Dynamic self-adaptive calibration method and system for cable eccentricity

Abstract

The application relates to the technical field of cable insulation layer extrusion eccentricity control, and discloses a cable eccentricity rate dynamic self-adaptive calibration method and system, which comprises the following steps: collecting traction speed time sequence data and cross-section eccentricity detection data, calculating a dynamic transmission delay to generate a dynamic delay sequence, performing dynamic timestamp alignment and constructing a speed segmented eccentricity state queue, constructing an eccentricity rate oscillation detection ring buffer to calculate an eccentricity rate oscillation index, performing branch calibration gain calculation according to double threshold value comparison, and generating adaptive calibration gain to drive die position calibration; through dynamic delay sequence and speed segmented self-adaptive calibration, the application solves the problems of phase lag and adjustment oscillation in closed-loop control.

Description

Technical Field

[0001] This invention relates to the field of cable insulation layer extrusion eccentricity control technology, and particularly to a dynamic adaptive calibration method and system for cable eccentricity. Background Technology In the continuous extrusion process for producing cable insulation layers, an automatic eccentricity adjustment technique based on real-time monitoring data is commonly used to control the uniformity of the insulation layer thickness. This technique measures the eccentricity of the insulation layer after extrusion and feeds the calculation results back to the adjustment mechanism to compensate and correct the die position or process parameters, thereby maintaining the concentricity of the insulation layer and the conductor. The control objective of this type of adjustment system is to dynamically maintain the eccentricity within the allowable range based on changes in traction speed and extrusion volume. Its key feature is the use of a closed-loop feedback circuit to achieve rapid deviation correction, thus meeting the basic requirement of thickness consistency in continuous production.

[0002] In actual production lines, during acceleration start-up or deceleration traction phases, operators often observe the following phenomenon: after the adjustment system performs a correction based on eccentricity detection data, the side with thinner insulation does not return to the center position but instead shifts to the opposite side in the next detection cycle, causing the eccentricity to amplify in the opposite direction. Subsequently, the system detects again and performs reverse correction, resulting in another excessive offset in the opposite direction. Thus, during continuous cable routing, the eccentricity of the cable cross-section exhibits a wave-like periodic oscillation, with the thinnest points of the insulation alternating longitudinally. This phenomenon causes continuous localized thinning areas in the cable insulation layer. During subsequent shielding attenuation testing or high-frequency network transmission performance testing, these areas cause sudden spikes in the return loss curve, resulting in momentary discontinuities in transmission impedance, severely reducing the yield rate of finished cables and signal transmission quality. Therefore, how to reduce overcompensation in eccentricity adjustment during dynamic changes in traction speed is a problem that urgently needs to be solved in this field. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a dynamic adaptive calibration method for cable eccentricity. This method calculates the dynamic transmission delay using traction speed time-series data and performs dynamic timestamp alignment on cross-sectional eccentricity detection data. By combining a speed-segmented eccentricity state queue and an eccentricity oscillation detection ring buffer, adaptive calibration gain calculation and closed-loop calibration are achieved. This method effectively compensates for detection transmission delay, suppresses eccentricity oscillation, and improves the calibration accuracy of the extrusion die when the production line speed changes.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a dynamic adaptive calibration method for cable eccentricity, comprising: Traction speed time-series data and cross-sectional eccentricity detection data are collected from the extrusion production line. The dynamic transmission delay between the detection end and the extrusion end is calculated based on the traction speed time-series data to generate a dynamic delay sequence. The cross-sectional eccentricity detection data is dynamically timestamped using the dynamic delay sequence to generate a time-aligned eccentricity data table; a speed segmented eccentricity state queue is constructed based on the time-aligned eccentricity data table and the traction speed time series data. An eccentricity oscillation detection ring buffer is constructed using the speed segmented eccentricity state queue, the eccentricity oscillation index is calculated, and branch calibration gain calculation is performed based on the comparison relationship between the eccentricity oscillation index and the dual thresholds to generate an adaptive calibration gain. The adaptive calibration gain and oscillation lock flag are used to drive the generation of extruder die position calibration instructions, execute closed-loop calibration actions, and write back the calibration results to the speed segment eccentricity state queue.

[0005] Furthermore, the acquisition of traction speed timing data and cross-sectional eccentricity detection data from the extrusion production line includes: A speed sensor is deployed at the extruder traction mechanism, and a smart projector is deployed at the detection end. The traction speed time series data is collected synchronously at a fixed sampling period, and the cross-sectional eccentricity detection data is collected using the time of cross-sectional image acquisition by the smart projector as the timestamp. The real-time traction speed at the sampling moment is obtained by the pulse signal output by the speed sensor within a fixed sampling period; The intelligent projector images the cable cross-section and uses image processing to identify the center of the outer circle of the insulation layer and the center of the conductor outline to extract the eccentricity direction angle and eccentricity as the cross-section eccentricity detection data.

[0006] Further, the step of calculating the dynamic transmission delay between the detection end and the extrusion end based on the traction speed time-series data, and generating a dynamic delay sequence, includes: Based on the fixed physical distance between the detection end and the extrusion end, the single-step transmission delay corresponding to each sampling moment in the traction speed time series data is calculated using the real-time traction speed at that moment. The single-step transmission delay is equal to the fixed physical distance divided by the real-time traction speed. The single-step transmission delay at each sampling time is arranged in the sampling time sequence to generate the dynamic delay sequence.

[0007] Further, the step of performing dynamic timestamp alignment on the cross-sectional eccentricity detection data using the dynamic delay sequence to generate a time-aligned eccentricity data table includes: The actual time of extrusion corresponding to the eccentricity detection data is obtained by subtracting the single-step transmission delay of the corresponding moment in the dynamic delay sequence from the acquisition timestamp of each record in the cross-sectional eccentricity detection data. Each record in the cross-sectional eccentricity detection data, along with the actual time of occurrence of the extrusion, the eccentricity direction angle, and the eccentricity amount, is written into the time-aligned eccentricity data table.

[0008] Furthermore, the step of constructing a speed segmented eccentricity state queue based on the time-aligned eccentricity data table and the traction speed time-series data includes: The production process is divided into three speed ranges: acceleration, constant speed, and deceleration, based on the real-time traction speed in the traction speed time series data. Each speed range corresponds to all records in the time alignment eccentricity data table that fall within the time range of that range. The eccentric record sets corresponding to the three types of speed ranges are arranged in chronological order to form eccentric sub-queues for acceleration, uniform speed, and deceleration, respectively. The three are then merged into the speed segment eccentric state queue.

[0009] Furthermore, the division of the production process into three speed ranges—acceleration, constant speed, and deceleration—based on the real-time traction speed in the traction speed time-series data includes: The traction acceleration at that sampling moment is obtained by dividing the difference between the real-time traction speeds at two adjacent sampling times in the traction speed time series data by a fixed sampling period. If the traction acceleration is greater than the positive acceleration threshold, then the sampling moment is classified as the acceleration phase; If the traction acceleration is less than the negative acceleration threshold, then the sampling moment is classified as the deceleration phase; If the traction acceleration is between the negative acceleration threshold and the positive acceleration threshold, then the sampling moment is classified as a uniform speed segment.

[0010] Furthermore, the step of constructing an eccentricity oscillation detection circular buffer using the segmented eccentricity state queue at the stated speed, and calculating the eccentricity oscillation exponent, includes: With a fixed capacity as the length of the circular buffer, the eccentricity is written into the eccentricity oscillation detection circular buffer according to the time sequence of the segmented eccentricity state queue at the speed, and the newly written record overwrites the oldest record. After each write operation, the mean absolute value of the adjacent differences of all fixed-capacity eccentricity records in the eccentricity oscillation detection ring buffer is calculated and denoted as the eccentricity oscillation index.

[0011] Furthermore, the dual thresholds include a strong oscillation threshold and a weak oscillation threshold; Both the strong oscillation threshold and the weak oscillation threshold are dynamically updated based on the sliding standard deviation of the eccentricity in the current velocity segment sub-queue. When the sliding standard deviation increases, the two thresholds are adjusted upwards simultaneously, and when the sliding standard deviation decreases, the two thresholds are adjusted downwards simultaneously. The sliding standard deviation is calculated by taking the standard deviation of the latest sliding window length bar eccentricity record in the current speed segment sub-queue of the speed segment eccentricity state queue at the actual time of extrusion.

[0012] Furthermore, the calculation of the branch calibration gain includes: If the eccentricity oscillation index is less than or equal to the weak oscillation threshold, the mean of the most recent eccentricity records in the current velocity segment sub-queue is multiplied by the reference gain coefficient to obtain the adaptive calibration gain. If the eccentricity oscillation index is greater than the weak oscillation threshold and less than the strong oscillation threshold, then a damping correction amount with the eccentricity oscillation index as the weight is superimposed on the reference gain coefficient to obtain the adaptive calibration gain. If the eccentricity oscillation index is greater than or equal to the oscillation strength threshold, the adaptive calibration gain is set to the fixed proportional lower limit of the reference gain coefficient, and the oscillation lock flag is output to the triggered state.

[0013] Furthermore, the generation of the drive extruder die position calibration command and the execution of the closed-loop calibration action include: Read the eccentricity direction angle and eccentricity amount of the latest record in the time alignment eccentricity data table, determine the mold head adjustment direction by the eccentricity direction angle, calculate the mold head displacement command amount by multiplying the eccentricity amount by the adaptive calibration gain, and generate a calibration command. Check the status of the oscillation lock flag; If the oscillation lock flag is not triggered, the calibration command will be directly sent to the extruder die head actuator. If the oscillation lock flag is triggered, a forced waiting interval is inserted before the calibration command is issued. During the waiting period, cross-sectional eccentricity detection data is continuously collected and the eccentricity oscillation detection ring buffer is updated. The calibration command is issued only after the oscillation lock flag is re-determined to be untriggered.

[0014] Furthermore, the process of writing back the calibration results to the velocity segment eccentricity state queue includes: After the extruder die head actuator completes the die head displacement, the actual execution completion time is used as the timestamp, and the execution time is recorded as the calibration action timestamp at the end of the corresponding speed interval sub-queue of the speed segment eccentricity state queue. Simultaneously, subsequent cross-sectional eccentricity detection data are collected and written into the corresponding sub-queue. The natural convergence of subsequent eccentricity is used as the calibration result to form a reverse constraint closed loop for the calibration parameters.

[0015] Secondly, the present invention provides a dynamic adaptive calibration system for cable eccentricity, comprising: The dynamic transmission delay calculation module is used to collect traction speed time-series data and cross-sectional eccentricity detection data from the extrusion production line, calculate the dynamic transmission delay between the detection end and the extrusion end according to the traction speed time-series data, and generate a dynamic delay sequence. The data table and queue construction module is used to perform dynamic timestamp alignment on the cross-sectional eccentricity detection data using the dynamic delay sequence to generate a time-aligned eccentricity data table; and to construct a speed segmented eccentricity state queue based on the time-aligned eccentricity data table and the traction speed time series data. An adaptive calibration gain generation module is used to construct an eccentricity oscillation detection ring buffer using the segmented eccentricity state queue at the speed, calculate the eccentricity oscillation index, perform branch calibration gain calculation based on the comparison relationship between the eccentricity oscillation index and the dual thresholds, and generate an adaptive calibration gain. The closed-loop calibration execution module is used to drive the generation of extruder die head position calibration instructions with the adaptive calibration gain and oscillation lock flag, execute closed-loop calibration actions, and write back the calibration results to the speed segment eccentricity state queue.

[0016] Compared to existing technologies, the advantages of this invention are as follows: By constructing a dynamic delay sequence and a speed-segmented eccentricity state queue, this invention effectively solves the problems of eccentricity adjustment phase lag and periodic overcompensation oscillation caused by dynamic changes in traction speed during cable extrusion production. This method calculates the single-step transmission delay using the real-time traction speed at each sampling moment, ensuring that the eccentricity detection data for each section is accurately traced back to the corresponding actual time of occurrence in the extruder. This significantly reduces the cumulative timing error compared to the systematic phase deviation introduced by a fixed delay constant during acceleration and deceleration. Simultaneously, by dividing the eccentricity records into independent sub-queues according to acceleration, constant speed, and deceleration stages, the weight dilution of cross-speed interval mixed statistics is avoided from the grouping source, ensuring that the calibration parameter calculation matches the eccentricity distribution characteristics under different operating conditions. Based on this segmented queue, the fixed capacity of the annular buffer for eccentricity oscillation detection is adaptively narrowed by the constraint of historical eccentricity variance. The strong and weak oscillation thresholds are dynamically updated by the sliding standard deviation. Together, these two factors enable the eccentricity oscillation index to respond quickly in the high-fluctuation range while maintaining a judgment benchmark that fluctuates with operating conditions. Damping correction intervenes earlier before the oscillation develops into overcompensation. Furthermore, in strong oscillation states, by temporarily delaying the issuance of calibration commands and introducing a waiting window that expands and contracts with the traction speed, the positive feedback path of oscillation superposition is cut off. In addition, by writing back the calibration action execution time to the speed-segmented eccentricity state queue, the sliding standard deviation naturally narrows as the calibration history converges, and the two thresholds are adjusted synchronously, thereby automatically improving the system's sensitivity to micro-oscillations. This overcomes the deficiency of static thresholds in detecting small oscillations after calibration convergence without requiring additional parameter adjustment modules. This solution improves the timing matching accuracy between the eccentric calibration command and the dynamic operating state of the extruder, enhances the closed-loop stability of the cable insulation layer eccentric adjustment under variable speed conditions, and makes the adaptive control system more capable of identifying the eccentric distribution and adjusting the gain parameters under different operating conditions. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the dynamic adaptive calibration method for cable eccentricity in this invention; Figure 2 This is a schematic diagram of the deployment of the smart projector and speed sensor in an embodiment of the present invention; Figure 3 This is a schematic diagram of time backtracking mapping alignment in an embodiment of the present invention; Figure 4This is a schematic diagram of the construction of the velocity segmented eccentric state queue in an embodiment of the present invention; Figure 5 This is a schematic diagram of writing to the eccentricity oscillation detection ring buffer in an embodiment of the present invention; Figure 6 This is a schematic diagram illustrating the calculation of branch calibration gain determination in an embodiment of the present invention; Figure 7 This is a schematic diagram illustrating forced waiting interval insertion in an embodiment of the present invention; Figure 8 This is a schematic diagram of the calibration action timestamp write-back mechanism in an embodiment of the present invention; Figure 9 This is a functional block diagram of the dynamic adaptive calibration system for cable eccentricity in this invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1: Please see Figure 1 As shown, this embodiment provides a dynamic adaptive calibration method for cable eccentricity, including: S1: Collect traction speed time-series data and cross-sectional eccentricity detection data from the extrusion production line, calculate the dynamic transmission delay between the detection end and the extrusion end based on the traction speed time-series data, and generate a dynamic delay sequence.

[0021] This step addresses the phase lag issue in eccentricity adjustment during cable insulation extrusion. When the extruder is in continuous production and the traction speed changes (acceleration or deceleration), the cross-sectional eccentricity data collected by the detection end corresponds to a cable segment extruded from the extrusion end some time prior. There is a spatial transmission time between the detection end and the extrusion end, which is related to the traction speed. Existing eccentricity control systems use a fixed delay constant for feedback calculation, failing to dynamically update the transmission time with changes in traction speed. This results in a misalignment between the detection data and the current operating state of the extruder in terms of timestamps, leading to phase lag in calibration commands. This step aims to calculate the single-step transmission delay between the detection end and the extrusion end using the real-time traction speed at each sampling moment, generating a dynamic delay sequence to provide a time reference for dynamic correction at each moment in subsequent steps.

[0022] Further, step S1 includes: S11: Deploy a speed sensor at the extruder traction mechanism and a smart projector at the detection end to synchronously collect traction speed time-series data with a fixed sampling period, and collect cross-sectional eccentricity detection data with the time of cross-sectional image acquisition by the smart projector as the timestamp.

[0023] This step aims to obtain the two types of raw data required for the eccentric adjustment closed loop and ensure that the two types of data are recorded synchronously with timestamps under a unified sampling clock, providing raw input for S12 to calculate the single-step transmission delay and S21 to execute the time backtracking mapping.

[0024] Specifically, the speed sensor is an incremental rotary encoder, which is deployed at the end of the traction wheel shaft of the extruder traction mechanism. The traction wheel directly drives the cable to move axially. The angular displacement of the traction wheel shaft is proportional to the axial displacement of the cable. Deploying an incremental rotary encoder at the end of the traction wheel shaft can directly collect pulse signals related to the axial movement speed of the cable.

[0025] The real-time traction speed is calculated as follows: multiply the pulse count output by the incremental rotary encoder within a fixed sampling period by the circumference of the traction wheel, and then divide by the product of the number of pulses per revolution of the incremental rotary encoder and the fixed sampling period to obtain the real-time traction speed at that sampling moment, in meters per second.

[0026] The intelligent projector used is a small-cable intelligent projector, model SCP-AE-Z5020, deployed at the detection end after the outlet of the extrusion cooling water tank. A fixed physical distance exists between the detection end and the extruder die outlet along the cable path. The intelligent projector images the cable cross-section and identifies the center of the outer contour of the insulation layer and the center of the conductor contour through image processing. The offset distance of the conductor contour center relative to the outer contour center of the insulation layer is used as the eccentricity, measured in micrometers. The angle between the vector pointing from the outer contour center of the insulation layer to the conductor contour center and the horizontal reference direction is used as the eccentricity direction angle, ranging from 0 degrees to 360 degrees. The cross-sectional eccentricity detection data includes the eccentricity direction angle and the eccentricity.

[0027] The method for determining the fixed sampling period is as follows: take a sampling period in which the axial movement distance of the cable between two adjacent frames does not exceed the preset axial resolution length. The preset axial resolution length is determined according to the detection accuracy requirements of the insulation layer thickness uniformity of the finished cable. For example, the preset axial resolution length can be set to 0.1 meters. When the maximum traction speed is 2 meters per second, the fixed sampling period can be set to 50 milliseconds.

[0028] Synchronous acquisition refers to the speed sensor and the smart projector using the same sampling clock signal as a reference, completing data sampling and recording the same timestamp within the same sampling period boundary. Synchronous acquisition ensures that the traction speed timing data and the cross-sectional eccentricity detection data are strictly aligned on the time axis, so that the real-time traction speed used by S12 to calculate the single-step transmission delay corresponds one-to-one with the actual motion state of the cable at that moment. See also Figure 2 This is a schematic diagram of the deployment of the smart projector and speed sensor provided in an embodiment of this application. Figure 2 As shown, the front end of the extruder is the die outlet. The extruded cable is transmitted over a fixed physical distance to the detection end after the outlet of the extrusion cooling water tank. The detection end is equipped with an intelligent projector for cross-sectional imaging. At the end traction mechanism, the cable wraps around the traction wheel, and a speed sensor for collecting pulse signals is deployed at the shaft end of the traction wheel. Both types of sensors are deployed and synchronously collect traction speed time-series data and cross-sectional eccentricity detection data at a fixed sampling period. This ensures that the two types of data are strictly aligned on the time axis and recorded with the same timestamp. This synchronous acquisition operation ensures that the real-time traction speed used in subsequent calculations of single-step transmission delay corresponds one-to-one with the actual motion state of the cable at that moment, and provides the original input for performing time-backtracking mapping.

[0029] S12: Based on the fixed physical distance L between the detection end and the extrusion end, calculate the single-step transmission delay corresponding to each sampling moment using the real-time traction speed v in the traction speed time series data. The single-step transmission delay is equal to the fixed physical distance L divided by the real-time traction speed v. Arrange the single-step transmission delays of each sampling moment in the sampling time series order to generate a dynamic delay sequence.

[0030] This step aims to transform the dynamic changes in traction speed into a transmission delay estimate at each sampling time, so that the delay estimate is continuously updated with the traction speed, and the phase lag is significantly reduced compared to the systematic phase deviation introduced by a fixed delay constant during acceleration and deceleration.

[0031] Specifically, the method for determining the fixed physical distance L is as follows: measure the physical length along the cable path between the extruder die outlet and the imaging position of the smart projector section. For example, the fixed physical distance L can be set to 2.0 meters.

[0032] The expression for calculating single-step transmission delay is: in, Let L be the single-step transmission delay corresponding to the i-th sampling moment in the traction speed time series data, in seconds; L is the fixed physical distance determined by S12. This refers to the real-time traction speed at the i-th sampling moment in the traction speed time series data collected by S11.

[0033] The design principle of the calculation expression for single-step transmission delay is as follows: the numerator, a fixed physical distance L, represents the spatial distance that the cable segment needs to travel from the extrusion end to the detection end, and the denominator is the real-time traction speed. The axial movement speed of the cable at the sampling moment is characterized. Dividing the two values ​​yields an estimate of the time taken for the cable segment to travel a fixed physical distance L, i.e., an estimate of the time interval between the cross-section currently sampled at the detection end and the actual time of extrusion at the extrusion end. When the real-time traction speed... Increased single-step transmission delay Reduce, when the real-time traction speed Reduce single-step transmission delay The increase in speed makes the delay estimate respond continuously in the opposite direction to the dynamic changes in traction speed, which is consistent with the physical law of the spatial transmission time of the cable segment changing with speed. Under engineering conditions where the extruder acceleration and deceleration are smooth and the fixed sampling period is much smaller than the single-step transmission delay, the instantaneous speed approximation at each moment can effectively follow the dynamic changes in speed, and significantly reduce the delay estimation error relative to the systematic phase deviation introduced by the fixed delay constant in the acceleration and deceleration stages.

[0034] The single-step transmission delays calculated at each sampling time are arranged sequentially according to the sampling time sequence to form a dynamic delay sequence. The dynamic delay sequence is a one-dimensional numerical sequence indexed by the sampling time sequence, and the sequence length is equal to the number of sampling points of the traction speed time sequence data.

[0035] For example, when the fixed physical distance L is 2.0 meters and the real-time traction speed at a certain sampling time is 2.0 meters per second, the single-step transmission delay at that time is 2.0 divided by 2.0 equals 1.0 seconds; when the production line accelerates to the real-time traction speed at that sampling time of 4.0 meters per second, the single-step transmission delay at that time is 2.0 divided by 4.0 equals 0.5 seconds. The single-step transmission delay is shortened from 1.0 seconds to 0.5 seconds as the traction speed increases.

[0036] The dynamic delay sequence calculates the single-step transmission delay at real-time traction speed, ensuring continuous updates to the delay estimate as the traction speed dynamically changes. During acceleration or deceleration, the actual time taken for the cable segment to traverse a fixed physical distance L changes continuously with the traction speed. If a fixed delay constant is used to uniformly convert the entire cross-sectional eccentricity detection data, the actual transmission time in the acceleration phase is less than the fixed constant, while the actual transmission time in the deceleration phase is greater than the fixed constant. The conversion error accumulates in a form proportional to the speed change, causing the extrusion end time traced back to by the detection data to deviate from the actual extrusion time. The single-step transmission delay calculated at each time step ensures that the delay estimate at each sampling moment continuously corresponds to the cable's motion state at that moment. Compared to the systematic phase deviation introduced by the fixed delay constant in the acceleration and deceleration phases, this significantly reduces the conversion error, providing a continuously corrected time reference for the time backtracking mapping of S21.

[0037] S2: Perform dynamic timestamp alignment on the cross-sectional eccentricity detection data using a dynamic delay sequence to generate a time-aligned eccentricity data table; construct a speed segmented eccentricity state queue based on the time-aligned eccentricity data table and the traction speed time series data.

[0038] This step addresses the timing accumulation error caused by the misalignment between the timestamps of the data collected at the detection end and the extrusion end's operating status. The cross-sectional eccentricity detection data collected at the detection end lags behind the actual extrusion action at the extrusion end. If the collection timestamp is directly associated with the current operating parameters of the extruder, the calibration command will be applied to the incorrect extrusion end time, causing regulation oscillations in the closed-loop control. This step aims to use the dynamic delay sequence generated in S12 to back-map each cross-sectional eccentricity detection data back to its corresponding actual extrusion time, and to divide the eccentricity records into different speed condition sub-queues according to the traction speed, providing condition-adaptive statistical input for S3 to independently calculate calibration parameters according to speed condition.

[0039] Further, step S2 includes: S21: Subtract the single-step transmission delay of the corresponding moment in the dynamic delay sequence from the acquisition timestamp of each record in the cross-sectional eccentricity detection data to obtain the actual extruder occurrence time corresponding to the eccentricity detection data. Write each record in the cross-sectional eccentricity detection data, together with the actual extruder occurrence time, eccentricity direction angle, and eccentricity amount, into the time-aligned eccentricity data table.

[0040] This step aims to perform a time-backtracking mapping on each eccentricity detection record, so that each record corresponds to the actual time of extrusion at the level of a single sampling cycle, which significantly reduces the time alignment error compared to the phase lag caused by uniform conversion with a fixed delay constant.

[0041] Specifically, the calculation method for the actual occurrence time of the extruder is as follows: take the collection timestamp of each record in the cross-sectional eccentricity detection data collected by S11, subtract the single-step transmission delay corresponding to the same sampling time as the collection timestamp in the dynamic delay sequence generated by S12, and obtain the actual occurrence time of the extruder corresponding to the eccentricity detection data.

[0042] The time-aligned eccentricity data table is organized by record entries. Each record contains three fields: the actual time the eccentricity occurred in the extruder, the eccentricity direction angle, and the eccentricity amount. The number of record entries equals the number of records in the cross-sectional eccentricity detection data. See also... Figure 3 This is a schematic diagram of time backtracking mapping alignment provided in an embodiment of this application. For example... Figure 3As shown, the top section contains cross-sectional eccentricity detection data, including three fields: acquisition timestamp, eccentricity direction angle, and eccentricity amount. The acquisition timestamp is extracted and subtracted from the single-step transmission delay of the external input. This is used to perform a time-backtracking mapping operation to calculate the actual occurrence time of the extruder. This calculated occurrence time, along with the directly acquired eccentricity direction angle and eccentricity amount, is combined and output downwards, and written into the underlying time-aligned eccentricity data table. This time-backtracking mapping operation makes the actual occurrence time of the extruder corresponding to each eccentricity detection record accurate to the single-step sampling period level. Compared to the phase lag caused by uniform calculation using a fixed delay constant, this significantly reduces time alignment errors.

[0043] For example, if the acquisition timestamp of a certain cross-sectional eccentricity detection record is 100.000 seconds and the single-step transmission delay corresponding to this acquisition time is 1.0 second, then the actual extruder occurrence time corresponding to this eccentricity detection data is 100.000 minus 1.0 equals 99.000 seconds. When the production line is in the acceleration phase and the single-step transmission delay corresponding to the adjacent acquisition time is shortened to 0.5 seconds, the actual extruder occurrence time obtained by backtracking the record with an acquisition timestamp of 100.050 seconds is 100.050 minus 0.5 equals 99.550 seconds.

[0044] S21 performs time-backtracking mapping on the cross-section eccentricity detection data using a single-step transmission delay recorded line by line. This ensures that the actual extruder occurrence time corresponding to each eccentricity detection record is accurate to the single-step sampling period level, rather than using a fixed delay constant for uniform conversion. During dynamic changes in traction speed, the fixed delay constant applies the same conversion amount to the acceleration and deceleration records, causing the backtracked extrusion end time to lead in the acceleration phase and lag in the deceleration phase. The conversion error accumulates in the same direction among consecutive records and eventually enters the closed loop in the form of phase lag. Line-by-line backtracking mapping ensures that the conversion amount of each record strictly corresponds to the single-step transmission delay at the time of record acquisition. Phase lag is significantly suppressed during record generation, providing a time-based dynamic correction eccentricity record for the speed segmentation in S22.

[0045] S22: The production process is divided into three speed intervals: acceleration, constant speed, and deceleration, based on the real-time traction speed in the traction speed time series data. Each speed interval corresponds to all records in the time-aligned eccentricity data table that fall within the time range of that interval. The eccentricity record sets corresponding to each of the three speed intervals are arranged in time sequence to form the eccentricity sub-queues of the acceleration, constant speed, and deceleration segments, respectively. The three are then merged into a speed segment eccentricity state queue.

[0046] This step refines the time-aligned eccentricity data table into groups based on speed conditions, enabling the S3 calibration parameter calculation to independently read the statistical characteristics of the corresponding sub-queue according to speed conditions. This avoids the large eccentricity in the acceleration phase diluting the calibration gain in the constant speed phase when statistics are mixed across speed ranges.

[0047] Specifically, the speed range is divided as follows: the difference between the real-time traction speeds of two adjacent sampling times in the traction speed time series data is divided by the fixed sampling period to obtain the traction acceleration at that sampling time; if the traction acceleration is greater than the positive acceleration threshold, the sampling time is classified into the acceleration segment; if the traction acceleration is less than the negative acceleration threshold, the sampling time is classified into the deceleration segment; if the traction acceleration is between the negative acceleration threshold and the positive acceleration threshold, the sampling time is classified into the constant speed segment.

[0048] The method for determining the positive acceleration threshold and the negative acceleration threshold is as follows: statistically analyze the standard deviation of the traction acceleration during the uniform production stage in the historical traction speed time series data, take three times the positive standard deviation as the positive acceleration threshold, and take three times the negative standard deviation as the negative acceleration threshold. For example, the positive acceleration threshold can be set to 0.05 m / s², and the negative acceleration threshold can be set to -0.05 m / s².

[0049] For each type of speed range, the time-aligned eccentricity data table contains all records whose actual extrusion time falls within that speed range. The eccentricity records within the acceleration phase time range are arranged in ascending order of the actual extrusion time to form an acceleration phase eccentricity sub-queue. The eccentricity records within the constant speed phase time range are arranged in ascending order of the actual extrusion time to form a constant speed phase eccentricity sub-queue. The eccentricity records within the deceleration phase time range are arranged in ascending order of the actual extrusion time to form a deceleration phase eccentricity sub-queue. The three sub-queues are merged to form a speed segment eccentricity state queue.

[0050] S22 refines the time-aligned eccentricity data table into three sub-queues using a speed segmentation mechanism. During the accelerated start-up phase, the cable segment's dwell time at the die head is shortened, resulting in a generally larger and more volatile overall eccentricity in the insulation extrusion. During the constant-speed production phase, the overall eccentricity is smaller and the volatility is lower. If all eccentricity records are mixed and statistically analyzed without distinguishing speed conditions, the large eccentricity in the accelerated phase will raise the mean and variance benchmark of the mixed statistics, causing the calibration gain corresponding to the constant-speed phase to be diluted by the accelerated phase data and deviate from the true eccentricity distribution of the constant-speed condition. The speed segmentation mechanism allows S3 to automatically switch to the eccentricity distribution statistics specific to that speed range during the accelerated start-up or deceleration phases, avoiding weight dilution from cross-speed range mixed statistics at the grouping source, and providing an independent statistical benchmark for S3 to calculate the calibration gain according to the operating condition. See also Figure 4 This is a schematic diagram of the construction of a speed segmented eccentric state queue provided in an embodiment of this application. Figure 4As shown, the time-aligned eccentricity data table is divided according to real-time traction speed, and eccentricity records are independently written to different subordinate categories. Through interval definition, the records are divided into acceleration segment eccentricity sub-queues, constant speed segment eccentricity sub-queues, and deceleration segment eccentricity sub-queues. Finally, the three sub-queues combine to form a larger-scale speed segmented eccentricity status queue. This speed segmentation mechanism can automatically refine the grouping according to speed conditions, and can automatically switch to dedicated eccentricity distribution statistics during acceleration start-up or deceleration phases. This avoids the weight dilution phenomenon caused by cross-speed interval mixed statistics from the source of grouping, and provides a non-interfering statistical benchmark for the system to independently calculate calibration gain.

[0051] S3: Construct an eccentricity oscillation detection ring buffer using a velocity-segmented eccentricity state queue, calculate the eccentricity oscillation index, perform branch calibration gain calculation based on the comparison relationship between the eccentricity oscillation index and the dual thresholds, and generate an adaptive calibration gain.

[0052] When the traction speed changes dynamically, the accumulated timing error in the closed-loop control causes the side with the thinner insulation layer to shift excessively to the other side after calibration, resulting in eccentricity oscillation amplification within a certain length. This step aims to calculate the eccentricity oscillation index in real time through an eccentricity oscillation detection ring buffer, and to classify the eccentricity oscillation index using a dual threshold coupled with the statistical characteristics of the speed segmented sub-queue. Damping correction or lock-in calibration is intervened in advance before the oscillation develops into overcompensation, suppressing oscillation amplification from both oscillation detection and gain adjustment stages.

[0053] Further, step S3 includes: S31: With a fixed capacity K as the length of the circular buffer, write the eccentricity into the eccentricity oscillation detection circular buffer according to the temporal sequence of the segmented eccentricity state queues based on speed. Newly written records overwrite the oldest records. After each write, calculate the mean of the absolute values ​​of the adjacent differences of all K eccentricity records in the eccentricity oscillation detection circular buffer, and record it as the eccentricity oscillation index.

[0054] This step aims to measure the degree of eccentricity oscillation by the differential fluctuation of adjacent eccentricities, and to adaptively narrow the oscillation detection window length as the historical eccentricity variance of the current velocity segment sub-queue increases, thereby improving the oscillation response sensitivity in the high-fluctuation velocity range.

[0055] Specifically, the eccentricity oscillation detection ring buffer is a one-dimensional numerical buffer structure with fixed-length cyclic writing. The buffer length is equal to the fixed capacity K. When the number of written records reaches the fixed capacity K, the newly written eccentricity record overwrites the earliest record in the buffer that actually occurred during the extrusion.

[0056] The formula for calculating the eccentricity oscillation index is: Where OI is the eccentricity oscillation exponent; K is the fixed capacity of the annular buffer for eccentricity oscillation detection; The j-th eccentricity record is arranged in time sequence in the ring buffer for eccentricity oscillation detection. The eccentricity record comes from the sub-queue corresponding to the current velocity segment in the velocity segment eccentricity state queue generated by S22.

[0057] The expression for calculating the fixed capacity K is: in, Base capacity; Minimum capacity; The historical eccentricity variance of the sub-queue corresponding to the current speed segment is calculated from all eccentricity records in the current speed segment sub-queue according to the definition of variance; This is the variance adjustment coefficient. , and The method for determining it is as follows: Find the minimum number of records required for the eccentricity oscillation index to converge stably under uniform production conditions. The minimum number of records required to guarantee that the mean of the absolute values ​​of adjacent differences is statistically significant is selected. Take the fixed capacity K corresponding to the highest historical eccentricity variance to narrow to The coefficient value, for example, It can be set to 32. It can be set to 8. It can be set to 0.5.

[0058] The design principle of the calculation expression for a fixed capacity K is: In the denominator... Item with historical eccentricity variance The fixed capacity K decreases as the historical eccentricity variance increases, reflecting the engineering requirement for a shorter detection window to quickly respond to sudden changes in eccentricity during periods of high fluctuation. Adding 1 to the denominator ensures that the denominator is not zero when the historical eccentricity variance is zero, and that the fixed capacity K does not exceed the baseline capacity. Round down to make the fixed capacity K an integer that can be used for buffer addressing; and the minimum capacity. Taking the maximum value ensures that the detection window is not too short, thus losing the statistical significance of the average absolute value of adjacent differences. The overall design goal of the fixed capacity K calculation expression is to keep the oscillation detection window relatively long in the low-fluctuation uniform speed range to filter out random disturbances, and to automatically narrow it in the high-fluctuation acceleration or deceleration range to improve the sensitivity to the rapid fluctuation of eccentricity.

[0059] For example, when the baseline capacity The variance adjustment coefficient is 32. The historical eccentricity variance of the current speed segmented sub-queue is 0.5. When the value is 2, the fixed capacity K is 32 divided by 1 plus 0.5 multiplied by 2, which equals 32 divided by 2, which equals 16. Rounded down, this is 16. When the production line enters the acceleration phase, the historical eccentricity variance... When the value is increased to 6, the fixed capacity K becomes 32 divided by 1 plus 0.5 multiplied by 6, which equals 32 divided by 4, which equals 8. The oscillation detection window narrows from 16 lines to 8 lines. See also... Figure 5 This is a schematic diagram of writing to the eccentricity oscillation detection ring buffer provided in an embodiment of this application. Figure 5 As shown, eccentricity recording units are extracted from the velocity segmentation eccentricity state queue on the left according to the time sequence and transmitted to the right as input to the eccentricity oscillation detection annular buffer. This fixed-length annular buffer has a fixed capacity attribute and performs a replacement operation of overwriting the oldest record with a new record when the write load is too high. The final output is the eccentricity oscillation index calculation result by calculating the adjacent eccentricities within it. This structural design uses adjacent differential fluctuations to measure the degree of oscillation and allows the oscillation detection window length to adaptively narrow with the historical eccentricity variance; it automatically shortens the long detection band, not only filtering out random disturbances in the low-fluctuation uniform speed range, but also significantly improving the system's sensitivity to rapid fluctuations in eccentricity in the high-fluctuation acceleration or deceleration range.

[0060] S32: Set strong oscillation threshold and weak oscillation threshold. Both strong oscillation threshold and weak oscillation threshold are dynamically updated based on the sliding standard deviation of the eccentricity in the current velocity segment sub-queue. When the sliding standard deviation increases, the two thresholds are adjusted upwards simultaneously, and when the sliding standard deviation decreases, the two thresholds are adjusted downwards simultaneously.

[0061] This step aims to make the oscillation discrimination boundary dynamically float with the level of eccentric fluctuation under the current speed condition, so as to avoid the static threshold from producing a fixed misjudgment boundary under different speed conditions or different production stages.

[0062] Specifically, the sliding standard deviation is calculated as follows: Take the latest sliding window length W eccentricity records from the current speed segment sub-queue within the speed segment eccentricity status queue generated by S22 at the actual time of extrusion. Calculate the standard deviation of these W eccentricity records according to the definition of standard deviation, and use this as the sliding standard deviation. The sliding window length W is determined by taking the number of eccentricity records required to cover a complete eccentricity adjustment cycle. For example, the sliding window length W can be set to 20.

[0063] The weak oscillation threshold is calculated as follows: the weak oscillation threshold equals the weak threshold coefficient multiplied by the moving standard deviation. The strong oscillation threshold is calculated as follows: the strong oscillation threshold equals the strong threshold coefficient multiplied by the moving standard deviation. The weak threshold coefficient is less than the strong threshold coefficient. The weak and strong threshold coefficients are determined as follows: The distribution of the ratio of the eccentricity oscillation index to the moving standard deviation under historical normal production conditions is statistically analyzed. The median of this distribution is taken as the weak threshold coefficient, and the upper quartile of this distribution is taken as the strong threshold coefficient. For example, the weak threshold coefficient can be set to 0.8, and the strong threshold coefficient can be set to 1.5.

[0064] When the sliding standard deviation increases, the weak oscillation threshold and the strong oscillation threshold are adjusted upwards synchronously according to their respective coefficients; when the sliding standard deviation decreases, the weak oscillation threshold and the strong oscillation threshold are adjusted downwards synchronously according to their respective coefficients.

[0065] S33: Based on the relationship between the eccentricity oscillation index and the strong and weak oscillation thresholds, perform branch calculations on the adaptive calibration gain; if the eccentricity oscillation index is less than or equal to the weak oscillation threshold, multiply the average of the M most recent eccentricity records in the current velocity segment sub-queue by the reference gain coefficient to obtain the adaptive calibration gain; if the eccentricity oscillation index is greater than the weak oscillation threshold but less than the strong oscillation threshold, add a damping correction weighted by the eccentricity oscillation index to the reference gain coefficient to obtain the adaptive calibration gain; if the eccentricity oscillation index is greater than or equal to the strong oscillation threshold, set the adaptive calibration gain to the fixed proportional lower limit of the reference gain coefficient, and simultaneously output the oscillation lock flag to S4.

[0066] This step aims to perform differentiated calculations on the calibration gain based on the strength of the eccentric oscillation, calibrating sufficiently in the low oscillation state, damping and suppressing in the medium oscillation state, and locking and limiting in the strong oscillation state, thereby cutting off the positive feedback path of oscillation amplification from the gain output stage.

[0067] Specifically, the method for determining the reference gain coefficient is as follows: The extruder die head position actuator is calibrated with a unit eccentricity correction displacement, and the die head displacement corresponding to the unit eccentricity is taken as the reference gain coefficient. For example, the reference gain coefficient can be set to 0.8. The method for determining the number of most recent records M is as follows: The number of records covering one sampling level of eccentricity fluctuation is taken. For example, the number of most recent records M can be set to 5. The fixed ratio lower limit is equal to the reference gain coefficient multiplied by the fixed ratio lower limit coefficient. The method for determining the fixed ratio lower limit coefficient is as follows: The maximum allowable ratio that prevents the die head single-step correction displacement from causing reverse overshoot is taken. For example, the fixed ratio lower limit coefficient can be set to 0.2.

[0068] The first branch is: when the eccentricity oscillation index is less than or equal to the weak oscillation threshold, take the average of the latest M eccentricity records in the current speed segment sub-queue generated by S22 at the actual time of extrusion, multiply it by the reference gain coefficient, and obtain the adaptive calibration gain.

[0069] The second branch is as follows: When the eccentricity oscillation index is greater than the weak oscillation threshold and less than the strong oscillation threshold, the adaptive calibration gain is calculated using the following formula: Where G is the adaptive calibration gain; The reference gain coefficient is OI; the eccentricity oscillation index is calculated by S31. The weak oscillation threshold obtained from S32; The oscillation strength threshold obtained from S32 is given by the formula. This is the damping correction value, which ranges from 0 to 1. It increases with the increase of the eccentricity oscillation exponent, causing the adaptive calibration gain to decrease as the eccentricity oscillation exponent increases. This reflects the damping physical law that the stronger the eccentricity oscillation, the more the calibration action is suppressed.

[0070] The third branch is: when the eccentricity oscillation index is greater than or equal to the oscillation strength threshold, the adaptive calibration gain is set to the fixed ratio lower limit value obtained by multiplying the reference gain coefficient by the fixed ratio lower limit coefficient, and the oscillation lock flag is set to triggered and output to S4.

[0071] For example, when the reference gain coefficient is 0.8, the weak oscillation threshold is 3, the strong oscillation threshold is 8, and the eccentricity oscillation index is 5.5, the eccentricity oscillation index is between the weak and strong oscillation thresholds. The damping correction is 5.5 minus 3 divided by 8 minus 3 equals 2.5 divided by 5 equals 0.5, and the adaptive calibration gain is 0.8 multiplied by 1 minus 0.5 equals 0.4. When the eccentricity oscillation index rises to 8.5, which is greater than or equal to the strong oscillation threshold of 8, the adaptive calibration gain is set to 0.8 multiplied by 0.2 equals 0.16, and the oscillation lock flag is set to triggered. See also Figure 6 This is a schematic diagram of the branch calibration gain determination calculation provided in an embodiment of this application. Figure 6As shown, the core structure introduces the eccentricity oscillation index from the left input and compares it with the three-layered strong and weak oscillation thresholds. This leads to the system's computation path being divided into three branches: the first, second, and third. Finally, after calculation and aggregation, the three branches output an adaptive calibration gain to the right. When the index exceeds the strong threshold, an additional connection is drawn from the third branch to trigger the oscillation lock flag to the external program. Through this three-branch discrimination logic, the system performs differentiated gain calculations based on the strength of the oscillation: sufficient calibration in low oscillation, damping suppression in medium oscillation, and amplitude locking and limiting in strong oscillation, thereby cutting off the positive feedback path of oscillation amplification from the external gain output stage.

[0072] S31 to S33 construct a parameter coupling loop consisting of the eccentricity oscillation exponent, dual thresholds, and adaptive calibration gain. This addresses the problem that oscillation detection with a fixed window and static threshold lags behind the actual occurrence of oscillations under high-speed dynamic conditions, and amplifies overcompensation in the closed loop. The fixed capacity K of the eccentricity oscillation detection ring buffer is constrained by the historical eccentricity variance of the current velocity segment subqueue. The strong oscillation threshold and weak oscillation threshold are dynamically updated by the sliding standard deviation of the current subqueue. Both constraint parameters originate from the velocity segment eccentricity state queue generated in S22, forming a cross-step parameter inverse dependency. In the three-branch calculation structure, the mean ratio of the first branch, the damping superposition of the second branch, and the fixed lower limit of the third branch correspond to three non-overlapping calculation paths. The discrimination conditions of any two paths are numerically mutually exclusive and cannot be simultaneously true. If there is only variance adaptation with a fixed capacity K but no standard deviation update with a double threshold, the oscillation detection window will narrow with the variance, but the discrimination boundary will remain a fixed value. In the low-fluctuation uniform range, normal small fluctuations will be misjudged as oscillations, resulting in unnecessary suppression calibration. If there is only standard deviation update with a double threshold but no variance adaptation with a fixed capacity K, the discrimination boundary will fluctuate with the fluctuation. However, in the high-fluctuation acceleration range, the detection window will be too long, causing the rapid fluctuation of eccentricity to be smoothed by the mean within the window. The eccentricity oscillation index will rise with a lag behind the actual occurrence of the oscillation, and the damping correction will not be able to intervene before overcompensation occurs. The two constraints act on the detection window length and the discrimination boundary, respectively. Their combined effect enables the eccentricity oscillation index to respond quickly in the high fluctuation range and has a discrimination benchmark that fluctuates with the operating conditions. When the historical eccentricity variance is large during the accelerated start-up phase, causing the fixed capacity K to narrow, the sensitivity of the eccentricity oscillation index to the rapid fluctuation of eccentricity is simultaneously improved. This allows the damping correction to intervene earlier before the oscillation develops into overcompensation. Logically, this sensitivity improvement cannot be attributed to either the narrowing of the fixed capacity K or the dual threshold update mechanism. The absence of either mechanism will cause the eccentricity oscillation index or the discrimination boundary to lose its match with the speed operating conditions, and the overcompensation suppression path will fail.

[0073] S4: The extruder die head position calibration command is generated by driving the adaptive calibration gain and oscillation lock flag, the closed-loop calibration action is executed and the calibration result is written back to the speed segment eccentricity status queue.

[0074] This step focuses on the final issuance and execution of the eccentric calibration command. After the calibration gain and oscillation state are determined in the previous steps, they must be translated into specific displacement actions of the extruder die head. Furthermore, the issuance must be temporarily suspended while the oscillation is locked to avoid the superposition of new calibration disturbances before the oscillation subsides. This step aims to convert the adaptive calibration gain and eccentricity direction angle into die head displacement command quantities. The oscillation lock flag determines whether the calibration command is issued directly or with a delay, and the execution time of the calibration action is written back to the speed segmented eccentricity state queue. The natural convergence of subsequent eccentricity quantities forms a closed loop of inverse constraints on the calibration parameters based on the calibration results.

[0075] Further, step S4 includes: S41: Read the latest record in the time-aligned eccentricity data table for the eccentricity direction angle and eccentricity amount. Use the eccentricity direction angle to determine the mold head adjustment direction, and use the eccentricity amount to multiply by the adaptive calibration gain to calculate the mold head displacement command amount and generate a calibration command.

[0076] This step converts the eccentricity detection results and adaptive calibration gain into displacement actions that can be directly executed by the extruder die head actuator, thus determining the direction and magnitude of the calibration.

[0077] Specifically, the method for determining the die head adjustment direction is as follows: take the latest record of the extruder's actual occurrence time in the time-aligned eccentricity data table generated by S21, and take the direction after adding 180 degrees to the eccentricity direction angle as the die head adjustment direction, so that the die head moves in the opposite direction of the conductor offset direction and pushes the conductor back to the center of the outer circle of the insulation layer.

[0078] The method for calculating the die head displacement command is as follows: take the latest record of the extruder's actual occurrence time in the time-aligned eccentricity data table generated by S21, multiply it by the adaptive calibration gain generated by S33, and obtain the die head displacement command.

[0079] The calibration command is a command structure containing two fields: the mold head adjustment direction and the mold head displacement command amount.

[0080] For example, when the latest record in the time alignment eccentricity data table has an eccentricity of 40 micrometers, an eccentricity direction angle of 90 degrees, and an adaptive calibration gain of 0.4, the mold head adjustment direction is 90 degrees plus 180 degrees, which equals 270 degrees. The mold head displacement command is 40 multiplied by 0.4, which equals 16 micrometers. The calibration command is to move 16 micrometers along the 270-degree direction.

[0081] S42: Check the status of the oscillation lock flag; if the oscillation lock flag is not triggered, the calibration command is directly sent to the extruder die head actuator; if the oscillation lock flag is triggered, a forced waiting interval is inserted before the calibration command is sent. The duration of the forced waiting interval is equal to the single-step transmission delay corresponding to the current moment in the dynamic delay sequence multiplied by the preset delay multiplier. During the waiting period, cross-sectional eccentricity detection data is continuously collected and the eccentricity oscillation detection annular buffer is updated. The calibration command is sent after the oscillation lock flag is re-determined as not triggered by S3.

[0082] This step aims to determine the timing of issuing calibration commands based on the oscillation lock flag, postpone issuing commands during strong oscillations to avoid superimposing new calibration disturbances before the oscillations subside, and make the waiting window adaptively expand and contract with the traction speed.

[0083] Specifically, the state of the oscillation lock flag is output by S33, with values ​​of "not triggered" or "triggered".

[0084] When the oscillation lock flag is not triggered, the calibration command generated by S41 is directly sent to the extruder die head actuator.

[0085] When the oscillation lock flag is triggered, a forced waiting interval is inserted before the calibration command is issued. The forced waiting interval is calculated as follows: the single-step transmission delay corresponding to the current moment in the dynamic delay sequence generated in S12 is multiplied by a preset delay multiplier to obtain the duration of the forced waiting interval. The preset delay multiplier is determined as follows: the transmission delay multiplier required for a complete eccentricity oscillation cycle to be fully observed by the detection end is taken as the required transmission delay multiplier. For example, the preset delay multiplier can be set to 2.5. During the waiting period, the intelligent projector continuously collects cross-sectional eccentricity detection data and writes the eccentricity amount into the eccentricity oscillation detection ring buffer according to S31. The eccentricity oscillation index is recalculated from S31 to S33, and the oscillation lock flag is re-determined. The calibration command is then issued to the extruder die head actuator only after the oscillation lock flag is re-determined as not triggered.

[0086] The duration of the forced waiting interval is coupled with the current single-step transmission delay in the dynamic delay sequence, allowing the waiting window to adaptively expand and contract with changes in traction speed. During the deceleration phase, as the single-step transmission delay increases, the waiting window extends synchronously, enabling the detection end to observe the cable segment eccentricity after a longer spatial transmission time within a longer waiting period. This avoids the problem of prematurely releasing the oscillation lock due to insufficient waiting time at low speeds, and issuing new calibration commands before the oscillations have subsided, as is the case with fixed waiting durations. This effectively cuts off the path to oscillation superposition at the timing of command issuance. See also Figure 7 This is a schematic diagram of forced waiting interval insertion provided in an embodiment of this application. For example... Figure 7As shown, the issued calibration command is intercepted by checking the oscillation lock flag status: if the branch is determined to be untriggered, it is allowed to proceed directly; if the branch is determined to be triggered, the forced waiting interval module is used to suspend it by combining the single-step transmission delay input from the side with the preset delay multiplier. During this suspension waiting period, a cyclic judgment section is introduced, and another round of re-judgment is triggered by the dotted feedback loop that continuously collects cross-sectional eccentricity detection data until the lock is released. This logical combination method avoids the operation of adding new calibration commands when the strong oscillation state has not subsided. At the same time, the coupling of the single-step transmission delay characteristic allows the waiting window to expand and contract adaptively, solving the problems of premature release and misoperation under low-speed conditions, and cutting off the generation path of oscillation superposition at the issuance timing stage.

[0087] S43: After the extruder die head actuator completes the die head displacement, the actual execution completion time is used as the timestamp. The execution time is recorded as the calibration action timestamp at the end of the corresponding speed interval sub-queue of the speed segment eccentricity state queue. At the same time, S11 continues to collect subsequent cross-sectional eccentricity detection data and writes it into the corresponding sub-queue via S21 to S22. After the calibration action takes effect, the subsequent changes in eccentricity will naturally be reflected in the sub-queue, which can be read by S3 when updating the sliding standard deviation and dual thresholds in the next sampling cycle, forming a reverse constraint closed loop of calibration results on calibration parameters.

[0088] Write back the execution time of the calibration action to the velocity segment eccentricity state queue so that the calibration history node corresponds to the subsequent eccentricity sequence in time. Drive the dynamic update of the threshold of S3 through the natural convergence of the subsequent eccentricity, forming a closed loop of reverse constraint on the calibration parameters by the calibration execution result.

[0089] Specifically, after the extruder die head actuator completes the die head displacement according to the calibration command issued in S42, the actual completion time of the die head displacement is used as a timestamp. This execution time is then marked at the end of the corresponding speed interval sub-queue in the speed segment eccentricity state queue generated in S22, according to the speed interval to which the execution time belongs. This serves as the calibration action timestamp, used by S32 to identify the calibration action occurrence node when calculating the sliding standard deviation. Subsequently, the cross-sectional eccentricity detection data continuously collected by S11 is written into the corresponding sub-queue after time backtracking mapping in S21. The reduction in eccentricity after the calibration action takes effect will naturally be reflected in the subsequent records of the sub-queue. The method for determining the speed interval to which the execution time belongs is consistent with the speed interval division method in S22, using the comparison between the traction acceleration at the sampling time corresponding to the execution time and the positive and negative acceleration thresholds to determine the speed interval.

[0090] S43 writes back the execution time of the calibration action to the speed segment eccentricity state queue, addressing the hidden defect of the static threshold being sluggish in sensing small oscillations after calibration convergence—a defect not previously linked to the write-back mechanism. The design goal of S43 is to ensure that calibration history nodes correspond temporally to subsequent eccentricity sequences, forming a complete causal chain: after multiple consecutive small calibration actions continuously act on the production line during a long period of uniform speed, the eccentricity subsequently collected and written to the sub-queue by the detection end naturally decreases as the system calibration converges. The fluctuation amplitude of the pure eccentricity sequence in the sub-queue decreases accordingly, and the sliding standard deviation calculated by S32 based on the sub-queue eccentricity narrows accordingly. The strong oscillation threshold and the weak oscillation threshold are simultaneously lowered along with the sliding standard deviation. After the dual thresholds are lowered, the micro-amplitude eccentricity oscillation index, which was originally below the static weak threshold and unable to trigger damping correction, can now reach the second branch above the weak oscillation threshold, allowing S33's damping correction to intervene earlier before the micro-oscillations caused by a new round of speed changes develop into overcompensation. This automatic enhancement of the sensitivity to detect micro-amplitude oscillations stems from the mathematical structure where the sliding standard deviation naturally narrows as the calibration history converges. This occurs spontaneously without the need for an independent sensitivity adjustment module or additional threshold parameters, overcoming the inherent limitation of static thresholds, which, after calibration convergence, are slow to detect small oscillations due to fixed discrimination boundaries, and are only detected after the oscillations have amplified. See also Figure 8 This is a schematic diagram of the calibration action timestamp write-back mechanism provided in the embodiments of this application. Figure 8 As shown, the actual completion time of the extruder die displacement driven by the calibration command at the top is extracted and converted into a calibration action timestamp. This timestamp is written down through a dashed path and into the bottom wrapping layer, i.e., the velocity segment eccentricity state queue, becoming the last record of the corresponding velocity interval sub-queue within that region. Subsequently, a macroscopic reverse constraint closed-loop link is formed by the reverse dashed path that encircles the entire graph. The above-mentioned write-back mechanism associates the calibration history nodes with subsequent continuous sequences. The natural convergence of the subsequent eccentricity in the sub-queue can drive the synchronous reduction of the sliding standard deviation and dual thresholds. Without introducing additional parameters, it overcomes the inherent defect of static thresholds being insensitive to small oscillations and only detecting them after the oscillations have been amplified.

[0091] Example 2: This embodiment, based on Embodiment 1, provides a dynamic adaptive calibration system for cable eccentricity, such as... Figure 9 As shown, it includes: The dynamic transmission delay calculation module is used to collect traction speed time-series data and cross-sectional eccentricity detection data from the extrusion production line, calculate the dynamic transmission delay between the detection end and the extrusion end according to the traction speed time-series data, and generate a dynamic delay sequence. The data table and queue construction module is used to perform dynamic timestamp alignment on the cross-sectional eccentricity detection data using the dynamic delay sequence to generate a time-aligned eccentricity data table; and to construct a speed segmented eccentricity state queue based on the time-aligned eccentricity data table and the traction speed time series data. An adaptive calibration gain generation module is used to construct an eccentricity oscillation detection ring buffer using the segmented eccentricity state queue at the speed, calculate the eccentricity oscillation index, perform branch calibration gain calculation based on the comparison relationship between the eccentricity oscillation index and the dual thresholds, and generate an adaptive calibration gain. The closed-loop calibration execution module is used to drive the generation of extruder die head position calibration instructions with the adaptive calibration gain and oscillation lock flag, execute closed-loop calibration actions, and write back the calibration results to the speed segment eccentricity state queue.

Claims

1. A dynamic adaptive calibration method for cable eccentricity, characterized in that, include: Traction speed time-series data and cross-sectional eccentricity detection data are collected from the extrusion production line. The dynamic transmission delay between the detection end and the extrusion end is calculated based on the traction speed time-series data to generate a dynamic delay sequence. The cross-sectional eccentricity detection data is dynamically timestamped using the dynamic delay sequence to generate a time-aligned eccentricity data table; a speed segmented eccentricity state queue is constructed based on the time-aligned eccentricity data table and the traction speed time series data. An eccentricity oscillation detection ring buffer is constructed using the speed segmented eccentricity state queue, the eccentricity oscillation index is calculated, and branch calibration gain calculation is performed based on the comparison relationship between the eccentricity oscillation index and the dual thresholds to generate an adaptive calibration gain. The adaptive calibration gain and oscillation lock flag are used to drive the generation of extruder die position calibration instructions, execute closed-loop calibration actions, and write back the calibration results to the speed segment eccentricity state queue.

2. The dynamic adaptive calibration method for cable eccentricity according to claim 1, characterized in that, The traction speed timing data and cross-sectional eccentricity detection data collected from the extrusion production line include: A speed sensor is deployed at the extruder traction mechanism, and a smart projector is deployed at the detection end. The traction speed time series data is collected synchronously at a fixed sampling period, and the cross-sectional eccentricity detection data is collected using the time of cross-sectional image acquisition by the smart projector as the timestamp. The real-time traction speed at the sampling moment is obtained by the pulse signal output by the speed sensor within a fixed sampling period; The intelligent projector images the cable cross-section and uses image processing to identify the center of the outer circle of the insulation layer and the center of the conductor outline to extract the eccentricity direction angle and eccentricity as the cross-section eccentricity detection data.

3. The dynamic adaptive calibration method for cable eccentricity according to claim 1, characterized in that, The step of calculating the dynamic transmission delay between the detection end and the extrusion end based on the traction speed time-series data, and generating a dynamic delay sequence, includes: Based on the fixed physical distance between the detection end and the extrusion end, the single-step transmission delay corresponding to each sampling moment in the traction speed time series data is calculated using the real-time traction speed at that moment. The single-step transmission delay is equal to the fixed physical distance divided by the real-time traction speed. The single-step transmission delay at each sampling time is arranged in the sampling time sequence to generate the dynamic delay sequence.

4. The dynamic adaptive calibration method for cable eccentricity according to claim 1, characterized in that, The step of performing dynamic timestamp alignment on the cross-sectional eccentricity detection data using the dynamic delay sequence to generate a time-aligned eccentricity data table includes: The actual time of extrusion corresponding to the eccentricity detection data is obtained by subtracting the single-step transmission delay of the corresponding moment in the dynamic delay sequence from the acquisition timestamp of each record in the cross-sectional eccentricity detection data. Each record in the cross-sectional eccentricity detection data, along with the actual time of occurrence of the extrusion, the eccentricity direction angle, and the eccentricity amount, is written into the time-aligned eccentricity data table.

5. The dynamic adaptive calibration method for cable eccentricity according to claim 1, characterized in that, The step of constructing a speed segmented eccentricity state queue based on the time-aligned eccentricity data table and the traction speed time-series data includes: The production process is divided into three speed ranges: acceleration, constant speed, and deceleration, based on the real-time traction speed in the traction speed time series data. Each speed range corresponds to all records in the time alignment eccentricity data table that fall within the time range of that range. The eccentric record sets corresponding to the three types of speed ranges are arranged in chronological order to form eccentric sub-queues for acceleration, uniform speed, and deceleration, respectively. The three are then merged into the speed segment eccentric state queue.

6. The dynamic adaptive calibration method for cable eccentricity according to claim 5, characterized in that, The process is divided into three speed ranges—acceleration, constant speed, and deceleration—based on the real-time traction speed in the traction speed time series data. The traction acceleration at that sampling moment is obtained by dividing the difference between the real-time traction speeds at two adjacent sampling times in the traction speed time series data by a fixed sampling period. If the traction acceleration is greater than the positive acceleration threshold, then the sampling moment is classified as the acceleration phase; If the traction acceleration is less than the negative acceleration threshold, then the sampling moment is classified as the deceleration phase; If the traction acceleration is between the negative acceleration threshold and the positive acceleration threshold, then the sampling moment is classified as a uniform speed segment.

7. The dynamic adaptive calibration method for cable eccentricity according to claim 1, characterized in that, The construction of an eccentricity oscillation detection circular buffer using the segmented eccentricity state queue at the aforementioned speed, and the calculation of the eccentricity oscillation index, include: With a fixed capacity as the length of the circular buffer, the eccentricity is written into the eccentricity oscillation detection circular buffer according to the time sequence of the segmented eccentricity state queue at the speed, and the newly written record overwrites the oldest record. After each write operation, the mean absolute value of the adjacent differences of all fixed-capacity eccentricity records in the eccentricity oscillation detection ring buffer is calculated and denoted as the eccentricity oscillation index.

8. The dynamic adaptive calibration method for cable eccentricity according to claim 1, characterized in that, The dual thresholds include: Strong oscillation threshold and weak oscillation threshold; Both the strong oscillation threshold and the weak oscillation threshold are dynamically updated based on the sliding standard deviation of the eccentricity in the current velocity segment sub-queue. When the sliding standard deviation increases, the two thresholds are adjusted upwards simultaneously, and when the sliding standard deviation decreases, the two thresholds are adjusted downwards simultaneously. The sliding standard deviation is calculated by taking the standard deviation of the latest sliding window length bar eccentricity record in the current speed segment sub-queue of the speed segment eccentricity state queue at the actual time of extrusion.

9. The dynamic adaptive calibration method for cable eccentricity according to claim 1, characterized in that, The calculation of the execution branch calibration gain includes: If the eccentricity oscillation index is less than or equal to the weak oscillation threshold, the mean of the most recent eccentricity records in the current velocity segment sub-queue is multiplied by the reference gain coefficient to obtain the adaptive calibration gain. If the eccentricity oscillation index is greater than the weak oscillation threshold and less than the strong oscillation threshold, then a damping correction amount with the eccentricity oscillation index as the weight is superimposed on the reference gain coefficient to obtain the adaptive calibration gain. If the eccentricity oscillation index is greater than or equal to the oscillation strength threshold, the adaptive calibration gain is set to the fixed proportional lower limit of the reference gain coefficient, and the oscillation lock flag is output to the triggered state.

10. A dynamic adaptive calibration system for cable eccentricity, characterized in that, include: The dynamic transmission delay calculation module is used to collect traction speed time-series data and cross-sectional eccentricity detection data from the extrusion production line, calculate the dynamic transmission delay between the detection end and the extrusion end according to the traction speed time-series data, and generate a dynamic delay sequence. The data table and queue construction module is used to perform dynamic timestamp alignment on the cross-sectional eccentricity detection data using the dynamic delay sequence to generate a time-aligned eccentricity data table; and to construct a speed segmented eccentricity state queue based on the time-aligned eccentricity data table and the traction speed time series data. An adaptive calibration gain generation module is used to construct an eccentricity oscillation detection ring buffer using the segmented eccentricity state queue at the speed, calculate the eccentricity oscillation index, perform branch calibration gain calculation based on the comparison relationship between the eccentricity oscillation index and the dual thresholds, and generate an adaptive calibration gain. The closed-loop calibration execution module is used to drive the generation of extruder die head position calibration instructions with the adaptive calibration gain and oscillation lock flag, execute closed-loop calibration actions, and write back the calibration results to the speed segment eccentricity state queue.

Images