A method for measuring the thickness of cable insulation

Abstract

The application discloses a cable insulation thickness metering method, and belongs to the field of cable insulation on-line metering and extrusion process quality monitoring. The method collects cold and hot cross-section data after insulation extrusion and cooling and setting, so as to eliminate the scale cross-section geometric fingerprint under the length candidate interval and the same material cross-section registration under the monotone stretching and contraction constraint, correct the rotation angle through the circumferential phase and the shape correlation, and calculate the shrinkage rate and the cold equivalent thickness. The thickness value and the material position uncertainty are estimated respectively, the conservative thickness envelope and the extended section boundary are formed, the under-thickness section, the evidence label and the process time window are output, and continuous non-destructive metering, positioning and tracing are realized.

Description

Technical Field

[0001] This invention relates to the field of online measurement of cable insulation and quality monitoring of the extrusion process, and particularly to a method for measuring cable insulation thickness. Background Technology

[0002] Cable insulation is typically continuously applied to the conductor's periphery using extrusion equipment and then shaped in a subsequent cooling unit. During cooling, the insulation material shrinks, and center misalignment may occur between the conductor and the insulation layer. Variations in traction speed, extrusion volume, and die wear further alter the circumferential thickness distribution of the cross-section. Evaluating insulation quality solely based on the instantaneous outer diameter or average thickness at a single station makes it difficult to distinguish the effects of cooling shrinkage, conductor eccentricity, and process fluctuations on local thickness. Furthermore, it is challenging to accurately map thermal measurements to the same material cross-section during variable-speed production.

[0003] Current insulation thickness testing methods mostly employ offline slicing, requiring the cutting of cable samples before measuring the cross-sectional thickness. This method suffers from drawbacks such as sample destruction, limited testing frequency, and delayed results. Because offline samples contain undetected continuous lengths, localized areas of insufficient thickness may not be detected in a timely manner. Even if online measurements detect thickness anomalies, without material coordinate registration, circumferential angle alignment, and evaluation of quantitative and positional uncertainties, it is difficult to provide reliable start and end meters, conservative boundaries, and corresponding production times, thus affecting the traceability of reel numbers, batches, and process causes.

[0004] Therefore, there is a need for a method for measuring cable insulation thickness that can overcome the shortcomings of the existing technology. Summary of the Invention

[0005] One objective of this invention is to propose a method for measuring cable insulation thickness. Addressing the problems of difficulty in matching hot and cold cross-sections due to cable insulation cooling shrinkage, conductor eccentricity, production speed variations, and mold wear, as well as the difficulty in continuously locating localized thin sections and the damage to samples caused by traditional slicing inspection, this invention proposes a method for acquiring cross-sectional boundary, circumferential thickness, speed, and length data at both hot and cold workstations. It utilizes the internal cross-sectional geometric fingerprint, which eliminates scale, to complete cross-sectional registration with the same material under candidate zones, sequential monotonicity, and local scaling constraints. Then, it performs circumferential angle correction, shrinkage measurement, and cold-state equivalent thickness calculation. Finally, it propagates the uncertainty in thickness measurement and material position uncertainty to conservative thickness determination and section boundaries, respectively. This invention can generate thin section results with meter value, circumferential position, confidence information, conservative boundaries, evidence labels, and process time windows under continuous production conditions.

[0006] This invention provides a method for measuring cable insulation thickness, comprising: S1. After insulation extrusion and cooling and shaping, the conductor boundary, insulation outer boundary, circumferential thickness sequence, speed and cumulative length are collected to form a cold and hot section sequence. S2. Eliminate the overall scale of the circumferential thickness sequence and generate an internal cross-sectional geometric fingerprint based on the average thickness, circumferential shape, and cross-sectional center geometric features; S3. Based on the range of cumulative length and velocity, determine the candidate intervals for the same material cross section. Under the conditions that the cross section order is not reversed and the local expansion and contraction rate is limited, monotonically match the geometric fingerprint of the internal cross section to obtain the pairing index, material coordinates and registration confidence. S4. Determine the circumferential rotation angle between measurement stations based on the circumferential shape phase and correlation of the paired cross sections, transform the cold circumferential thickness sequence to the same material angular coordinates, and obtain the partition shrinkage rate and cold equivalent thickness. S5. Determine the thickness uncertainty of the cold equivalent thickness and the material position uncertainty of the material coordinates based on the calibration data, measurement status and matching quality. Obtain the upper limit of the thickness uncertainty according to the coverage level of the target side. Subtract this upper limit from the cold equivalent thickness to obtain the conservative thickness lower envelope. S6. Compare the conservative thickness envelope with the insulation thickness acceptance limit to determine the under-thickness point, merge the under-thickness points on the material coordinate and expand the section boundary according to the material position uncertainty, determine the evidence label according to the correlation between the cold and hot thickness residual and the production process quantity, and output the under-thickness section report.

[0007] Optionally, S1 includes: the hot measurement station is set after the cable insulation extrusion head, the cold measurement station is set after the cooling and shaping device, each measurement station uses a through-type measurement device to obtain the conductor boundary and insulation outer boundary of the same cross section, and generates a circumferential thickness sequence from the radial distance along the circumferential direction between the conductor boundary and the insulation outer boundary; Non-contact speed, cumulative length, and sampling time are collected synchronously at each measurement station, and arranged according to the sampling time to form the hot cross-section sequence and the cold cross-section sequence.

[0008] Optionally, S2 includes: normalizing the corresponding circumferential thickness sequence with the average value of the circumferential thickness sequence to eliminate the overall scale; The normalized circumferential thickness sequence is subjected to circumferential harmonic decomposition. The ratio of the amplitude of each harmonic to the average value of the circumferential thickness sequence is determined as the amplitude ratio of the corresponding order. The phase of each harmonic is extracted, where higher-order harmonics are harmonics of the second order and above. The conductor center offset is determined based on the offset of the conductor boundary center relative to the insulation outer boundary center. The ellipticity is determined based on the major and minor axes of the insulation outer boundary. Local extreme positions in the continuous cross section where the circumferential position change does not exceed the position stability threshold are extracted. The average thickness, amplitude ratio, phase, conductor center offset, ellipticity, and local extreme value positions are arranged in a predetermined field order to form the internal cross-sectional geometric fingerprint. The position stability threshold is determined by the upper quantile of the local extreme value position error obtained from repeated measurements of a standard part.

[0009] Optionally, S3 includes: determining the travel time range of the material cross section from the hot measurement station to the cold measurement station based on the equipment path length between the two measurement stations and the lower and upper bounds of the non-contact speed; The sampling time of the hot section is added to the travel time range to obtain the cold candidate sampling time range, and the cumulative length of the cold section whose sampling time is within the cold candidate sampling time range is determined as the candidate cumulative length range; Using the sum of normalized differences of each field of the internal cross-section geometric fingerprint as the matching cost, and under the constraints that neither the hot cross-section index nor the cold cross-section index of the matching path decreases and the ratio of material coordinate increments between adjacent paired cross-sections is within the allowable range of local expansion and contraction rate, the matching path with the minimum cumulative matching cost is obtained. The matching path determines the pairing index and material coordinates, and the registration confidence is determined based on the interval between the minimum and the second minimum cumulative matching costs, the candidate interval width, and the local slope deviation of the matching path. The allowable interval of the local stretching rate is determined by the lower and upper quantiles of the ratio of cold to hot length increments of the corresponding cable specification in the verification slice data.

[0010] Optionally, S4 includes: obtaining the thermal first harmonic amplitude ratio and the cold first harmonic amplitude ratio of the paired cross sections respectively, and calculating the first harmonic phase difference; When both the thermal first harmonic amplitude ratio and the cold first harmonic amplitude ratio reach the phase usability threshold, the circumferential rotation angle is determined by the first harmonic phase difference; When at least one of the hot-state first harmonic amplitude ratio and the cold-state first harmonic amplitude ratio does not reach the phase availability threshold, the peak values ​​of the correlation coefficients between the circumferential morphological sequences reconstructed by second-order and above harmonics are arranged in descending order of value. When the difference between the first peak value and the second peak value reaches the peak value discrimination threshold, the circumferential rotation angle is determined by the angular displacement corresponding to the first peak value. When the ratio of the first harmonic amplitude in the hot state or the ratio of the first harmonic amplitude in the cold state does not reach the phase available threshold, and the difference between the first peak value and the second peak value does not reach the peak value discrimination threshold, a circumferential registration invalid state is generated and the output of the cold equivalent thickness of the corresponding cross section is stopped. The available phase threshold is determined by the lower limit of the amplitude ratio corresponding to the first harmonic phase variance of repeated measurements of the standard part not exceeding the preset upper limit of the phase variance, and the peak discrimination threshold is determined by the lower quantile of the difference between the correct angular displacement peak and the other angular displacement peaks in repeated measurements of the standard part. The cold circumferential thickness sequence is resampled according to the circumferential rotation angle. Under the same material angular coordinate, the shrinkage rate of each circumferential zone is determined according to the ratio of the difference between the hot and cold thicknesses of the paired section to the hot thickness. The cold thickness that is paired with the hot section and has been resampled is determined as the cold equivalent thickness.

[0011] Optionally, S5 includes: grouping the residuals obtained by subtracting the reference value of the standard part from the thickness measurement value of the traceable standard part or the residuals obtained by subtracting the reference value of the slice from the thickness measurement value of the verification slice according to the cable specification and production condition label, and determining the thickness error sample by combining the calibration error and drift amount of the hot measurement station and the cold measurement station. The position error samples are determined based on the cumulative length candidate interval width, the interval between the minimum cumulative matching cost and the second minimum cumulative matching cost, the angular width of the circumferential morphological correlation coefficient peak, and the registration residual. The current measurement status is determined by whether the calibration data is valid, whether the current drift exceeds the drift limit determined by the upper quantile of the drift within the calibration cycle, whether the current production condition label is covered by the calibration data, and whether a valid pairing index and circumferential rotation angle status combination is generated. When the current measurement state is valid and the cold equivalent thickness is obtained, select the thickness error sample and position error sample with the same current measurement state from the calibration data. Determine the thickness uncertainty by the quantile corresponding to the absolute value of the thickness error sample at the target double-sided coverage level. Determine the material position uncertainty by the quantile corresponding to the absolute value of the position error sample at the target double-sided coverage level. Then, truncate all thickness error samples at the target one-sided coverage level by setting the lower limit to zero to determine the upper bound of the thickness uncertainty. If any of the following occurs: the calibration data is invalid, the current drift exceeds the drift limit, the current production condition label is not covered by the calibration data, no valid pairing index or circumferential rotation angle is generated, no error sample with the same current measurement state is obtained, or there is an invalid circumferential registration state, the current measurement state will be determined as an invalid value state and the calculation of the conservative thickness envelope will be stopped. Only when the current measurement state is valid, the conservative lower envelope of the thickness is obtained by subtracting the upper bound of the thickness uncertainty from the cold-state equivalent thickness; Furthermore, a current production condition label is formed based on the current cable specifications, material grade, extrusion temperature range, cooling conditions, and linear speed range. The current production condition label is then compared with the set of condition labels covered by the calibration data to determine whether the current condition is within the calibration range. The current drift amount is determined based on the difference between the standard part retest result and the most recent calibration result, and the current drift amount is compared with the drift limit. When the calibration data is invalid, the current operating condition is outside the calibration domain, the current drift exceeds the drift limit, or the current measurement state is invalid, the output of a qualified conclusion is prohibited. Regardless of whether a thin section is formed, the measurement result record containing the invalid value flag and the manual verification flag will be output. After recalibration or verification is completed, the deterministic metrological conclusion will be restored only when the calibration data is valid, the current operating condition is within the calibration domain, the current drift amount does not exceed the drift limit, and the current measurement state is not an invalid value state.

[0012] Optionally, S6 includes: for the cross section with an effective conservative thickness envelope only, in the same material coordinates and material angular coordinates, subtracting the hot reference thickness from the cold equivalent thickness, which is converted from the hot thickness according to the circumferential average shrinkage rate of the paired cross section, to obtain the cold and hot thickness residual, wherein a positive residual indicates that the cold equivalent thickness is greater than the hot reference thickness. The upper quantile of the time delay between the change in production process quantity and the thickness response in the verification sample without process anomalies is multiplied by the arithmetic mean of the hot and cold non-contact velocities of the paired cross sections to obtain the synchronization material window length, and the range of adjacent material coordinates of the corresponding length is used as the synchronization material window. When the first harmonic phase of the cold and hot thickness residual within the synchronous material window is consistent with the direction of conductor center offset, and the correlation coefficient between the first harmonic amplitude and the conductor center offset reaches the eccentric synchronous correlation threshold, conductor eccentricity evidence is generated. The eccentric synchronous correlation threshold is determined by the upper quantile of the corresponding correlation coefficient of the sample without conductor eccentricity verification. When the first harmonic component corresponding to the conductor center offset is removed from the cold and hot thickness residual, and the difference between the residuals of each circumferential zone reaches the non-uniform shrinkage threshold, evidence of non-uniform shrinkage is generated. The non-uniform shrinkage threshold is determined by the upper quantile of the difference in the residuals of the zoned residuals of the uniform cooling verification sample. The conductor eccentricity evidence and the non-uniform shrinkage evidence are allowed to be generated simultaneously when their respective conditions are met. When the correlation coefficient between the residual cold and hot thickness within the synchronous material window and the ratio of the traction speed derivative to the extrusion amount reaches the traction synchronization correlation threshold, traction fluctuation evidence is generated. When the regression slope of the cold and hot thickness residuals with the cumulative running length of the mold reaches the wear slope threshold, mold wear evidence is generated. The traction synchronization correlation threshold and the wear slope threshold are determined by the upper quantiles of the corresponding statistics of the non-traction abnormal samples and the new mold benchmark samples, respectively. The section whose envelope under the conservative thickness does not reach the qualified limit of insulation thickness is identified as the insufficient thickness point, and the insufficient thickness points are merged in the order of material coordinates to form the insufficient thickness section. The evidence is used as the evidence label of the insufficient thickness section. Furthermore, the under-thickness section report includes the start and end meters of the under-thickness section, the circumferential angle at which the lower envelope of the thickness does not reach the qualified limit of insulation thickness, the reel number, the batch, the registration confidence level, the process time window, the evidence label, the point estimation boundary, and the conservative boundary. The process time window is determined based on the hot and cold measurement times corresponding to the paired indexes within the thin section, and the tray number and batch to which the thin section belongs are determined based on the mapping relationship between tray number and cumulative length. Furthermore, points with material coordinate intervals no greater than the sum of the material position uncertainties of the two corresponding points with insufficient thickness are identified as continuous points with insufficient thickness. The material coordinates of the first and last thinnest points obtained after merging consecutive thinnest points are used as the starting point and ending point of point estimation, respectively. The starting point of point estimation is extended forward by the material position uncertainty of the first thinnest point to obtain a conservative starting point, and the ending point of point estimation is extended backward by the material position uncertainty of the last thinnest point to obtain a conservative ending point. When the conservative boundaries of two thin sections within the same disk number intersect, the two thin sections are merged, and the estimated point boundary before merging and the conservative boundary after merging are recorded simultaneously in the thin section report.

[0013] The beneficial effects of this invention are: 1. By combining the internal cross-sectional geometric fingerprint of the dual-station system with restricted monotonic matching, the hot and cold measurement data are mapped to the same material coordinates, reducing the risk of cross-sectional mismatch caused by changes in traction speed and local expansion and contraction.

[0014] 2. By determining the rotation angle between workstations through circumferential morphological phase and correlation, the cold thickness is transformed to a unified material angular coordinate, enabling consistent measurement of partition shrinkage, conductor eccentricity, and circumferential underthickness positions.

[0015] 3. By calculating the thickness uncertainty and material location uncertainty separately, a conservative thickness envelope is constructed and the segment boundary is expanded, so that the under-thickness report can cover both thickness error and location mismatch risk, and can be traced by combining reel number, batch, process time window and evidence label. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a cable insulation thickness measurement method according to the present invention.

[0017] Figure 2 This is a flowchart of the S3 material cross-section registration step of the present invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figures 1-2 A method for measuring cable insulation thickness, comprising: S1. After insulation extrusion and cooling and shaping, the conductor boundary, insulation outer boundary, circumferential thickness sequence, speed and cumulative length are collected to form a cold and hot section sequence. S2. Eliminate the overall scale of the circumferential thickness sequence and generate an internal cross-sectional geometric fingerprint based on the average thickness, circumferential shape, and cross-sectional center geometric features; S3. Based on the range of cumulative length and velocity, determine the candidate intervals for the same material cross section. Under the conditions that the cross section order is not reversed and the local expansion and contraction rate is limited, monotonically match the geometric fingerprint of the internal cross section to obtain the pairing index, material coordinates and registration confidence. S4. Determine the circumferential rotation angle between measurement stations based on the circumferential shape phase and correlation of the paired cross sections, transform the cold circumferential thickness sequence to the same material angular coordinates, and obtain the partition shrinkage rate and cold equivalent thickness. S5. Determine the thickness uncertainty of the cold equivalent thickness and the material position uncertainty of the material coordinates based on the calibration data, measurement status and matching quality. Obtain the upper limit of the thickness uncertainty according to the coverage level of the target side. Subtract this upper limit from the cold equivalent thickness to obtain the conservative thickness lower envelope. S6. Compare the conservative thickness envelope with the insulation thickness acceptance limit to determine the under-thickness point, merge the under-thickness points on the material coordinate and expand the section boundary according to the material position uncertainty, determine the evidence label according to the correlation between the cold and hot thickness residual and the production process quantity, and output the under-thickness section report.

[0020] In this specific embodiment, S1 includes: This specific implementation takes a continuously extruded 10 kV cross-linked polyethylene insulated cable as the measurement object. The hot measurement station is installed 1.2 m behind the insulation extrusion head, and the cold measurement station is installed behind the outlet of the water-cooled shaping device. The equipment path length of the two stations along the actual cable travel path is verified to be 28.0 m by the stop line measuring rope and laser ranging. Each station uses a temperature-compensated penetrating X-ray measuring device to reconstruct the conductor boundary and the outer boundary of the insulation, and establishes a dual-station measurement task with the same production order number, reel number, equipment clock version and measurement station identifier. Each penetration measurement device acquires attenuation data in four projection directions with a sampling period of 20 ms. First, the projection coordinates are corrected for gain, zero point and geometric magnification according to the known inner and outer diameters of the standard circular ring. Then, the cross-sectional grayscale array is obtained by algebraic reconstruction. Sub-pixel positioning is performed on the grayscale transition edges between conductor and insulation and between insulation and air, forming 72 conductor boundary points and 72 insulation outer boundary points arranged at 5° material angle intervals. The coordinates of the boundary points are all based on the intersection of the measurement optical axis as the origin and in mm. For the The section of the workstation and the first The metering controller calculates the circumferential thickness along the same ray pointing from the center of the conductor boundary to the outer boundary of the insulation at a circumferential sampling angle. , Indicates the first The cross section at the workstation is at the first The insulation thickness at each circumferential sampling angle is measured in mm. This represents the cross-sectional indices arranged in ascending order of sampling time and taking positive integer values. This indicates the circumferential sampling angle index, which ranges from 0 to 71. This represents the radial distance from the outer boundary of the insulation to the center of the conductor boundary, in mm. This represents the radial distance from the conductor boundary to the center of the conductor boundary, in mm. When the effective ray ratio of a certain projection direction is less than 95% or the boundary fitting residual exceeds the 99th percentile of the repeated measurement residual of the standard ring, the controller sets the boundary valid flag of the cross section record to 0 and retains the original projection, without sending the record to the subsequent geometric fingerprint calculation. Other effective cross sections are written into the cross section record consisting of the cross section index, sampling time, conductor boundary point set, insulation outer boundary point set, circumferential thickness sequence, boundary fitting residual and boundary valid flag. Two workstations are each equipped with a laser Doppler non-contact velocimeter. The velocimeter and the penetration measuring device share a microsecond-level clock calibrated with IEEE 1588. The controller reads the instantaneous velocity and sampling time at each cross-sectional sampling and then... Perform a three-point moving average. Indicates the first The non-contact velocity of each cross section, in m / s. This represents the initial non-contact velocity of the corresponding cross-section, in m / s; only one valid initial velocity is obtained after startup. Obtain two valid original speed commands The three-point moving average is executed starting from the third valid original velocity. Cross sections that have not yet obtained a valid original velocity are marked as velocity not ready and the cumulative length is not updated. The effective velocity range in this specific embodiment is set to 0.5 m / s to 2.0 m / s. Records exceeding this range are written with a velocity anomaly flag but the cumulative length is not updated with the anomaly value. The controller starts from the most recent valid cumulative length and follows... Update the cumulative length for each workstation. Indicates the first The cumulative length of each cross section at the corresponding workstation, in meters. This represents the cumulative length of the previous effective cross section, in meters (m). and These represent the sampling times of the current cross section and the previous effective cross section, respectively, in seconds. The product of velocity and time difference directly yields m. When there is no effective velocity for three consecutive sampling periods, the update is paused and the length state is set to invalid. The effective cross-section records of the hot and cold workstations are written into the hot cross-section sequence and the cold cross-section sequence in ascending order of sampling time, respectively. Each record simultaneously saves the speed, cumulative length, disk number, batch, order number, equipment clock version and calibration version. The sequence uses the sampling time as the primary key and the cross-section index as the secondary key. For duplicate primary key records, only the one with the lowest boundary fitting residual value is retained. Finally, the hot and cold cross-section sequences and their validity flags are written into the dual-workstation cross-section buffer and used as the input of S2.

[0021] In this specific embodiment, S2 includes: The geometric fingerprint calculation module reads each valid record in the S1 dual-station cross-section cache sequentially, and performs the calculation on the first record. The average thickness was calculated using a 72-point circumferential thickness sequence across a cross section. , Indicates the first The average circumferential thickness of each cross section, in mm. This represents the total number of circumferential sampling points, and in this specific implementation, it is taken as 72. Using the circumferential thickness defined in S1, Using the circumferential sampling angle index defined in S1, cross sections with an average thickness less than 20% of the nominal thickness of the cable specification are marked as boundary reconstruction anomalies and fingerprint generation is stopped. Calculate the normalized circumferential thickness for the inspected sections. , Represents the first dimension after eliminating the overall thickness scale. The first section The circumferential dimensionless thickness is set to an effective range of 0.5 to 1.5. and Following the aforementioned definition, a single point outside the valid range is replaced by a linear interpolation of two adjacent valid points, and if more than three consecutive outlier points are found, the fingerprint status of that section is set to invalid. The module performs discrete circumferential harmonic decomposition on the normalized circumferential thickness, according to and Obtain the sine and cosine coefficients. and They represent the first The first section The dimensionless cosine and sine coefficients of the first harmonic. Indicates the harmonic order and takes values ​​from 1 to 6. , and Following the aforementioned definition, It represents pi and is a dimensionless constant, and can be retained up to the sixth order to cover stable forms of conductors with eccentricity, ellipse and local circumferential undulation. Amplitude calculated from harmonic coefficients and phase , Indicates the first The first section The dimensionless amplitude ratio of the first harmonic to the average thickness, with a value ranging from 0 to 1. Indicates the corresponding harmonic phase, in rad and normalized to 0. , Indicates the first parameter Second parameter The two-parameter arctangent function that jointly determines the quadrant returns a value mapped to... , Pi is a dimensionless constant representing the value of a circle. , , and Following the aforementioned definition, second to sixth order harmonics simultaneously serve as inputs for S4 higher-order morphological reconstruction; The module performs least-squares circle center fitting on the conductor boundary point set and the insulation outer boundary point set respectively, and calculates the conductor center offset. , Indicates the first The offset of the conductor center relative to the center of the outer boundary of the insulation for each cross section, in mm. and This represents the coordinates of the conductor boundary fitting center, in mm. and This represents the coordinates of the fitting center of the outer boundary of the insulation, in mm, and the offset direction is recorded as 0 to 100 mm according to the material angular coordinates. rad; Perform elliptic least-squares fitting on the outer boundary of the insulation and calculate the ellipticity. , Indicates the first The dimensionless ellipticity of each cross section, with values ​​ranging from 0 to 1. This represents the diameter of the major axis of the fitted ellipse, in mm. The ellipticity field represents the diameter of the minor axis of the fitted ellipse in mm. The fingerprint record is also written along the major axis. The ellipticity field is not used for matching when the fitting residual exceeds the 99th quantile of the calibration sample. In the normalized circumferential thickness sequence, the module uses the change of the first-order difference sign between two adjacent points as a candidate condition for local extrema, and tracks the candidate angular positions using five consecutive effective cross sections as windows. In this specific implementation, the position stability threshold of 3° is obtained from the 99th quantile of the absolute value of the local extrema angular error obtained by repeated measurements of the standard circular ring eccentric part. Only the maximum change in the angular position within the window does not exceed 3° and the amplitude exceeds the extreme value of the upper quantile of the repeated measurement noise amplitude is written into the stable extrema field. Other extrema do not participate in the subsequent registration. Finally, the internal cross-sectional geometric fingerprint is generated in a fixed order according to the average thickness, the amplitude ratio of the first to sixth orders, the phase of the first to sixth orders, the conductor center offset and direction, the ellipticity and major axis direction, the stable local extreme angle and amplitude, and the effective mask of the field. The fingerprint is written into the geometric fingerprint table along with the cross-section index, station identifier, sampling time, and cumulative length. The average thickness retains the value information, and the calibration scale of each field is used for normalization when matching. At the beginning of each shift and after reinstallation at any station, the identification mark of the same standard eccentric ring is set to 0°, 90°, 180°, and 270° respectively. The circumferential mean of the difference in direction measured at the hot and cold stations is calculated as the fixed installation angle offset. When the circumferential standard deviation of the four differences does not exceed 2°, only the coordinate system of the hot and cold measurement device is changed to the hot position angular coordinate; otherwise, the cross-position effective flag of the direction sensitive field is set to 0. The fixed installation angle offset does not represent the cross-section rotation angle generated by the material section along the equipment path. This table serves as the unique fingerprint input of S3 and retains the original direction sensitive field without cross-section material rotation angle correction to both S3 and S4. S4 still independently calculates the final circumferential rotation angle of each material section.

[0022] In this specific embodiment, S3 includes: The material registration module reads the S2 geometric fingerprint table and the S1 dual-station cross-section sequence, using the hot cross-section index as the query primary key. It employs the equipment path length between the two stations (28.0 m) confirmed by stop-line measurement, and the lower limit of the full-stroke effective speed, solidified through equipment type testing and safety interlocks. m / s and upper limit of velocity m / s, divide 28.0 m by the upper and lower speed limits respectively to obtain the conservative travel time range. , and These represent the shortest and longest travel times of the material cross-section from the hot position to the cold position, respectively, in seconds. and These represent the lower and upper bounds of the effective speed of the device covering the entire 28.0 m travel distance, respectively, in m / s, and satisfy the following conditions: When the production speed setting range or interlock parameters change, re-verify and expand the candidate time range, and do not narrow the range before the verification is completed; For sampling time is The hot cross section, the module As a candidate cold-state sampling time range, the minimum and maximum cumulative lengths of the cold-state cross-sections within this range are read to form a candidate cumulative length range. Indicates the first The sampling time for each hot cross-section is measured in seconds. The module establishes a registration status for each hot cross-section. Upon initial entry, it waits for cold data, with an empty candidate range and the current time not exceeding [time value missing]. Please wait. This indicates that the cold data entry and network transmission allowance is allowed, and the upper quantile of the 99th entry delay is taken as 0.5 seconds. If there are still no candidates after this time, it is closed as the no-candidate final state. If there are candidates but they are all excluded by the field gate, it is closed as the non-registerable final state. If a path is obtained but the confidence is lower than 0.65, it is closed as the low-confidence final state. Only when the threshold is reached is it closed as the valid registration final state. All invalid final states, along with the reason code and candidate range, are directly sent to S5 to form invalid value and manual review records. If the candidate range width exceeds the 99th upper quantile of the verification data of this specification, the candidate is retained but the confidence of subsequent registration is reduced. The module first uses rotation-invariant fields such as average thickness, first to sixth order amplitude ratio, conductor center offset, ellipticity, and stable extreme value amplitude to form candidate pair initial costs; for thermal fingerprints... With candidate cold fingerprints Establish a candidate corner set , This indicates that the area covers one week and the interval between adjacent zones is 1. The set of 72 candidate material rotation angles in rad, with the positive direction consistent with the positive direction of the S2 material angle; in The joint search aims to determine the rotation angle of candidate materials from the hot material orientation to the cold material orientation. , Indicates hot fingerprint With candidate cold fingerprints The angle between candidate materials is expressed in rad; according to S2... The phase convention, the cold state of the first The first harmonic phase transformation is After the hot state Phase comparison, Indicates the angle in radians Addition and subtraction And uniquely mapped to Angle normalization operation; transforming the cold conductor offset direction and stable extreme angle into... Press back Periodic comparison, among which This indicates the corresponding cold-state direction or stable extreme angle, in rad; subtract the major axis direction of the cold-state ellipse. Press back The period takes the smallest included angle; For each number from 1 to 6, calculate the non-negative normalized sum of differences for all direction-sensitive fields, and take the angle with the smallest sum of differences as the value. Minimum difference and at least distance from candidate corners Only when the second smallest difference in rad reaches the 1% lower quantile (0.08) of the peak difference of the correct rotation angle of the standard eccentric ring component is the result considered acceptable. If the rectified direction-sensitive field is added to the match, otherwise all direction-sensitive fields of the candidate pair are disabled; if the number of participating fields is less than 70% of the total number of predetermined fields after disabling, the candidate is closed and becomes an unregisterable final state. The module defines a non-negative difference for each participating matching field that has completed the candidate corner shifting process. The average thickness, amplitude ratio, conductor center offset, ellipticity, and stable extremum amplitude are taken as the absolute value of the difference between the hot and cold scalar values; the harmonic phase is taken as the minimum angle of the 360° circle after order transformation; the conductor offset direction and the stable extremum angle are taken as the minimum angle of the 360° circle; the major axis direction of the ellipse is taken as the minimum angle of the 180° period; the stable extremum set is first subjected to Hungarian matching and then the non-negative average angular distance is taken; subsequently, the field matching cost is calculated. , Represents the hot cross section and cold cross section The non-negative dimensionless matching cost between them indicates that the smaller the cost, the more similar they are. Indicates the first Non-negative differences in each field Indicates the number of valid fingerprint fields participating in the matching. Indicates field index, Indicates the first The dimensionless weights of each field are determined by a grid search on the correct-pair validation set that minimizes the mismatch rate, and the sum of the weights is 1. This indicates the interquartile range of the field in the same specification of the verification samples and has the same unit as the corresponding field; when Not greater than the repeatability resolution of this field At this point, the field should be removed from the summation and the remaining weights renormalized. This indicates that the value is obtained by taking three times the standard deviation of repeated measurements of the standard part and is in the same unit as the corresponding field; when the fixed installation angle offset of S2 is invalid, all direction-sensitive fields are removed from the cross-station matching fields; the candidate angle shift only serves the S3 matching and leaves a trace with the candidate record; S4 still calculates and outputs the material circumferential rotation angle based on the original harmonic field of the final paired section. The difference in the fingerprint phase field is calculated using the minimum circumferential angle. Stable local extrema are calculated using the average angle difference after Hungarian matching. Fields without valid markers are not included in the summation, and the remaining weights are re-normalized. If the number of participating fields is less than 70% of the total number of predetermined fields, the candidate pairing is prohibited. The average thickness field is only normalized to the same workstation calibration scale and participates in low-weight auxiliary matching, thereby avoiding the overall scale difference from thermal shrinkage from dominating the registration. Within the candidate band, a labeled three-state dynamic programming approach is employed. Each label stores the current hot state index, cold state index, indices on both sides of the most recent true matching anchor, consecutive skipped hot state count, consecutive skipped cold state count, cumulative cost, path signature, and predecessor pointer. For labels with the same current index, most recent anchor, skip count, and end state, the true matching cost is calculated according to... Update, skip hot state cost according to Update, skipping the cold state cost according to Update, of the three formulas , and These represent the dimensionless cumulative costs for the corresponding labels ending with a true match, skipping hot states, and skipping cold states, respectively. Using the aforementioned definition and only adding this cost to the true matching state, The dimensionless penalty for skipping a section is 0.15, as indicated by the frame loss verification sequence; labels with different nearest anchor points or different skip counts must not be merged when taking the minimum value; Before each label is enqueued, the local scaling ratio is calculated using the label's most recent real matching anchor point and the current candidate point. , Indicates the first The dimensionless ratio of the cold and hot length increments of adjacent real matching anchor point segments This indicates the actual matching anchor segment index. and Represents the coordinates of adjacent hot materials, in meters. and Represents the coordinates of adjacent cold-state materials, in meters; the first true match only registers the anchor point and does not calculate the expansion rate, subsequent matches are only made when the coordinate increments of both sides are greater than the corresponding cumulative length resolution and The data is located at the 0.5% and 99.5% quantiles of the validation slice data. Enqueuing is only allowed when the state is within the specified time; otherwise, the transfer is discarded. Skipping a state does not update the anchor point, and tags that skip more than three frames on either side are not enqueued. During initialization, the cumulative cost of the virtual starting point tag is 0, and other states are unreachable. The endpoint is selected from all constrained reachable tags at the end of the candidate band, and predecessors are no longer deleted after backtracking. To obtain the global primary and secondary paths, for each tag group with the same current index, nearest anchor point, skip count, and ending state, the two smallest and second smallest tags with different path signatures are retained according to the cumulative cost. The path signature is determined by all real matching index pairs connected in sequence; any different index pair is considered a different path. After summarizing all constrained reachable tags, the first two complete paths with different path signatures are obtained in ascending order of cumulative cost, and are respectively recorded as the primary path and secondary path. When the cost is equal, the path is selected in descending order of real matching count, ascending order of skip count, and ascending order of paired index lexicographical order. If there is no second complete path, the hot cross-section is closed to a low-confidence final state and the cost interval ratio is not calculated. If there is a second path, the cost interval ratio is calculated. , This represents the dimensionless separation degree of the primary path relative to the secondary path, and the larger the value, the more reliable the path. and Let these represent the minimum and second smallest dimensionless cumulative costs in the complete set of feasible paths, respectively, and satisfy the following conditions: , This indicates that when the second smallest cumulative cost approaches zero, the cost interval is kept greater than a continuously computable dimensionless regularity, and the value is taken as... During backtracking, only the actual matching status of the main path is output as the paired index; Simultaneously, the candidate cumulative length range width is divided by the 99th percentile width of the validation samples and truncated to 0-1 to obtain the candidate width penalty. The absolute value of each local scaling factor deviating from the validation median is divided by the allowable interval width, the average is taken, and truncated to 0-1 to obtain the slope deviation penalty. Obtain the registration confidence level. This represents a dimensionless registration confidence level ranging from 0 to 1, with larger values ​​indicating greater reliability. This represents a candidate width penalty of 0 to 1. This represents the local slope deviation penalty for a path from 0 to 1. This means input The coefficients are truncated to a closed interval containing endpoints 0 and 1, and are calibrated by order-preserving logistic regression of known paired validation sequences. This specific implementation method only considers the main path The hot and cold state indices in the state are written into the pairing index table. The hot cumulative length is used as the unified material coordinate and the corresponding cold cumulative length, primary and secondary cumulative costs, candidate range width, local scaling factor, candidate angular shift, registration confidence, and registration residual are recorded. The registration residual is defined as the non-negative field matching cost of the final paired candidate. It is a dimensionless quantity; pairings with a registration confidence level below 0.65 are closed as low-confidence final states and sent to S5 but not to S4. Pairings with a confidence level of 0.65 and above are written to the material registration table in ascending order of material coordinates and sent to S4. The registration registration table uses the hot section index as the unique key to store five mutually exclusive states: waiting, no candidates, unregisterable, low confidence, or valid registration. Records that have not formed a final state are prohibited from being silently discarded.

[0023] In this specific embodiment, S4 includes: The circumferential registration module reads the S3 material registration table line by line and obtains the thermal first harmonic amplitude ratio and cold first harmonic amplitude ratio of the paired section from the S2 geometric fingerprint table. At the same time, it obtains the amplitude ratio and phase of the second to sixth harmonics of hot and cold. During reading, the pairing index, station identifier and geometric fingerprint calibration version are used as the joint key. If any pairing record is missing the cold and hot harmonic fields of the same version, an invalid circumferential registration state is generated and no angle inference is performed. The module calculates the phase difference of the first harmonic. , Indicates the first The phase difference between the cold-state first harmonic and the hot-state first harmonic of each paired cross section, in rad, and limited to... , and These represent the first harmonic phases of the corresponding cold and hot cross sections, respectively, in rad. Indicates addition and subtraction The angle normalization rule that is then mapped to the interval; The phase threshold can be established by recording the results of 200 repeated measurements of a standard eccentric ring at each station. The phase variance is selected based on the first harmonic amplitude divided by a bin, and the threshold value is set to no more than [value missing]. The minimum amplitude ratio is used as the threshold, which is calibrated to 0.015 in this specific implementation. When the amplitude ratios of both the hot and cold first harmonics are not lower than 0.015, according to... Convert the radian phase difference into circumferential rotation angles in degrees. The negative sign is used to ensure that subsequent cold-state sequences are within the range of... The value at that point is related to the angle of the hot material. Align the elements and denote the corner source field as the first harmonic; It represents pi and is a dimensionless constant. Subsequently, the circumferential angle in higher-order related branches, angle resampling, cold equivalent thickness table and report is uniformly expressed in degrees. When at least one first harmonic amplitude ratio is less than 0.015, the module reconstructs the higher-order circumferential shape after removing the average thickness and first harmonic using second to sixth order harmonics, and calculates the normalized cyclic correlation coefficient over 72 discrete angular displacements. , Indicates the first The paired section in the first The dimensionless correlation coefficient for each angular displacement ranges from -1 to 1. Indicates the angular displacement index from 0 to 71. and These represent the higher-order reconstruction forms of cold and hot, respectively. and These represent the circumferential mean values ​​of the corresponding reconstructed forms. and Following the definition of S2, Modulus Take the modulo operation and uniquely map any integer index to 0. Before calculation, check the sum of squared deviations of the cold and hot reconstruction sequences relative to their respective means. If any sum of squared deviations is not greater than the 99th quantile of the variance of the repeated measurement noise of the standard part, an invalid state of circumferential registration is generated and it will not enter the peak sorting. The module obtains the first peak and a second peak at least 10° away from it according to the correlation coefficient from largest to smallest. The peak discrimination threshold is taken as the 1% lower quantile of the difference between the correct angular displacement peak and the other peaks in the repeated measurements of the standard part. In this specific implementation, it is calibrated to 0.08. When the difference between the first peak and the second peak is not less than 0.08, the circumferential rotation angle is taken as the angle corresponding to the first peak multiplied by 5° and corrected by parabolic three-point interpolation. The rotation angle source field is recorded as higher-order correlation. For higher-order correlation branches, the sampling points in the cyclic correlation sequence whose values ​​are not greater than the adjacent sampling points are first identified as local valleys. The median of all local valleys is taken as the correlation baseline. , Indicates the first For each paired section of the dimensionless correlation baseline, if there are fewer than two local valleys, the circumferential registration state is set to invalid; then according to... Determine the half-peak threshold. Indicates the first Dimensionless half-peak threshold of paired cross sections The first correlation peak is represented as a dimensionless quantity; a quantity containing the first peak and satisfying... The angle covered by the largest continuous set of angular displacements is taken as the angular width of the circumferential correlation peak. Following the aforementioned definition and For angular displacement indexing, the boundary uses linear interpolation of two adjacent five-degree sampling points with width in degrees; for the first harmonic branch, the circumferential correlation peak angular width is taken as twice the standard deviation of the circumferential phase error of the standard eccentric ring in the fractional box, multiplied by the same amplitude. The unit is also degrees; if any branch cannot obtain the corresponding width, the circumferential registration state will be set to invalid. When the first harmonic amplitude condition is not met and the peak difference is less than 0.08, the module generates a circumferential registration invalid state, writes the pairing index, material coordinates, first harmonic amplitude ratio, two related peak values ​​and failure reason into the invalid record, and stops outputting the cold equivalent thickness of the section. Subsequently, S5 incorporates this state into the value validity judgment. For a cross section that yields an effective circumferential rotation angle, the module follows... Periodic cubic interpolation is performed on the cold circumferential thickness sequence. This represents the cold-state thickness after transformation to hot-state material angular coordinates, in mm. Indicates the first The circumferential thickness sequence of 72 points on each paired cold-state section, in mm. Indicates the first The material angle is equal to Multiply by 5° This indicates the effective circumferential rotation angle of the paired cross-section, in degrees. A cubic interpolation rule indicating continuous first and last periods; The circumference was divided into 12 consecutive 30° zones, and the shrinkage rate was calculated for six sampling points in each zone. , Indicates the first The first paired section Dimensionless shrinkage rate of each circumferential zone, This represents a circumferential partition index from 0 to 11. This represents the average thermal thickness of the zone, in mm. This represents the resampled cold-state average thickness of the same material corner zone, in mm. If the average thickness of the hot-state zone is not greater than 0, the zone is set to invalid. The module completes the angle resampling. The cold equivalent thickness is directly determined as the same material coordinate and material angular coordinate. The material coordinate, material angle, circumferential rotation angle in degrees, rotation angle source, relevant peak value, circumferential relevant peak angle width in degrees, zone shrinkage rate, cold equivalent thickness and circumferential registration validity flag are written into the cold equivalent thickness table. This table and invalid records are sent together to S5.

[0024] In this specific embodiment, S5 includes: The uncertainty assessment module reads the S4 cold equivalent thickness table, the S3 material registration table, and the calibration database. The calibration database is tagged with production conditions according to cable specifications, material grade, extrusion temperature range, cooling conditions, and linear speed range. It saves standard part records, verification slice records, drift records, and position registration verification records using the condition label, measurement station, calibration version, and validity period as the joint key. In this specific implementation, standard part retesting is performed at the beginning of each shift and after material grade switching. Verification slices are performed once every 50 km of cumulative production. For each circumferential location of a traceable standard or verification slice, the module calculates a thickness error sample. , Indicates the first A thickness error sample, in mm, with positive values ​​indicating that the measured value is higher than the reference value. This indicates the sample index within the same production condition label. This represents the thickness measurement value after correction for calibration errors and re-measurement drift at the corresponding workstation, in mm. The standard part certificate value or the microscopic measurement reference value of the slice is represented in mm. The standard part sample of the single station is only written into the station measurement error layer and the station identifier is saved. The verification slice sample with real cold and hot correspondence mark and completed S3 and S4 full-link processing is written into the full-link error layer, and the registration confidence bin and the source of the rotation angle are additionally saved. The calibration error, drift correction amount and reference value are all written into the sample record along with the traceability number. The position registration verification record uses tiny identification marks or stop line marks on the verification slice to provide the true material coordinates, and the module calculates position error samples. , Indicates the first Number of position error samples, in meters (m). This indicates the material coordinates given by the S3 pairing index, in meters. The reference material coordinates for the same marker are in meters. Each record simultaneously stores the candidate cumulative length range width, primary and secondary cumulative cost interval, circumferential correlation peak angle width, registration residual, and local scaling factor, thereby filtering position error samples based on records with similar matching quality. Before calibration data is entered into the database, duplicate records are removed by traceability number, sample number, circumferential angle and measurement time. Only records whose measurement units can be converted to mm or m, whose reference certificates are valid, and whose original measurement values ​​and calibration versions are complete are retained. Thickness error samples are not deleted according to error values, so that the quantiles on the target side still cover all valid signed errors. If a position record is missing real marker coordinates or any of the four matching quality fields, it is only prohibited from entering the position error sample set. When selecting positional error samples with similar matching quality for the current paired record, the module reads the candidate cumulative length range width, primary and secondary cumulative cost interval, and dimensionless registration residual from the S3 material registration table, and the circumferential correlation peak angle width in degrees from the S4 cold equivalent thickness table. The module then divides the candidate cumulative length range width into 0.5 m increments, the primary and secondary cumulative cost interval into 0.05 increments, the circumferential correlation peak angle width into 5° increments, and the registration residual into 0.02 increments. Each increment uses a left-closed, right-open interval, with the last valid increment including the right endpoint. The selection predicate is based on all four fields falling within the current record's increment. Verification records and the current record must use the same fields to generate the caliber and calibration version. If fewer than 30 samples are obtained, the module expands to adjacent increments by one increment, up to a maximum of two expansions. If fewer than 30 samples are still obtained, the current measurement status is set to invalid. The current production condition label is compared field by field with the calibration database label. The numerical range field must fall within the calibrated closed range, and the category field must be completely consistent. If no joint key is found, it is determined to be outside the calibration domain. The current drift amount is calculated according to... calculate, This indicates the maximum thickness drift of the current standard part relative to the most recent calibration result, expressed in mm. Using the circumferential sampling angle index defined in S1, This indicates the current thickness being remeasured, in mm. This indicates the thickness of the same circumferential angle during the most recent calibration, in mm. The drift limit is taken as the 99th percentile of the drift within the corresponding calibration period and calibrated as 0.030 mm. The module takes the following conditions as valid conditions for the current measurement state: the calibration data is within the validity period, the current drift amount does not exceed 0.030 mm, the current working condition is within the calibration domain, S3 has generated a valid pairing index and S4 has generated a valid circumferential rotation angle and obtained the cold equivalent thickness. If any condition is not met or there are fewer than 30 error samples in the same state, the current measurement state is set to invalid and the calculation of the envelope under the conservative thickness is stopped. When the current measurement status is valid, thickness error samples are preferentially selected from the end-to-end verification slice records with the same working condition label, the same registration confidence bin, and the same rotation angle source; the registration confidence bin is fixed as follows. , , and For the first three boxes, the left endpoint is included, but the right endpoint is excluded. For the last box, both endpoints are included. The box division rule version is recorded as CONF-2026-01 and written into both the verification record and the current measurement record. Sample merging is prohibited if the versions are inconsistent. When there are fewer than 30 signed thickness error samples in the same state across the entire chain, single-station samples lacking registration confidence or corner sources are not forcibly merged into the same set. The 95th quantile of the single-station measurement error under the same working condition can be used to form a double-sided conservative thickness coverage for state verification only, which is calibrated by the sum of the 95th quantile of the registration and corner additional errors calibrated by no less than 30 full-chain verification records. However, since the thickness error distribution on one side of the target in the same state cannot be formed, the current measurement state must be set to invalid and the formation of the conservative thickness lower envelope must be stopped. Calculation continues only when there are no fewer than 30 signed samples in the same state across the entire chain. Location errors are still filtered from end-to-end records with actual markers, according to... and Determine the thickness uncertainty and material location uncertainty corresponding to 95% horizontal coverage on both sides of the target, respectively. This indicates the uncertainty in thickness measurement, in mm. This indicates the uncertainty of material location, expressed in meters (m). This means sorting the samples by their absolute values ​​in ascending order and then taking their serial numbers. Empirical 95th percentile calculation for numerical values. This indicates the number of error samples selected in the same layer, and it must be no less than 30. and These represent the filtered sets of thickness error samples and position error samples, respectively. The module also follows Forming a state verification ratio, The value represents the dimensionless state verification ratio, and the larger the ratio, the wider the coverage range of the current value or location relative to the benchmark. and Following the aforementioned definition, This indicates the nominal insulation thickness of the current cable specification, in mm. The material position reference length is represented by the median of the true marker spacing obtained from the registration verification records of the same working conditions without abnormal positions. The unit is m, and in this specific implementation, it is taken as 1.0 m. When the state verification ratio exceeds the 99th quantile of the calibration database, the uncertainty result is retained but a manual verification mark is added. To avoid misclassifying actual insufficient thickness as acceptable due to positive measurement deviations, the module only calculates the thickness error for at least 30 signed thickness error samples from the entire supply chain, based on the same operating conditions, registration confidence levels, and rotation angles. , This represents the upper bound of the thickness uncertainty after the target side is 99% covered horizontally and a zero lower bound is applied, in mm. This represents the 99th percentile operation of an empirical sample. This represents the set of signed thickness error samples in the same state across the entire link. The larger the positive error, the greater the overestimation of the true thickness by the measured thickness. When the samples are insufficient, the aforementioned invalid final state of the quantity value is used, and the upper bound of one side is not replaced by the double-sided coverage amount. Material coordinates and materials corner Calculation of effective cold equivalent thickness at the location , This represents the envelope under conservative thickness, in mm. This represents the cold-state equivalent thickness obtained from S4, in mm. Indicates the relationship with the first The upper bound of the thickness uncertainty corresponding to the current state of each paired section, in mm. Represents material coordinates in meters. The material angle is expressed in degrees. The module writes the cold equivalent thickness, thickness uncertainty, material location uncertainty, upper limit of thickness uncertainty, lower envelope of conservative thickness, calibration version, working condition label, registration confidence and current measurement status into the uncertainty measurement record. Records with valid status are sent to S6. Records with invalid status are prohibited from outputting a qualified conclusion regardless of whether a low thickness measurement value occurs, and are written with a value invalidation flag and a manual review flag. After recalibration or supplementary verification is completed, the module re-verifies the calibration validity period, operating condition label coverage, drift limit, pairing index, circumferential rotation angle, and number of samples in the same state. Only when all conditions are restored to validity will the manual review flag be cleared and the deterministic measurement conclusion be restored. The restoration event, along with the original invalid record flag, is written to the non-overwhelmable measurement audit log.

[0025] In this specific embodiment, S6 includes: The underthickness determination module only processes records where the current measurement state of S5 is valid and a conservative thickness envelope has been formed, reads the shrinkage rate of each circumferential zone in S4, and defines the set of valid zones. and number of effective partitions and in accordance with The circumferential average shrinkage rate was obtained. Indicates the first A set of valid circumferential partition indexes for paired sections This indicates the number of its elements, and this specific implementation requires... , Indicates the first Dimensionless circumferential average shrinkage of paired cross sections This indicates a circumferential partitioned index. Using the partition shrinkage rate defined in S4, evidence calculation is stopped when invalid partitions are not included in the set and summation, and when there are fewer than 10 valid partitions. Module calculates cold and hot thickness residuals , Indicates the first The first paired section The cold and hot thickness residuals of each material corner, in mm, with a positive value indicating that the cold equivalent thickness is greater than the hot reference thickness. This represents the cold-state equivalent thickness, in mm. This indicates the hot-state thickness of the paired hot-state cross-section at the same material angle, in mm. The aforementioned definition will be used; The module obtains the cross-correlation peak time delays of traction speed, extrusion rate, cooling water temperature, and thickness response from the verification samples without process anomalies, and uses the 95th percentile of the absolute value of the time delay as the upper quantile. As the process response coverage time, according to Determine the length of the synchronous material window. This indicates the length of the synchronized material window, in meters (m). This represents the 95th upper quantile of the time delay from the change in process quantity to the thickness response, expressed in seconds. and These represent the hot and cold non-contact velocities of the paired cross-sections, respectively, in m / s, and are given in the order of the current material coordinates before and after the current cross-sections. Establish a synchronous material window within the scope; for each process sampling time, the module reads the speed from the same device clock version and converts it into hot unified material coordinates according to the S1 integration rule, and then shifts the process quantity time series forward or backward according to the sign of the cross-correlation peak delay, so that the process sampling that causes the response is aligned with the residual material coordinates; Before entering the public grid, the module performs the following checks: Evidence of material coordinate definition process using dedicated scalar residuals , Indicates the first The weighted average of the effective material angular thickness residuals for all paired sections, in mm. This represents the set of material corner indices that are valid for S4 circumferential registration and for both hot and cold thicknesses. Indicate the effective material angle quantity and requirements , Using the aforementioned definition of two-dimensional cold and hot thickness residuals; invalid material angles are not included in the set and denominator; if there are fewer than 60 valid angles, the cross section is not included in the process evidence sequence. Only used for traction correlation and mold wear regression; conductor eccentricity and non-uniform shrinkage still use the method that retains the circumferential dimension. ; The module uses the hot material coordinates of the effective paired anchor points of S3 as a common grid. It first samples repeated processes within the same grid cell and takes a time-weighted average. Then, it performs linear interpolation on adjacent effective process samples, without crossing intervals with consecutive missing lengths greater than 0.20 m; scalar residuals... Only samples that simultaneously fall into the common grid constitute a pair of sequences, including the sequence and the time-delayed translation sequence. and When calculating traction evidence Pick In the The value of each common grid point Take the normalization process quantity, and unify the two according to Calculate the Pearson correlation coefficient. This represents the dimensionless correlation coefficient between two sequences, ranging from -1 to 1. Indicates the paired sample index within the public grid. This indicates the number of common valid samples within the window. and Indicates the first A pair of samples, and This represents the mean of the corresponding sequence when the number of common valid samples is less than 20 and the common coverage length is insufficient. Or, no relevant evidence is generated when the variance of any sequence is zero; The module performs the same circumferential harmonic decomposition on the cold and hot thickness residual sequences as S2, and calculates the phase angle. , Indicates the first The minimum circumferential angle between the residual first harmonic phase of each cross section and the conductor center offset direction, in rad. This represents the first harmonic phase of the residual, in rad. The direction of the center offset of conductor S2 is indicated in rad. When the phase angle does not exceed 3° and the correlation coefficient between the residual first harmonic amplitude and the center offset of the conductor within the synchronous material window reaches 0.82, conductor eccentricity evidence is generated. 0.82 is taken as the 99th upper quantile of the correlation coefficient corresponding to the sample without conductor eccentricity verification. The module subtracts the first harmonic reconstruction component corresponding to the conductor center offset direction from the cold and hot thickness residuals of each cross section, and calculates... , Indicates the first The non-uniform shrinkage difference of each cross section, in mm. This indicates the number of harmonics removed after removing the first harmonic. The average residual of each circumferential zone is in mm. When the non-uniform shrinkage difference reaches 0.12 mm, non-uniform shrinkage evidence is generated. 0.12 mm is taken as the 99th upper quantile of the residual difference of the uniform cooling verification sample zone. When the conductor eccentricity condition and the non-uniform shrinkage condition are met at the same time, two evidence labels are retained. The module calculates the derivative of traction speed using a 100 ms differential period. , Indicates the first The traction acceleration at each sampling point in the process, in m / s² 2 , and This indicates the traction speed between adjacent sampling points, in m / s. and This indicates the corresponding sampling time, in seconds; the module first checks the synchronous extrusion volume. The normalized process quantity is calculated only if it is not lower than the effective lower limit of the equipment and is a positive value. , This represents the change in traction normalized to the extrusion volume during the same period, expressed in m / (kg·s). This indicates the extrusion volume during the same period, in kg / s. If this condition is not met, the sampling of this process will be marked as invalid and the division will not be performed. The judgment of this window is stopped when there are invalid extrusion points or insufficient common valid samples within the synchronous material window; otherwise, it is proceeded according to... Execution of evidence determination This represents the dimensionless correlation coefficient between the cold and hot thickness residuals and the normalized process quantity within the synchronous material window. The threshold for traction synchronization is calibrated to 0.78. When the absolute value of the correlation coefficient reaches the threshold, evidence of traction fluctuation is generated. 0.78 is taken as the 99th upper quantile of the absolute value of the correlation coefficient corresponding to the verification sample without traction anomalies. For records of the same mold version with valid metering status and acceptable common material mesh coverage, the module first checks the residuals of the process evidence-specific scalar within each synchronized material window. The arithmetic mean is used to obtain the unique window residual. Then, based on the cumulative running length of the mold, at most one window mean is retained for every 1 km to form the total number of regression samples. And require The least squares regression slope was calculated using the cumulative running length of the mold as the independent variable and the window scalar residual as the dependent variable. , This represents the total number of valid windows for the same mold version participating in the regression. The slope of the scalar residual for process evidence as a function of the cumulative running length of the mold, expressed in millimeters per kilometer. Indicates the first The cumulative running length of the mold for each window, in km. This represents the average cumulative running length, in km. Indicates the first Within each window The average value, in mm. This represents the mean scalar residual of the corresponding window, in mm. Module according to Perform mold wear evidence assessment. Following the aforementioned definition, The 99th upper quantile of the absolute value of the regression slope of the new mold baseline sample is defined as a change of 0.004 mm per kilometer of production. When the absolute value of the slope reaches this threshold, mold wear evidence is generated and the mold version and cumulative running length are recorded. The module compares the conservative thickness envelope with the insulation thickness limit of the cable specification in the acceptable limit table point by point. The acceptable limit table uses voltage level, nominal conductor cross-sectional area, insulation material grade, and applicable standard version as the joint key. The value fields include thickness limit, unit, validity date, and approval number. The thickness limit matched in this task is 3.85 mm. If no unique valid record is found, the measurement status is changed to invalid value, and the insufficient thickness judgment is not performed. The established material coordinates and material angles are recorded as insufficient thickness points. This represents the envelope under conservative thickness, in mm. This indicates the acceptable limit for insulation thickness, in mm. Calculate the thickness gap for each insufficient thickness point. , Indicates the first The first section The severity of insufficient thickness at each material corner is expressed in mm, with larger values ​​indicating more severe thickness loss. This indicates the insulation thickness limit obtained from the acceptable limit table, and for this task it is 3.85 mm. Using the conservative thickness envelope defined in S5, the module first uses the conservative thickness envelope being lower than the acceptable limit as the screening predicate, and then determines the review order within the section according to the severity of the thickness deficiency in descending order, material coordinates in ascending order, and material angle in ascending order. Points with the same severity of thickness deficiency are determined to be reviewed according to the material coordinates in ascending order. Points with insufficient thickness are arranged in ascending order according to material coordinates, and then... Determine whether two adjacent points are continuous. and This represents the material coordinates of two adjacent under-thickness points, in meters (m). and This represents the uncertainty of the material location at the corresponding under-thickness point, in meters. If this inequality is satisfied, the material is classified into the same segment; otherwise, a new segment is created from the next under-thickness point. The material coordinates of the first and last insufficient thickness points in the section are used as the starting and ending points for point estimation, respectively, according to... and Get section Conservative starting point and conservative ending point, and Each represents a segment The conservative starting point and conservative ending point, in meters. and These represent the material coordinates of the first and last under-thickness points, in meters (m). and These represent the material position uncertainty at the first and last points, respectively, in meters (m). For segments within the same disk number, arranged in ascending order from a conservative starting point. and the next adjacent segment ,when When merging two sections, Indicates the conservative endpoint of the previous segment, in meters. The conservative starting point of the next segment is indicated in meters. The merged record retains the estimated boundary of each point before the merge and the conservative boundary after the merge. Points that cross the disk number boundary are segmented separately even if they are consecutive. The module is based on the paired index within the segment according to Forming a process time window, Indicates a section The process time window, in seconds. and They represent the first and second digits within the segment, respectively. The hot and cold measurement times corresponding to each pairing index are in seconds. Then, the disk number and batch are determined according to the disk number, batch, start cumulative length, end cumulative length and effective time fields of the disk number and cumulative length mapping table. When the mapping boundary conflicts, the segment is marked for manual review at the same time. Finally according to Obtain the minimum conservative thickness of the section. Indicates a section The minimum conservative thickness of the lower envelope, in mm. This indicates that the under-thickness point corresponding to the cross section and material angle belongs to the section. The evidence generated under the corresponding conditions is used as the evidence label for the under-thickness section. The report records the estimated start and end meters, conservative start and end meters, circumferential angle set that does not reach the thickness limit, minimum conservative thickness of the section, plate number, batch, registration confidence range, process time window, and evidence label for each section. It also associates the thickness uncertainty, material location uncertainty, calibration version, and metrological status of S5. The report is written into the quality traceability database and outputs a prohibition on release or manual review instruction to the production control console.

[0026] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0027] This invention is based on a dual-station cold and hot section sequence. First, it eliminates the influence of overall dimensional changes on section identification through internal geometric features. Then, it uses cumulative length, speed range, section sequence, and local expansion / contraction constraints to define the matching path, ensuring that cold and hot sections in continuous variable-speed production correspond on the material coordinate system. The matching results, after circumferential rotation correction, are used to calculate the zoned shrinkage rate and cold-state equivalent thickness, thus providing a unified material angular coordinate basis for judging local thickness deficiencies.

[0028] This invention propagates calibration residuals, measurement status, and matching quality into thickness uncertainty and material position uncertainty, forming a conservative thickness lower envelope with one side covering the horizontal, and using position uncertainty for section boundary expansion; when the working condition or measurement status does not meet the valid conditions, the output of the qualified conclusion is stopped and the process is transferred to manual review. Thus, the measurement results not only provide an estimate of the under-thickness point, but also a conservative range for the coverage value error and position mismatch risk.

Claims

1. A method for measuring cable insulation thickness, characterized in that, include: S1. After insulation extrusion and cooling and shaping, the conductor boundary, insulation outer boundary, circumferential thickness sequence, speed and cumulative length are collected to form a cold and hot section sequence. S2. Eliminate the overall scale of the circumferential thickness sequence and generate an internal cross-sectional geometric fingerprint based on the average thickness, circumferential shape, and cross-sectional center geometric features; S3. Based on the range of cumulative length and velocity, determine the candidate intervals for the same material cross section. Under the conditions that the cross section order is not reversed and the local expansion and contraction rate is limited, monotonically match the geometric fingerprint of the internal cross section to obtain the pairing index, material coordinates and registration confidence. S4. Determine the circumferential rotation angle between measurement stations based on the circumferential shape phase and correlation of the paired cross sections, transform the cold circumferential thickness sequence to the same material angular coordinates, and obtain the partition shrinkage rate and cold equivalent thickness. S5. Determine the thickness uncertainty of the cold equivalent thickness and the material position uncertainty of the material coordinates based on the calibration data, measurement status and matching quality. Obtain the upper limit of the thickness uncertainty according to the coverage level of the target side. Subtract this upper limit from the cold equivalent thickness to obtain the conservative thickness lower envelope. S6. Compare the conservative thickness envelope with the insulation thickness acceptance limit to determine the under-thickness point, merge the under-thickness points on the material coordinate and expand the section boundary according to the material position uncertainty, determine the evidence label according to the correlation between the cold and hot thickness residual and the production process quantity, and output the under-thickness section report.

2. The cable insulation thickness measurement method according to claim 1, characterized in that, S1 includes: The hot measurement station is set after the cable insulation extrusion head, and the cold measurement station is set after the cooling and shaping device. Each measurement station uses a through-type measurement device to obtain the conductor boundary and the insulation outer boundary of the same cross section, and generates a circumferential thickness sequence from the radial distance along the circumferential direction between the conductor boundary and the insulation outer boundary. Non-contact speed, cumulative length, and sampling time are collected synchronously at each measurement station, and arranged according to the sampling time to form a hot cross-section sequence and a cold cross-section sequence.

3. The cable insulation thickness measurement method according to claim 1, characterized in that, S2 include: The corresponding circumferential thickness sequence is normalized by the average value of the circumferential thickness sequence to eliminate the overall scale. The normalized circumferential thickness sequence is subjected to circumferential harmonic decomposition. The ratio of the amplitude of each harmonic to the average value of the circumferential thickness sequence is determined as the amplitude ratio of the corresponding order. The phase of each harmonic is extracted, where higher-order harmonics are harmonics of the second order and above. The conductor center offset is determined based on the offset of the conductor boundary center relative to the insulation outer boundary center. The ellipticity is determined based on the major and minor axes of the insulation outer boundary. Local extreme positions in the continuous cross section where the circumferential position change does not exceed the position stability threshold are extracted. The average thickness, amplitude ratio, phase, conductor center offset, ellipticity, and local extreme value positions are arranged in a predetermined field order to form the internal cross-sectional geometric fingerprint. The position stability threshold is determined by the upper quantile of the local extreme value position error obtained from repeated measurements of a standard part.

4. The cable insulation thickness measurement method according to claim 1, characterized in that, S3 include: Based on the equipment path length between the two measurement stations and the lower and upper bounds of the non-contact speed, determine the range of travel time for the material cross section to travel from the hot measurement station to the cold measurement station. The sampling time of the hot section is added to the travel time range to obtain the cold candidate sampling time range, and the cumulative length of the cold section whose sampling time is within the cold candidate sampling time range is determined as the candidate cumulative length range; Using the sum of normalized differences of each field of the internal cross-section geometric fingerprint as the matching cost, and under the constraints that neither the hot cross-section index nor the cold cross-section index of the matching path decreases and the ratio of material coordinate increments between adjacent paired cross-sections is within the allowable range of local expansion and contraction rate, the matching path with the minimum cumulative matching cost is obtained. The matching path determines the pairing index and material coordinates, and the registration confidence is determined based on the interval between the minimum and the second minimum cumulative matching costs, the candidate interval width, and the local slope deviation of the matching path. The allowable interval of the local stretching rate is determined by the lower and upper quantiles of the ratio of cold to hot length increments of the corresponding cable specification in the verification slice data.

5. The cable insulation thickness measurement method according to claim 3, characterized in that, S4 include: The amplitude ratios of the first harmonic in the hot state and the cold state of the first harmonic in the paired cross sections are obtained respectively, and the phase difference of the first harmonic is calculated. When both the thermal first harmonic amplitude ratio and the cold first harmonic amplitude ratio reach the phase usability threshold, the circumferential rotation angle is determined by the first harmonic phase difference; When at least one of the hot-state first harmonic amplitude ratio and the cold-state first harmonic amplitude ratio does not reach the phase availability threshold, the peak values ​​of the correlation coefficients between the circumferential morphological sequences reconstructed by second-order and above harmonics are arranged in descending order of value. When the difference between the first peak value and the second peak value reaches the peak value discrimination threshold, the circumferential rotation angle is determined by the angular displacement corresponding to the first peak value. When the ratio of the first harmonic amplitude in the hot state or the ratio of the first harmonic amplitude in the cold state does not reach the phase available threshold, and the difference between the first peak value and the second peak value does not reach the peak value discrimination threshold, a circumferential registration invalid state is generated and the output of the cold equivalent thickness of the corresponding cross section is stopped. The available phase threshold is determined by the lower limit of the amplitude ratio corresponding to the first harmonic phase variance of repeated measurements of the standard part not exceeding the preset upper limit of the phase variance, and the peak discrimination threshold is determined by the lower quantile of the difference between the correct angular displacement peak and the other angular displacement peaks in repeated measurements of the standard part. The cold circumferential thickness sequence is resampled according to the circumferential rotation angle. Under the same material angular coordinate, the shrinkage rate of each circumferential zone is determined according to the ratio of the difference between the hot and cold thicknesses of the paired section to the hot thickness. The cold thickness that is paired with the hot section and has been resampled is determined as the cold equivalent thickness.

6. The cable insulation thickness measurement method according to claim 5, characterized in that, S5 include: The residuals obtained by subtracting the reference value of the standard part from the thickness measurement value of the traceable standard part or the residuals obtained by subtracting the reference value of the slice from the thickness measurement value of the verification slice are grouped according to the cable specification and production condition label. The thickness error samples are determined by combining the calibration error and drift of the hot and cold measurement stations. The position error samples are determined based on the cumulative length candidate interval width, the interval between the minimum cumulative matching cost and the second minimum cumulative matching cost, the angular width of the circumferential morphological correlation coefficient peak, and the registration residual. The current measurement status is determined by whether the calibration data is valid, whether the current drift exceeds the drift limit determined by the upper quantile of the drift within the calibration cycle, whether the current production condition label is covered by the calibration data, and whether a valid pairing index and circumferential rotation angle status combination is generated. When the current measurement state is valid and the cold equivalent thickness is obtained, select the thickness error sample and position error sample with the same current measurement state from the calibration data. Determine the thickness uncertainty by the quantile corresponding to the absolute value of the thickness error sample at the target double-sided coverage level. Determine the material position uncertainty by the quantile corresponding to the absolute value of the position error sample at the target double-sided coverage level. Then, truncate all thickness error samples at the target one-sided coverage level by setting the lower limit to zero to determine the upper bound of the thickness uncertainty. If any of the following occurs: the calibration data is invalid, the current drift exceeds the drift limit, the current production condition label is not covered by the calibration data, no valid pairing index or circumferential rotation angle is generated, no error sample with the same current measurement state is obtained, or there is an invalid circumferential registration state, the current measurement state will be determined as an invalid value state and the calculation of the conservative thickness envelope will be stopped. Only when the current measurement state is valid, the conservative lower envelope of the thickness is obtained by subtracting the upper bound of the thickness uncertainty from the cold-state equivalent thickness.

7. The cable insulation thickness measurement method according to claim 6, characterized in that, S6 include: For sections with an effective conservative thickness envelope, the cold equivalent thickness is subtracted from the hot reference thickness, which is calculated from the hot thickness according to the circumferential average shrinkage rate of the paired sections, under the same material coordinates and material angular coordinates. The cold and hot thickness residuals are obtained, where a positive residual indicates that the cold equivalent thickness is greater than the hot reference thickness. The upper quantile of the time delay between the change in production process quantity and the thickness response in the verification sample without process anomalies is multiplied by the arithmetic mean of the hot and cold non-contact velocities of the paired cross sections to obtain the synchronization material window length, and the range of adjacent material coordinates of the corresponding length is used as the synchronization material window. When the first harmonic phase of the cold and hot thickness residual within the synchronous material window is consistent with the direction of conductor center offset, and the correlation coefficient between the first harmonic amplitude and the conductor center offset reaches the eccentric synchronous correlation threshold, conductor eccentricity evidence is generated. The eccentric synchronous correlation threshold is determined by the upper quantile of the corresponding correlation coefficient of the sample without conductor eccentricity verification. When the first harmonic component corresponding to the conductor center offset is removed from the cold and hot thickness residual, and the difference between the residuals of each circumferential zone reaches the non-uniform shrinkage threshold, evidence of non-uniform shrinkage is generated. The non-uniform shrinkage threshold is determined by the upper quantile of the difference in the residuals of the zoned residuals of the uniform cooling verification sample. The conductor eccentricity evidence and the non-uniform shrinkage evidence are allowed to be generated simultaneously when their respective conditions are met. When the correlation coefficient between the residual cold and hot thickness within the synchronous material window and the ratio of the traction speed derivative to the extrusion amount reaches the traction synchronization correlation threshold, traction fluctuation evidence is generated. When the regression slope of the cold and hot thickness residuals with the cumulative running length of the mold reaches the wear slope threshold, mold wear evidence is generated. The traction synchronization correlation threshold and the wear slope threshold are determined by the upper quantiles of the corresponding statistics of the non-traction abnormal samples and the new mold benchmark samples, respectively. The section whose envelope under the conservative thickness does not reach the qualified limit of insulation thickness is identified as the insufficient thickness point, and the insufficient thickness points are merged in the order of material coordinates to form the insufficient thickness section. The evidence is used as the evidence label of the insufficient thickness section.

8. The cable insulation thickness measurement method according to claim 7, characterized in that, The report on the insufficient thickness section includes the start and end meters of the insufficient thickness section, the circumferential angle at which the lower envelope of the thickness does not reach the qualified limit of insulation thickness, the reel number, the batch number, the registration confidence level, the process time window, the evidence label, the point estimation boundary, and the conservative boundary. The process time window is determined based on the hot and cold measurement times corresponding to the paired indexes within the thin section, and the reel number and batch to which the thin section belongs are determined based on the mapping relationship between the reel number and the cumulative length.

9. The cable insulation thickness measurement method according to claim 6, characterized in that, The current production condition label is composed based on the current cable specifications, material grade, extrusion temperature range, cooling conditions, and linear speed range. The current production condition label is then compared with the set of condition labels covered by the calibration data to determine whether the current condition is within the calibration range. The current drift amount is determined based on the difference between the standard part retest result and the most recent calibration result, and the current drift amount is compared with the drift limit. When the calibration data is invalid, the current operating condition is outside the calibration domain, the current drift exceeds the drift limit, or the current measurement state is invalid, the output of a qualified conclusion is prohibited. Regardless of whether a thin section is formed, the measurement result record containing the invalid value flag and the manual verification flag will be output. After recalibration or verification is completed, the deterministic metrological conclusion will be restored only when the calibration data is valid, the current operating condition is within the calibration domain, the current drift amount does not exceed the drift limit, and the current measurement state is not an invalid value state.

10. The cable insulation thickness measurement method according to claim 8, characterized in that, Points with material coordinate intervals no greater than the sum of the material position uncertainties of the two corresponding points are identified as continuous points with insufficient thickness. The material coordinates of the first and last thinnest points obtained after merging consecutive thinnest points are used as the starting point and ending point of point estimation, respectively. The starting point of point estimation is extended forward by the material position uncertainty of the first thinnest point to obtain a conservative starting point, and the ending point of point estimation is extended backward by the material position uncertainty of the last thinnest point to obtain a conservative ending point. When the conservative boundaries of two thin sections within the same disk number intersect, the two thin sections are merged, and the estimated point boundary before merging and the conservative boundary after merging are recorded simultaneously in the thin section report.

Images