A multi-layer twisting control method and system for a copper core cross-linked polyethylene insulated cable

Abstract

The application belongs to the technical field of cross-linked polyethylene insulated cables, and particularly relates to a copper core cross-linked polyethylene insulated cable multi-layer twisting control method and system, which comprises the following steps: collecting multi-source heterogeneous data of key stations of a twisting machine, using a virtual queue mapping mechanism to perform space-time alignment on the data collected at different physical positions; based on a material work hardening mechanism and a thermal softening effect, a thermal hardening coupled cumulative factor is constructed, and a dynamic springback correction coefficient is constructed in combination with a die wear trend; the above features are input into a random forest model to predict the springback increment of a conductor, and the traction tension and the twisting speed are adjusted in a closed loop according to the prediction; and the model failure drift index is used to monitor the prediction residual, and the model is triggered to adaptively update. Through the fusion of physical mechanism features and machine learning, the application realizes adaptive and robust control in the whole life cycle, and eliminates the hidden danger of insulation eccentricity.

Description

Technical Field

[0001] This invention relates to the field of cross-linked polyethylene insulated cable technology. More specifically, this invention relates to a method and system for controlling the multilayer stranding of copper core cross-linked polyethylene insulated cables. Background Technology

[0002] Copper core cross-linked polyethylene (XLPE) insulated cables are the core carrier of modern high-voltage power transmission and distribution networks, and their manufacturing quality is directly related to the safe and stable operation of the power grid. In the cable manufacturing process, the conductor stranding process is the key link that determines the electrical performance and mechanical structure stability of the cable.

[0003] Currently, large-section cable conductors generally adopt a compaction stranding process, which involves stranding multiple copper wires together in a stranding cage and then subjecting them to forced plastic deformation through a compaction die. This process increases the conductor's fill factor and produces a smooth, round surface, thus providing a good foundation for subsequent insulation extrusion.

[0004] However, existing stranding control technologies face the technical challenge of uncontrollable dynamic springback effects when dealing with long-distance, continuous production tasks. Copper conductors, as a typical elasto-plastic material, undergo a certain degree of elastic recovery after being forcibly compressed and deformed by the mold, resulting in the final measured outer diameter of the conductor being slightly larger than the mold aperture. This springback behavior is not constant but is subject to dynamic interference from multiple nonlinear factors: on the one hand, the copper lattice is distorted due to stress during stranding, resulting in a cumulative effect of work hardening. With the slight fluctuations in stranding pitch and traction tension, the hardness of the material changes in real time; on the other hand, the intense friction between the mold and the copper wire generates heat, causing thermal expansion of the mold aperture and softening of the copper material during annealing, which alters the elastic modulus of the material.

[0005] Existing control strategies mostly rely on PID feedback or regression models trained based on historical static data. They cannot detect the micro-wear of the mold as the production length increases. After continuous production exceeds a certain length, traditional control strategies often produce systematic deviations because they cannot adapt to the slow changes in boundary conditions.

[0006] The aforementioned problems can cause the conductor's outer diameter to fluctuate beyond the process tolerance range, which in turn can lead to eccentricity in the subsequent XLPE insulation extrusion process, resulting in uneven insulation thickness. In severe cases, this can lead to electric field concentration, significantly reducing the cable's breakdown voltage and service life. Summary of the Invention

[0007] To address the technical problems of uncontrollable conductor springback, failure of long-distance production models, and insulation layer eccentricity in the prior art, the present invention provides solutions in the following aspects.

[0008] In a first aspect, the present invention provides a method for controlling the multilayer stranding of copper core cross-linked polyethylene insulated cables, comprising: collecting power parameters, state parameters, and conductor quality feedback data of key stations of the stranding machine; using a virtual queue mapping mechanism to perform spatiotemporal alignment of data collected from different physical locations and perform cleaning preprocessing; based on the aligned data, by analyzing the microstructure of the material and the wear trend of the die, calculating the thermo-hardening coupling accumulation factor characterizing the effective deformation energy of the material and the dynamic springback correction coefficient characterizing the physical boundary drift of the die; constructing a random forest regression model, using the power parameters, state parameters, thermo-hardening coupling accumulation factor, and dynamic springback correction coefficient as input features to predict the springback increment of the conductor; summing the cold-state die aperture with the predicted springback increment to obtain the model predicted outer diameter of the conductor; adjusting the traction tension and traction speed in a closed loop according to the difference between the model predicted outer diameter and the ideal process outer diameter; calculating the residual sequence between the model predicted outer diameter and the measured outer diameter, constructing a model failure drift index to evaluate the current predictive stability of the model; and updating the random forest regression model using high-confidence historical data when the model failure drift index exceeds a set drift threshold.

[0009] This invention constructs a hybrid drive control mode that combines physical mechanism characteristics with machine learning models, derives the thermo-hardening coupling accumulation factor, integrates the mechanical stress hardening and thermal softening effects using the root mean square form, and introduces a dynamic rebound correction coefficient that evolves with production mileage. Through the model failure drift index, the system is endowed with self-evolution capability in long-distance continuous production. By using a random forest model to predict the rebound increment and combining it with the drift index for adaptive updates, precise closed-loop control of the conductor outer diameter in long-distance continuous production is achieved. This not only ensures the concentricity and electrical performance of the cable insulation layer, but also significantly reduces the energy consumption and wear caused by frequent acceleration and deceleration of the traction motor and winch, and extends the equipment maintenance cycle.

[0010] Preferably, the power parameters include the rotational speed of the winch, the traction speed of the traction machine, and the traction tension of the single-wire feeding frame; the state parameters include the conductor surface temperature; and the conductor quality feedback data includes the measured outer diameter of the conductor.

[0011] Preferably, the step of using a virtual queue mapping mechanism to perform spatiotemporal alignment of data collected from different physical locations includes: aligning the data according to the traction speed of the tractor. Calculate transmission latency ,in, The distance between the laser diameter gauge and the mold exit; the time... Collected conductor quality feedback data and historical moments The dynamic parameters and state parameters are matched to complete the spatiotemporal alignment.

[0012] Preferably, the formula for calculating the thermo-hardening coupling accumulation factor is: In the formula: for The thermal hardening coupling accumulation factor at any given time; The length of the sliding time window; The index of the moment within the sliding time window; for The traction tension at any moment; Design the cross-sectional area for the conductor; for The rotational speed of the winch at any given moment; for The traction speed of the tractor at any given moment; for The surface temperature of the conductor at that moment; This is the reference temperature constant for the dynamic recrystallization of copper.

[0013] This invention identifies the dynamic interplay between the work hardening effect caused by tensile stress and stranding shear deformation and the annealing softening effect caused by temperature rise. The obtained thermo-hardening coupling cumulative factor enables the model to understand the real-time elastic modulus changes of materials under different process conditions, thereby more accurately predicting springback behavior.

[0014] Preferably, the formula for calculating the dynamic rebound correction coefficient is: In the formula: for The dynamic rebound correction coefficient at any given time; for The thermal hardening coupling accumulation factor at any given time; This is the baseline hardening value; for Cumulative production length at any given moment; Design life mileage for the mold; It is a natural constant; It is the hyperbolic tangent function.

[0015] This invention introduces a mold wear compensation mechanism in the time dimension. By simulating the small physical boundary drift of the mold aperture as production progresses through the relationship between cumulative production length and mold life, it solves the problem of the significant decrease in prediction accuracy caused by mold wear in the later stages of long-distance production in traditional static models, and ensures the control stability throughout the entire life cycle.

[0016] Preferably, the step of adjusting the traction tension and traction speed in a closed loop based on the difference between the model-predicted outer diameter and the ideal process outer diameter includes: calculating... Target control deviation at any time In the formula: Ideal process outer diameter; for The model predicts the outer diameter at time 10:00. This refers to the diameter of the cold mold hole. for Predicted rebound increment at any given time; if the target control deviation... The system reduces the traction tension of the single-wire pay-off frame or increases the traction speed of the traction machine; if The system should increase the traction tension of the single-wire pay-off frame or reduce the traction speed of the traction machine; if the target control deviation... The system maintains the traction tension of the single-line wire feeding frame and the traction speed of the traction machine.

[0017] Based on the predicted deviation, this invention adjusts the traction tension or speed accordingly. This dynamic adjustment mechanism can counteract material rebound fluctuations in real time, keeping the conductor's outer diameter within the ideal process tolerance range and preventing quality defects such as excessive compression or excessive rebound.

[0018] Preferably, the formula for calculating the model failure drift index is: In the formula: for The model failure drift exponent at time t; To evaluate the length of the window in real time, and to smooth out instantaneous fluctuations; These are the weighting coefficients; for The outer diameter residual at time, , for The measured outer diameter of the conductor at any given time; for The model predicts the outer diameter at time [time], and , For cold mold aperture, for The predicted rebound increment at any given moment; To evaluate the mean of the residual sequence within the real-time evaluation window, the residual sequence consists of the outer diameter residuals at all times within the real-time evaluation window.

[0019] This invention integrates the mean and volatility of residuals to calculate the model failure drift index. This index can proactively identify systematic deviations, assess in real time whether the current model is suitable for the current production environment, and provide a basis for triggering model self-updates.

[0020] Preferably, the cleaning pretreatment is achieved using the interquartile range method.

[0021] Preferably, updating the random forest regression model using high-confidence historical data includes: setting a drift threshold. ;when If the duration exceeds 1 minute, the current model is deemed invalid; the most recently stored... A set of high-confidence historical data is used to incrementally train or fully retrain the random forest regression model, updating the decision tree parameters; among which... The number of samples is the preset number; the high-confidence historical data is historical data processed by the interquartile range method.

[0022] This invention utilizes a real-time monitoring and proactive updating mechanism for the model failure drift index. The system is able to keenly capture process drift caused by mold wear or environmental changes. This full lifecycle adaptive capability ensures that the control model maintains high control accuracy throughout the early, middle, and late stages of the mold's lifespan.

[0023] In a second aspect, the present invention provides a multi-layer stranding control system for copper core cross-linked polyethylene insulated cables, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned multi-layer stranding control method for copper core cross-linked polyethylene insulated cables is implemented.

[0024] By adopting the above technical solution, a computer program is generated from the above-mentioned method for controlling the multi-layer stranding of copper core cross-linked polyethylene insulated cables, and stored in a memory for loading and execution by a processor. Terminal equipment is then manufactured based on the memory and processor for convenient use.

[0025] The beneficial effects of this invention are as follows: This invention constructs a hybrid drive control mode that combines physical mechanism characteristics with machine learning models, derives the thermo-hardening coupling accumulation factor, integrates the mechanical stress hardening and thermal softening effects using the root mean square form, and introduces a dynamic rebound correction coefficient that evolves with production mileage. Through the model failure drift index, the system is endowed with self-evolution capability in long-distance continuous production. By using a random forest model to predict the rebound increment and combining it with the drift index for adaptive updates, precise closed-loop control of the conductor outer diameter in long-distance continuous production is achieved. This not only ensures the concentricity and electrical performance of the cable insulation layer, but also significantly reduces the energy consumption and wear caused by frequent acceleration and deceleration of the traction motor and winch, and extends the equipment maintenance cycle. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a multilayer stranding control method for a copper core cross-linked polyethylene insulated cable according to the present invention; Figure 2 This is a schematic diagram illustrating the real-time monitoring and proactive correction process of the model's state; Figure 3 This is a schematic diagram showing the comparison of conductor outer diameter consistency during long-distance production. Detailed Implementation

[0027] 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, not all, of the embodiments of the present invention. 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.

[0028] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0029] This invention discloses a method for controlling the multilayer stranding of copper core cross-linked polyethylene insulated cables, referring to... Figure 1 This includes steps S1-S4: S1: Collect power parameters, status parameters and conductor quality feedback data of key stations of the stranding machine, use the virtual queue mapping mechanism to perform spatiotemporal alignment of data collected from different physical locations, and perform cleaning and preprocessing.

[0030] On the cable stranding production line, since the laser diameter gauge is usually installed at a certain distance behind the compaction mold, the conductor quality feedback data is delayed relative to the current power parameters. In order to solve this problem, this step constructs a full-dimensional data acquisition and alignment mechanism.

[0031] First, deploy a high-frequency sensor network on the stranding machine, setting the sampling frequency to 50Hz, and collect the following data: (1) Power parameters: The rotational speed of the winch is collected by encoder and force sensor. (Unit: rpm) Traction speed of the tractor (Unit: m / min) and the traction tension of the single-wire pay-off frame. (Unit: N).

[0032] (2) Status parameters: The surface temperature of the conductor at the inlet of the compaction mold is collected by an infrared thermometer. (Unit: °C)

[0033] (3) Conductor quality feedback data: via data installed behind the mold exit A laser diameter gauge at a distance of meters collects the measured outer diameter of the conductor. (Unit: mm), where, =2 meters.

[0034] Next, a virtual queue mapping mechanism is used to perform spatiotemporal alignment of data collected from different physical locations; specifically, the system constructs a dynamically variable-length virtual FIFO queue in memory, adopting a first-in-first-out (FIFO) approach, by mapping the time... Collected conductor quality feedback data With historical moments The dynamic parameters and state parameters are paired to achieve spatiotemporal alignment; only the aligned data rows will be sent to the subsequent processing flow.

[0035] in, The transmission delay time is determined based on the traction speed of the traction machine. Dynamic calculation, and ,in, This refers to the distance between the laser diameter gauge and the mold exit, in meters. The traction speed of the tractor; when At that time, transmission lag time .

[0036] Finally, the interquartile range (IQR) method within a sliding window is used to clean the aligned data, remove noise caused by instantaneous device fluctuations, and normalize the maximum and minimum values ​​of all physical quantities.

[0037] In this way, the virtual queue mapping mechanism solves the problem of data time synchronization caused by the dispersed physical locations of sensors, ensures the spatiotemporal consistency of model training data, eliminates noise interference caused by data misalignment, and lays a high-quality data foundation for accurate prediction.

[0038] S2. Based on the aligned data, by analyzing the material microstate and mold wear trend, the thermo-hardening coupling accumulation factor characterizing the effective deformation energy of the material and the dynamic springback correction coefficient characterizing the physical boundary drift of the mold are calculated respectively.

[0039] It should be noted that since simple physical quantities are insufficient to describe the internal microscopic state of copper conductors, and the springback behavior of conductors is affected by the dynamic game between work hardening and thermal softening effects, and the mold will experience boundary drift due to wear during long-distance production; therefore, this invention deeply analyzes the metal processing mechanism and extracts the thermo-hardening coupling accumulation factor that can characterize the microscopic energy state of the material and the dynamic springback correction coefficient that characterizes the wear state of the mold.

[0040] Specifically, combining lengths of Calculate the thermo-hardening coupling accumulation factor using a sliding time window. This factor determines the effective deformation energy accumulated per unit volume of copper material, after thermal effect correction. The specific calculation formula is as follows:

[0041] In the formula: for The thermal hardening coupling accumulation factor at any given time; The length of the sliding time window determines the model's memory of historical states. If it's too small, it cannot capture the cumulative effect; if it's too large, computational lag becomes significant. The value range is [5, 20]. In this embodiment, Set to 10; The index of the moment within the sliding time window; for The traction tension at any moment; Design the cross-sectional area for the conductor; for The rotational speed of the winch at any given moment; for The traction speed of the tractor at any given moment; for The surface temperature of the conductor at that moment; This is the reference temperature constant for the dynamic recrystallization of copper.

[0042] in, Characterizes the contribution of tensile stress; Characterizes the degree of shear deformation caused by the hinge pitch; Characterizing the weakening effect of temperature increase on the hardening process, i.e., softening; the overall form using the square root of the sum of squares enhances the sensitivity to instantaneous high-energy deformation; when Or, as the stranding shear increases, hardening intensifies, and the thermo-hardening coupling accumulation factor increases. Increase; when As the temperature rises, the softening effect intensifies, and the thermo-hardening coupling accumulation factor increases. Decrease.

[0043] Furthermore, the cumulative production length is introduced to calculate the dynamic springback correction coefficient. This is used to compensate for boundary drift caused by mold wear. The specific calculation formula is as follows:

[0044] In the formula: for The dynamic rebound correction coefficient at any given time; for The thermal hardening coupling accumulation factor at any given time; The baseline hardening value is used to normalize the hardening level. In this embodiment, the baseline hardening value is... Set to 5; for Cumulative production length at any given moment; Design life mileage for the mold; It is a natural constant; It is the hyperbolic tangent function.

[0045] in, This reflects a wear trend that increases non-linearly with increasing production length; hyperbolic tangent function. Limit the correction factor to a reasonable range; with the cumulative production length The increase in dynamic rebound correction coefficient It gradually increases, and is higher in the hardened state, i.e., the thermo-hardening coupling accumulation factor. When the wear and tear is significant, the negative effects of wear and tear are amplified.

[0046] Thus, by constructing composite features that incorporate physical mechanisms, it is possible to understand the hardening and softening mechanisms of materials and the wear trends of molds, thereby maintaining a precise ability to perceive the outer diameter springback during long-cycle production.

[0047] S3. Construct a random forest regression model, using dynamic parameters, state parameters, thermosetting coupling accumulation factor, and dynamic rebound correction coefficient as input features to predict the rebound increment of the conductor; adjust the traction tension and traction speed in a closed loop based on the rebound increment.

[0048] It should be noted that, due to the complex nonlinear relationship between the process parameters and conductor springback during stranding, traditional linear control is difficult to achieve high-precision prediction. Therefore, this invention utilizes the powerful nonlinear mapping capability of the random forest regression model to predict the springback increment of the conductor and implements closed-loop feedback control based on this.

[0049] It should be further noted that the prediction accuracy of the random forest regression model is highly dependent on the quality of the training data and the completeness of the features. In order for the model to accurately learn the nonlinear mapping relationship between dynamic parameters, state parameters and conductor rebound behavior, a high-quality training dataset containing physical mechanism features must be constructed based on historical production data.

[0050] Specifically, based on historical production data, the coil rotation speed, traction speed of the traction machine, traction tension, conductor surface temperature, thermosetting coupling accumulation factor, and dynamic springback correction coefficient at each moment are combined to form a feature vector of a sample. For each sample, calculate its corresponding true rebound increment as the training target label for the model. The actual rebound increment is equal to the measured outer diameter of the conductor. With cold mold aperture The difference.

[0051] Furthermore, the constructed paired sample data The original dataset was constructed, and outliers caused by equipment startup / shutdown and sensor malfunctions were removed using the interquartile range (IQR) method to obtain a high-confidence training dataset. This high-confidence training dataset was then input into a random forest regression algorithm for training. During training, multiple subsets were generated using the Bootstrap sampling method, and 100 decision trees were constructed for each subset. When splitting at a node, each decision tree was analyzed from the feature vector... The optimal split point search is performed by randomly selecting some features from all dimensions, and finally averaging the prediction results of all decision trees to obtain a pre-trained random forest regression model.

[0052] Furthermore, targeting Time, building eigenvectors at time step ,in, for The rotation speed of the winch at any given moment, for The traction speed of the tractor at any given moment. for The tension of time for The conductor surface temperature at time t. for The thermal hardening coupling accumulation factor at time. for The dynamic rebound correction coefficient at any given time; Feature vector at time step Input into the trained random forest regression model to obtain Predicted rebound increment of the conductor at time That is, the measured outer diameter of the conductor. Relative to the diameter of the cold mold hole The amount of expansion.

[0053] Further calculation Target control deviation at any time The specific calculation formula is as follows:

[0054] In the formula: for Target control deviation at any given time; Ideal process outer diameter; This refers to the diameter of the cold mold hole. for The predicted rebound increment at any given moment.

[0055] Finally, the traction tension and traction speed are adjusted in a closed loop based on the target control deviation, including: if the target control deviation... This indicates that the predicted outer diameter is smaller than the target value, i.e., the ideal process outer diameter. This indicates excessive compression; the system should reduce the traction tension of the single-wire pay-off frame or increase the traction speed of the traction machine. This indicates that the predicted outer diameter is greater than the target value, suggesting excessive springback. The system should increase the traction tension of the single-wire pay-off frame or reduce the traction speed of the traction machine. If the target control deviation... This indicates that the predicted outer diameter equals the target value, i.e., the ideal process outer diameter. The system maintains the traction tension of the single-line wire feeding frame and the traction speed of the traction machine.

[0056] Thus, by introducing the random forest algorithm and closed-loop feedback control, the outer diameter of the conductor can be precisely adjusted, ensuring that the conductor size is always within the target tolerance range, and providing a high-quality geometric basis for subsequent processes.

[0057] S4. Calculate the residual sequence between the predicted outer diameter and the measured outer diameter, construct the model failure drift index to evaluate the current predictive stability of the model, and update the random forest regression model using high-confidence historical data when the model failure drift index exceeds the set drift threshold.

[0058] It should be noted that, due to the gradual changes in the production environment, the model will gradually age and fail, and a fixed model cannot adapt to dynamic working conditions in the long term; therefore, this invention constructs a model failure drift index to monitor the health status of the model in real time and implement proactive updates.

[0059] Specifically, maintain a length of The real-time evaluation window calculates the model failure drift index. The specific calculation formula is as follows:

[0060] In the formula: for The model failure drift exponent at time t; To ensure real-time evaluation of the window length and to smooth out instantaneous fluctuations, it is set to 5 in this embodiment; is a weighting coefficient used to balance the weights of the deviation and fluctuation terms, with a value range of [0.4, 0.8], and is set to 0.6 in this embodiment; for The outer diameter residual at time, , for The measured outer diameter of the conductor at any given time; for The model predicts the outer diameter at time [time], and , For cold mold aperture, for The predicted rebound increment at any given moment; To evaluate the mean of the residual sequence within the real-time evaluation window, the residual sequence consists of the outer diameter residuals at all times within the real-time evaluation window.

[0061] The first term calculates the absolute mean of the residuals, reflecting whether there is a systematic bias in the model; the second term calculates the standard deviation of the residuals, reflecting the stability of the model.

[0062] Furthermore, a drift threshold is set. The model failure drift index comprehensively reflects the absolute value and volatility of the model prediction bias, and the drift threshold. It is the critical line for determining whether the model is distorted. Setting the value too low can lead to frequent model retraining due to even minor measurement noise, consuming computational resources and potentially introducing overfitting oscillations; if... If the drift threshold is set too high, the system will be unable to detect the systematic deviation caused by mold wear for a long time, resulting in the production of defective products; therefore, the drift threshold should be set appropriately. The recommended value range is 1% to 10% of the allowable deviation of the process tolerance; in this embodiment, considering the high precision requirements of the high voltage cable conductor, the drift threshold is set to 3% of the allowable deviation of the process tolerance.

[0063] Finally, when If the duration exceeds 1 minute, the current model is determined to be invalid, and the system immediately starts a background thread to utilize the most recently stored... A set of high-confidence historical data is used to incrementally train or fully retrain the random forest regression model, updating the decision tree parameters; among which... The number of samples is the preset number; the high-confidence historical data is historical data processed by the interquartile range method.

[0064] In this embodiment, a preset number of samples is used. Setting it to 1000 means selecting the 1000 most recently accumulated sets of high-confidence historical data.

[0065] For example, Figure 2 This diagram illustrates the real-time monitoring and proactive correction process of the model's state. The model failure drift index curve is used to determine the cumulative deviation between the current model and the actual physical process. As production progresses, the model failure drift index slowly rises due to environmental changes and wear, indicating that the model's accuracy is declining. When the model failure drift index touches a set threshold, a red inverted triangle appears at the corresponding position, triggering a model update. The curve then instantly drops vertically to a low level, demonstrating that after the system detects model failure, it automatically triggers background retraining and resets the error, ensuring that the control model remains in optimal condition throughout the entire production cycle.

[0066] For example, Figure 3This is a comparison chart of conductor outer diameter consistency during long-distance production. Existing static control methods show acceptable control in the early stages of production, but as the production length increases, the outer diameter deviation exhibits a significant and continuous upward trend due to the inability to detect systematic drift caused by mold wear and heat accumulation. At approximately 1500 meters of production, the deviation curve exceeds the acceptable tolerance range and severely exceeds the limit in the later stages, indicating that existing technology cannot maintain quality stability in long-distance production. In contrast, the adaptive control method of this invention, thanks to the introduction of dynamic features and an adaptive update mechanism, ensures that the outer diameter deviation consistently fluctuates closely around the 0 axis. Even in the later stages of production, under the same mold wear conditions, the outer diameter deviation remains steadily within the acceptable tolerance range, demonstrating the invention's ability to suppress nonlinear springback and process drift.

[0067] In this way, through real-time monitoring and adaptive updating of the model failure drift index, the system can overcome the long-cycle drift problem caused by mold aging and environmental changes, achieve robust control throughout the entire life cycle, ensure product quality consistency, and significantly extend the effective duration of continuous production.

[0068] This invention also discloses a multi-layer stranding control system for copper core cross-linked polyethylene insulated cables, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a multi-layer stranding control method for copper core cross-linked polyethylene insulated cables according to the present invention.

[0069] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method for controlling the multilayer stranding of copper core cross-linked polyethylene insulated cables, characterized in that, include: The system collects power parameters, status parameters, and conductor quality feedback data from key stations of the stranding machine. It then uses a virtual queue mapping mechanism to perform spatiotemporal alignment of the data collected from different physical locations and performs cleaning and preprocessing. Based on the aligned data, by analyzing the material microstate and mold wear trend, the thermo-hardening coupling accumulation factor characterizing the effective deformation energy of the material and the dynamic springback correction coefficient characterizing the physical boundary drift of the mold are calculated respectively. A random forest regression model is constructed, using dynamic parameters, state parameters, thermosetting coupling accumulation factor, and dynamic springback correction coefficient as input features to predict the springback increment of the conductor; the model predicted outer diameter of the conductor is obtained by summing the cold mold aperture with the predicted springback increment; the traction tension and traction speed are adjusted in a closed loop based on the difference between the model predicted outer diameter and the ideal process outer diameter. The residual sequence between the predicted outer diameter and the measured outer diameter is calculated, and the model failure drift index is constructed to evaluate the current predictive stability of the model. When the model failure drift index exceeds the set drift threshold, the random forest regression model is updated using high-confidence historical data. The formula for calculating the thermo-hardening coupling accumulation factor is: ； In the formula: for The thermal hardening coupling accumulation factor at any given time; The length of the sliding time window; The index of the moment within the sliding time window; for The traction tension at any moment; Design the cross-sectional area for the conductor; for The rotational speed of the winch at any given moment; for The traction speed of the tractor at any given moment; for The surface temperature of the conductor at that moment; This is the reference temperature constant for the dynamic recrystallization of copper; The formula for calculating the dynamic rebound correction coefficient is: ； In the formula: for The dynamic rebound correction coefficient at any given time; for The thermal hardening coupling accumulation factor at any given time; This is the baseline hardening value; for Cumulative production length at any given moment; Design life mileage for the mold; It is a natural constant; It is the hyperbolic tangent function; The formula for calculating the model failure drift index is: ； In the formula: for The model failure drift exponent at time t; To evaluate the length of the window in real time, and to smooth out instantaneous fluctuations; These are the weighting coefficients; for The outer diameter residual at time, , for The measured outer diameter of the conductor at any given time; for The model predicts the outer diameter at time [time], and , For cold mold aperture, for The predicted rebound increment at any given moment; To evaluate the mean of the residual sequence within the real-time evaluation window, the residual sequence consists of the outer diameter residuals at all times within the real-time evaluation window.

2. The method for controlling the multilayer stranding of a copper core cross-linked polyethylene insulated cable according to claim 1, characterized in that, The power parameters include the rotational speed of the winch, the traction speed of the traction machine, and the traction tension of the single-wire feeding frame; the state parameters include the conductor surface temperature; and the conductor quality feedback data includes the measured outer diameter of the conductor.

3. The method for controlling the multilayer stranding of a copper core cross-linked polyethylene insulated cable according to claim 2, characterized in that, The method of using a virtual queue mapping mechanism to perform spatiotemporal alignment of data collected from different physical locations includes: According to the traction speed of the traction machine Calculate transmission latency ,in, The distance between the laser diameter gauge and the mold exit; the time... Collected conductor quality feedback data and historical moments The dynamic parameters and state parameters are matched to complete the spatiotemporal alignment.

4. The method for controlling the multilayer stranding of a copper core cross-linked polyethylene insulated cable according to claim 2, characterized in that, The method of adjusting the traction tension and traction speed in a closed loop based on the difference between the predicted outer diameter and the ideal process outer diameter includes: calculate Target control deviation at any time In the formula: Ideal process outer diameter; for The model predicts the outer diameter at time 10:

00. This refers to the diameter of the cold mold hole. for The predicted rebound increment at any given moment; If the target control deviation The system reduces the traction tension of the single-wire pay-off frame or increases the traction speed of the traction machine; if The system should increase the traction tension of the single-wire pay-off frame or reduce the traction speed of the traction machine; if the target control deviation... The system maintains the traction tension of the single-line wire feeding frame and the traction speed of the traction machine.

5. The method for controlling the multilayer stranding of a copper core cross-linked polyethylene insulated cable according to claim 1, characterized in that, The cleaning pretreatment is achieved using the interquartile range method.

6. The method for controlling the multilayer stranding of a copper core cross-linked polyethylene insulated cable according to claim 1, characterized in that, The method of updating the random forest regression model using high-confidence historical data includes: Set drift threshold ;when If the duration exceeds 1 minute, the current model is deemed invalid; the most recently stored... A set of high-confidence historical data is used to incrementally train or fully retrain the random forest regression model, updating the decision tree parameters; among which... The number of samples is the preset number; the high-confidence historical data is historical data processed by the interquartile range method.

7. A multi-layer stranding control system for copper core cross-linked polyethylene insulated cables, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a multilayer stranding control method for copper core cross-linked polyethylene insulated cables according to any one of claims 1-6.

Images