient Fresnel reflection jitter waveform remains relatively smooth, and the end-face structure scattering entropy is close to the ideal, unbiased state, falling within a low range. Connectors with end-face contamination, represented by connectors with known-sized microparticles attached to their end faces, are simulated by the artificial attachment of standard microparticles of a specific diameter to mimic actual contamination conditions during operation and maintenance. The static reference insertion loss of this type of sample increases due to particle obstruction and scattering, and discrete high-frequency disturbances appear in the transient Fresnel reflection jitter waveform. The end-face structure scattering entropy value increases to the medium range. Connectors with end-face damage, represented by connectors with scratches on their end faces, are simulated by artificially created scratches of a specific depth and length to mimic physical damage. The static reference insertion loss of this type of sample increases significantly due to scattering loss caused by the scratches, and the transient Fresnel reflection jitter waveform exhibits continuous, chaotic high-frequency fluctuations. The end-face structure scattering entropy value increases to the highest range. Four types of prefabricated standard samples are used to cover the main state types that may be encountered in actual operation and maintenance. The end faces of the ideal unbiased state and the pure mechanical position deviation state are both clean, and the end face structure scattering entropy values ​​of the two are in a similar low range. The difference between them depends on the difference in the insertion loss value. The end face contamination state and the end face damage state are both end face physical defects. The end face structure scattering entropy values ​​of the two are higher than those of the clean state, but the structure scattering entropy value of the damage state is higher than that of the contamination state. The difference between them depends on the further subdivision of the scattering entropy values.

[0056] The determination of the discrimination boundary thresholds employs statistical analysis methods to process the characteristic parameter distributions of various types of samples. For the end-face structure scattering entropy, the statistical upper limit of the scattering entropy for two types of clean samples—ideal unbiased state and purely mechanical position deviation state—is calculated as the scattering entropy cleanliness threshold. This threshold is used to distinguish whether the end face is clean. The scattering entropy distribution boundary for two types of samples—end-face contamination state and end-face damage state—is calculated as the scattering entropy damage threshold. This threshold is used to further distinguish between contamination and damage when the end face is not clean. For the insertion loss value, the statistical upper limit of the insertion loss for samples in the ideal unbiased state is calculated as the insertion loss deviation threshold. This threshold is used to determine whether position deviation exists when the end face is clean. The relative positions of the three thresholds in the numerical space are as follows: the scattering entropy cleanliness threshold is lower than the scattering entropy damage threshold. This relationship reflects that the scattering entropy of a clean end face is necessarily lower than that of an end face with contamination or damage. The insertion loss deviation threshold and the scattering entropy threshold are in different physical dimension spaces; the former measures the degree of optical power attenuation, while the latter measures waveform randomness. They independently play a discriminative role in their respective dimensions. The static reference insertion loss value, transient Fresnel reflection jitter waveform, and end-face structure scattering entropy corresponding to each category of samples are combined to form the feature vector of the corresponding category. Simultaneously, fiber physical parameters such as the mode field diameter of the single-mode fiber are stored together with the discrimination boundary threshold as a multi-dimensional optical transmission feature reference library. This reference library serves as the comparison basis in subsequent real-time docking status discrimination. The real-time acquired feature parameters are compared with the thresholds in the reference library, and the category to which the current docking status belongs is determined through hierarchical discrimination logic.

[0057] Step S10 introduces time-dimensional optical response features by constructing a multi-dimensional optical transmission feature benchmark library that includes transient time-domain characteristics. This builds upon traditional fiber optic splicing detection techniques that rely solely on static optical power thresholds, providing the system with the information basis to distinguish between end-face physical defects and mechanical positional deviations. The multi-dimensional optical transmission feature benchmark library contains both static benchmark insertion loss values ​​and end-face structure scattering entropy. The former reflects optical energy transmission efficiency, while the latter reflects the microstructural characteristics of the end face. These two features are physically independent yet complementary; neither can accurately distinguish anomaly types using either feature alone. Their combined use creates a synergistic enhancement effect, improving the reliability of the discrimination results. The multi-dimensional optical transmission feature benchmark library also includes fiber optic physical parameters such as the mode field diameter of single-mode fibers. These parameters provide the necessary physical constants to support the calculation of the theoretical additional loss of the longitudinal gap and the lateral misalignment in step S30. The introduction of transient Fresnel reflection jitter waveforms extends traditional steady-state single-point measurement to dynamic full-process monitoring. The optical response characteristics of the end-face microstructure generated at the moment of contact are fully captured; these characteristics are unavailable in steady-state measurements after docking. The system thus obtains information sources unavailable through traditional detection methods. The end-face structure scattering entropy, as the information entropy of high-frequency detail components, compresses the complex optical response of the end-face micromorphology into a single numerical index. This index is negatively correlated with the cleanliness of the end-face and positively correlated with the degree of end-face damage, enabling efficient threshold discrimination in a low-dimensional numerical space. The selection of four types of pre-fabricated standard samples covers the main state types in actual operation and maintenance, and the statistical determination method for the discrimination boundary thresholds ensures that the distinction boundaries between each category are statistically significant. The system can accurately identify anomaly types during real-time docking, avoiding the risk of misjudging end-face contamination as positional deviation and triggering mechanical correction actions. When physical defects exist on the end face, the system can promptly generate cleaning or maintenance alarms and terminate invalid correction processes, avoiding both unnecessary wear of mechanical components and preventing secondary problems such as repeated adjustments that may exacerbate end-face damage. When the anomaly is indeed due to positional deviation, the system can accurately trigger subsequent precise detection and correction processes, making mechanical correction actions clearly targeted and effective. The establishment of a multi-dimensional optical transmission characteristic benchmark library provides standardized discrimination criteria for the entire detection method, so that the detection results no longer depend on the operator's experience judgment or temporarily set thresholds. The detection process is repeatable and consistent, and detection tasks performed at different times and locations can be discriminated based on the same benchmark.

[0058] Step S20: Collect the real-time transient Fresnel reflection jitter waveform during the on-site docking process and calculate the real-time end-face structure scattering entropy. Perform feature matching between the real-time end-face structure scattering entropy and the multi-dimensional optical transmission feature reference library, and perform decoupling branch determination of anomaly types. The anomaly types include end-face physical defects and mechanical geometric position deviations. Generate optical link data packets to be corrected for mechanical geometric position deviations.

[0059] Further, step S20 includes:

[0060] Step S21: During the time window of the on-site fiber optic connector under test performing the docking operation, collect the real-time transient Fresnel reflection jitter waveform; at the same time, collect the real-time attenuation curve of the current full optical path and the real-time insertion loss value at the current docking port.

[0061] The fiber optic connector under test in the field refers to the fiber optic connector that needs to be connected in the actual operation and maintenance scenario of the power optical cable network. Unlike the standard reference patch cord used to establish the benchmark library in step S10, the end face state and connection position of the fiber optic connector under test in the field are unknown quantities to be tested. When the power optical cable network management platform issues a light path scheduling command, the adaptive fiber optic connection robotic arm in the intelligent optical distribution terminal responds to the command and performs a connection operation on the fiber optic connector under test in the field. The multi-functional test module simultaneously starts the data acquisition process while the robotic arm starts the connection action. The start and end times of the acquisition time window and the sampling frequency are completely consistent with the parameter settings when performing data acquisition on the standard reference patch cord in step S11, ensuring that the real-time acquired data has the same time scale and sampling density as the data in the benchmark library, thus ensuring the effectiveness of subsequent feature matching and comparison. The optical time domain reflectometer in the multi-functional test module continuously acquires the real-time return loss time series within the acquisition time window, which is the real-time transient Fresnel reflection jitter waveform. The physical formation mechanism of the real-time transient Fresnel reflection jitter waveform is exactly the same as that of the transient Fresnel reflection jitter waveform in step S11. Both originate from the Fresnel reflection and interference effects caused by the reflection cavity structure formed by the air gap during the approach of the fiber end face. When there are dust particles or scratches on the end face of the fiber connector under test, the diffuse reflection and scattering components generated by the incident light on the micro-irregular structure are superimposed on the interference signal, causing the real-time transient Fresnel reflection jitter waveform to exhibit high-frequency burrs and sawtooth characteristics that are different from those of a clean end face.

[0062] While acquiring the real-time transient Fresnel reflection jitter waveform, the optical time-domain reflectometer in the multi-functional test module also measures and records the real-time attenuation curve of the current entire optical path, and the optical power meter measures and records the real-time insertion loss value at the current docking port. The real-time attenuation curve is a distribution map of optical power attenuation measured by the optical time-domain reflectometer along the entire length of the optical fiber. The horizontal axis represents the optical fiber length position, and the vertical axis represents the optical power attenuation at that position. The docking port is represented by a reflection peak on the real-time attenuation curve, and the amplitude and shape of the reflection peak are related to the physical state of the connection point. The real-time insertion loss value is the difference between the optical power value measured by the optical power meter in steady state after docking and the transmitted optical power, reflecting the degree of optical energy attenuation in the current docking state. The purpose of acquiring the real-time attenuation curve is to provide the original waveform data of the connection point reflection peak for accurate quantification of the position deviation in step S30. If the current docking state is determined to be a purely mechanical position deviation state, the original waveform data of the connection point reflection peak in the real-time attenuation curve will be packaged into the optical link data packet to be corrected for subsequent iterative deconvolution signal reconstruction. The purpose of acquiring real-time insertion loss values ​​is to provide a basis for the determination of the second-layer clean branch in step S22, and to provide the total loss value for the analysis of lateral misalignment deviation in step S30. The synchronous acquisition of real-time attenuation curves and real-time insertion loss values ​​enables the system to obtain all the required optical measurement data in a single docking operation, avoiding end-face wear and position drift that may be caused by repeated docking operations, improving detection efficiency and ensuring the time consistency of data acquisition.

[0063] Step S22: Process the real-time transient Fresnel reflection jitter waveform, calculate the real-time end-face structure scattering entropy, match the real-time end-face structure scattering entropy with the multi-dimensional optical transmission feature reference library, determine the category to which the current docking state belongs, and perform decoupling branch determination of the abnormal type according to the category to which the current docking state belongs.

[0064] Step S22, which processes the real-time transient Fresnel reflection jitter waveform, is identical to step S12. The same wavelet transform algorithm is used to decompose the waveform into low-frequency trend components and high-frequency detail components. The same information entropy calculation method is used to calculate the entropy value of the high-frequency detail components, yielding the real-time end-face structure scattering entropy for the current docking process. The consistency of the wavelet transform algorithm ensures that the decomposition boundaries of the low-frequency trend components and high-frequency detail components remain the same between the real-time data and the reference data. The consistency of the information entropy calculation method ensures that the amplitude distribution interval division and probability statistics method remain the same between the real-time data and the reference data. Therefore, the calculated real-time end-face structure scattering entropy is comparable to the end-face structure scattering entropy in the reference library, and the threshold discrimination result accurately reflects the degree of deviation of the current end-face state from the reference state.

[0065] The process of matching the real-time end-face structure scattering entropy with a multi-dimensional optical transmission feature benchmark library is implemented using hierarchical discrimination logic. Three discrimination boundary thresholds are read from the constructed multi-dimensional optical transmission feature benchmark library: scattering entropy cleanliness threshold, scattering entropy damage threshold, and insertion loss deviation threshold. The scattering entropy cleanliness threshold is an upper limit determined based on the statistical distribution of end-face structure scattering entropy for two types of clean samples: ideal unbiased state and purely mechanical position deviation state. It is used to distinguish whether the end face is in a clean state. The scattering entropy damage threshold is a boundary value determined based on the distribution boundary of end-face structure scattering entropy for two types of samples: end-face contamination state and end-face damage state. It is used to further distinguish between contamination and damage when the end face is not clean. The insertion loss deviation threshold is an upper limit determined based on the statistical distribution of insertion loss for samples in the ideal unbiased state. It is used to determine whether position deviation exists when the end face is clean. The methods for determining the three thresholds are all based on the statistical analysis of the four types of prefabricated standard samples in step S13. For example, the scattering entropy cleanliness threshold can be taken as the mean of the scattering entropy distribution of the end face structure of the two types of samples, namely the ideal unbiased state and the pure mechanical position deviation state, plus 3 times the standard deviation. The scattering entropy damage threshold can be taken as the midpoint between the mean of the scattering entropy distribution of the end face structure of the sample with end face contamination plus 3 times the standard deviation and the mean of the scattering entropy distribution of the end face structure of the sample with end face damage minus 3 times the standard deviation. The insertion loss deviation threshold can be taken as the mean of the insertion loss distribution of the sample with ideal unbiased state plus 3 times the standard deviation.

[0066] See Figure 4 The hierarchical discrimination logic is executed in two layers. The first layer determines the cleanliness of the end face, comparing the real-time end face structure scattering entropy with the scattering entropy cleanliness threshold. If the real-time end face structure scattering entropy is less than or equal to the scattering entropy cleanliness threshold, it indicates that the randomness of the high-frequency detail components of the current end face is within the normal range of a clean sample, and the end face is determined to be clean, proceeding to the second layer cleanliness branch discrimination. If the real-time end face structure scattering entropy is greater than the scattering entropy cleanliness threshold, it indicates that the randomness of the high-frequency detail components of the current end face exceeds the normal range of a clean sample, with additional scattering caused by dust particle adhesion or scratch damage, and the end face is determined to have surface anomalies, proceeding to the second layer anomaly branch discrimination. The first layer discrimination utilizes the end face structure scattering entropy's ability to characterize the irregularity of the end face's microstructure, dividing the end face state into two major categories: clean and unclean, laying the foundation for subsequent detailed discrimination.

[0067] The second-layer cleanliness branch discrimination is for position deviation determination. It is performed for cases where the first layer determines the state to be clean. The real-time insertion loss value is compared with the insertion loss deviation threshold: if the real-time insertion loss value is less than or equal to the threshold, it indicates that the optical energy attenuation in the current docking state is within the normal range of an ideal, deviation-free state sample, the end face is clean, and the fiber core is well aligned; the current docking state is determined to be an ideal, deviation-free state. If the real-time insertion loss value is greater than the threshold, it indicates that the optical energy attenuation in the current docking state exceeds the normal range of an ideal, deviation-free state sample. Since the end face has been determined to be clean, the additional loss can only come from the decrease in coupling efficiency caused by fiber core misalignment; the current docking state is determined to be a purely mechanical position deviation state. The second-layer cleanliness branch discrimination utilizes the insertion loss value's ability to characterize optical energy transmission efficiency, determining whether the position is aligned based on the loss value, assuming the end face is clean. The second-level anomaly branch determines the anomaly type, specifically for cases where the first level identifies surface anomalies. It compares the real-time end-face structure scattering entropy with the scattering entropy damage threshold. If the real-time end-face structure scattering entropy is less than or equal to the scattering entropy damage threshold, it indicates that the randomness of the high-frequency detail components of the current end-face is within the distribution range of contaminated samples, and the current docking state is determined to be end-face contaminated. If the real-time end-face structure scattering entropy is greater than the scattering entropy damage threshold, it indicates that the randomness of the high-frequency detail components of the current end-face exceeds the distribution range of contaminated samples and enters the distribution range of damaged samples, and the current docking state is determined to be end-face damaged. The second-level anomaly branch utilizes the ability of end-face structure scattering entropy to distinguish between contamination and damage. Contamination is typically caused by discretely distributed dust particles, resulting in localized scattering, with a corresponding scattering entropy at a moderate level. Damage is typically caused by continuously distributed scratches, resulting in large-area scattering, with a corresponding scattering entropy at a high level.

[0068] The hierarchical discrimination logic decomposes the determination of four docking states into two levels with three branches. Each level of discrimination requires only one numerical comparison, resulting in low computational complexity and a clear discrimination path. The hierarchical discrimination logic prioritizes scattering entropy over insertion loss because: scattering entropy directly reflects the microstructural characteristics of the end face and can distinguish the physical state of the end face at the beginning of the discrimination process; insertion loss is affected by both the end face state and the positional state. Directly judging based on insertion loss without confirming end face cleanliness will lead to ambiguity. If the end face is contaminated and there is also a positional deviation, the insertion loss value alone cannot distinguish the source of loss. Placing scattering entropy judgment before insertion loss judgment ensures that the system determines the positional deviation based on insertion loss only after confirming end face cleanliness, avoiding the risk of increased insertion loss due to end face contamination being misjudged as positional deviation. The relationship between the judgment results of the hierarchical discrimination logic and the threshold comparison is shown in Table 1.

[0069] Table 1. Comparison of the judgment results and thresholds of the hierarchical discrimination logic.

[0070]

[0071] Based on the category of the current docking status output by the hierarchical discrimination logic, a decoupled branch judgment for the anomaly type is executed. If the current docking status is an ideal, unbiased state, it indicates that the end faces are clean and the positions are well aligned, and the docking has reached its optimal state, requiring no correction operations. The system records the final optical parameters of the current docking, including real-time insertion loss and return loss values, marks this docking task as successfully completed, and ends the docking inspection process. The determination result of an ideal, unbiased state directly triggers process termination, avoiding unnecessary subsequent processing steps and improving inspection efficiency.

[0072] If the current docking status is characterized by end-face contamination or damage, the anomaly is determined to be a physical defect rather than a mechanical geometric position deviation. Physical defects cannot be eliminated through mechanical position adjustments. If the system performs position correction on a contaminated or damaged end-face, it will not only fail to improve optical transmission performance but may also exacerbate end-face wear or cause permanent damage by repeatedly adjusting the robotic arm. The system immediately generates an end-face cleaning or maintenance alarm, including the anomaly type (contamination or damage) and the port number where the anomaly occurs. The alarm is sent to the maintenance personnel's terminal via the power fiber optic network management platform. The system forcibly terminates subsequent position deviation calculation and correction processes to prevent the adaptive fiber optic docking robotic arm from performing ineffective position adjustments on the defective end-face, thus avoiding wear on mechanical components or further end-face damage. The system marks this docking task as requiring manual intervention, waiting for maintenance personnel to complete end-face cleaning or connector replacement before re-initiating the docking operation. End-face anomaly branch handling isolates end-face physical defects, anomalies requiring manual intervention, from the automated detection and correction process through timely alarms and process termination. This allows maintenance personnel to perform targeted end-face cleaning or connector replacement, preventing the automated system from blindly operating on anomalies that are not suitable for automatic handling.

[0073] If the current docking state is a purely mechanical position deviation, the anomaly type is determined to be a mechanical geometric position deviation. Since the end face itself is clean and intact, it can be corrected through mechanical adjustment. The system packages the real-time insertion loss value and real-time attenuation curve collected in step S21 to generate a data packet for the optical link to be corrected. The real-time insertion loss value provides information on the total amount of optical energy attenuation in the current docking state, used to calculate the lateral misalignment in step S30; the real-time attenuation curve provides the original waveform data of the reflection peak at the docking port, used to perform iterative deconvolution signal reconstruction in step S30. The generation of the data packet for the optical link to be corrected encapsulates the measurement data collected in step S21 in a structured format, facilitating data extraction and processing in step S30. The system triggers entry into step S30, initiating the precise detection and correction process for the position deviation. The position deviation branch processing, through data packaging and process triggering, guides the purely mechanical position deviation—an anomaly type suitable for automatic correction—to the subsequent precise quantification and correction process, enabling the adaptive fiber optic docking robotic arm to perform precise position adjustments based on the quantified deviation data.

[0074] Step S20 achieves precise decoupling of anomaly types through hierarchical discrimination logic, enabling the system to distinguish between two types of anomalies that appear similar in traditional detection methods: end-face physical defects and mechanical geometric position deviations. Traditional fiber optic splice detection methods rely solely on static optical power thresholds for discrimination. Both end-face contamination and position deviations lead to increased insertion loss, and the system cannot differentiate between the two. This may result in misjudging contamination as deviation and performing incorrect position correction actions, failing to eliminate the loss caused by contamination and causing unnecessary wear on mechanical components. Step S20 introduces end-face structural scattering entropy as a discrimination dimension independent of the insertion loss value. Utilizing the ability of scattering entropy to characterize the irregularity of the end-face microstructure, it distinguishes between clean and unclean end-face states at the beginning of the discrimination process, isolating end-face physical defects from the automatic correction process. Subsequent correction processes are only triggered for cases where the end-face is clean but position deviations exist. The hierarchical discrimination logic's priority order of scattering entropy ensures that position deviation judgment is based on the premise of end-face cleanliness, avoiding the misjudgment of increased loss caused by end-face contamination as position deviation. The three processing logics of the decoupled branch judgment correspond to three situations: ideal state, end face defects, and position deviation. Each situation has a clear follow-up processing action, and the system behavior is deterministic and predictable. Step S20 accurately classifies the real-time docking status into four categories and performs differentiated follow-up processing based on the classification results. The system can accurately identify end face physical defects and generate alarms to notify maintenance personnel in a timely manner, avoiding the adaptive fiber optic docking robot arm from performing ineffective position adjustment actions on end faces with dirt or damage. The service life of mechanical components is extended, and the risk of aggravated end face damage is avoided. The system can accurately identify pure mechanical position deviations and trigger a precise correction process. The correction action is clearly targeted, and correction resources are rationally allocated. The judgment result of the ideal deviation-free state directly triggers the process termination, omitting unnecessary follow-up processing steps and improving detection efficiency. The dual-dimensional collaborative discrimination of end face structure scattering entropy and insertion loss value makes the classification accuracy higher than that of single-dimensional discrimination. The two dimensions are physically independent and complementary. Scattering entropy reflects the microstructure characteristics of the end face, and insertion loss value reflects the optical energy transmission efficiency. The combination of the two covers the main factors affecting docking quality. Decoupling branch determination clearly associates the anomaly type with subsequent processing actions, enabling the system to adopt the most appropriate processing strategy for different anomaly types. End face defects guide manual intervention, while position deviations guide automatic correction, ensuring the rationality of resource allocation.

[0075] Step S30: Based on the optical link data packets to be corrected, perform iterative deconvolution signal reconstruction to obtain the connection point feature signal, parse the longitudinal gap deviation value and the lateral misalignment deviation vector from the connection point feature signal, summarize the longitudinal gap deviation value and the lateral misalignment deviation vector to generate a three-dimensional position correction command and perform closed-loop verification.

[0076] Further, step S30 includes:

[0077] Step S31: Extract the real-time attenuation curve from the optical link data packet to be calibrated, locate the original waveform data of the reflection peak at the docking port in the real-time attenuation curve, and establish the measurement impulse response function.

[0078] The optical link data packet to be calibrated is a structured data encapsulation generated in step S22 when the docking state is determined to be a purely mechanical position deviation state. It contains two measurement data points: the real-time insertion loss value and the real-time attenuation curve collected in step S21. The real-time insertion loss value characterizes the total attenuation of optical energy in the current docking state. The real-time attenuation curve is an optical power attenuation distribution map measured along the entire length of the optical fiber by an optical time domain reflectometer. The entire curve completely records the spatial distribution characteristics of energy loss of the optical signal along the transmission path. The position of the docking port in the real-time attenuation curve can be determined by the distance calibration function of the optical time domain reflectometer. During the installation and commissioning phase, the intelligent optical distribution terminal has recorded the physical distance of each docking port relative to the measurement starting point of the optical time domain reflectometer. The system locates the corresponding vertical coordinate region on the real-time attenuation curve based on the pre-stored port position information. The docking port is represented by a reflection peak on the real-time attenuation curve. The formation of the reflection peak is due to Fresnel reflection caused by the discontinuity of refractive index at the optical fiber connection point. The raw waveform data of the connection point reflection peak refers to a local data segment of the attenuation curve extending forward and backward within a preset distance range centered on the docking port location. This data segment completely includes the rising edge, peak region, and falling edge of the reflection peak. For example, if the physical distance between the docking port and the measurement starting point of the optical time domain reflectometer is 500 meters, and the preset distance range is 20 meters forward and backward, then the raw waveform data of the connection point reflection peak is all the sampling point data within the interval of 480 meters to 520 meters on the real-time attenuation curve. Extracting the raw waveform data of the connection point reflection peak, rather than the entire real-time attenuation curve, for subsequent processing is to concentrate computational resources on the local area of ​​the docking port, reduce the computational overhead of irrelevant data, and improve signal processing efficiency.

[0079] The measurement impulse response function is established based on the hardware characteristics of the optical time-domain reflectometer (OTDR). The OTD works by injecting a narrow-pulse laser signal into an optical fiber and obtaining the attenuation distribution information of the fiber by detecting the backscattered and reflected light returning along the fiber. The laser pulse has a certain time width, and the photodetector's response to the returned light signal also has a certain bandwidth limitation; these two factors together determine the spatial resolution of the OTD. The measurement impulse response function is a mathematical model describing the response characteristics of the OTD to an ideal point reflection source. Its physical meaning is: assuming the existence of an ideal reflection point with an infinitely narrow location and unit reflection intensity, the waveform of this reflection point on the attenuation curve after measurement by the OTD is the measurement impulse response function. Actual connection points are not ideal point reflection sources, but their spatial scale is much smaller than the spatial range corresponding to the pulse width of the OTD. Therefore, the measured waveform of the reflection peak at the connection point can be regarded as the result of the convolution of the actual connection point characteristics and the measurement impulse response function. The method for establishing the measurement impulse response function is as follows: The emitted pulse width of the laser and the response bandwidth of the photodetector are read from the hardware parameters of the optical time-domain reflectometer in the multi-functional test module of the intelligent optical distribution terminal. The time-domain waveform of the laser pulse can typically be approximated as a Gaussian or rectangular distribution. The frequency response of the detector can be modeled as a low-pass filter characteristic. The measurement impulse response function is obtained by convolving the laser pulse waveform with the detector response. For example, if the laser pulse width is 10 ns and approximately Gaussian, and the detector response bandwidth is 100 MHz, then the measurement impulse response function is the result of convolving the Gaussian pulse with the impulse response of a first-order low-pass filter. The full width at half maximum (FWHM) of this function is approximately 11 ns, corresponding to a spatial resolution of approximately 1.1 meters. The purpose of establishing the measurement impulse response function is to provide the kernel function required for deblurring in subsequent deconvolution operations, enabling the system to recover a sharpened signal that more closely approximates the characteristics of the actual connection points from the blurred measurement waveform.

[0080] Step S32: Extract the real-time insertion loss value from the data packet of the optical link to be corrected, control the robotic arm to perform micro-motion detection and record the change of insertion loss value at the docking port to calculate the optical power displacement sensitivity, and perform iterative deconvolution operation on the original waveform data of the connection point reflection peak based on the measured impulse response function and with the optical power displacement sensitivity as a constraint to obtain the connection point feature signal;

[0081] Micro-motion detection refers to a robotic arm performing a small-amplitude position adjustment based on its current docking position. This adjustment is much smaller than the physical dimensions of the connector but sufficient to cause a measurable change in optical power. The micro-motion detection process is as follows: While maintaining the docking state, the adaptive fiber optic docking robotic arm performs a small displacement along a preset direction. The displacement can be set to be on the order of one-thousandth of the diameter of the connector's ceramic ferrule. An optical power meter continuously measures the optical power at the docking port during the micro-motion, recording the data sequence of optical power changes with displacement. The optical power displacement sensitivity is defined as the ratio of the change in optical power to the change in displacement, characterizing the degree of optical power change caused by a unit displacement. The optical power displacement sensitivity is calculated by performing a linear regression fit between the optical power data sequence recorded during micro-motion detection and the corresponding displacement data sequence; the slope of the regression line is the optical power displacement sensitivity. Since the robotic arm can move in three-dimensional space, micro-motion detection needs to be performed independently along the optical axis, in the horizontal direction perpendicular to the optical axis, and in the vertical direction perpendicular to the optical axis, obtaining optical power displacement sensitivity components in three directions. Optical power displacement sensitivity quantifies the response of optical coupling efficiency to positional changes under the current docking state: when there is a lateral misalignment between the cores of the two fibers, the optical power displacement sensitivity in the lateral displacement direction is high because a small lateral adjustment can significantly change the overlap area of ​​the cores, thus causing a large change in optical power; when there is a longitudinal gap between the end faces of the two fibers, the optical power displacement sensitivity in the longitudinal displacement direction reflects the impact of gap changes on divergence loss. Optical power displacement sensitivity provides constraints for subsequent iterative deconvolution operations, ensuring that the convergence direction of the deconvolution result is consistent with the actual positional deviation direction.

[0082] The goal of iterative deconvolution is to recover a sharpened signal that more closely approximates the true characteristics of the connection points from the original waveform data of the connection point reflection peaks. The mathematical principle of deconvolution is based on the convolution theorem: if the measured waveform is equal to the convolution of the true signal and the measured impulse response function, then theoretically, the true signal can be recovered by performing a deconvolution operation on the measured waveform. Direct deconvolution faces numerical stability issues; measurement noise is amplified during the deconvolution process, leading to spurious high-frequency oscillations in the recovered result. Iterative deconvolution overcomes the noise amplification problem by progressively approximating the measured signal. Each iteration corrects the current estimated signal, gradually improving its match with the measured waveform. The iterative deconvolution operation is implemented using the constrained Richardson-Lucy iterative algorithm, a commonly used image restoration method in optical imaging. Its iterative process is as follows: the initial estimated signal is set as the original waveform data of the connection point reflection peak itself. In each iteration, the current estimated signal is convolved with the measured impulse response function to obtain a simulated measured waveform. The ratio of the measured waveform to the simulated measured waveform is calculated as a correction coefficient. The correction coefficient is then convolved with the inverted form of the measured impulse response function and multiplied by the current estimated signal to obtain a new estimated signal. Optical power displacement sensitivity is introduced as a constraint during the iteration process. This is implemented as follows: after each iteration, the offset of the reflection peak edge position is calculated based on the current estimated signal. This offset is compared with the optical power displacement sensitivity data. If the optical power change corresponding to the offset is inconsistent with the measured optical power displacement sensitivity, the estimated signal is corrected to make them consistent. The optical power displacement sensitivity is introduced as a constraint because: optical time-domain reflectometers are limited by pulse width, and their spatial resolution is typically only on the order of meters, making it impossible to directly detect sub-micron level docking deviations; while optical power meters are highly sensitive to changes in coupling efficiency caused by minute displacements, their output is a scalar value and does not contain spatial position information. Optical power displacement sensitivity links the sensitivity advantage of the optical power meter with displacement information, guiding the edge position of the estimated signal to converge towards the true position during iterative deconvolution, thus achieving the fusion of information from the two measurement methods. The iteration termination condition is set when the change in the estimated signal between two adjacent iterations is less than a preset convergence threshold, or when the number of iterations reaches a preset upper limit. For example, the convergence threshold can be set to one ten-thousandth of the estimated signal energy, and the upper limit of the number of iterations can be set to one hundred. The signal obtained after iterative deconvolution is defined as the connection point feature signal. This signal has a narrower peak width and steeper edges compared to the original waveform data of the connection point reflection peak, and is closer to the reflection characteristics of the true connection point.

[0083] Step S32 integrates the spatial distribution measurement capability of the optical time-domain reflectometer (OTDR) with the high-sensitivity measurement capability of the optical power meter at the algorithm level. The ODR provides the overall waveform profile information of the reflection peak at the connection point, while the optical power meter provides the quantitative relationship between position change and optical power change through micro-motion detection. Iterative deconvolution operation sharpens the reflection peak waveform with optical power displacement sensitivity as a constraint, ensuring that the deconvolution result not only meets the convolution consistency with the measured waveform but also the physical consistency with the optical power response characteristics. At the software level, the system breaks through the physical resolution limit of single hardware. The spatial resolution of the connection point feature signal is no longer limited by the pulse width of the ODR but reaches the sub-micron level, comparable to the displacement detection sensitivity of the optical power meter, laying a signal quality foundation for subsequent accurate quantification of three-dimensional position deviation.

[0084] Step S33: Perform feature analysis on the connection point feature signals to extract the longitudinal gap deviation value and the lateral misalignment deviation vector, respectively;

[0085] The method for feature analysis of the connection point feature signal includes: performing spectral analysis on the connection point feature signal, detecting periodic interference modulation components, and calculating the longitudinal gap deviation value; calculating the theoretical additional loss corresponding to the longitudinal gap based on the longitudinal gap deviation value, calculating the difference between the real-time insertion loss value and the static reference insertion loss value to obtain the excess loss, subtracting the theoretical additional loss from the excess loss to obtain the remaining loss component, estimating the lateral misalignment based on the remaining loss component, and decomposing the lateral misalignment into components in two orthogonal directions to obtain the lateral misalignment deviation vector.

[0086] See Figure 5 This is a schematic diagram of longitudinal gap deviation provided in the embodiments of this application. Figure 5 The diagram illustrates the transmitting and receiving optical fibers arranged sequentially along the optical axis. A longitudinal gap and corresponding air gap region are formed between the end faces of the two fibers. The internal core structures of the transmitting and receiving fibers, as well as the optical axis direction for fiber signal transmission, are also labeled. Longitudinal gap deviation refers to the air gap between the end faces of the two fibers along the optical axis, while lateral misalignment deviation refers to the positional misalignment of the fiber cores in a plane perpendicular to the optical axis. Longitudinal gap deviation and lateral misalignment deviation together constitute the three-dimensional positional deviation of the fiber splicing. The extraction of the longitudinal gap deviation value is based on the double reflection superposition effect formed by the end face gap. When there is a tiny air gap between the end faces of the two fibers, such as... Figure 5 The transmitter fiber end face shown and as follows Figure 5 The fiber end facets at the receiving end, as shown, each generate an independent Fresnel reflection. These two reflections correspond to two spatially close reflection response components on the distance axis of the signal measured by the optical time-domain reflectometer, with their spatial separation distance being... Figure 5The longitudinal gap thickness is directly related to the distance between the two reflections. Since light travels back and forth within the gap, the spatial separation distance is equal to twice the longitudinal gap thickness multiplied by the ratio of the speed of light to the sampling frequency. When the longitudinal gap size is smaller than the inherent spatial resolution of the optical time-domain reflectometer, the two reflection response components are difficult to distinguish directly in the original attenuation curve. However, the connection point feature signal after iterative deconvolution processing in step S32 can more clearly present this double reflection superposition feature. Performing spectral analysis on the connection point feature signal can detect whether there is a periodic interference modulation component caused by the superposition of double reflections: performing a Fourier transform on the connection point feature signal, two spatially close reflection responses will introduce periodic interference modulation features in the spatial frequency domain. The modulation period is inversely proportional to the spatial separation distance between the two reflection responses. Searching for frequency peaks significantly higher than the noise floor in the frequency domain, if such frequency peaks are detected, it indicates the presence of a periodic interference modulation component caused by the superposition of double reflections. The spatial modulation frequency corresponding to the peak is inversely proportional to the spatial separation distance between the two reflection responses. The longitudinal gap deviation is calculated using the physical conversion relationship between the spatial spectrum modulation frequency and the gap thickness: the reciprocal of the spatial modulation frequency is equal to the spatial separation distance between the two reflection responses. Dividing the spatial separation distance by 2 and then by the ratio of the speed of light to the sampling frequency yields the result. Figure 5 The thickness of the longitudinal gap is indicated in the figure.

[0087] The extraction of the lateral misalignment deviation vector is based on the loss decomposition principle and the Gaussian beam coupling model. The currently measured total insertion loss is contributed by multiple factors, including divergence loss caused by longitudinal gap and mode field mismatch loss caused by lateral misalignment. Given the longitudinal gap deviation value, the theoretical additional loss corresponding to the longitudinal gap can be calculated. The calculation method utilizes the single-mode fiber mode field diameter parameters stored in the multidimensional optical transmission characteristic reference library and the divergence characteristics of Gaussian beams in free space: after light is emitted from the transmitting fiber, it diverges and propagates in the gap as a Gaussian beam. When it reaches the receiving fiber, the spot size becomes larger, and there is a size mismatch between the mode field and the receiving fiber. The degree of mismatch determines the amount of coupling loss. The theoretical additional loss corresponding to the longitudinal gap can be calculated using the Gaussian beam coupling efficiency formula. This formula is based on the principle of mode field overlap integral between two Gaussian beams. When a longitudinal gap exists between two single-mode fibers, the Gaussian beam emitted from the transmitting fiber diverges and propagates within the gap. Upon reaching the receiving fiber, the beam spot radius increases, resulting in a size mismatch with the mode field radius of the receiving fiber. The result of the mode field overlap integral is the coupling efficiency, and the negative logarithm of the coupling efficiency is the theoretical additional loss caused by the longitudinal gap. This formula is a well-known technique in the field of optical fiber communication. First, the static reference insertion loss value corresponding to the ideal unbiased state is read from the multidimensional optical transmission characteristic reference library. The difference between the real-time insertion loss value and the static reference insertion loss value is calculated to obtain the excess loss. Then, the theoretical additional loss corresponding to the longitudinal gap is subtracted from the excess loss to obtain the remaining loss component. Since the end face has been determined to be in a clean state in step S20, the remaining loss component is mainly caused by mode field mismatch due to lateral misalignment. The mode field diameter parameters of single-mode optical fibers are read from a multidimensional optical transmission characteristic reference library. Using the correspondence between Gaussian beam coupling efficiency and lateral misalignment, the value of the lateral misalignment is inversely calculated from the residual loss component. The Gaussian beam coupling efficiency formula shows that when there is lateral misalignment between two single-mode optical fibers, the coupling efficiency monotonically decreases as the misalignment increases. The ratio of the misalignment to the mode field diameter determines the value of the coupling efficiency; therefore, the misalignment can be inversely calculated from the coupling efficiency. Lateral misalignment occurs in a two-dimensional plane perpendicular to the optical axis. The residual loss component can only determine the magnitude of the lateral misalignment, not the direction of the misalignment. The direction of the misalignment is determined by the difference in optical power displacement sensitivity in different directions during the micro-motion detection of the robotic arm in step S32: if the optical power displacement sensitivity is positive when the robotic arm is micro-moved in a certain direction, it indicates that the direction is a direction that reduces the misalignment; if the optical power displacement sensitivity is negative, it indicates that the direction is a direction that increases the misalignment. Comparing the difference in optical power displacement sensitivity in the horizontal and vertical directions determines the direction angle of the misalignment vector in the two-dimensional plane. Combining the magnitude of the lateral misalignment with the direction angle yields the complete lateral misalignment deviation vector, which can be decomposed into two orthogonal components: a horizontal component and a vertical component.For example, if the lateral misalignment modulus obtained by back-calculation of the remaining loss component is 2 micrometers, and the micro-motion detection shows that the horizontal optical power displacement sensitivity is positive and the value is 0.1 dB per micrometer, while the vertical optical power displacement sensitivity is positive and the value is 0.2 dB per micrometer, then the tangent of the misalignment direction angle is 0.2 divided by 0.1 equals 2, the direction angle is approximately 63 degrees, the horizontal component of the lateral misalignment deviation vector is approximately 0.9 micrometers, and the vertical component is approximately 1.8 micrometers.

[0088] Step S33 analyzes the connection point feature signal into two independent deviation components: the longitudinal gap deviation value and the lateral misalignment deviation vector. The positional deviation problem in three-dimensional space is decomposed into two relatively independent sub-problems: longitudinal and lateral. Each sub-problem is processed using a specific analytical method. The longitudinal gap deviation is extracted through interference modulation frequency analysis, and the lateral misalignment deviation is extracted through loss decomposition and optical power response analysis. The system obtains complete quantitative data characterizing the three-dimensional positional deviation, providing clear deviation direction and magnitude information for subsequent generation of precise correction commands, enabling the robotic arm to perform precise amplitude adjustments in the correct direction.

[0089] Step S34: Summarize the longitudinal clearance deviation value and the lateral misalignment deviation vector to generate a three-dimensional position correction command. Execute the correction action in the order of longitudinal priority over lateral priority. After the correction is completed, return to step S21 to re-acquire the corrected real-time transient Fresnel reflection jitter waveform, real-time attenuation curve, and real-time insertion loss value for closed-loop verification to form closed-loop feedback control.

[0090] Step S34 summarizes the longitudinal gap deviation value and lateral misalignment deviation vector obtained from step S33 to generate a three-dimensional position correction command. The three-dimensional position correction command consists of two parts: a longitudinal correction component and a lateral correction component. The longitudinal correction component indicates the displacement and direction of the robotic arm along the optical axis; its value is equal to the longitudinal gap deviation value, and its direction is the direction of compressing the gap, i.e., moving closer to the opposite end face. The lateral correction component indicates the displacement and direction of the robotic arm in a plane perpendicular to the optical axis; its value is equal to the magnitude of the lateral misalignment deviation vector, and its direction is the direction of reducing misalignment, i.e., the opposite direction to the lateral misalignment deviation vector. The execution of the three-dimensional position correction command prioritizes the longitudinal direction over the lateral direction. The technical consideration behind this priority setting is that the longitudinal gap affects the measurement and correction of the lateral misalignment. When the longitudinal gap is large, the beam diverges within the gap, resulting in a larger spot size. In this case, the measured lateral misalignment deviation may differ from the actual misalignment deviation when the end faces are fully aligned. If the longitudinal gap is corrected first to ensure full alignment of the two end faces, and then the lateral misalignment is measured and corrected, the measurement of the lateral misalignment is more accurate, and the correction action is more precise.

[0091] The execution process of the three-dimensional position correction command is as follows: First, it determines whether the longitudinal gap deviation exceeds a preset contact threshold. The contact threshold is determined by measuring the statistical distribution of the residual longitudinal gap deviation under the condition that the standard reference jumper is fully fitted and connected, and taking the upper limit of the distribution as the contact threshold. For example, the contact threshold can be set to 0.5 micrometers. If the longitudinal gap deviation is greater than the contact threshold, it indicates that the two end faces are not fully fitted. The three-dimensional position correction command drives the adaptive fiber optic docking robot arm to feed along the optical axis, with the feed amount equal to the longitudinal gap deviation, compressing the longitudinal gap to below the contact threshold. After the longitudinal gap deviation drops below the contact threshold, it determines whether the magnitude of the lateral misalignment deviation vector exceeds the allowable upper limit of lateral misalignment. The allowable upper limit of lateral misalignment is determined by inversely calculating the corresponding maximum lateral misalignment based on the target insertion loss requirement and the Gaussian beam coupling efficiency formula. For example, if the target insertion loss requirement is below 0.3 dB and the mode field diameter is 10 micrometers, the allowable upper limit of lateral misalignment is approximately 1.2 micrometers. If the magnitude of the lateral misalignment deviation vector is greater than the upper limit of the allowable lateral misalignment, the three-dimensional position correction command drives the adaptive fiber optic docking robot arm to perform position fine-tuning in a plane perpendicular to the optical axis. The fine-tuning direction is opposite to the direction of the lateral misalignment deviation vector, and the fine-tuning amount is equal to the magnitude of the lateral misalignment deviation vector.

[0092] After the adaptive fiber optic docking robotic arm completes its correction action, the system automatically triggers a return to step S21 to re-acquire the corrected real-time transient Fresnel reflection jitter waveform, real-time attenuation curve, and real-time insertion loss value. Step S22 processes the newly acquired data, calculates the corrected real-time end-face structure scattering entropy and real-time insertion loss value, and compares them with the discrimination boundary threshold in the multi-dimensional optical transmission feature reference library to determine the category of the corrected docking state. If the corrected docking state still belongs to a purely mechanical position deviation state, the system continues to execute the position deviation quantification and correction process in step S30, generating a new round of three-dimensional position correction commands and driving the robotic arm to perform correction actions. The above cycle continues, with the position deviation gradually decreasing in each cycle, and the system state gradually approaching the ideal deviation-free state. The loop termination condition is: the real-time end-face structure scattering entropy is less than or equal to the scattering entropy cleanliness threshold, and the real-time insertion loss value is less than or equal to the insertion loss deviation threshold. At this time, the hierarchical discrimination logic determines that the current docking status is an ideal and deviation-free state, indicating that the end face is clean and the position is well aligned. The system determines that this automatic docking correction is successfully completed, ends the correction process, and records the final docking parameters, including the final insertion loss value and the final return loss value.

[0093] Step S34 achieves iterative convergence correction of position deviation through closed-loop feedback control. After each correction action, the correction effect is verified by re-acquisition and discrimination. If the ideal deviation-free state is not reached, the correction continues, forming a closed-loop feedback control loop. The system has adaptive correction capability. Even if a single correction action fails to completely eliminate the deviation due to mechanical precision limitations or load disturbances, subsequent iterative correction rounds can further reduce the residual deviation until the judgment condition of the ideal deviation-free state is met. The closed-loop verification mechanism ensures that the system will not blindly determine success after the correction action is executed, but confirms the correction effect based on actual measurement data, avoiding the risk of correction failure caused by actuator error being falsely reported as success.

[0094] Step S30 achieves sub-micron-level 3D position deviation quantization and correction through multi-source signal fusion. Optical time-domain reflectometers (OTDRs), limited by laser pulse width, typically have a spatial resolution on the meter level, making them unable to directly detect sub-micron-level docking deviations. While optical power meters (OPMs) are highly sensitive to coupling efficiency changes caused by minute displacements, their output is a scalar value, lacking directional and dimensional component information of the deviation. Using the attenuation curve data from the OTM as the base waveform and the displacement sensitivity data from the OPM as constraints, super-resolution signal reconstruction is achieved at the software level through iterative deconvolution operations. Longitudinal gap deviation and lateral misalignment deviation are analyzed from the reconstructed connection point feature signals to obtain complete 3D position deviation quantization information. Precise correction commands are generated based on the quantized deviation information to drive the robotic arm to perform targeted position adjustments. Closed-loop verification ensures that the correction effect meets the expected requirements. Independent analysis of longitudinal clearance deviation and lateral misalignment deviation gives the correction action clear directionality and quantitative accuracy, avoiding the inefficiency and mechanical wear caused by blind trial-and-error adjustments. The longitudinal-priority correction sequence eliminates the interference of longitudinal clearance on lateral misalignment measurement, improving the accuracy of lateral correction. The closed-loop feedback control mechanism enables the system to have adaptive iterative correction capability, relaxing the requirements for mechanical execution accuracy. The execution error of a single correction action can be compensated by subsequent iterations. If step S30 is missing, after step S20 determines that the state is a pure mechanical position deviation, there will be a lack of precise deviation quantification. The robotic arm cannot obtain clear adjustment direction and adjustment amount information, and can only perform position adjustment by blind trial and error. The adjustment efficiency is low, and repeated ineffective adjustments may lead to wear of mechanical parts. The accuracy and efficiency of adaptive correction will decrease.

[0095] Example 2:

[0096] This embodiment, based on Embodiment 1, provides a fiber optic connector mating position deviation detection system, such as... Figure 6 As shown, it includes:

[0097] Feature reference construction module: Perform physical connection on standard reference jumper, collect static reference insertion loss value at the connection port and transient Fresnel reflection jitter waveform throughout the connection action cycle, extract end face structure scattering entropy from transient Fresnel reflection jitter waveform, and construct a multi-dimensional optical transmission feature reference library;

[0098] Anomaly decoupling determination module: used to acquire real-time transient Fresnel reflection jitter waveforms during the on-site docking process and calculate the real-time end-face structure scattering entropy, match the real-time end-face structure scattering entropy with the multi-dimensional optical transmission feature reference library, and perform decoupling branch determination of anomaly types, including end-face physical defects and mechanical geometric position deviations, and generate optical link data packets to be corrected for mechanical geometric position deviations;

[0099] Deviation quantization correction module: Based on the optical link data packets to be corrected, iterative deconvolution signal reconstruction is performed to obtain the connection point feature signal. The longitudinal gap deviation value and the lateral misalignment deviation vector are analyzed from the connection point feature signal. The longitudinal gap deviation value and the lateral misalignment deviation vector are summarized to generate a three-dimensional position correction command and perform closed-loop verification.

[0100] Furthermore, in the feature benchmark construction module, the method for extracting the scattering entropy of the end-face structure includes:

[0101] The transient Fresnel reflection jitter waveform is decomposed into low-frequency trend components and high-frequency detail components by performing wavelet transform, and the information entropy of the high-frequency detail components is calculated to obtain the end face structure scattering entropy.

[0102] The method for constructing the multidimensional optical transmission feature reference library includes:

[0103] The static reference insertion loss value, transient Fresnel reflection jitter waveform, and end-face structure scattering entropy corresponding to each type of prefabricated standard sample in multiple types of prefabricated standard samples are combined into the feature vector of the corresponding category. The discrimination boundary threshold between each category is determined, and the feature vector of each category and its discrimination boundary threshold are stored as a multi-dimensional optical transmission feature reference library. The discrimination boundary threshold includes the scattering entropy cleanliness threshold, the scattering entropy damage threshold, and the insertion loss deviation threshold.

[0104] Furthermore, in the anomaly decoupling determination module, the method for matching the real-time end-face structure scattering entropy with the multi-dimensional optical transmission feature reference library includes: reading the scattering entropy cleanliness threshold, scattering entropy damage threshold, and insertion loss deviation threshold from the multi-dimensional optical transmission feature reference library;

[0105] The first layer of discrimination is performed, including: if the real-time end-face structure scattering entropy is less than or equal to the scattering entropy cleanliness threshold, the end face is determined to be in a clean state and the second layer of cleanliness branch discrimination is entered; if the real-time end-face structure scattering entropy is greater than the scattering entropy cleanliness threshold, the end face is determined to have surface anomalies and the second layer of anomaly branch discrimination is entered.

[0106] The execution method for the second-layer clean branch discrimination is as follows: if the real-time insertion loss value is less than or equal to the insertion loss deviation threshold, the current docking state is determined to be an ideal no-deviation state; if the real-time insertion loss value is greater than the insertion loss deviation threshold, the current docking state is determined to be a pure mechanical position deviation state.

[0107] The execution method for the second-level abnormal branch discrimination is as follows: if the real-time end-face structure scattering entropy is less than or equal to the scattering entropy damage threshold, the current docking state is determined to be an end-face contamination state; if the real-time end-face structure scattering entropy is greater than the scattering entropy damage threshold, the current docking state is determined to be an end-face damage state; the scattering entropy cleanliness threshold is lower than the scattering entropy damage threshold.

[0108] The method for determining the decoupled branch of the execution exception type includes:

[0109] Collect the real-time attenuation curve of the current full optical path and the real-time insertion loss value at the current docking port;

[0110] If the current docking status is an ideal, deviation-free state, the docking is considered successful, and the docking inspection process ends. If the current docking status is an end-face contamination or end-face damage state, the current anomaly type is determined to be an end-face physical defect, an end-face cleaning or maintenance alarm is generated, and the subsequent position deviation correction process is terminated. If the current docking status is a purely mechanical position deviation state, the current anomaly type is determined to be a mechanical geometric position deviation, and the real-time insertion loss value and real-time attenuation curve are packaged to generate an optical link data packet to be corrected.

[0111] The methods and systems of this application may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the method is for illustrative purposes only, and the steps of the method of this application are not limited to the order specifically described above, unless otherwise specifically stated.

[0112] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0113] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting misalignment of fiber optic connector mating positions, characterized in that, The method includes: Physically connect the standard reference jumper fiber, collect the static reference insertion loss value at the connection port and the transient Fresnel reflection jitter waveform throughout the entire connection operation cycle, perform wavelet transform on the transient Fresnel reflection jitter waveform to decompose it into low-frequency trend component and high-frequency detail component, calculate the information entropy of the high-frequency detail component to obtain the end face structure scattering entropy, and construct a multi-dimensional optical transmission feature reference library; The system collects real-time transient Fresnel reflection jitter waveforms during the on-site docking process and calculates the real-time end-face structure scattering entropy. It then performs feature matching between the real-time end-face structure scattering entropy and a multi-dimensional optical transmission feature benchmark library, and performs decoupling branch determination of anomaly types, including end-face physical defects and mechanical geometric position deviations. For mechanical geometric position deviations, it generates optical link data packets to be corrected. Based on the optical link data packets to be corrected, iterative deconvolution signal reconstruction is performed to obtain the connection point feature signal. The longitudinal gap deviation value and the lateral misalignment deviation vector are analyzed from the connection point feature signal. The longitudinal gap deviation value and the lateral misalignment deviation vector are summarized to generate a three-dimensional position correction command and perform closed-loop verification.

2. The method for detecting optical fiber connector mating position deviation according to claim 1, characterized in that, The method for constructing the multidimensional optical transmission feature reference library includes: The static reference insertion loss value, transient Fresnel reflection jitter waveform, and end-face structure scattering entropy corresponding to each type of prefabricated standard sample in multiple types of prefabricated standard samples are combined into the feature vector of the corresponding category. The discrimination boundary threshold between each category is determined, and the feature vector of each category and its discrimination boundary threshold are stored as a multi-dimensional optical transmission feature reference library. The discrimination boundary threshold includes the scattering entropy cleanliness threshold, the scattering entropy damage threshold, and the insertion loss deviation threshold.

3. The method for detecting optical fiber connector mating position deviation according to claim 2, characterized in that, The various prefabricated standard samples include four connector types: standard connectors representing an ideal, unbiased state; connectors representing a purely mechanical positional deviation state; connectors representing a contaminated end face state; and connectors representing a damaged end face state.

4. The method for detecting optical fiber connector mating position deviation according to claim 3, characterized in that, The method for matching the real-time end-face structure scattering entropy with the multi-dimensional optical transmission feature reference library includes: reading the scattering entropy cleanliness threshold, scattering entropy damage threshold, and insertion loss deviation threshold from the multi-dimensional optical transmission feature reference library; The first layer of discrimination is performed, including: if the real-time end-face structure scattering entropy is less than or equal to the scattering entropy cleanliness threshold, the end face is determined to be in a clean state and the second layer of cleanliness branch discrimination is entered; if the real-time end-face structure scattering entropy is greater than the scattering entropy cleanliness threshold, the end face is determined to have surface anomalies and the second layer of anomaly branch discrimination is entered.

5. The method for detecting optical fiber connector mating position deviation according to claim 4, characterized in that, The execution method for the second-layer clean branch discrimination is as follows: if the real-time insertion loss value is less than or equal to the insertion loss deviation threshold, the current docking state is determined to be an ideal no-deviation state; if the real-time insertion loss value is greater than the insertion loss deviation threshold, the current docking state is determined to be a pure mechanical position deviation state. The execution method for the second-level abnormal branch discrimination is as follows: if the real-time end face structure scattering entropy is less than or equal to the scattering entropy damage threshold, then the current docking state is determined to be an end face contamination state. If the real-time end-face structure scattering entropy is greater than the scattering entropy damage threshold, the current docking state is determined to be an end-face damage state; the scattering entropy cleanliness threshold is lower than the scattering entropy damage threshold.

6. The method for detecting optical fiber connector mating position deviation according to claim 5, characterized in that, The method for determining the decoupled branch of the execution exception type includes: Collect the real-time attenuation curve of the current full optical path and the real-time insertion loss value at the current docking port; If the current docking status is an ideal, deviation-free state, the docking is considered successful, and the docking inspection process ends. If the current docking status is an end-face contamination or end-face damage state, the current anomaly type is determined to be an end-face physical defect, an end-face cleaning or maintenance alarm is generated, and the subsequent position deviation correction process is terminated. If the current docking status is a purely mechanical position deviation state, the current anomaly type is determined to be a mechanical geometric position deviation, and the real-time insertion loss value and real-time attenuation curve are packaged to generate an optical link data packet to be corrected.

7. The method for detecting optical fiber connector mating position deviation according to claim 6, characterized in that, The method for obtaining the connection point feature signal includes: Extract the real-time attenuation curve from the data packets of the optical link to be calibrated, locate the original waveform data of the reflection peak at the docking port in the real-time attenuation curve, and establish the measurement impulse response function; The real-time insertion loss value is extracted from the data packets of the optical link to be calibrated. The robotic arm is controlled to perform micro-motion detection and record the change of insertion loss value at the docking port to calculate the optical power displacement sensitivity. Based on the measured impulse response function and constrained by the optical power displacement sensitivity, iterative deconvolution operation is performed on the original waveform data of the reflection peak at the connection point to obtain the connection point characteristic signal.

8. The method for detecting misalignment of fiber optic connector mating positions according to claim 7, characterized in that, The method for resolving longitudinal gap deviation values ​​and lateral misalignment deviation vectors from connection point feature signals includes: Perform spectral analysis on the characteristic signal of the connection point, detect the periodic interference modulation component, and calculate the longitudinal gap deviation value; The theoretical additional loss corresponding to the longitudinal gap is calculated based on the longitudinal gap deviation value. The difference between the real-time insertion loss value and the static reference insertion loss value is calculated to obtain the excess loss. The theoretical additional loss is subtracted from the excess loss to obtain the remaining loss component. The lateral misalignment is calculated based on the remaining loss component. The lateral misalignment is decomposed into components in two orthogonal directions to obtain the lateral misalignment deviation vector.

9. A fiber optic connector mating position deviation detection system, used to implement the fiber optic connector mating position deviation detection method according to any one of claims 1-8, characterized in that, The system includes: Feature reference construction module: Perform physical connection on standard reference jumper, collect static reference insertion loss value at the connection port and transient Fresnel reflection jitter waveform throughout the connection action cycle, extract end face structure scattering entropy from transient Fresnel reflection jitter waveform, and construct a multi-dimensional optical transmission feature reference library; Anomaly decoupling determination module: used to acquire real-time transient Fresnel reflection jitter waveforms during the on-site docking process and calculate the real-time end-face structure scattering entropy, match the real-time end-face structure scattering entropy with the multi-dimensional optical transmission feature reference library, and perform decoupling branch determination of anomaly types, including end-face physical defects and mechanical geometric position deviations, and generate optical link data packets to be corrected for mechanical geometric position deviations; Deviation quantization correction module: Based on the optical link data packets to be corrected, iterative deconvolution signal reconstruction is performed to obtain the connection point feature signal. The longitudinal gap deviation value and the lateral misalignment deviation vector are analyzed from the connection point feature signal. The longitudinal gap deviation value and the lateral misalignment deviation vector are summarized to generate a three-dimensional position correction command and perform closed-loop verification.

Images