Full-automatic optical fiber winding tension intelligent control system

Abstract

The application discloses a kind of full-automatic optical fiber winding tension intelligent regulation system, the system includes: state acquisition module, for collecting pay-off end speed, take-up end speed, tension detection value, tension roller displacement, wire arrangement mechanism position, servo motor current, volume change and ambient temperature in the process of optical fiber winding, form current control period state data;Precursor characterization module is used to extract angular velocity difference, phase shift, tension roller microvibration, servo current pulsation and wire arrangement commutation impact according to the current control period state data, generates microslip precursor index, the present application relates to optical fiber winding technical field.This kind of full-automatic optical fiber winding tension intelligent regulation system, tension risk trend value and target tension correction amount are simultaneously applied to pay-off brake, take-up torque, tension roller position and translation wire arrangement commutation action, avoid the regulation lag and control incoordination problem caused by single actuator compensation.

Description

Technical Field

[0001] This invention relates to the field of optical fiber winding, and more specifically, to a fully automatic intelligent tension control system for optical fiber winding. Background Technology

[0002] In optical fiber production, winding is a crucial process. During winding, tension control remains a key factor affecting fiber quality and production efficiency due to variations in material properties and process conditions. Traditional fiber winding tension control relies heavily on mechanical means or conventional sensors for monitoring and adjustment. However, this method often cannot respond in real-time to complex process changes, leading to large tension fluctuations that impact fiber quality and production stability. Furthermore, fiber winding is frequently affected by various factors, such as changes in winding diameter, ambient temperature fluctuations, and servo motor current variations, further increasing the difficulty of tension control. Existing tension control methods often fail to adequately consider the influence of these dynamic factors and lack effective predictive and adaptive adjustment mechanisms.

[0003] Therefore, an intelligent, fully automated fiber optic winding tension control system is needed. This system can collect multiple process parameters in real time, adjust the tension in real time through multi-dimensional data analysis and prediction, and precisely control tension fluctuations, thereby improving the quality stability and production efficiency of the fiber optic winding process. Summary of the Invention

[0004] The purpose of this invention is to provide a fully automatic intelligent tension control system for optical fiber winding. This system solves the problem that existing tension control methods often rely on mechanical means or conventional sensors for monitoring and adjustment, which often cannot respond to complex process changes in real time, resulting in large tension fluctuations that affect the quality and production stability of optical fibers. Furthermore, the optical fiber winding process is often affected by various factors, such as changes in winding diameter, fluctuations in ambient temperature, and changes in the current of servo motors, which increases the difficulty of tension control and fails to meet the application requirements.

[0005] This invention achieves the above objective through the following technical solution: a fully automatic intelligent control system for optical fiber winding tension, the system comprising: Status acquisition module, early warning characterization module, layer memory module, coupled decision-making module, linkage control module, and feedback adaptive module; The status acquisition module is used to acquire the speed of the pay-off end, the speed of the take-up end, the tension detection value, the displacement of the tension roller, the position of the wire laying mechanism, the servo motor current, the change in winding diameter, and the ambient temperature during the optical fiber winding process, so as to form the status data of the current control cycle. The precursor characterization module is used to extract angular velocity difference, phase offset, tension roller micro-vibration, servo current pulsation and cable reversal impact based on the current control cycle state data, and generate micro-slip precursor index; The layer memory module is used to construct a layer residual stress memory matrix for the completed layers; The coupled decision module is used to perform coupled analysis on the microslip precursor index and the layer residual stress memory matrix, and output the tension risk trend value and the target tension correction amount; The linkage control module is used to generate wire release brake correction, wire take-up torque correction, tension roller balance position correction and wire reversal advance based on the tension risk trend value and the target tension correction amount, so as to implement cross-layer predictive collaborative control of the optical fiber winding process. The feedback adaptive module is used to update the micro-slippage precursor threshold, the layer memory weight, and the target tension correction coefficient based on the tension response and layer forming results after linkage compensation.

[0006] Furthermore, the status acquisition module includes: Multi-source sampling unit, time-stamp alignment unit, and preprocessing unit; The multi-source sampling unit is used to acquire data from each sensing channel according to a unified control cycle; The time-scale alignment unit is used to map rotation speed signals, tension signals, displacement signals, current signals and temperature signals at different sampling frequencies to the same time axis; The preprocessing unit is used to remove missing sampling points and abnormal jump points, and outputs a standardized state data sequence corresponding to the control cycle.

[0007] Furthermore, the precursor characterization module includes: Differential analysis unit, frequency domain filtering unit, impulse identification unit, and fusion calculation unit; The differential analysis unit is used to generate the angular velocity difference characteristics and phase offset characteristics between the pay-off end and the take-up end; The frequency domain filtering unit is used to separate the micro-vibration characteristics of the tension roller and the servo current pulsation characteristics. The impact recognition unit is used to identify the impact characteristics at the moment of cable reversal; The fusion calculation unit is used to perform weighted fusion of each feature according to preset weights to generate a micro-slip precursor index that reflects the degree of early anomalies in tension instability.

[0008] Furthermore, the layered memory module includes: Layer segmentation unit, parameter extraction unit, matrix construction unit, and historical storage unit; The layer segmentation unit is used as a memory unit for the complete winding process of a single layer. The parameter extraction unit is used to extract the average tension, tension fluctuation amplitude, wire distribution status, roll diameter increment, curvature change parameters, and local overburden status parameters of the ring layer. The matrix construction unit is used to form a concentric residual stress memory matrix in a fixed dimensional order; The historical storage unit is used to write the memory matrix of the residual stress of each layer into the historical memory bank in the order of winding.

[0009] Furthermore, the layered memory module also includes: Weight allocation unit and calling unit; The weight allocation unit is used to assign corresponding layer memory weights to the residual stress memory matrix of each layer in the historical memory bank according to the layer winding sequence, the winding diameter growth state and the stress accumulation degree. The calling unit is used to call the target layer residual stress memory matrix that matches the current roll diameter range and wiring state within the current control cycle, so as to improve the characterization accuracy of interlayer inheritance effect.

[0010] Furthermore, the coupled decision module includes: Coupled operation unit, normalization processing unit and correction generation unit; The coupling operation unit is used to perform correlation operations between the microslip precursor index and the layer residual stress memory matrix; The normalization processing unit is used to generate tension risk trend values ​​by combining matrix norm and layer memory weights; The correction generation unit is used to generate a target tension correction amount based on the tension risk trend value, the current tension deviation, and the cumulative deviation of the control window, so as to achieve joint decision-making of trend prediction and error compensation.

[0011] Furthermore, the coupled decision module also includes: Tension prediction model unit; The tension prediction model unit is used to train a diffusion autoregressive transformer model based on historical winding multi-source time series data, which includes normal winding state, tension fluctuation state, micro-slip state and abnormal winding state. The tension prediction model unit is used to output the tension prediction sequence for the future control window, and to generate the target tension correction amount by fusing the tension prediction sequence with the tension risk trend value.

[0012] Furthermore, the linkage control module includes: Braking distribution unit, torque distribution unit, position distribution unit, and commutation advance unit; The braking distribution unit is used to generate a line-laying braking correction amount based on the target tension correction amount; The torque distribution unit is used to generate a take-up torque correction amount based on the target tension correction amount; The position allocation unit is used to generate a tension roller balance position correction amount based on the target tension correction amount; The reversal advance unit is used to generate a reversal advance amount based on the tension risk trend value and the real-time linear speed of the optical fiber, so as to achieve synchronous compensation of the wire feeding mechanism, the wire taking-up mechanism, the tension roller adjustment mechanism and the wire laying mechanism.

[0013] Furthermore, the feedback adaptive module includes: Response acquisition unit, deviation evaluation unit, and parameter update unit; The response acquisition unit is used to acquire actual tension response, layer formation state and subsequent fluctuation data within a fixed evaluation window; The deviation assessment unit is used to calculate the compensation deviation result based on the actual tension value and the target tension value; The parameter update unit is used to perform incremental updates on the micro-slip precursor threshold, the layer memory weight, and the target tension correction coefficient based on the compensation deviation result, so as to keep the control parameters of subsequent control cycles and subsequent layers adaptively matched.

[0014] Furthermore, the tension prediction model unit also includes an online incremental training subunit; The online incremental training subunit is used to write the winding data into the incremental training set after each optical fiber winding is completed, freeze the backbone parameters of the diffusion autoregressive transformer model, fine-tune only the top fully connected layer and attention weights, and perform a limited number of iterations with a preset small learning rate to maintain the continuous stability of the future control window tension prediction accuracy under different working conditions.

[0015] The beneficial effects of this invention are as follows: 1. By jointly characterizing angular velocity difference, phase shift, tension roller micro-vibration, servo current pulsation, and cable reversal impact, a micro-slippage precursor index is generated, which can identify instability trends before tension anomalies become apparent and improve anomaly prediction capabilities.

[0016] 2. By constructing a residual stress memory matrix for each layer and establishing a historical memory library, this invention can incorporate the average tension, tension fluctuation amplitude, roll diameter increment, curvature change, and local overburden state in the completed layers into the current control decision. This overcomes the problem in the prior art that the control is based solely on the current cycle data and ignores the influence of interlayer stress accumulation, thereby improving the control targeting in multi-layer continuous winding scenarios.

[0017] 3. The tension risk trend value and the target tension correction amount are applied simultaneously to the wire release brake, wire take-up torque, tension roller position and translation wire reversing action to achieve synchronous linkage control of multiple mechanisms. This avoids the adjustment lag and control incoordination problems caused by compensation of a single actuator, and helps to suppress the forming deviation caused by tension peak, tension drop and reversing disturbance.

[0018] 4. By incrementally updating the micro-slip precursor threshold, layer memory weight, and target tension correction coefficient based on the actual tension response and layer forming results after linkage compensation, this invention enables the system parameters to be adjusted dynamically according to changes in fiber type, winding diameter, and working condition drift, thereby enhancing the adaptability and stability during long-term continuous winding.

[0019] 5. By introducing a tension prediction model and combining it with tension risk trend values ​​for joint decision-making, this invention can more accurately predict the tension change trend within the future control window and convert the prediction results into the linkage compensation amount of the specific actuator, thereby helping to reduce the tension fluctuation amplitude, improve the coil forming state, reduce interlayer slippage and local overlay risks, and improve the fiber winding quality and overall process consistency. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating the overall system architecture of the present invention; Figure 2 This is a flowchart of the status acquisition module of the present invention; Figure 3 This is a flowchart of the precursor characterization module calculation of the present invention; Figure 4 This is a flowchart of the coupled decision-making and linkage control of the present invention. Detailed Implementation

[0021] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0022] Example 1: Please see Figure 1-4 This invention provides a technical solution: a fully automatic intelligent control system for optical fiber winding tension, the system comprising: Status acquisition module, early warning characterization module, layer memory module, coupled decision-making module, linkage control module, and feedback adaptive module; The status acquisition module is used to collect real-time data on the speed of the pay-off end, the speed of the take-up end, the tension detection value, the displacement of the tension roller, the position of the wire laying mechanism, the current of the servo motor, the change in winding diameter, and the ambient temperature parameters during the optical fiber winding process, forming a status dataset for the current control cycle. Among these parameters, the pay-off rotation speed, measured in revolutions per minute (rpm), reflects the speed at which the fiber is released; the take-up rotation speed, also measured in rpm, determines the speed at which the fiber is wound; the tension detection value, measured in real-time by a tension sensor, indicates the magnitude of the tension experienced by the fiber during winding, typically measured in Newtons (N); the tension roller displacement, the amount of positional movement of the tension roller during tension adjustment, reflects the tension adjustment process; and the position of the fiber winding mechanism determines the spatial arrangement of the fiber on the reel. Servo motor current: The amount of current consumed by the servo motor driving the take-up end and other equipment during operation. Changes in current can reflect the load and operating status of the motor. Coil diameter change: The change in the diameter of the fiber coil on the reel during fiber winding. As the fiber is continuously wound, the coil diameter gradually increases. Ambient temperature parameter: The temperature of the environment at the fiber winding site. Temperature changes may affect the physical properties of the fiber and the performance of the winding equipment. Current control cycle status dataset: Within a specific control cycle, a dataset is formed by combining all the parameters collected above, such as the pay-off end speed and the take-up end speed, for subsequent analysis and processing. The precursor characterization module is used to perform time-series analysis on the state dataset, extract the angular velocity difference between the pay-off end and the take-up end, phase offset, tension roller micro-vibration characteristics, servo current pulsation characteristics, and impact characteristics at the moment of wire reversal, and generate the corresponding micro-slip precursor index. Among these, time-series analysis involves analyzing state datasets arranged chronologically to study the patterns and trends of data changes over time in order to extract useful feature information; angular velocity difference refers to the difference in angular velocity between the pay-off and take-up ends, as angular velocity is the angle rotated per unit time, and the angular velocity difference reflects the difference in rotational speed at both ends; phase offset refers to the difference in phase between the pay-off and take-up ends during rotation, as phase is a physical quantity describing the state of periodic motion, and the phase offset affects the synchronization of fiber winding; and tension roller micro-vibration characteristics refer to the characteristics of the minute vibrations generated by the tension roller during tension adjustment, such as the frequency and amplitude of the vibrations, which may reflect the tension... Unstable factors during the adjustment process; servo current pulsation characteristics, the characteristics of servo motor current during fluctuations, such as the frequency and amplitude of pulsation, which may be related to factors such as motor operating status and load changes; impact characteristics at the moment of cable reversal, the characteristics of the impact phenomenon generated by the cable winding mechanism at the moment of changing the cable winding direction, such as the magnitude and duration of the impact force, which may affect the fiber arrangement quality and winding stability; microslip precursor index, an index obtained by comprehensively analyzing and calculating the various characteristics extracted above, used to characterize the degree of precursor of microslip, a phenomenon that may occur during fiber winding and affects the winding quality; The layer memory module is used to extract the average tension, tension fluctuation amplitude, wire distribution status, roll diameter increment, curvature change parameters and local overburden state parameters of each layer after the winding is completed, construct the layer residual stress memory matrix and store it in the historical memory library. Among these, the components are: a coil, which is the structure formed by winding the optical fiber one turn on the reel, with each coil having specific winding parameters and states; average tension, which is the average of the tension measured at all times during the winding process of a coil, reflecting the overall tension level of that coil; tension fluctuation amplitude, which is the difference between the maximum and minimum tension measured during the winding process of a coil, reflecting the degree of tension fluctuation within that coil; cable distribution, which is the distribution of the optical fiber within a coil, including its tightness and uniformity, and the cable distribution affects the quality and performance of the optical fiber roll; and coil diameter increment, which is the relative increase in the diameter of the reel after completing one coil of winding. The parameters include: the amount added before winding; curvature variation parameters, which describe the change in the degree of fiber bending within a layer, and curvature variation may affect the stress and performance of the fiber; local cladding state parameters, which reflect the local cladding of the fiber within a layer, and local cladding may cause fiber damage or affect winding quality; layer residual stress memory matrix, which is a matrix constructed according to certain rules using the above parameters such as the average tension and tension fluctuation amplitude of a layer, used to record the residual stress related information of the layer; and a history memory bank, which is a database used to store the layer residual stress memory matrices of multiple layers, providing historical data support for subsequent coupling operations. The coupling decision module is used to couple the micro-slip precursor index of the current control cycle with the layer residual stress memory matrix in the historical memory bank to obtain the tension risk trend value and target tension correction amount within the future control window. The coupling operation involves performing a mathematical operation between the micro-slip precursor index of the current control cycle and the layer residual stress memory matrix in the historical memory bank to comprehensively consider the current precursor information and the historical residual stress information. The tension risk trend value within the future control window, obtained through the coupling operation, is used to predict the tension risk during fiber winding within a certain control window in the future, such as the degree of tension being too high or too low. The target tension correction amount, calculated based on the tension risk trend value, is the amount that needs to be adjusted to the currently set target tension to avoid tension risks. The linkage control module is used to generate corresponding wire release brake correction, take-up torque correction, tension roller balance position correction, and wire arrangement reversal advance based on the tension risk trend value and target tension correction amount; and sends the control commands to the wire release brake mechanism, take-up servo drive mechanism, tension roller adjustment mechanism, and wire arrangement mechanism to perform cross-layer predictive linkage control of the optical fiber winding process in order to suppress tension peaks, sudden drops, and interlayer cumulative errors in advance. Among them, the pay-off braking correction amount, generated based on the target tension correction amount, is used to adjust the braking force of the pay-off end braking mechanism to control the pay-off speed, thereby regulating the tension; the take-up torque correction amount, generated based on the target tension correction amount, is used to adjust the torque of the take-up end servo drive mechanism to control the take-up speed and tension; the tension roller balance position correction amount, generated based on the target tension correction amount, is used to adjust the balance position of the tension roller to change the tension regulation effect of the tension roller on the optical fiber; and the cable reversal advance amount, generated based on the target tension correction amount, is used to change the cable direction in advance. The amount of reversing time of the fiber winding mechanism is adjusted to optimize the winding process and reduce the impact on tension. Control commands, such as the correction amount of the wire release brake and the correction amount of the wire take-up torque, are sent to the corresponding wire release brake mechanism, take-up servo drive mechanism and other equipment to control the fiber winding process. Cross-layer predictive linkage control comprehensively considers the information of the current control cycle and the information of the historical layers to predict the fiber winding process of multiple control cycles and layers in the future, and adjusts the tension in advance through linkage control of various related mechanisms to suppress tension peaks, sudden drops and inter-layer cumulative errors. The feedback adaptive module is used to collect the actual tension response, the formation state of the ring, and subsequent fluctuation data after linkage compensation, evaluate the current control effect, and obtain the compensation deviation result. Based on the compensation deviation result, the micro-slip precursor threshold, the ring memory weight, and the target tension correction coefficient are adaptively updated to form a continuous optimization control mechanism for subsequent control cycles and subsequent rings. Among them, the actual tension response, after linkage compensation, is the actual tension change of the optical fiber, which is collected again by the tension sensor; the coil forming state, after linkage compensation, is the shape and arrangement of the current coiled optical fiber after winding; the subsequent fluctuation data, after linkage compensation, is the tension fluctuation data over a period of time; the compensation deviation result is the result of comparing and analyzing the actual tension response, coil forming state, and subsequent fluctuation data with the expected target to obtain the deviation between the control effect and the expectation; and the microslip precursor threshold is a critical value used to determine whether microslip precursors have appeared. When the microslip precursor index exceeds this threshold... When the value is set, micro-slippage is considered to be possible; the layer memory weight, in the coupling operation, is the weight value assigned to the residual stress memory matrix of different layers in the historical memory bank, reflecting the degree of influence of historical information of different layers on the current decision; the target tension correction coefficient is the coefficient used to calculate the target tension correction amount, and it is adaptively updated according to the compensation deviation result to optimize the correction effect of the target tension; the continuous optimization control mechanism is a mechanism that enables the system to continuously optimize the control performance in subsequent control cycles and subsequent layers by continuously collecting actual data, evaluating the control effect, and updating relevant parameters through the feedback adaptive module.

[0023] It should be noted that during use, the status acquisition module comprehensively acquires various parameters in real time, providing a rich data foundation for precise control. The precursor characterization module extracts key features through time series analysis to generate a precursor index, which can detect potential problems in advance. The layer memory module constructs and stores a residual stress memory matrix, accumulating historical experience. The coupling decision module combines current precursors with historical memory calculations to scientifically predict tension risks and determine correction amounts. The linkage control module generates multi-dimensional correction amounts based on the decision results and links and controls various mechanisms to achieve cross-layer predictive control, effectively suppressing tension anomalies and inter-layer errors. The feedback adaptive module collects data to evaluate the effect, adaptively updates parameters, and forms a continuous optimization mechanism, enabling the system to continuously improve as it runs, improving the quality and stability of fiber winding, reducing scrap rate, and increasing production efficiency and economic benefits.

[0024] In one embodiment, the calculation process of the microslip precursor index is as follows: The state dataset is subjected to time-series difference and frequency domain filtering to remove interference components caused by ambient temperature and mechanical noise. The angular velocity difference characteristic values ​​between the pay-off end and the take-up end are calculated separately. Phase offset characteristic value between speed signals High-frequency micro-vibration amplitude of tension roller Servo motor current ripple coefficient Impact coefficient during cable reversal ; Microslip precursor index A multi-feature weighted fusion calculation is used to comprehensively reflect the degree of early anomaly in tension instability:

[0025] in, To meet the preset weighting coefficients based on the winding process and fiber type. ; Basis for determining weights: , Determined based on the sensitivity of angular velocity difference and phase difference to microslip; Determined based on the strong correlation between tension roller vibration and tension fluctuation; Determined based on the linear relationship between servo current and torque output; The weights are determined based on the degree to which commutation shock induces interlayer slip; the weights are calibrated by the variance contribution rate and Pearson correlation coefficient of multiple sets of winding fault samples.

[0026] This design first processes the state dataset to remove interference, then calculates various feature values ​​separately, and finally fuses the weighted features to obtain the microslip precursor index. By processing the data to remove interference, the accuracy of feature extraction is ensured. The fusion of multiple features can comprehensively reflect early anomalies of tension instability, considering the possibility of microslip from different aspects, such as angular velocity difference and phase difference reflecting the synchronicity of the two ends, and tension roller vibration related to tension fluctuations. The preset weights are determined based on multiple factors, making the calculation more scientific and reasonable, accurately assessing the risk of microslip, providing a reliable basis for subsequent control, preventing tension anomalies in advance, and improving the quality of optical fiber winding.

[0027] In one embodiment, the layer residual stress memory matrix The construction process is as follows: Using a single winding layer as a memory unit, the key process and state parameters of the entire winding process of that layer are extracted and arranged dimensionally to form a one-dimensional memory matrix, which is used to record the residual stress distribution and forming quality information of the winding layer. in, For the average tension throughout the entire circle, This represents the amplitude of tension fluctuation. The parameter represents the uniformity of the cable distribution. This is the increase in the diameter after winding. The parameter represents the variation in the curvature of the concentric circles. These are parameters representing the localized interlayer cladding state of optical fibers. Matrix parameter basis: Each parameter is obtained from statistical analysis of sampled data within a complete winding cycle of a single turn. , These are time-domain statistical values. , Quantified values ​​for spatial distribution , These are the eigenvalues ​​of geometric deformation.

[0028] This design, using a single coil as a unit, extracts key parameters and forms a matrix according to dimensions. With each coil as a memory unit, it can accurately record information for each coil, facilitating the analysis of different coil conditions. The extracted key parameters cover multiple aspects, from average tension and fluctuation amplitude to wire distribution and geometric deformation, comprehensively recording the residual stress distribution and forming quality of the coil. The statistical methods for each parameter are reasonable, combining time-domain, spatial, and geometric characteristic values ​​to accurately reflect the characteristics of the coil. This provides rich and accurate data for subsequent coupling calculations, helping to predict tension risks and optimize the winding process.

[0029] In one embodiment, the tension prediction model based on a diffusion autoregressive transformer is first trained, and the training steps are as follows: Dataset construction involves collecting multi-source time-series data from historical winding processes to construct a labeled dataset that includes normal winding, tension fluctuations, micro-slippage, and abnormal wire laying. This dataset is then divided into training, validation, and test sets in a 7:2:1 ratio. Data preprocessing involves normalizing the input time-series data. The normalization formula is as follows:

[0030] in, For the first Dimensional data mean Standard deviation; Fragment embedding divides normalized time-series data into subsequences of fixed length, maps them to a high-dimensional space through an embedding layer, and obtains the embedding features. ; Encoder diffusion processing: The autoregressive encoder captures temporal dependencies through causal masking and applies cosine scheduling noise to the encoded features to generate a self-supervised signal. in, , , ; Decoder denoising and reconstruction: The decoder removes noise using an inverse diffusion process and fuses global and local features through cross-attention to output a tension prediction sequence. Loss function calculation, total loss ,in

[0031] Model optimization was performed using the AdamW optimizer with a learning rate of [missing information]. Weight decay batchsize=16 / 32, iteration epoch= ; The stopping condition is that the loss on the validation set does not decrease for 10 consecutive rounds, and the tension prediction error on the test set... Training should be stopped immediately. After training is completed, the current cycle microslip precursor index will be adjusted. Residual stress memory matrix of corresponding concentric layers in the historical memory bank Perform vector dot product operation and normalize using matrix norm to obtain the tension risk trend value within the future control window; Then, the target tension correction is calculated using a PID control law to achieve real-time trend prediction and error compensation.

[0032] in, For the circle memory weight coefficient, This is the proportional adjustment coefficient. This is the integral adjustment coefficient. This is the integral term of the tension risk trend value within the control window; Parameter determination rules: The weights are assigned based on the order in which the layers are wound and the degree of stress accumulation, with newer layers having a higher weight. , Based on the servo system response bandwidth and tension control steady-state error calibration, the results were determined through the critical proportionality method and step response test.

[0033] This design first constructs and partitions a labeled dataset, then performs data preprocessing and fragment embedding, followed by encoder diffusion and decoder denoising and reconstruction. The loss is calculated and the model is optimized, and training stops when a certain condition is met. By constructing labeled datasets for various winding conditions, the model can learn comprehensively. Data preprocessing and fragment embedding improve data quality and feature representation, encoder diffusion and decoder denoising and reconstruction enhance the model's ability to process time-series data, reasonable loss calculation and optimization improve the model's prediction accuracy, and setting stopping conditions avoids overfitting. The trained model can accurately predict tension, providing a foundation for subsequent calculation of tension risk trend values, and realizing real-time trend prediction and error compensation.

[0034] In one embodiment, the line-laying brake correction amount Torque correction amount Tension roller balance position correction amount Lead time for cable reversal The calculation process is as follows: Using the target tension correction amount as the core control variable, the calculation is distributed according to the response characteristics and transmission ratio of each actuator. At the same time, the cable reversal advance is determined by combining the tension risk trend value and the real-time linear velocity of the optical fiber, so as to achieve multi-dimensional synchronous correction.

[0035]

[0036]

[0037] in, The line-laying braking response coefficient is... The winding torque response coefficient is... The position response coefficient of the tension roller. This is the correction factor for commutation advance. The real-time winding speed of the optical fiber; Basis for determining coefficients: , , The calibration is based on the measured mechanical transmission ratio, torque constant, and displacement gain of the actuator. It is determined based on the reversing stroke of the cabling, mechanical inertia, and system control delay time.

[0038] This design, with the target tension correction amount as the core, allocates calculations according to the characteristics of each actuator, and determines the lead time for cable reversal by combining risk trends and linear velocity. With the target tension correction amount as the core, it ensures that the control revolves around tension control. The allocation according to the response characteristics and transmission ratio of each actuator can give full play to the role of each mechanism and achieve precise control. The lead time for cable reversal is determined by combining the tension risk trend value and the real-time linear velocity of the optical fiber. Taking into account multiple factors, the cable reversal is more reasonable. Multi-dimensional synchronous correction can comprehensively adjust the optical fiber winding process, suppress tension peaks and other problems in advance, and improve winding stability and quality.

[0039] In one embodiment, the compensation deviation result The calculation process is as follows: Within a fixed-length evaluation window, multiple sets of actual tension and target tension sampling data are collected. The root mean square error is used to calculate the compensation deviation, and the execution effect of the control command and the tension control accuracy are quantitatively evaluated. in, To evaluate the first within the window The actual tension value at each sampling point For the target tension value at the corresponding sampling point, To evaluate the total number of sampling points within the window; Evaluation window criteria: Based on the control cycle and winding speed, the window is determined to ensure that it covers at least one complete wire reversal action.

[0040] This design collects data within a fixed evaluation window and uses root mean square error (RMSE) to calculate and compensate for deviations. The fixed evaluation window ensures the stability and consistency of the evaluation, covers at least one complete cable reversal action, and comprehensively considers the control effect. The use of RMSE calculation can accurately quantify the deviation between the actual tension and the target tension, intuitively reflecting the execution effect of the control command and the tension control accuracy. By calculating the compensation deviation results, problems in the control can be identified in a timely manner, providing a basis for subsequent adaptive parameter updates, so as to adjust the control strategy and continuously improve the accuracy of tension control during fiber winding.

[0041] In one embodiment, the calculation process for adaptive parameter updates is as follows: Using the compensation deviation result as a feedback signal, an incremental update rule is adopted to adaptively correct the control parameters online, so that the parameters adapt to the changes in the winding process state and continuously improve the control effect of subsequent cycles and layers:

[0042]

[0043] in, This is the threshold for microslip precursors. For circle memory weight, The target tension correction factor. For adaptive learning rate, , , For parameters before the update, , , These are the updated parameters; Threshold and learning rate determination rules: Based on the microslip precursor index during fault-free winding Quantiles are determined; The parameters are selected based on system stability and convergence speed to ensure that parameter updates are free of overshoot and converge quickly.

[0044] This design uses the compensation deviation result as feedback and incremental update rules to correct control parameters. The compensation deviation result serves as the feedback signal, reflecting the gap between the control effect and the expectation in real time. This makes parameter updates targeted. The use of incremental update rules avoids large parameter fluctuations, ensuring system stability. Through online adaptive parameter correction, the parameters can adapt to changes in the winding process, continuously improving the control effect of subsequent cycles and layers. Reasonably determining the threshold and learning rate ensures that parameter updates are free of overshoot and converge quickly, allowing the system to maintain good control performance under different operating conditions and improving the quality of fiber winding.

[0045] In one embodiment, the online incremental training steps of the diffusion autoregressive transformer model are as follows: After each fiber optic winding is completed, the winding data is added to the incremental training set. Freeze the core parameters of the model and only fine-tune the top fully connected layer and attention weights; Use a small learning rate epoch per iteration = This enables online adaptive updating of the model, maintaining stable long-term winding accuracy.

[0046] This design adds data to the incremental training set after each winding is completed, freezes the core parameters, and fine-tunes the top layer. By adding data after each winding is completed, the model can be updated in a timely manner using new data, allowing the model to adapt to the constantly changing winding process. Freezing the core parameters of the model and only fine-tuning the top layer avoids excessive changes in the overall model structure that could lead to performance instability, and ensures that the effective features learned by the model are not destroyed. By using a small learning rate and fewer iterations, the model can be updated online adaptively. This maintains long-term stable winding accuracy without affecting the normal operation of the system, thereby improving the quality and production efficiency of optical fiber winding.

[0047] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0048] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A fully automatic intelligent tension control system for optical fiber winding, characterized in that, The system includes: Status acquisition module, early warning characterization module, layer memory module, coupled decision-making module, linkage control module, and feedback adaptive module; The status acquisition module is used to acquire the speed of the pay-off end, the speed of the take-up end, the tension detection value, the displacement of the tension roller, the position of the wire laying mechanism, the servo motor current, the change in winding diameter, and the ambient temperature during the optical fiber winding process, so as to form the status data of the current control cycle. The precursor characterization module is used to extract angular velocity difference, phase offset, tension roller micro-vibration, servo current pulsation and cable reversal impact based on the current control cycle state data, and generate micro-slip precursor index; The layer memory module is used to construct a layer residual stress memory matrix for the completed layers; The coupled decision module is used to perform coupled analysis on the microslip precursor index and the layer residual stress memory matrix, and output the tension risk trend value and the target tension correction amount; The linkage control module is used to generate wire release brake correction, wire take-up torque correction, tension roller balance position correction and wire reversal advance based on the tension risk trend value and the target tension correction amount, so as to implement cross-layer predictive collaborative control of the optical fiber winding process. The feedback adaptive module is used to update the micro-slippage precursor threshold, the layer memory weight, and the target tension correction coefficient based on the tension response and layer forming results after linkage compensation.

2. The fully automatic fiber optic winding tension intelligent control system according to claim 1, characterized in that, The status acquisition module includes: Multi-source sampling unit, time-stamp alignment unit, and preprocessing unit; The multi-source sampling unit is used to acquire data from each sensing channel according to a unified control cycle; The time-scale alignment unit is used to map rotation speed signals, tension signals, displacement signals, current signals and temperature signals at different sampling frequencies to the same time axis; The preprocessing unit is used to remove missing sampling points and abnormal jump points, and outputs a standardized state data sequence corresponding to the control cycle.

3. The fully automatic fiber optic winding tension intelligent control system according to claim 2, characterized in that, The precursor characterization module includes: Differential analysis unit, frequency domain filtering unit, impulse identification unit, and fusion calculation unit; The differential analysis unit is used to generate the angular velocity difference characteristics and phase offset characteristics between the pay-off end and the take-up end; The frequency domain filtering unit is used to separate the micro-vibration characteristics of the tension roller and the servo current pulsation characteristics. The impact recognition unit is used to identify the impact characteristics at the moment of cable reversal; The fusion calculation unit is used to perform weighted fusion of each feature according to preset weights to generate a micro-slip precursor index that reflects the degree of early anomalies in tension instability.

4. The fully automatic fiber optic winding tension intelligent control system according to claim 1, characterized in that, The layered memory module includes: Layer segmentation unit, parameter extraction unit, matrix construction unit, and historical storage unit; The layer segmentation unit is used as a memory unit for the complete winding process of a single layer. The parameter extraction unit is used to extract the average tension, tension fluctuation amplitude, wire distribution status, roll diameter increment, curvature change parameters, and local overburden status parameters of the ring layer. The matrix construction unit is used to form a concentric residual stress memory matrix in a fixed dimensional order; The historical storage unit is used to write the memory matrix of the residual stress of each layer into the historical memory bank in the order of winding.

5. The fully automatic optical fiber winding tension intelligent control system according to claim 4, characterized in that, The layered memory module also includes: Weight allocation unit and invocation unit; The weight allocation unit is used to assign corresponding layer memory weights to the residual stress memory matrix of each layer in the historical memory bank according to the layer winding sequence, the winding diameter growth state and the stress accumulation degree. The calling unit is used to call the target layer residual stress memory matrix that matches the current roll diameter range and wiring state within the current control cycle, so as to improve the characterization accuracy of interlayer inheritance effect.

6. The fully automatic fiber optic winding tension intelligent control system according to claim 1, characterized in that, The coupled decision module includes: Coupled operation unit, normalization processing unit and correction generation unit; The coupling operation unit is used to perform correlation operations between the microslip precursor index and the layer residual stress memory matrix; The normalization processing unit is used to generate tension risk trend values ​​by combining matrix norm and layer memory weights; The correction generation unit is used to generate a target tension correction amount based on the tension risk trend value, the current tension deviation, and the cumulative deviation of the control window, so as to achieve joint decision-making of trend prediction and error compensation.

7. The fully automatic optical fiber winding tension intelligent control system according to claim 6, characterized in that, The coupled decision module also includes: Tension prediction model unit; The tension prediction model unit is used to train a diffusion autoregressive transformer model based on historical winding multi-source time series data, which includes normal winding state, tension fluctuation state, micro-slip state and abnormal winding state. The tension prediction model unit is used to output the tension prediction sequence for the future control window, and to generate the target tension correction amount by fusing the tension prediction sequence with the tension risk trend value.

8. The fully automatic optical fiber winding tension intelligent control system according to claim 1, characterized in that, The linkage control module includes: Braking distribution unit, torque distribution unit, position distribution unit, and commutation advance unit; The braking distribution unit is used to generate a line-laying braking correction amount based on the target tension correction amount; The torque distribution unit is used to generate a take-up torque correction amount based on the target tension correction amount; The position allocation unit is used to generate a tension roller balance position correction amount based on the target tension correction amount; The reversal advance unit is used to generate a reversal advance amount based on the tension risk trend value and the real-time linear speed of the optical fiber, so as to achieve synchronous compensation of the wire feeding mechanism, the wire taking-up mechanism, the tension roller adjustment mechanism and the wire laying mechanism.

9. The fully automatic optical fiber winding tension intelligent control system according to claim 8, characterized in that, The feedback adaptive module includes: Response acquisition unit, deviation evaluation unit, and parameter update unit; The response acquisition unit is used to acquire actual tension response, layer formation state and subsequent fluctuation data within a fixed evaluation window; The deviation assessment unit is used to calculate the compensation deviation result based on the actual tension value and the target tension value; The parameter update unit is used to perform incremental updates on the micro-slip precursor threshold, the layer memory weight, and the target tension correction coefficient based on the compensation deviation result, so as to keep the control parameters of subsequent control cycles and subsequent layers adaptively matched.

10. The fully automatic optical fiber winding tension intelligent control system according to claim 7, characterized in that: The tension prediction model unit also includes an online incremental training subunit; The online incremental training subunit is used to write the winding data into the incremental training set after each optical fiber winding is completed, freeze the backbone parameters of the diffusion autoregressive transformer model, fine-tune only the top fully connected layer and attention weights, and perform a limited number of iterations with a preset small learning rate to maintain the continuous stability of the future control window tension prediction accuracy under different working conditions.

Images