A method and system for evaluating the contamination level of a fiber coupling end face of a fiber lead frame

Abstract

The present application relates to the field of optical fiber connection and optical detection technology, and discloses a kind of optical fiber coupling end face pollution degree evaluation method and system of optical fiber lead frame, method includes: establishing the coordinate system of optical fiber lead frame and optical fiber end face, collecting end face multi-source image data and obtaining pollutant information;According to mode field distribution data and pollution attenuation response, calculate current functional pollution degree;Filter key pollutants and apply controlled micro disturbance below cleaning threshold and damage threshold, obtain pollutant migration response data;Based on actual coupling working condition and contact evolution calibration data, predict the evolution state of pollutant and calculate the functional pollution degree after evolution;According to current functional pollution degree, the functional pollution degree after evolution and the maximum contact evolution risk calculated in advance, generate comprehensive evaluation result.The present application can realize pollution influence evaluation and coupling risk prediction without damaging optical fiber end face, improve pollution detection accuracy and disposal reliability.

Description

Technical Field

[0001] This invention relates to the field of optical fiber connection and optical inspection technology, and in particular to a method and system for assessing the contamination degree of the optical fiber coupling end face of an optical fiber lead frame. Background Technology

[0002] The surface condition of the fiber optic coupling endface is a crucial factor affecting the quality of fiber optic connections. The presence of particles, fiber debris, oil, or other contaminants on the fiber coupling endface can cause problems such as optical signal attenuation, changes in reflection, and unstable transmission. During the process of fiber endfaces approaching or coming into contact, some contaminants may also change their original position or shape, causing damage to the fiber coupling endface and thus affecting the performance and long-term reliability of the fiber optic assembly. Therefore, the contamination status of the fiber coupling endface needs to be detected and assessed before assembling or coupling fiber optic assemblies.

[0003] Existing fiber optic coupling end-face inspection methods typically rely on identifying anomalous regions from acquired end-face images and determining whether the fiber optic coupling end-face meets requirements based on the size, number, location, or grayscale characteristics of these anomalous regions. While this type of inspection can reflect the surface state of the fiber optic coupling end-face at the time of inspection, the results depend primarily on the visible morphology of contaminants in the image and cannot comprehensively reflect the actual impact of factors such as the location, morphology, adhesion state of contaminants, and the distribution of the fiber's optical field.

[0004] For multiple fiber end faces mounted on a fiber optic lead frame, there may be deviations in the clamping position of the fiber optic lead frame and the actual position and orientation of the fiber end faces. There may also be assembly differences between different fiber end faces, making it difficult to reproduce the same detection position stably during batch testing, thus affecting the correspondence and comparability between different test results.

[0005] Furthermore, the contamination state of the fiber optic coupling end face is not constant throughout the assembly and coupling process. Some contaminants may move, spread, break, or become embedded during transport, vibration, airflow, or end face contact. The small impact of contaminants at the initial detection location does not mean they will not affect optical signal transmission or damage the fiber coupling end face during subsequent coupling. Judging solely based on the static surface state at the moment of detection is insufficient to identify the potential risks posed by subsequent changes in the contaminant's state.

[0006] Furthermore, variations in the height, volume, material properties, and relative position of contaminants to the fiber coupling end face can lead to different optical or mechanical effects. Contaminants of similar size or area in a two-dimensional image may cause different risks of optical signal attenuation, end face indentation, or scratches during actual coupling. Therefore, assessing the degree of contamination at the fiber coupling end face using only a single surface geometric feature may result in discrepancies between the detection results and the actual fiber coupling state.

[0007] Therefore, existing fiber optic coupling end-face contamination detection still needs further improvement in areas such as batch detection location reproduction, assessment of the actual functional impact of contaminants, identification of contaminant state changes, and risk assessment of the coupling process. Summary of the Invention

[0008] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for assessing the contamination level of the fiber coupling end face of an optical fiber lead frame. This method comprehensively considers the coupling relationship between contaminants and the mode field, as well as the evolutionary impact of contaminants during the actual coupling process, to achieve an accurate assessment of the contamination risk of the fiber coupling end face.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a method for assessing the contamination level of the fiber coupling end face of an optical fiber lead frame, comprising: Step S1: Acquire a global image of the optical fiber lead frame, establish a frame coordinate system, determine the end face coordinates of the optical fiber coupling end face based on the frame coordinate system, and establish a local end face coordinate system. Step S2: Acquire multi-source image data of the target fiber coupling end face in the local coordinate system of the end face, and register the multi-source image data to obtain end face image data; Step S3: Process the end face image data to obtain a set of pollutant instances and pollutant morphology data; calculate the current functional pollution level based on the pre-acquired pattern field distribution data and pollution attenuation response data. Step S4: Determine the key pollutants based on the current functional pollution level, the pollutant instance set, and the pollutant morphology data; Step S5: Apply controlled micro-perturbation to the target optical fiber coupling end face containing the key pollutant, and calibrate the controlled micro-perturbation according to the actual action parameters at the end face; Step S6: Apply controlled micro-perturbations of different levels to the key pollutant and acquire end-face images after perturbation. Based on the changes in end-face images before and after perturbation, obtain the migration response data of the key pollutant. Step S7: Based on the migration response data and the pre-established contact evolution calibration data, predict the evolution results of the key pollutant under optical fiber coupling conditions, and obtain the evolved pollutant distribution data and damage probability data. Step S8: Based on the evolved contamination distribution data, the damage probability data, and the mode field distribution data, calculate the evolved functional contamination degree, and perform a comprehensive calculation based on the current functional contamination degree, the evolved functional contamination degree, and the pre-calculated maximum contact evolution risk to obtain the fiber coupling end face contamination assessment result.

[0010] Furthermore, the registration of the multi-source image data in step S2 specifically includes: Based on the end face coordinate information in the frame coordinate system, the multi-source image data obtained by different acquisition methods are converted to the end face local coordinate system; Based on the pollutant feature regions in different multi-source image data, pollutant instances corresponding to different acquisition methods are associated so that the same pollutant instance maintains a corresponding relationship in different multi-source image data. A unified dataset for end faces is established based on the multi-source image data after pollutant instance association is completed.

[0011] Further, step S3 calculates the current functional contamination level, specifically including: Based on the mode field distribution data of the target fiber coupling end face under the working wavelength in the clean state, the mode field energy distribution corresponding to different spatial locations is obtained, and mode field energy weight data is established based on the mode field energy distribution. Based on the light power attenuation data corresponding to different pollutant categories, pollutant sizes, and pollutant morphology parameters, pollution attenuation response data are established. Based on the spatial overlap between the pollutant instance coverage area and the model field energy distribution area, combined with the model field energy weight data and the pollution attenuation response data, the functional pollution contribution value corresponding to each pollutant instance is calculated, and the current functional pollution level is obtained.

[0012] Further, step S4 identifies the key pollutant, specifically including: Sort the pollutant instances according to their current functional pollution contribution values ​​in the pollutant instance set. Obtain pollutant morphology data corresponding to each pollutant instance and the positional relationship between each pollutant instance and the model field region; Based on the current functional pollution contribution value, the pollutant morphology data, and the location relationship, the key pollutants used for migration response characterization are selected from the pollutant instance set.

[0013] Furthermore, applying the controlled micro-perturbation in step S5 specifically includes: The actual action parameters at the location where the micro-perturbation is applied are collected, and the correspondence between the actual action parameters and the driving parameters of the micro-perturbation application device is established. The driving parameters of the micro-perturbation application device are adjusted according to the correspondence to ensure that the controlled micro-perturbation action applied to the key pollutant is within a preset range; The controlled micro-perturbation action is less than the end-face cleaning threshold and the end-face damage threshold, so as to keep the key contaminants in an unremoved state.

[0014] Further, obtaining the migration response data in step S6 specifically includes: Different levels of controlled micro-perturbations were applied to the key pollutant along at least two non-collinear directions, and end-face images were acquired after each perturbation. The location change data, morphological change data, and orientation change data of the key pollutant are determined based on the end face images before and after the disturbance. Based on the relationship between the controlled perturbation action at different levels and the position change data, the morphological change data, and the orientation change data, the migration sensitivity of the key pollutant is determined, and the migration response data is generated.

[0015] Furthermore, step S7, predicting the evolution of the key pollutants, specifically includes: Establish contact evolution calibration data, which includes the correspondence between contaminant type, contaminant size parameters, contaminant height parameters, and fiber optic coupling condition parameters and contaminant migration, morphological changes, and end-face damage status. The optical fiber coupling operating parameters include at least the end face gap parameter, the contact force parameter, and the lateral deviation parameter. Based on the migration response data, migration sensitivity, and current fiber coupling parameters corresponding to the key pollutants, the corresponding pollutant evolution characteristic parameters are determined in the contact evolution calibration data. Based on the pollutant evolution characteristic parameters, the evolution results of the key pollutants under the optical fiber coupling condition are predicted, and the post-evolution pollutant distribution data and the damage probability data are generated.

[0016] Further, step S8 calculates the post-evolutionary functional contamination degree, specifically including: The spatial distribution information of pollutants after evolution is determined based on the evolved pollution distribution data; Determine end-face damage risk information based on the damage probability data; The evolved contamination distribution data and the damage probability data are mapped to the spatial region corresponding to the model field distribution data, and the evolved functional contamination degree is calculated based on the model field distribution data.

[0017] Further, the fiber coupling end-face contamination assessment result obtained in step S8 specifically includes: Obtain the current fiber coupling operating condition parameters, which include at least a predetermined end face gap, a predetermined contact force, a predetermined contact speed, and a predetermined lateral deviation; Based on the migration response data of the key pollutants, the contact evolution calibration data, and the current fiber coupling operating parameters, the contact evolution risk corresponding to each key pollutant is calculated, and contact evolution risk data is generated. The maximum contact evolution risk is determined from the contact evolution risk data; The comprehensive contamination level is obtained by comprehensively calculating the current functional contamination level, the evolved functional contamination level, and the maximum exposure evolution risk. Based on the overall contamination level and the contact evolution risk data, the risk status and contamination level of the optical fiber coupling end face are determined; Based on the risk status, a corresponding handling result is generated, and the handling result includes at least one of the following: direct passage, re-inspection after cleaning, prohibition of coupling, and maintenance. Based on the comprehensive contamination level, the contact evolution risk data, the risk status, the contamination level, and the treatment results, the fiber optic coupling end face contamination assessment result is generated.

[0018] A system for assessing the contamination level of the fiber coupling end face of a fiber optic lead frame, used to perform the method for assessing the contamination level of the fiber coupling end face of a fiber optic lead frame as described above, comprising: The frame positioning module is used to acquire a global image of the optical fiber lead frame, establish a frame coordinate system, determine the end face coordinates of the optical fiber coupling end face based on the frame coordinate system, and establish a local end face coordinate system. A multi-source acquisition module, connected to the frame positioning module, is used to acquire multi-source image data of the target fiber-coupled end face in the local coordinate system of the end face, and to register the multi-source image data to obtain end face image data. The pollution identification and assessment module, connected to the multi-source acquisition module, is used to process the end-face image data to obtain a set of pollutant instances and pollutant morphology data; and to calculate the current functional pollution level based on the pre-acquired pattern field distribution data and pollution attenuation response data. A critical pollutant determination module, connected to the pollution identification and assessment module, is used to determine critical pollutants based on the current functional pollution level, the pollutant instance set, and the pollutant morphology data. The micro-perturbation control module, connected to the key pollutant determination module, is used to apply controlled micro-perturbations to the target optical fiber coupling end face containing the key pollutant, and to calibrate the controlled micro-perturbations according to the actual action parameters at the end face. The migration response analysis module is connected to the micro-perturbation control module and is used to apply controlled micro-perturbations of different levels to the key pollutant, and to collect end-face images after perturbation. Based on the changes in the end-face images before and after perturbation, the migration response data of the key pollutant is obtained. The evolution prediction module, connected to the migration response analysis module, is used to predict the evolution results of the key pollutant under optical fiber coupling conditions based on the migration response data and pre-established contact evolution calibration data, and to obtain the pollution distribution data and damage probability data after evolution. The evolution assessment module is connected to the evolution prediction module and the contamination identification and assessment module, respectively. It is used to calculate the functional contamination degree after evolution based on the contamination distribution data after evolution, the damage probability data, and the pattern field distribution data. It also performs a comprehensive calculation based on the current functional contamination degree, the functional contamination degree after evolution, and the pre-calculated maximum contact evolution risk to obtain the fiber coupling end face contamination assessment result.

[0019] The beneficial effects of this invention are: This invention correlates the spatial distribution, morphological characteristics, and pollution attenuation response of pollutants with the distribution of the model field, which can reflect the actual impact of different locations and types of pollutants on the optical fiber coupling performance and improve the accuracy of the current pollution status assessment.

[0020] This invention obtains migration response data of key pollutants by applying controlled micro-perturbations calibrated with actual action parameters, and can identify adhesion stability and migration risks that are difficult to reflect in static images.

[0021] This invention combines contact evolution calibration data to predict the evolution of contaminants under optical fiber coupling conditions, and correlates the contaminant distribution and end-face damage risk after evolution with the mode field distribution for calculation, which can identify the possible optical performance degradation and end-face damage that may occur during the coupling process in advance.

[0022] This invention provides a comprehensive assessment based on the current level of functional contamination, the level of functional contamination after evolution, and the maximum risk of contact evolution. It can provide a basis for the release, cleaning and re-inspection, prohibition of coupling, and maintenance of optical fiber coupling end faces, thereby improving the reliability of contamination assessment and disposal decisions. Attached Figure Description

[0023] Figure 1 This is a flowchart of the steps in the method for assessing the contamination level of the fiber coupling end face of the fiber optic lead frame in this invention. Figure 2 This is a schematic diagram of the fiber optic lead frame fiber coupling end face contamination assessment system in this invention.

[0024] Attached reference numerals: 1. Frame positioning module; 2. Multi-source acquisition module; 3. Pollution identification and assessment module; 4. Key pollutant identification module; 5. Micro-disturbance control module; 6. Migration response analysis module; 7. Evolution prediction module; 8. Evolution assessment module. Detailed Implementation

[0025] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0026] Example 1, as Figure 1 As shown in the figure, this embodiment discloses a method for assessing the contamination level of an optical fiber coupling end face. Without damaging the optical fiber coupling end face, the influence of the current state of the contaminant on the mode field, the adhesion and migration characteristics characterized by controlled micro-perturbation, the three-dimensional morphology of the contaminant, the contact evolution under predetermined optical fiber coupling conditions, and the influence of the mode field after contact evolution are sequentially correlated to form a closed loop for assessing the contamination level of the optical fiber coupling end face. This loop is used to determine at least one treatment outcome among direct passage, post-cleaning re-inspection, prohibited coupling, and maintenance.

[0027] Working principle: This embodiment sequentially performs framework-level positioning, multi-source data acquisition and unification, current functional pollution assessment, key pollutant screening, controlled micro-perturbation characterization, contact evolution prediction, back projection of evolution results, and comprehensive classification.

[0028] First, establish the frame coordinate system using the positioning holes and stamping reference edge on the fiber optic lead frame. And establish a local coordinate system for each fiber coupling end face. Frame coordinate system Used to determine the batch detection positions of multiple fiber optic coupling end faces, end face local coordinate system Used to unify bright-field images, dark-field images, axial focal plane images, mode field images, contaminant locations, and controlled micro-perturbation directions.

[0029] Secondly, the pixel-level existence probability of pollutants, pollutant category, pollutant local height, pollutant morphology parameters, and pollutant spatial location are spatially correlated with the mode field energy surface density and current pollutant attenuation response at the working wavelength to obtain the current functional pollution contribution value of each pollutant instance and the current functional pollution of the target fiber coupling end face.

[0030] Next, key pollutants are screened based on their current functional pollution contribution, pollutant morphology parameters, and positional relationship between the pollutants and high-energy regions of the model field. Controlled micro-perturbations below the end-face cleanliness and end-face damage thresholds are applied to the key pollutants along at least two non-collinear directions to obtain positional, morphological, and directional change data, further determining migration sensitivity and preferred migration directions.

[0031] Subsequently, a standard sample with the same or equivalent end face material, end face curvature, end face coating and coupling mechanism as the fiber under test was used for actual controlled coupling to establish contact evolution calibration data between pollutant type, size, volume, height, migration sensitivity and fiber coupling condition parameters and pollutant migration, spreading, fragmentation, embedding, indentation and scratching.

[0032] Furthermore, based on the migration response data, contact evolution calibration data, and current fiber coupling parameters of each key pollutant, the contact evolution risk corresponding to each key pollutant is calculated, and the contact evolution risks corresponding to each key pollutant are combined into contact evolution risk data; the maximum contact evolution risk is determined from the contact evolution risk data.

[0033] Finally, based on the measured characteristics of key pollutants and the predicted pollution probability map and damage probability map after evolution under predetermined fiber coupling conditions, the evolution results are mapped to the model field distribution area to obtain the functional pollution degree after evolution. The current functional pollution degree, the functional pollution degree after evolution, and the maximum exposure evolution risk are combined into a comprehensive pollution degree to determine the pollution level and treatment results.

[0034] The hardware configuration includes a frame positioning unit, a multi-source imaging unit, a mode field acquisition unit, a three-dimensional topography acquisition unit, a controlled micro-perturbation unit, an endpoint action detection unit, a controlled coupling unit, and an industrial computer.

[0035] The multi-source imaging unit includes a global shutter monochrome industrial camera with a resolution of no less than 2448×2048 pixels, a pixel size of no more than 3.45μm, and a bit depth of no less than 12 bits, as well as a microscope objective with a magnification of 20x, a numerical aperture of no less than 0.40, and a working distance of no less than 8mm. This configuration achieves an end-face image sampling accuracy of 0.18μm / pixel.

[0036] The lighting components include a coaxial bright field light source with a center wavelength of 530nm±20nm and a brightness stability of better than 1%, and a four-quadrant dark field light source with four independently controlled paths, adjacent incident azimuths differing by 90°, and incident angles adjustable within the range of 35° to 70°.

[0037] The mode field acquisition unit includes an InGaAs near-infrared camera with a response range covering 900–1700 nm, a 1310 nm semiconductor light source and a 1550 nm semiconductor light source with an output power of 1–5 mW and a power fluctuation of no more than 0.05 dB over 30 minutes, an optical switch, a beam splitter, a reference photodetector, a detection photodetector, and an optical path switching component.

[0038] The 1310nm and 1550nm semiconductor light sources select their operating wavelengths via an optical switch. The selected optical signal is split into a detection optical path and a reference optical path by a beam splitter. The reference optical path is connected to a reference photodetector to acquire the real-time reference power of the light source; the detection optical path is connected to the end of the single-mode fiber under test facing away from the fiber coupling endface, allowing the optical signal to be output from the fiber coupling endface. The output light from the fiber coupling endface is projected onto an InGaAs near-infrared camera via a microscopic imaging optical path to acquire mode field images. During transmitted light power measurement, the optical path switching component guides the output light to the detection photodetector, and the transmitted light power is corrected based on the ratio between the output of the detection photodetector and the output of the reference photodetector.

[0039] The frame positioning unit includes an XYθ precision platform and a Z-axis scanning platform. The XY-axis resolution of the XY-axis in the XYθ precision platform is no greater than 0.1 μm, and the repeatability error is no greater than 1 μm; the θ-axis resolution is no greater than 0.001°. The minimum step size of the Z-axis scanning platform is no greater than 0.1 μm, and the bidirectional repeatability error is no greater than 0.3 μm.

[0040] The controlled micro-perturbation unit includes a micro-airflow nozzle with an inner diameter of 0.20 mm and a pulse valve with a response time of no more than 0.5 ms and an effective pulse width adjustment range of 0.5–50 ms. The micro-airflow nozzle can change its orientation around the normal of the fiber coupling end face to form at least two non-collinear controlled micro-perturbation directions.

[0041] The endpoint action detection unit uses a dynamic pressure sensor with a range of not less than 20 kPa, a response time of not more than 0.05 ms, a sampling frequency of not less than 100 kHz, and a measurement error of not more than 1.5% of the full scale.

[0042] The contact force adjustment range of the controlled coupling unit is 0–1 N, with a force resolution of no more than 1 mN and a displacement resolution of no more than 0.1 μm. The industrial computer includes a processor with no fewer than 8 cores, no fewer than 32 GB of RAM, and a graphics processor with no fewer than 8 GB of video memory.

[0043] The aforementioned devices can be replaced by devices with the same or higher performance.

[0044] Basic operating parameters: The end-face image sampling accuracy was set to 0.18 μm / pixel. The scanning range of the axial focal plane stack was set to -15 μm to +15 μm, with a scanning step size of 1 μm, and a total of 31 layers of images were acquired. After repeated clamping of the fiber optic lead frame, the repeatability error of the same fiber coupling end face was no greater than 1 μm.

[0045] The distance between the micro-airflow nozzle and the fiber optic coupling end face was set to 8.0 mm, and the angle between the nozzle axis and the end face normal was set to 60°. The ambient temperature was maintained at 23℃±2℃, and the relative humidity was maintained at 45%~55%. The residual difference after registration of images obtained by different acquisition methods was no greater than 0.5μm.

[0046] Frame positioning and end face coordinate calibration: The fiber optic lead frame is installed in the platform fixture. The translation position and rotation angle are determined by two non-collinear positioning holes. The mirror image and orientation ambiguity are eliminated by stamping the reference edge, and the frame coordinate system is established. .

[0047] The nominal coordinates of the twelve fiber coupling end faces were obtained based on the design documents of the fiber optic lead frame. After the detection system moved to the vicinity of each nominal coordinate, edge extraction and circle fitting were performed on the fiber cladding circle, and the center of the fitted cladding circle was used as the local coordinate system of the corresponding end face. The origin of the frame coordinate system The horizontal axis is used as the local coordinate system of the end face. The initial transverse axis direction is determined and locally corrected based on the core boundary or mode field center.

[0048] The system saves the frame coordinate system. Local coordinate system of each end face Translation parameters, rotation parameters, and local correction parameters between them.

[0049] Repeat the clamping process 10 times for the same fiber optic lead frame. When the center reproduction error of any fiber coupling end face exceeds 1 μm, recalibrate the fixture and check the fit of the positioning holes until the repeat positioning error is no greater than 1 μm.

[0050] Model field, pollution attenuation, and altitude calibration: At least 10 fiber-coupled endfaces of the same type and in a clean state were selected, and 1310nm and 1550nm optical signals were input respectively to acquire mode field images. The mode field images were then subjected to dark current subtraction, bad pixel repair, saturation point removal, and background correction in sequence, and then normalized according to the total mode field energy. The normalized results of multiple clean endfaces at the same operating wavelength were averaged to obtain the mode field energy distribution data, and the sample differences were saved as the mode field uncertainty.

[0051] Calibration contaminants of known type, height, size, and coverage were set on the standard optical fiber coupling end face. These contaminants included hard particles, soft films or oil stains, adhesive residues, fiber debris, and salts or brittle particles. Transmitted light power was measured under both clean and contaminated conditions to establish current contamination attenuation response data between contaminant type, operating wavelength, local height, and morphological parameters and changes in light power.

[0052] When it is not possible to obtain calibration pollutants with continuous height or continuous size, multiple discrete standard points are used for calibration, and linear or monotonic interpolation is performed between adjacent standard points.

[0053] The Z-axis scanning position is calibrated using a standard sample with a known step height, within a calibration range of 1–10 μm. The height measurement error after calibration should not exceed 0.5 μm. If the height measurement error exceeds 0.5 μm, the Z-axis zero point and scanning step distance should be recalibrated.

[0054] Controlled perturbation action quantity and threshold calibration: During the offline calibration phase, the pressure-sensing surface of the dynamic pressure sensor is set in the spatial plane corresponding to the target fiber coupling end face, so that the center of the pressure-sensing surface coincides with the predetermined action point of the micro-airflow nozzle.

[0055] By changing the pulse valve opening, drive pulse width, and nozzle distance, pressure-time curves are collected, and the pressure within the effective action time is integrated to obtain the actual endpoint pressure impulse.

[0056] For each group of pulse valve openings and drive pulse widths, at least 10 measurements should be taken repeatedly, and the average value of the actual endpoint pressure impulse should be used as the calibration value of the corresponding drive parameter. If the coefficient of variation of 10 consecutive measurement results is greater than 5%, the corresponding calibration result should not be used.

[0057] After completing the offline calibration, remove the dynamic pressure sensor and install the fiber optic lead frame to be tested. While keeping the nozzle inner diameter, nozzle distance, nozzle angle, gas path structure, and environmental conditions unchanged, control the pulse valve according to the calibration relationship between the drive parameters and the actual endpoint pressure impulse.

[0058] When the dynamic pressure sensor is set at an adjacent position rather than the center of the target fiber coupling end face, a transfer coefficient is established by synchronous measurement of the end face center position sensor and the adjacent position sensor. The transfer coefficient is then used to convert the measurement results of the adjacent position into the actual end pressure impulse at the end face center position.

[0059] This embodiment uses three target pressure impulse levels: 45 Pa·s, 80 Pa·s, and 120 Pa·s. The allowable deviation for each level is ±5% of the target value, and the corresponding allowable ranges are 42.75–47.25 Pa·s, 76–84 Pa·s, and 114–126 Pa·s, respectively.

[0060] Before the start of each testing day, and after continuously testing 100 fiber optic coupling end faces or running continuously for 2 hours, the three target pressure impulse levels are reviewed. If any level exceeds the allowable range of ±5%, the calibration relationship between the drive parameters and the actual end-point pressure impulse is re-established.

[0061] The end-face cleaning threshold is defined as the minimum actual end-face pressure impulse when, under the same nozzle distance, nozzle angle, and environmental conditions, the same controlled micro-perturbation is repeatedly applied to a stable standard contaminant three times, and the contaminant detaches from the fiber-coupled end face, or the projected area of ​​the contaminant decreases by more than 10% compared to before the perturbation.

[0062] The end-face damage threshold is defined as follows: after applying the same controlled micro-perturbation three times to a clean standard end face, the damage must not exist before the perturbation and the projected area must be not less than 2 μm. 2 The minimum actual endpoint pressure impulse when there are scratches or indentations, or when the local height change exceeds 0.5μm.

[0063] The end-face cleaning threshold and end-face damage threshold were calibrated using at least 10 standard samples of the same material. The actual end-face pressure impulse was increased incrementally for each standard sample, with the difference between adjacent levels not exceeding 10 Pa·s. The 5th percentile of the pressure impulse that first met the cleaning criterion for each standard sample was taken as the end-face cleaning threshold, and the 5th percentile of the pressure impulse that first met the damage criterion for each clean standard end face was taken as the end-face damage threshold.

[0064] For the calibration of cleanliness thresholds, the stable standard contaminant, after being subjected to three consecutive controlled micro-perturbations at 120 Pa·s, exhibits a centroid displacement of no more than 0.5 μm and a projected area change rate of no more than 5%.

[0065] Under the nozzle structure, nozzle distance, nozzle angle, and environmental conditions of this embodiment, the end-face cleaning threshold is 150 Pa·s, and the end-face damage threshold is 180 Pa·s. The normal control limit for controlled micro-disturbances is 120 Pa·s, and considering a control tolerance of ±5%, the allowable limit is 126 Pa·s. The allowable limit is lower than both the end-face cleaning threshold and the end-face damage threshold.

[0066] When the end face material, end face coating, micro-disturbance medium, nozzle structure, or environmental conditions change, the end face cleaning threshold and end face damage threshold should be recalibrated.

[0067] Multi-source data acquisition and pollutant identification: Local coordinate system on the same end face The following steps were taken sequentially: coaxial bright field image, dark field images from four directions, 31-layer axial focal plane image, 1310nm mode field image, 1550nm mode field image, and reference optical power.

[0068] Dark current subtraction, flat-field correction, reflection suppression, and spatial registration are performed on multi-source images. First, based on the frame coordinate system... Data acquired using different acquisition methods are converted to the local coordinate system of the end face. Then, local corrections are made based on the cladding circle boundary, fiber core boundary, and end face fixed texture.

[0069] Using at least 10 identical clean fiber optic coupling end faces, establish clean reference mean images and clean reference standard deviation images corresponding to each acquisition method. For each pixel of the target fiber optic coupling end face, calculate the standardized absolute deviation of bright field grayscale, dark field response, and axial focal plane response relative to the clean reference data.

[0070] The standardized absolute deviation is calculated by dividing the absolute value of the difference between the target pixel value and the cleaning reference mean by the cleaning reference standard deviation. When the cleaning reference standard deviation is less than 0.5% of the corresponding full-scale gray value, 0.5% of the full-scale gray value is used as the denominator.

[0071] Bright-field normalized bias, dark-field normalized bias, and focal-area normalized bias are fused with weights of 0.4, 0.4, and 0.2, respectively. Each normalized bias is truncated to 0–6 before fusion and then divided by 6 to be converted to 0–1. The fusion result is used as the initial contamination probability of the corresponding pixel, thereby obtaining the initial contamination probability map.

[0072] Connected component segmentation is performed on adjacent pixels with an initial contamination probability of not less than 0.50 to obtain contaminant candidate instances. Isolated regions with an area of ​​less than 4 pixels that do not form an effective scattering response in any dark-field image are deleted as noise regions.

[0073] For each pollutant instance, the area, equivalent diameter, maximum height, average height, volume, aspect ratio, roundness, normalized edge sharpness, bright field grayscale features, dark field scattering features from all directions, and texture features are extracted to form a pollutant classification feature vector.

[0074] Five pollutant feature templates were established using standard pollutants of known categories. Each feature was standardized according to the mean and standard deviation of the database samples, and the standardized Euclidean distance between the pollutant instance to be classified and the feature centers of each category was calculated. For the k-th pollutant, the standardized Euclidean distance was denoted as... The corresponding index similarity is taken as Divide the index similarity of each category by the sum of the index similarities of the five categories to obtain the probability of the corresponding pollutant category. The category with the highest probability is taken as the pollutant category, and the corresponding probability is taken as the confidence level of the pollutant category.

[0075] Fiber debris typically exhibits a slender structure with an aspect ratio of at least 3; soft films or oil stains typically show a large coverage area, low edge sharpness, and a small average height; hard particles typically show high dark-field scattering, high edge sharpness, and localized protrusions; salt or brittle particles typically show irregular boundaries and multiple adjacent scattering peaks; adhesive residues typically show medium height, continuous coverage areas, and non-uniform texture. These characteristics are used to establish a classification template and are not considered as sole, exclusive classification criteria.

[0076] When a pollutant instance crosses the image boundary, coincides with a highly reflective saturated region, or has a local high-fit residual exceeding 0.5 μm, the output confidence of the corresponding pollutant instance is reduced and the relevant data is reacquired.

[0077] Current functional pollution assessment and screening of key pollutants: The initial contamination probability map is spatially registered with the mode field energy distribution data at two operating wavelengths. Current contamination attenuation response data, matching the contaminant category, operating wavelength, local height, and morphological parameters, are then retrieved to obtain the current functional contamination contribution value for each contaminant instance and the current functional contamination of the target fiber coupling end face. The specific calculation method is explained in Example 2.

[0078] The 90% cumulative mode field energy region is determined as follows: The mode field energy surface density corresponding to each pixel within the effective evaluation region is arranged in descending order. Starting from the highest energy pixel, the mode field energy is accumulated sequentially until the cumulative mode field energy reaches 90% of the total mode field energy for the corresponding operating wavelength. The pixel regions participating in the accumulation constitute the 90% cumulative mode field energy region for the corresponding operating wavelength. When multiple operating wavelengths are used, the union of the 90% cumulative mode field energy regions for each operating wavelength is taken as the key pollutant screening region.

[0079] Normalized edge sharpness is determined based on the average gradient magnitude at the boundary pixels of pollutant instances. The average gradient magnitude at the boundary of pollutant instances is divided by the 99th percentile of the gradient magnitude within the effective evaluation region of the same end face image, and the result is restricted to 0–1.

[0080] Any pollutant instance is identified as a critical pollutant if it meets at least one of the following conditions: the current functional pollution contribution value is not less than 0.08; the maximum height is not less than 1.5 μm; the shortest distance between the boundary of the pollutant instance and the 90% region of the cumulative model field energy is not greater than 10 μm; the pollutant category confidence is less than 0.85; the hard particle category confidence is not less than 0.80 and the normalized edge sharpness is not less than 0.70; the area covered by the pollutant instance overlaps with the 90% region of the cumulative model field energy; or the sum of the measurement errors of the maximum height of the pollutant and the 0.5 μm height reaches or exceeds the predetermined end face gap.

[0081] Undisturbed baseline images were acquired for each key pollutant. The micro-airflow nozzles were positioned along the frame coordinate system. Controlled micro-perturbations are applied to the horizontal and vertical axes, with the two directions being perpendicular to each other.

[0082] Three target pressure impulse levels of 45 Pa·s, 80 Pa·s, and 120 Pa·s were used sequentially in each direction. After each controlled micro-perturbation, the pulse valve was closed, and after waiting for no less than 0.5 s, bright-field images, dark-field images, and rapid axial focal plane images were acquired.

[0083] The contaminant instances before and after the perturbation are associated using the optimal matching method within a limited search range. For each contaminant instance before the perturbation, candidate contaminant instances are searched within a range of 20 μm centered on the original centroid.

[0084] The matching features include volume ratio similarity, equivalent diameter similarity, boundary shape similarity, texture cosine similarity, and neighboring pollutant topological similarity. Each similarity value is converted to a range of 0 to 1 and weighted with weights of 0.30, 0.20, 0.20, 0.20, and 0.10 respectively to obtain the instance matching confidence score. When multiple candidate pollutant instances correspond to the same instance before perturbation, a globally optimal matching method is used to maximize the sum of the matching confidence scores of all instances.

[0085] An instance match is considered valid if the confidence level is not lower than 0.85; otherwise, it is re-collected or manually reviewed.

[0086] Based on the images before and after the disturbance, obtain the centroid displacement, area change, principal axis rotation angle, number of fragments, grayscale change, maximum height change, and instance matching confidence of key pollutants.

[0087] When the centroid displacement is no greater than 0.5 μm, it is treated as undetected migration. When the centroid displacement is greater than 0.5 μm, and the angle between adjacent displacement directions corresponding to different pressure impulse levels is no greater than 30°, the migration response is considered to have directional consistency.

[0088] The system determines migration sensitivity based on position, shape, and direction changes under different application directions and pressure impulse levels, and identifies the direction with higher migration sensitivity as the preferred migration direction. The specific calculation method is explained in Example 3.

[0089] Contact evolution calibration and prediction: The contact evolution calibration data were established using standard samples that were identical or equivalent to the fiber under test in terms of coupling end face material, end face curvature, end face coating, and coupling mechanism.

[0090] The pollutant categories are stratified according to at least three height ranges, three volume ranges, three end face gap levels, and three contact force levels, with at least three standard samples set for each parameter combination.

[0091] The end face gaps include 0μm, 1.5μm and 3.0μm; the contact forces include 0.15N, 0.30N and 0.45N; the lateral deviations include 0μm, 2μm and 5μm; and the contact speeds include 0.1mm / s, 0.5mm / s and 1.0mm / s.

[0092] The contact evolution calibration data record pollutant type, equivalent diameter, volume, maximum height, initial position, migration sensitivity, preferred migration direction, end face gap, contact force, contact velocity and lateral deviation, and record the corresponding pollutant migration vector, spreading area, fragment particle distribution, embedding state, indentation state, scratch state and optical power loss.

[0093] During online prediction, based on measured data of key pollutants and predetermined fiber optic coupling conditions, matching pollutant evolution characteristic parameters are determined from contact evolution calibration data to generate post-evolution pollution probability maps and damage probability maps. The specific calculation method is explained in Example 3.

[0094] Evolutionary backprojection and comprehensive classification: The post-evolutionary contamination probability map and damage probability map are mapped to the model field distribution regions of 1310 nm and 1550 nm to obtain the post-evolutionary functional contamination degree. The comprehensive contamination degree is obtained based on the current functional contamination degree, the post-evolutionary functional contamination degree, and the maximum contact evolution risk.

[0095] When the overall pollution level is below 0.25, the maximum exposure evolution risk is below 0.35, and the output confidence level is not below 0.85, it is determined to be low risk, and the treatment result is direct pass.

[0096] If the overall pollution level is not less than 0.25 and less than 0.55, or the maximum exposure evolution risk is not less than 0.35 and less than 0.70, it is determined to be medium risk, and the treatment result is a re-inspection after cleaning.

[0097] If the overall pollution level is not less than 0.55, or the maximum risk of exposure and evolution is not less than 0.70, it is identified as high risk, and the disposal result is prohibition of coupling.

[0098] If the maximum probability of damage after cleaning is not less than 0.50, or if irreversible end-face damage is detected, it is determined as a damage risk, and the handling result is repair or scrapping.

[0099] If the output confidence level is below 0.85 and the conditions for damage risk or high risk are not met, it is determined to be an uncertain risk, and the handling result is re-collection, supplementary calibration or manual review.

[0100] When the same fiber coupling end face simultaneously meets two or more risk status conditions, the final contamination level is determined in the order of damage risk, high risk, uncertain risk, medium risk, and low risk. If the high risk threshold is reached but the output confidence level is below 0.85, the output coupling is prohibited and manual verification is required; otherwise, the output is not output and the product is passed directly or cleaned and re-inspected.

[0101] Inclusion criteria for contact evolution calibration data: The contact evolution calibration samples are divided into a library construction set and an independent validation set according to the pollutant preparation batch. The library construction set accounts for 80% of the calibration samples, and the independent validation set accounts for 20%. The same pollutant preparation batch cannot be included in both the library construction set and the independent validation set at the same time.

[0102] The library construction set is used to establish migration kernels, spreading kernels, fragmentation kernels, displacement kernels, damage kernels, material hazard coefficients, and model parameters. The independent validation set is used only to evaluate the library construction results and is not used to readjust kernel parameters, weights, risk thresholds, or model parameters.

[0103] Contact evolution calibration data can be used for online high-confidence prediction only when the corresponding pollutant category and fiber coupling operating condition range simultaneously meet the following conditions: the median crossover ratio between the predicted and actual polluted areas is not less than 0.75; the median crossover ratio between the predicted and actual damaged areas is not less than 0.70; the median absolute error of the predicted pollutant core migration distance is not greater than 2 μm; and the median error of the predicted migration direction is not greater than 20°.

[0104] If any condition is not met, the corresponding pollutant category and fiber optic coupling operating condition range will not be included in the high-confidence calibration data coverage.

[0105] The detection system passes repeatability admission when the coefficient of variation of the overall contamination degree is not greater than 10% after three consecutive measurements on the same fiber coupling end face, and the consistency rate of contamination level after repeated clamping of the fiber lead frame is not less than 95%.

[0106] The online contact evolution calibration data saves information on the batches of samples used for library construction, the batches of samples used for validation, kernel parameters, model parameters, access ranges, and version information. After independent validation is completed, the corresponding version is locked, and parameters are not adjusted for individual samples during online testing.

[0107] Technical effects: By transforming the coordinates between the frame coordinate system and the local coordinate system of the end face, the detection position and direction of action of the same fiber coupling end face can be reproduced after repeated clamping of the fiber lead frame, multi-source imaging, and multiple controlled micro-perturbations.

[0108] By correlating the pixel-level presence probability, spatial location, category, local height, and morphological parameters of pollutants with the mode field energy distribution and the current pollution decay response, it is possible to distinguish pollution states with similar geometric areas but different optical effects.

[0109] By measuring the pressure-time curve of the corresponding spatial position of the fiber optic coupling end face and obtaining the actual end pressure impulse, the micro-disturbance input error caused by differences in valve source pressure, gas path status, nozzle distance and pulse valve response can be reduced.

[0110] By applying controlled micro-perturbations of multiple levels along at least two non-collinear directions, weak attachment states and potential migration trends that are difficult to reflect in static images can be identified.

[0111] By establishing contact evolution calibration data using actual controlled coupling standard samples, the prediction of pollutant migration, spreading, fragmentation, indentation, and scratches can have a real physical data source.

[0112] By mapping the evolved contamination probability map and damage probability map to the mode field distribution region, the potential optical power loss and end-face damage risk after fiber coupling can be identified in advance.

[0113] Example 2, based on Example 1, further illustrates multi-source image data registration, pollutant instance association, establishment of a unified end-face dataset, calculation of current functional pollution level, and key pollutant screening methods. The tested object, detection system, environmental conditions, acquisition parameters, and controlled micro-perturbation threshold described in Example 1 are applicable to this example.

[0114] Working principle: This embodiment first uses the frame coordinate system The end-face coordinates in the model transform multi-source image data acquired through different acquisition methods into the corresponding end-face local coordinate system. Then, pollutant instances in different multi-source images are associated based on the pollutant feature regions, so that the same pollutant instance has a consistent instance identifier in different data.

[0115] After completing the pollutant instance association, the bright field response, dark field response, three-dimensional morphology, mode field location, reference optical power, and data quality parameters are combined into a unified end-face dataset.

[0116] The mode field energy surface density corresponding to different spatial locations is obtained based on the mode field distribution of the target fiber coupling end face under clean conditions; the current pollution attenuation response is obtained based on the optical power attenuation results corresponding to different pollutant types, operating wavelengths, local heights of pollutants, and pollutant morphology parameters.

[0117] The pixel-level existence probability of a contaminant instance, the energy surface density of the mode field, and the current contamination attenuation response are spatially integrated to obtain the current functional contamination contribution value of a single contaminant instance. The contribution values ​​of all contaminant instances are then nonlinearly synthesized to obtain the current functional contamination of the target fiber coupling end face.

[0118] Multi-source data acquisition: Local coordinate system on the same end face Next, coaxial bright field images, dark field images from at least two orientations, axial focal plane stack, mode field images at at least one working wavelength, and reference optical power are acquired sequentially.

[0119] When bright-field images, dark-field images, and axial focal plane images are insufficient to distinguish pollutant categories, polarization images, fluorescence images, or short-wave infrared images can also be acquired.

[0120] All data are saved, including acquisition time, end face coordinates, light source status, exposure parameters, and reference optical power, for data quality checks and cross-modal correlation.

[0121] Multi-source image registration: According to the frame coordinate system The end-face coordinate information is used to transform multi-source image data obtained by different acquisition methods into the end-face local coordinate system. .

[0122] After coordinate transformation, the overall position is corrected by using the cladding circle boundary, fiber core boundary or end face fixed texture, and then the local position is corrected according to the pollutant edge, pollutant core, texture features and scattering direction.

[0123] The registration residual threshold for multi-source image data is 0.5 μm. If the registration residual is no greater than 0.5 μm, proceed to the pollutant instance association step; if the registration residual is greater than 0.5 μm, re-execute end-face boundary extraction, local feature matching, or image acquisition.

[0124] Image registration confidence is determined based on the registration residual. When the registration residual is no greater than 0.25 μm, the image registration confidence is set to 1; when the registration residual is greater than 0.25 μm but no greater than 0.5 μm, the image registration confidence decreases linearly from 1 to 0.85; when the registration residual is greater than 0.5 μm, it is not included in subsequent calculations.

[0125] Pollutant instance association and end-face unified dataset: Based on the pollutant feature regions in different multi-source image data, pollutant instances corresponding to different acquisition methods are associated.

[0126] When the centroid distance between two pollutant feature regions is no greater than 0.5 μm, and the intersection-union ratio (IUU) of their projected regions is no less than 0.60, they are preferentially identified as the same pollutant instance. When the pollutant boundary varies significantly under different acquisition methods, a secondary judgment is made by combining the pollutant category probability, texture features, and height position.

[0127] The same pollutant instance is assigned the same instance identifier in different multi-source image data. The data after pollutant instance association is completed forms a unified dataset.

[0128] The unified dataset for the end face includes raw and corrected data from various acquisition methods, and the frame coordinate system. Local coordinate system of end face The coordinate transformation parameters, pollutant instance identifiers, pollutant instance boundaries, multimodal features, registration residuals, instance association confidence, mode field distribution data, and reference optical power are included.

[0129] Initial pollution probability map, pollutant categories, and 3D topography configuration: The initial pollution probability map was obtained in accordance with the method described in Example 1. The pollutant instance set, pollutant category, and pollutant category confidence level.

[0130] Initial contamination probability map Each pixel value in the graph represents the probability of a contaminant being present at the corresponding spatial location. Instance separation is performed on the initial contamination probability map to obtain... A collection of pollutant instances consisting of *i* pollutant instances. The coverage area corresponding to the *i*th pollutant instance is denoted as . .

[0131] The i-th pollutant instance is located at coordinates The pixel-level existence probability at a given location is denoted as . Initial contamination probability map This represents the union of the probabilities of all pollutant instances at the corresponding pixel.

[0132] When there is no spatial overlap between contaminant instances, the probability of the corresponding pixel in the initial contamination probability map is equal to the pixel-level existence probability of the contaminant instance to which that pixel belongs. When the candidate regions of two or more contaminant instances overlap, the initial contamination probability map is determined according to the union of the existence probabilities of each instance; in the instance-level integration, overlapping pixels are assigned to the contaminant instance with the highest posterior existence probability, so that the integration regions of each contaminant instance do not overlap, avoiding the same pixel being repeatedly counted in the current functional contamination degree.

[0133] The contaminant height map is obtained by analyzing the peak value of focus sharpness as a function of axial position. Alternatively, it can be obtained using structured light data or white light interferometry data. The i-th contaminant instance is located at coordinates... The local height at point is denoted as .

[0134] Based on the area covered by pollutants and local height The maximum height, average height, equivalent diameter, area, volume, roundness, normalized edge sharpness, and texture features of pollutant instances are obtained.

[0135] The morphology parameter vector of the i-th pollutant instance is denoted as... It includes at least the equivalent diameter, volume, roundness, normalized edge sharpness, and texture features.

[0136] When the high-fit residual is no greater than 0.25 μm, the confidence level of pollutant morphology reconstruction is set to 1; when the high-fit residual is greater than 0.25 μm but no greater than 0.5 μm, the confidence level of pollutant morphology reconstruction decreases linearly from 1 to 0.85; when the high-fit residual is greater than 0.5 μm, the axial focal plane image is reacquired or structured light or white light interferometry data is used for verification.

[0137] Mode field energy surface density: Mode field images at the operating wavelength were acquired on the coupling endface of a clean fiber of the same type. After subtracting the background signal, the mode field energy was normalized to obtain the first... Mode field energy surface density at each operating wavelength .

[0138] Mode field energy surface density The unit is And satisfy its requirements within the effective assessment area. The area integral within the area is equal to 1.

[0139] This embodiment uses two operating wavelengths, 1310nm and 1550nm. When the two operating wavelengths are given the same evaluation weight, the corresponding wavelength weights are... All are 0.5. When only one operating wavelength is used, the corresponding wavelength weight is 1. The sum of the weights of all wavelengths is 1.

[0140] Current pollution attenuation response: A calibration contaminant with known material, height, size, and coverage was placed on the standard optical fiber coupling end face, and the transmitted light power was measured in both clean and contaminated states.

[0141] Based on the change in optical power between the polluted and clean states, the current pollution attenuation response corresponding to different pollutant types, operating wavelengths, local heights, and morphological parameters is obtained. .

[0142] The current pollution decay response is defined as the natural logarithmic power decay response corresponding to a unit model field energy overlap after normalization according to the energy overlap between the calibrated pollutant coverage area and the model field, and it is a dimensionless quantity.

[0143] When the original calibration results are expressed in dB, first multiply the dB loss by... It is converted into natural logarithmic power decay, and then normalized according to the energy overlap between the calibrated pollutant coverage area and the model field.

[0144] The calibration is performed in layers based on pollutant type, operating wavelength, local height, equivalent diameter range, volume range, and morphological parameter range. During online calculation, linear or monotonic interpolation is used between adjacent calibration points to obtain the current pollution attenuation response.

[0145] The equivalent diameter, total volume, roundness, normalized edge sharpness, and texture features of pollutants are used to select the corresponding response calibration layer and interpolation interval, and are no longer repeatedly applied as multiplication factors to the spatial integration result. The pollutant coverage area is represented by the integration region, and the local height of pollutants is represented by the local height parameter.

[0146] Current functional contamination level: The current functional contamination contribution value of the i-th contaminant instance and the current functional contamination of the target fiber coupling end face are calculated according to formula (1): ; In the formula, The contribution value of the current functional pollution level to the i-th pollutant instance; Current level of contamination; The number of working wavelengths used in the evaluation; For the first Wavelength weights corresponding to each working wavelength; The coverage area of ​​the i-th pollutant instance is expressed in μm. 2 ; For the i-th pollutant instance at coordinates The probability of existence at the pixel level at that location; For the first At each operating wavelength, the target fiber coupling end face is in coordinate... The energy surface density of the mode field at the location, in units of ; Let i be the pollutant category of the i-th pollutant instance; For the first Each operating wavelength is measured in nm. This is the local height, in μm. This is a vector of topographic parameters; This is a dimensionless response to the current pollution decay. This represents the total number of pollutant instances.

[0147] The pixel-level existence probability, wavelength weight, and current contamination decay response are all dimensionless quantities, and the unit of mode field energy surface density is... The unit of area micro-element is μm. 2 Therefore, the exponent of the exponential function in formula (1) is a dimensionless quantity.

[0148] The current functional pollution contribution value and the current functional pollution level both range from 0 to 1. A current risk heatmap is generated based on the current functional pollution contribution distribution of each pollutant instance.

[0149] Screening of key pollutants: Key pollutants were identified based on the cumulative mode field energy 90% region, normalized edge sharpness, and screening threshold as described in Example 1.

[0150] The number of key pollutants obtained through screening is denoted as Key pollutant data includes instance identifier, end face coordinates, current functional pollution contribution value, pollutant category, pollutant category confidence level, maximum height, volume, equivalent diameter, normalized edge sharpness, and positional relationship with the 90% region of cumulative model field energy.

[0151] Controlled micro-perturbation input: The actual endpoint pressure impulse calibrated in Example 1 is used as the controlled perturbation action quantity, and the direction of action is represented by a unit direction vector.

[0152] The controlled micro-perturbation uses three target pressure impulse levels of 45 Pa·s, 80 Pa·s and 120 Pa·s, with corresponding allowable ranges of 42.75 to 47.25 Pa·s, 76 to 84 Pa·s and 114 to 126 Pa·s, respectively.

[0153] The end-face cleaning threshold is 150 Pa·s, and the end-face damage threshold is 180 Pa·s. The upper limit of the controlled micro-perturbation, 126 Pa·s, is lower than both the end-face cleaning and end-face damage thresholds.

[0154] Data output: This embodiment outputs a unified dataset of the end face and an initial contamination probability map. Pollutant instance set, pollutant category Pollutant morphology parameter vector Mode field energy surface density Current pollution attenuation response Current functional pollution contribution value Current functional contamination level Current risk heat map, key pollutant data, and actual endpoint pressure impulse and direction of action.

[0155] The above data serves as the input for migration sensitivity calculation and contact evolution prediction in Example 3.

[0156] Technical effects: By converting data obtained from different acquisition methods to the local coordinate system of the end face and performing cross-modal correlation on the same pollutant instance, it is possible to ensure that the bright field features, dark field features, three-dimensional morphology and mode field positional relationships have consistent spatial identification.

[0157] By employing the energy surface density of the model field, pollutants are assigned different spatial weights at different energy locations in the model field.

[0158] By employing the current contamination attenuation response normalized according to the mode field energy overlap, the total optical power loss of the standard sample can be avoided from being repeatedly distributed to each pixel.

[0159] By comprehensively calculating the pixel-level existence probability, spatial location, pollutant category, local height, and morphological parameters of pollutants using formula (1), it is possible to distinguish pollution states with the same geometric area but different optical effects.

[0160] By setting the current functional pollution contribution value, pollutant height, migration warning distance, pollutant category confidence level, and hard feature threshold, key pollutants with high current optical influence, which may cross the end face gap, or which pose a risk of damage can be screened.

[0161] Example 3, based on Examples 1 and 2, further illustrates the generation of migration response data, calculation of migration sensitivity, establishment of contact evolution calibration data, calculation of contact evolution risk, generation of post-evolution contamination probability map and damage probability map, calculation of post-evolution functional contamination degree, and comprehensive contamination assessment method.

[0162] The end-face local coordinate system, controlled perturbation action range, end-face cleaning threshold, end-face damage threshold, and fiber coupling operating condition parameters determined in Example 1, as well as the initial contamination probability map, contaminant instance set, contaminant category, contaminant morphology parameters, mode field energy surface density, current contamination attenuation response, current functional contamination level, and key contaminant data obtained in Example 2, are used as inputs for this example.

[0163] Working principle: In this embodiment, at least two levels of controlled micro-perturbations are applied to key pollutants along at least two non-collinear directions. The migration sensitivity and preferred migration direction of the key pollutants are determined based on the changes in position, shape and direction in the front and rear end face images before and after the perturbation.

[0164] Contact evolution calibration data are established using actual controlled coupling standard samples. Correspondence is established between pollutant type, pollutant size, pollutant volume, pollutant height, initial position, migration sensitivity, as well as end face gap, contact force, contact velocity and lateral deviation, and pollutant migration, spreading, fragmentation, embedding, indentation and scratch results.

[0165] During online prediction, based on the migration response data, migration sensitivity, and current fiber optic coupling conditions of key pollutants, the pollutant evolution characteristic parameters are determined from the contact evolution calibration data, and the evolved pollution probability map and damage probability map are generated.

[0166] The post-evolutionary contamination probability map and damage probability map are remapped to the model field distribution area to obtain the post-evolutionary functional contamination level. The current functional contamination level, the post-evolutionary functional contamination level, and the maximum exposure evolution risk are further synthesized into a comprehensive contamination level, which is used to determine the contamination level and treatment outcome.

[0167] Migration response data: For each critical pollutant, at least two levels of controlled micro-perturbations are applied along at least two non-collinear directions. This embodiment uses two mutually perpendicular application directions and three target pressure impulse levels of 45 Pa·s, 80 Pa·s, and 120 Pa·s.

[0168] After each controlled micro-perturbation, a new end-face image is acquired and transformed to the same local coordinate system of the end face. The image registration residual is no greater than 0.5 μm. If the registration residual exceeds 0.5 μm, the corresponding response data is not included in the migration sensitivity calculation.

[0169] The pollutant instance matching before and after the disturbance was completed in accordance with the method described in Example 1. For each key pollutant, the centroid coordinates, centroid displacement, principal axis direction, principal axis rotation angle, projected area, area change, number of pollutant instances, number of fragments, grayscale change, maximum height change, average height change, instance matching confidence, and displacement direction were obtained.

[0170] When the centroid displacement is no greater than 0.5 μm, it is treated as migration not detected. When the centroid displacement is greater than 0.5 μm, and the angle between adjacent displacement directions corresponding to different pressure impulse levels is no greater than 30°, the migration response is considered to have directional consistency.

[0171] Migration sensitivity: Actual endpoint pressure impulse Divide by reference pressure impulse The normalized pressure impulse is obtained. Reference pressure impulse The normalized pressure impulse is 120 Pa·s. The normalized pressure impulses are arranged in ascending order and include the baseline state without applied controlled micro-perturbations.

[0172] The dimensionless speciation index of the i-th key pollutant is marked as The morphological change index consists of the relative change in projected area, the normalized value of the principal axis rotation angle divided by 180°, the normalized value of the number of newly added fragments divided by 1 and the sum of the number of newly added fragments, the normalized value of the average gray level change divided by the standard deviation of the gray level of the clean reference image, and the normalized value of the maximum height change divided by 1 μm.

[0173] All five components are limited to a range of 0 to 1. In this embodiment, the five components are assigned the same weight, with each weight being 0.2. When a pollutant category does not have a corresponding morphological change characteristic, the weight of the corresponding component is reset to 0, and the weights of the remaining components are renormalized to keep the morphological change index within the range of 0 to 1.

[0174] The migration sensitivity of the i-th critical pollutant is calculated according to formula (2): ; In the formula, Let be the migration sensitivity of the i-th critical pollutant; For the set of directions of controlled perturbation actions; The unit direction vector of the controlled micro-perturbation; Normalized pressure impulse; The centroid coordinate vector of the i-th critical pollutant after applying controlled micro-perturbations, in μm; The centroid coordinate vector before the disturbance; This is the equivalent diameter, in μm. As an indicator of morphological changes; The weight for morphological changes is set to 0.3 in this embodiment; This is the directional consistency coefficient; The robust slope of the response quantity relative to the normalized pressure impulse; This indicates that the calculation result will be limited to 0 to 1.

[0175] The directional consistency coefficient is determined by the arithmetic mean of the cosine of the angle between the displacement directions of adjacent pressure impulse levels, and the result is limited to 0 to 1.

[0176] In formula (2), the ratio of centroid displacement to equivalent diameter, morphological change index, normalized pressure impulse and directional consistency coefficient are all dimensionless quantities, therefore the migration sensitivity is a dimensionless quantity.

[0177] Robust slopes are fitted using an unperturbed baseline and at least two non-zero pressure impulse levels. When three non-zero pressure impulse levels exist, Theil-Sen slope estimation or linear regression with outlier suppression capabilities is used.

[0178] Of the two non-collinear directions, the one with higher migration sensitivity is determined as the preferred migration direction. When the difference in migration sensitivity between the two directions is less than 0.05, the preferred migration direction is determined based on the resultant direction of the centroid displacement vectors in each direction.

[0179] A migration sensitivity below 0.20 is defined as low migration sensitivity; a migration sensitivity between 0.20 and 0.60 is defined as medium migration sensitivity; and a migration sensitivity above 0.60 is defined as high migration sensitivity.

[0180] Contact evolution calibration data: The contact evolution calibration data were established using standard samples that were identical or equivalent to the fiber under test in terms of coupling end face material, end face curvature, end face coating, and coupling mechanism.

[0181] Input features for contact evolution calibration data include contaminant type, contaminant equivalent diameter, contaminant area, contaminant volume, maximum contaminant height, initial contaminant location, migration sensitivity, preferred migration direction, end face gap, contact force, contact velocity, and lateral deviation.

[0182] The output of the contact evolution calibration data includes contaminant migration vectors, paving area, fragment particle distribution, embedding state, indentation area and depth, scratch length and direction, and optical power loss.

[0183] Based on the actual controlled coupling results, migration nuclei, spreading nuclei, fragmentation nuclei, dislocation nuclei, and damage nuclei were established. Each evolutionary nucleus was defined within the effective evaluation region of the end-face local coordinate system. Within, and uses dimensionless pixel-level conditional probability representation.

[0184] The initial contamination probability map, the post-evolutionary contamination probability map, the damage probability map, and each evolutionary kernel were all converted to the same 0.18 μm / pixel grid as the end face image. After rotation or scaling of the evolutionary kernels, bilinear interpolation was used for resampling.

[0185] Material risk factor The contact force index, contact velocity index, lateral deviation gain coefficient, and migration sensitivity gain coefficient were determined using a library set. Whether the standard sample exhibits indentation, scratches, or embedding was used as a binary objective, and normalized optical power loss was used as a continuous objective. Regression fitting was performed under the constraint that all parameters are non-negative, minimizing the sum of the cross-entropy loss between the contact evolution risk and the actual damage state, and the mean square error between the cross-entropy loss and the normalized optical power loss.

[0186] All model parameters are locked before independent validation begins and are not adjusted for individual online detection samples.

[0187] Contact evolution risk: The exposure evolution risk of the i-th critical pollutant is calculated according to formula (3): ; In the formula, The contact evolution risk of the i-th critical pollutant, and the contact evolution risk corresponding to each critical pollutant. Together, they form the contact evolution risk data, and the maximum value in the contact evolution risk data is denoted as the maximum contact evolution risk. ; The material hazard factor corresponding to the pollutant category; The volume of the pollutant is expressed in μm. 3 ; For volume reference value, 500 μm is used in this embodiment. 3 ; The maximum height of the pollutant, in μm; The predetermined end face gap is expressed in μm. The uncertainty in height measurement is set to 0.5 μm in this embodiment; The standard S-shaped contact probability function, , For the dimensionless independent variable of the S-shaped contact probability function; The predetermined contact force is expressed in Newton-hours (N). For the contact force reference value, it is taken as 0.30N in this embodiment; The predetermined contact speed is expressed in mm / s. For the reference value of contact speed, 0.5 mm / s is used in this embodiment; The predetermined lateral deviation is expressed in μm. The lateral deviation reference value is 2μm, which is taken in this embodiment; Contact force index; The contact speed index; This is the lateral deviation gain coefficient; This is the migration sensitivity gain coefficient; For transfer sensitivity, , , , , All of these are dimensionless parameters.

[0188] Maximum height of pollutants For the corresponding local height within the pollutant coverage area The maximum value.

[0189] In formula (3), the volume ratio, the normalized value of the height and gap difference, the contact force ratio, the contact speed ratio, the lateral deviation ratio, and the migration sensitivity gain term are all dimensionless quantities. Therefore, the exponent in the exponential function is a dimensionless quantity.

[0190] Exposure evolution risk of the i-th critical pollutant A value below 0.35 is considered low contact evolution risk; contact evolution risk A value not lower than 0.35 and lower than 0.70 is defined as medium contact evolution risk; contact evolution risk A value of 0.70 or higher indicates a high risk of contact evolution.

[0191] Neighbor selection and evolutionary kernel determination: Continuous features in the contact evolution calibration data are standardized. Each continuous feature is subtracted from the mean of the corresponding feature in the database set, and then divided by the standard deviation of the corresponding feature in the database set.

[0192] The continuous characteristics include the ratio of the maximum height of the contaminant to the predetermined end face gap, the volume of the contaminant, the equivalent diameter of the contaminant, the migration sensitivity, the ratio of the predetermined contact force to the contact force reference value, the ratio of the predetermined contact speed to the contact speed reference value, and the ratio of the lateral deviation to the lateral deviation reference value.

[0193] Labeled samples with inconsistent pollutant categories are not included in the neighbor selection. When categories are consistent, the Euclidean distance between standardized continuous features is used as the standardized feature distance.

[0194] Select 3 to 7 neighboring samples with the smallest standardized feature distance from the contact evolution calibration data. Preferably, the number of neighboring samples should be 5; if there are fewer than 5 but not less than 3 valid neighboring samples, all valid neighboring samples should be used; if there are fewer than 3 valid neighboring samples, it is considered to be outside the coverage of the contact evolution calibration data.

[0195] The initial weight of the j-th neighbor sample is taken as its standardized feature distance plus... The reciprocal of the initial weights is then used to divide the initial weights of all neighboring samples by the sum of the initial weights to obtain the normalized neighboring sample weights.

[0196] The migration nuclei in neighboring samples are rotated according to the angle between the preferred migration direction of neighboring samples and the preferred migration direction of the current critical pollutant. The scale of the nucleus in the end-face plane is adjusted according to the ratio of the equivalent diameter of the current critical pollutant to the equivalent diameter of pollutants in neighboring samples; the height correlation amplitude of the nucleus is adjusted according to the ratio of the maximum height of the current critical pollutant to the maximum height of pollutants in neighboring samples. The scale adjustment results are limited to the valid calibration range of the contact evolution calibration data.

[0197] The conditional weights corresponding to migration, spreading, fragmentation, and damage are determined by the sum of the normalized neighboring sample weights for the corresponding evolutionary states. When the sum of the conditional weights for migration, spreading, and fragmentation exceeds 1, normalization is performed according to the sum of the three conditional weights to make the sum of the three conditional weights equal to 1.

[0198] When the standardized nearest sample distance is no greater than 1.0, the coverage confidence of the contact evolution calibration data is set to 1; when the standardized nearest sample distance is greater than 1.0 but no greater than 1.5, the coverage confidence decreases linearly from 1 to 0.85; when the standardized nearest sample distance is greater than 1.5, the coverage confidence is lower than 0.85, and high confidence is not allowed.

[0199] Evolutionary state determination: When the displacement response slope corresponding to the unit normalized pressure impulse is less than 0.10 and the maximum height of the contaminant is less than the predetermined end face gap, it is preferentially determined to be in the holding state, and the contamination probability of the contaminant in its original position is retained.

[0200] When the centroid displacement is greater than 0.5 μm and has consistent direction, it is determined to be in a migration state. The migration nucleus is rotated to the preferred migration direction, and the corresponding probability is subtracted from the original position of the pollutant and the corresponding probability is added at the predicted target position.

[0201] When the pollutant is a soft film or adhesive residue, and the projected area increases with controlled micro-disturbance or contact, it is determined to be in a paving state. A paving core is used to increase the coverage area and reduce the average height, so that the volume of pollutants before and after paving is approximately the same.

[0202] When the pollutant is a salt or brittle particle, and the number of pollutant instances increases while the area of ​​a single instance decreases, it is determined to be in a fragmented state. Multiple sub-particles are generated using fragmentation nuclei, and the total volume before and after fragmentation is kept consistent within the measurement error range.

[0203] When the maximum height of the hard particles reaches or exceeds the predetermined end face gap, and the predetermined contact force is within the corresponding calibration range, an indentation risk is identified; when there is lateral deviation or tangential displacement, a scratch risk is further identified.

[0204] The weight of the dislocation condition is not determined as an independent evolutionary state weight. The weight of the dislocation condition is determined based on the proportion of contamination probability that actually leaves the original location during migration, spreading, and fragmentation.

[0205] When migration occurs only, the area integral of the displacement probability deduction kernel within the effective assessment area is equal to the area integral of the migration probability increment kernel. When paving or fragmentation occurs, the displacement probability deduction kernel is determined based on the reduction in the original site contamination probability in neighboring samples, and the contaminant volume is approximately conserved within the height measurement error range.

[0206] The spatial evolution of hard particles is included in the post-evolution contamination probability map, while indentations, scratches, or embeddings caused by hard particles are only included in the damage probability map.

[0207] Post-evolutionary contamination probability map and damage probability map: The evolved contamination probability map and damage probability map are generated according to formula (4): ; In the formula, This is the initial pollution probability map; This is a probability diagram of contamination after evolution; This is a damage probability map; The quantity of key pollutants; , , , and The conditional weights are, in order, the migration, spreading, fragmentation, displacement, and damage states; , , , and The nuclei are, in order, the migrating nucleus, the spreading nucleus, the fragmented nucleus, the displaced nucleus, and the damaged nucleus. In formula (4), the same characters as in formula (3) have the same meaning.

[0208] All conditional weights and evolutionary kernels are dimensionless. The sum of the conditional weights for migration, spreading, and fragmentation is no greater than 1. The value range is 0 to 1.

[0209] In the pure migration state, the resampled migration probability increment kernel and displacement probability deduction kernel are normalized to make their total pixel probabilities equal. In the spreading and fragmentation states, the kernel amplitude is adjusted based on the pollutant height map and the actual area of ​​a single pixel to ensure that the estimated volume before and after evolution remains consistent within the height measurement error range.

[0210] The evolution results exceed the effective evaluation area The portion of the record is the result of migration outside the end face and is deducted from the initial contamination probability map, and does not participate in the subsequent mode field space integration.

[0211] The truncation in formula (4) is only used to eliminate minor out-of-bounds errors caused by numerical interpolation, and is not used to replace the probability conservation constraints of the evolutionary kernel.

[0212] When the number of key pollutants When the value is 0, the probability of contamination after evolution is shown in the diagram. Equal to the initial pollution probability map Damage probability map It equals 0.

[0213] Post-evolutionary pollution attenuation response and damage attenuation response: Post-evolutionary pollution decay response Compared with the current pollution decay response in Example 2 Using the same normalization caliber, all responses represent the natural logarithmic power decay response corresponding to the energy overlap of a unit mode field.

[0214] Damage attenuation response Normalized according to the overlap between the indentation, scratch, or embedding region and the mode field energy, representing the natural logarithmic power decay response corresponding to a unit mode field energy overlap.

[0215] The current pollution attenuation response, the evolved pollution attenuation response, and the damage attenuation response all use the optical power of the clean fiber coupling end face as a common reference, and all are dimensionless quantities.

[0216] Post-evolutionary functional pollution level and overall pollution level: The functional pollution level and overall pollution level after evolution are calculated according to formula (5): ; In the formula, The degree of functional contamination after evolution; The overall pollution level; To effectively assess the area; This represents the response to pollution decay after evolution. This is a damage attenuation response; The damage weight is set to 1.2 in this embodiment; Current level of contamination; The maximum contact evolution risk determined from contact evolution risk data; , and The weights are, in order, the current functional contamination weight, the evolved functional contamination weight, and the maximum contact evolution risk weight. In formula (5), the same characters as in formula (1) have the same meaning.

[0217] When the number of key pollutants When greater than 0, the maximum contact evolution risk Take the contact evolution risk corresponding to each key pollutant from the contact evolution risk data. The maximum value in; when the number of key pollutants When equal to 0, the maximum contact evolution risk Take 0.

[0218] In this embodiment, , and We take values ​​of 0.25, 0.50, and 0.25 respectively. All three weights are greater than 0, and the sum of the three weights is 1.

[0219] The three overall weights are determined through the database set, under the constraint that all three weights are greater than 0 and the sum of the weights is 1, and are locked before independent validation begins. The overall weights are not adjusted for individual samples during online testing.

[0220] In formula (5), the unit of the energy surface density of the mode field is The unit of area micro-element is μm. 2 The remaining probabilities, decay responses, and weights are all dimensionless, therefore the exponent in the exponential function is also dimensionless.

[0221] Output confidence level, pollution level, and treatment results: The output confidence level is determined based on the confidence level of pollutant category, the confidence level of pollutant morphology reconstruction, the confidence level of image registration, the confidence level of instance matching before and after perturbation, and the confidence level of contact evolution calibration data coverage.

[0222] When multiple pollutant instances exist, the output confidence score is the minimum value among all the confidence scores corresponding to the pollutant instances that participated in the current functional pollution level calculation and contact evolution prediction.

[0223] When no key pollutants are identified but instances of common pollutants exist, the confidence scores for instance matching and contact evolution calibration data coverage are not used. Instead, the output confidence score is the minimum value among the pollutant category confidence score, pollutant morphology reconstruction confidence score, image registration confidence score, and image quality confidence score of all common pollutant instances.

[0224] When no contaminant instances are detected, the output confidence level is determined based on image exposure, image sharpness, registration residual, mode field image validity, and reference optical power stability. If any data quality item is below 0.85, a high-confidence pass result is not output.

[0225] When the overall pollution level is below 0.25, the maximum exposure evolution risk is below 0.35, and the output confidence level is not below 0.85, it is determined to be low risk, and the output can be directly passed.

[0226] If the overall contamination level is not less than 0.25 and less than 0.55, or the maximum exposure evolution risk is not less than 0.35 and less than 0.70, it is determined to be of medium risk, and a re-inspection is conducted after cleaning.

[0227] When the overall contamination level is not less than 0.55, or the maximum contact evolution risk is not less than 0.70, it is identified as high risk, and output coupling is prohibited.

[0228] After cleaning, calculate the maximum pixel probability within the effective assessment area of ​​the damage probability map, and record this maximum pixel probability as the maximum damage probability. If the maximum damage probability is not less than 0.50, or if irreversible indentations, scratches, or embeddings are detected, it is determined as a damage risk, and repair or scrapping is output.

[0229] If the output confidence level is below 0.85 and the conditions for damage risk or high risk are not met, it is determined to be an uncertain risk, and the output is re-acquired, recalibrated, or manually reviewed.

[0230] When the same fiber coupling end face simultaneously meets two or more risk status conditions, the final contamination level is determined in the order of damage risk, high risk, uncertain risk, medium risk, and low risk.

[0231] If the high-risk threshold is reached but the output confidence level is below 0.85, the output should not be coupled and should be manually reviewed. Alternatively, the output should be passed directly without output or cleaned and then re-inspected.

[0232] The system simultaneously outputs the current risk heatmap, the evolutionary risk heatmap, the post-evolutionary pollution probability map, the damage probability map, the highest risk pollutant, the maximum exposure evolution risk, the output confidence level, and the recommended treatment results.

[0233] Technical effects: By using at least two non-collinear directions and multiple pressure impulse levels, the possibility of missing migration trends caused by a single direction of action can be reduced.

[0234] By converting the relationship between centroid displacement and overall morphological changes relative to normalized pressure impulse into migration sensitivity, a unified comparison can be made of key pollutants of different sizes and categories.

[0235] By introducing a directional consistency coefficient, the impact of unstable directional responses caused by random image errors, local vibrations, or instance matching biases on migration judgment can be reduced.

[0236] By establishing migration, spreading, fragmentation, dislocation, and damage nuclei using actual controlled coupling standard samples, contact evolution prediction can be based on actual physical results.

[0237] By comprehensively considering the relationship between the maximum height of the contaminant and the predetermined end-face gap, the contaminant volume, material hazard factor, migration sensitivity, contact force, contact speed, and lateral deviation, the contact evolution risk under different fiber coupling conditions can be distinguished.

[0238] By using conditional evolution kernels and conditional weights, the evolution probability of the same pollutant in migration, spreading, and fragmentation states can be avoided from being repeatedly increased.

[0239] By recording the hard particle body and the end-face damage caused by the hard particle into the post-evolution contamination probability map and damage probability map respectively, the end-face damage can be avoided from being repeatedly included in the contamination decay and damage decay.

[0240] By remapping the evolved contamination probability map and damage probability map to the model field distribution area, the impact of contaminant migration, spreading, fragmentation, and end-face damage on fiber coupling performance can be identified in advance.

[0241] By comprehensively considering the current level of functional contamination, the level of functional contamination after evolution, and the maximum risk of contact evolution, we can avoid making release decisions based solely on the current static end-face status and provide traceable judgment criteria for direct passage, post-cleaning re-inspection, prohibited coupling, and maintenance.

[0242] A system for assessing the contamination level of the fiber coupling end face of a fiber optic lead frame is provided for performing the aforementioned method for assessing the contamination level of the fiber coupling end face of a fiber optic lead frame. Figure 2 As shown, it includes: Frame positioning module 1 is used to acquire a global image of the optical fiber lead frame, establish a frame coordinate system, determine the end face coordinates of the optical fiber coupling end face based on the frame coordinate system, and establish a local end face coordinate system. The multi-source acquisition module 2, connected to the frame positioning module 1, is used to acquire multi-source image data of the target fiber-coupled end face in the local coordinate system of the end face, and to register the multi-source image data to obtain end face image data. The pollution identification and assessment module 3 is connected to the multi-source acquisition module 2. It is used to process the end face image data to obtain a set of pollutant instances and pollutant morphology data. Based on the pre-acquired pattern field distribution data and pollution attenuation response data, it calculates the current functional pollution level of the pollutant instance set. The critical pollutant identification module 4 is connected to the pollution identification and assessment module 3 and is used to identify critical pollutants based on the current functional pollution level, pollutant instance set, and pollutant morphology data. The micro-perturbation control module 5 is connected to the key pollutant determination module 4. It is used to apply controlled micro-perturbation to the target optical fiber coupling end face containing key pollutants and to calibrate the controlled micro-perturbation according to the actual action parameters at the end face. The migration response analysis module 6 is connected to the micro-perturbation control module 5. It is used to apply controlled micro-perturbations of different levels to key pollutants and to collect end-face images after perturbation. Based on the changes in end-face images before and after perturbation, the migration response data of key pollutants are obtained. Evolution prediction module 7, connected to migration response analysis module 6, is used to predict the evolution results of key pollutants under optical fiber coupling conditions based on migration response data and pre-established contact evolution calibration data, and obtain pollution distribution data and damage probability data after evolution. The evolution assessment module 8 is connected to the evolution prediction module 7 and the contamination identification and assessment module 3 respectively. It is used to calculate the functional contamination degree after evolution based on the contamination distribution data, damage probability data and pattern field distribution data after evolution, and to perform a comprehensive calculation based on the current functional contamination degree, the functional contamination degree after evolution and the maximum contact evolution risk to obtain the fiber coupling end face contamination assessment result.

[0243] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for assessing the contamination level of the fiber coupling end face of an optical fiber lead frame, characterized in that, include: Step S1: Acquire a global image of the optical fiber lead frame, establish a frame coordinate system, determine the end face coordinates of the optical fiber coupling end face based on the frame coordinate system, and establish a local end face coordinate system. Step S2: Acquire multi-source image data of the target fiber coupling end face in the local coordinate system of the end face, and register the multi-source image data to obtain end face image data; Step S3: Process the end-face image data to obtain a set of contaminant instances and contaminant morphology data; calculate the current functional contamination level based on the pre-acquired mode field distribution data and contamination attenuation response data; specifically, calculating the current functional contamination level includes: obtaining the mode field energy distribution corresponding to different spatial locations based on the mode field distribution data of the target fiber coupling end face under the working wavelength in a clean state, and establishing mode field energy weight data based on the mode field energy distribution; establishing contamination attenuation response data based on the optical power attenuation data corresponding to different contaminant categories, contaminant sizes, and contaminant morphology parameters; calculating the functional contamination contribution value corresponding to each contaminant instance based on the spatial overlap relationship between the contaminant instance coverage area and the mode field energy distribution area, combined with the mode field energy weight data and the contamination attenuation response data, and obtaining the current functional contamination level. Step S4: Determine the key pollutants based on the current functional pollution level, the pollutant instance set, and the pollutant morphology data; Step S5: Apply controlled micro-perturbation to the target optical fiber coupling end face containing the key pollutant, and calibrate the controlled micro-perturbation according to the actual action parameters at the end face; Step S6: Apply controlled micro-perturbations of different levels to the key pollutant and acquire end-face images after perturbation. Based on the changes in end-face images before and after perturbation, obtain the migration response data of the key pollutant. Step S7: Based on the migration response data and the pre-established contact evolution calibration data, predict the evolution results of the key pollutant under optical fiber coupling conditions, and obtain the evolved pollutant distribution data and damage probability data. Step S8: Based on the evolved contamination distribution data, the damage probability data, and the mode field distribution data, calculate the evolved functional contamination degree, and perform a comprehensive calculation based on the current functional contamination degree, the evolved functional contamination degree, and the pre-calculated maximum contact evolution risk to obtain the fiber coupling end face contamination assessment result.

2. The method for assessing the contamination level of the fiber coupling end face of the fiber optic lead frame according to claim 1, characterized in that, Step S2 involves registering the multi-source image data, specifically including: Based on the end face coordinate information in the frame coordinate system, the multi-source image data obtained by different acquisition methods are converted to the end face local coordinate system; Based on the pollutant feature regions in different multi-source image data, pollutant instances corresponding to different acquisition methods are associated so that the same pollutant instance maintains a corresponding relationship in different multi-source image data. A unified dataset for the end face is established based on the multi-source image data after pollutant instance association is completed.

3. The method for assessing the contamination level of the fiber coupling end face of the fiber optic lead frame according to claim 1, characterized in that, Step S4 involves identifying the key pollutant, specifically including: Sort the pollutant instances according to their current functional pollution contribution values ​​in the pollutant instance set. Obtain pollutant morphology data corresponding to each pollutant instance and the positional relationship between each pollutant instance and the model field region; Based on the current functional pollution contribution value, the pollutant morphology data, and the location relationship, the key pollutants used for migration response characterization are selected from the pollutant instance set.

4. The method for assessing the contamination level of the fiber coupling end face of the fiber optic lead frame according to claim 1, characterized in that, Applying the controlled micro-perturbation in step S5 specifically includes: The actual action parameters at the location where the micro-perturbation is applied are collected, and the correspondence between the actual action parameters and the driving parameters of the micro-perturbation application device is established. The driving parameters of the micro-perturbation application device are adjusted according to the correspondence to ensure that the controlled micro-perturbation action applied to the key pollutant is within a preset range; The controlled micro-perturbation action is less than the end-face cleaning threshold and the end-face damage threshold, so as to keep the key contaminants in an unremoved state.

5. The method for assessing the contamination level of the fiber coupling end face of the fiber optic lead frame according to claim 1, characterized in that, Step S6 involves obtaining the migration response data, specifically including: Different levels of controlled micro-perturbations were applied to the key pollutant along at least two non-collinear directions, and end-face images were acquired after each perturbation. The location change data, morphological change data, and orientation change data of the key pollutant are determined based on the end face images before and after the disturbance. Based on the relationship between the controlled perturbation action at different levels and the position change data, the morphological change data, and the orientation change data, the migration sensitivity of the key pollutant is determined, and the migration response data is generated.

6. The method for assessing the contamination level of the fiber coupling end face of the fiber optic lead frame according to claim 5, characterized in that, Step S7, predicting the evolution of the key pollutants, specifically includes: Establish contact evolution calibration data, which includes the correspondence between contaminant type, contaminant size parameters, contaminant height parameters, and fiber optic coupling condition parameters and contaminant migration, morphological changes, and end-face damage status. The optical fiber coupling operating parameters include at least the end face gap parameter, the contact force parameter, and the lateral deviation parameter. Based on the migration response data, migration sensitivity, and current fiber coupling parameters corresponding to the key pollutants, the corresponding pollutant evolution characteristic parameters are determined in the contact evolution calibration data. Based on the pollutant evolution characteristic parameters, the evolution results of the key pollutants under the optical fiber coupling condition are predicted, and the post-evolution pollutant distribution data and the damage probability data are generated.

7. The method for assessing the contamination level of the fiber coupling end face of the fiber optic lead frame according to claim 1, characterized in that, Step S8 calculates the post-evolutionary functional contamination level, specifically including: The spatial distribution information of pollutants after evolution is determined based on the evolved pollution distribution data; Determine end-face damage risk information based on the damage probability data; The evolved contamination distribution data and the damage probability data are mapped to the spatial region corresponding to the model field distribution data, and the evolved functional contamination degree is calculated based on the model field distribution data.

8. The method for assessing the contamination level of the fiber coupling end face of the fiber optic lead frame according to claim 1, characterized in that, Step S8 obtains the fiber coupling end-face contamination assessment result, specifically including: Obtain the current fiber coupling operating condition parameters, which include at least a predetermined end face gap, a predetermined contact force, a predetermined contact speed, and a predetermined lateral deviation; Based on the migration response data of the key pollutants, the contact evolution calibration data, and the current fiber coupling operating parameters, the contact evolution risk corresponding to each key pollutant is calculated, and contact evolution risk data is generated. The maximum contact evolution risk is determined from the contact evolution risk data; The comprehensive contamination level is obtained by comprehensively calculating the current functional contamination level, the evolved functional contamination level, and the maximum exposure evolution risk. Based on the overall contamination level and the contact evolution risk data, the risk status and contamination level of the optical fiber coupling end face are determined; Based on the risk status, a corresponding handling result is generated, and the handling result includes at least one of the following: direct passage, re-inspection after cleaning, prohibition of coupling, and maintenance. Based on the comprehensive contamination level, the contact evolution risk data, the risk status, the contamination level, and the treatment results, the fiber optic coupling end face contamination assessment result is generated.

9. A system for assessing the contamination level of the fiber coupling end face of a fiber optic lead frame, used to execute the method for assessing the contamination level of the fiber coupling end face of a fiber optic lead frame according to any one of claims 1-8, characterized in that, include: The frame positioning module is used to acquire a global image of the optical fiber lead frame, establish a frame coordinate system, determine the end face coordinates of the optical fiber coupling end face based on the frame coordinate system, and establish a local end face coordinate system. A multi-source acquisition module, connected to the frame positioning module, is used to acquire multi-source image data of the target fiber-coupled end face in the local coordinate system of the end face, and to register the multi-source image data to obtain end face image data. A contamination identification and assessment module, connected to the multi-source acquisition module, processes the end-face image data to obtain a set of contaminant instances and contaminant morphology data. Based on pre-acquired mode field distribution data and contamination attenuation response data, it calculates the current functional contamination level for the set of contaminant instances. Specifically, calculating the current functional contamination level includes: acquiring the mode field energy distribution corresponding to different spatial locations based on the mode field distribution data of the target fiber coupling end-face under a clean state at the working wavelength, and establishing mode field energy weight data based on the mode field energy distribution; establishing contamination attenuation response data based on the optical power attenuation data corresponding to different contaminant categories, contaminant sizes, and contaminant morphology parameters; and calculating the functional contamination contribution value corresponding to each contaminant instance based on the spatial overlap between the contaminant instance coverage area and the mode field energy distribution area, combined with the mode field energy weight data and the contamination attenuation response data, to obtain the current functional contamination level. A critical pollutant determination module, connected to the pollution identification and assessment module, is used to determine critical pollutants based on the current functional pollution level, the pollutant instance set, and the pollutant morphology data. The micro-perturbation control module, connected to the key pollutant determination module, is used to apply controlled micro-perturbations to the target optical fiber coupling end face containing the key pollutant, and to calibrate the controlled micro-perturbations according to the actual action parameters at the end face. The migration response analysis module is connected to the micro-perturbation control module and is used to apply controlled micro-perturbations of different levels to the key pollutant, and to collect end-face images after perturbation. Based on the changes in the end-face images before and after perturbation, the migration response data of the key pollutant is obtained. The evolution prediction module, connected to the migration response analysis module, is used to predict the evolution results of the key pollutant under optical fiber coupling conditions based on the migration response data and pre-established contact evolution calibration data, and to obtain the pollution distribution data and damage probability data after evolution. The evolution assessment module is connected to the evolution prediction module and the contamination identification and assessment module, respectively. It is used to calculate the functional contamination degree after evolution based on the contamination distribution data after evolution, the damage probability data, and the pattern field distribution data. It also performs a comprehensive calculation based on the current functional contamination degree, the functional contamination degree after evolution, and the pre-calculated maximum contact evolution risk to obtain the fiber coupling end face contamination assessment result.

Images