Cable insulation layer extrusion temperature control method and system based on multi-parameter fusion

Abstract

The application relates to a cable insulation layer extrusion temperature control method and system based on multi-parameter fusion, relates to the technical field of cable extrusion temperature control, and comprises the following steps: collecting multi-source time sequence data of a cable insulation layer extrusion process; calculating an instantaneous viscosity characteristic parameter and determining a delay influence coefficient; combining the change rate of the instantaneous viscosity characteristic parameter in a sliding window and the delay influence coefficient to determine the occurrence time and drift amplitude of a viscosity drift event; extracting a melt temperature change rate, a melt pressure fluctuation amplitude and a cooling water inlet and outlet temperature difference change amount; inputting the melt temperature change rate, the melt pressure fluctuation amplitude and the cooling water inlet and outlet temperature difference change amount into a coupling compensation model to output a feedforward correction amount; layer-by-layer correcting temperature set values of heating zones and adjusting cooling water flow to obtain a stable melt temperature. The application solves the problems that a traditional temperature control mode adopts passive adjustment, temperature fluctuation overshoot and unstable flow are caused, and product size deviation and material thermal degradation are easily caused.

Description

Technical Field

[0001] This application relates to the field of temperature control for cable extrusion, and in particular to a method and system for temperature control of cable insulation extrusion based on multi-parameter fusion. Background Technology

[0002] As the cable manufacturing industry places increasing demands on product quality and molding stability, the temperature control precision during the cable insulation extrusion process has become a key technical requirement to ensure the core performance of the product.

[0003] Currently, traditional cable insulation extrusion temperature control methods adopt a passive adjustment mode, which only adjusts the temperature after it deviates from the set value. This delayed adjustment is prone to temperature fluctuation overshoot, causing unstable extrusion flow, and also brings the risk of uneven insulation layer size and material thermal degradation, thus reducing the forming quality of cable products. Summary of the Invention

[0004] This application provides a method and system for temperature control of cable insulation extrusion based on multi-parameter fusion, which improves the problems caused by the lag of traditional passive temperature control, such as melt temperature fluctuations, unstable extrusion flow, and easy occurrence of insulation layer size deviations and material thermal degradation, thereby improving the temperature control accuracy and molding stability of cable insulation extrusion.

[0005] This application discloses the following technical solution: In a first aspect, this application provides a method for temperature control of cable insulation extrusion based on multi-parameter fusion, the method comprising: Real-time acquisition of multi-source time-series data during the extrusion process of cable insulation layer, wherein the multi-source time-series data includes at least melt pressure, screw speed, cooling water inlet and outlet temperature difference, and melt temperature; Instantaneous viscosity characterization parameters are calculated based on the melt pressure and screw speed, and a delay influence coefficient is determined based on the transmission relationship of the influence of the cooling water inlet and outlet temperature difference on the melt pressure. By combining the rate of change of the instantaneous viscosity characterization parameter within the sliding window and the delay effect coefficient, the occurrence time and drift amplitude of the viscosity drift event can be determined. In response to the viscosity drift event, the melt temperature change rate within the first time window before the event occurs, and the melt pressure fluctuation amplitude and the change in cooling water inlet and outlet temperature difference within the second time window after the event occur are extracted. The melt temperature change rate, melt pressure fluctuation amplitude, and cooling water inlet and outlet temperature difference change are input into the pre-trained coupled compensation model, and the feedforward correction amount for the current temperature control setpoint is output. Based on the feedforward correction amount, the temperature setpoints of each heating zone during the extrusion process of the cable insulation layer are corrected in layers, and the cooling water flow rate is adjusted synchronously to obtain a stable melt temperature during the viscosity drift process.

[0006] Secondly, this application provides a cable insulation extrusion temperature control system based on multi-parameter fusion, the system comprising: A multi-source data acquisition module is used to acquire multi-source time-series data in real time during the extrusion process of cable insulation layer. The multi-source time-series data includes at least melt pressure, screw speed, cooling water inlet and outlet temperature difference, and melt temperature. The characteristic parameter calculation module is used to calculate the instantaneous viscosity characterization parameters based on the melt pressure and screw speed, and at the same time determine the delay influence coefficient based on the influence transmission relationship of the cooling water inlet and outlet temperature difference on the melt pressure; The viscosity event discrimination module is used to determine the occurrence time and drift amplitude of viscosity drift events by combining the rate of change of the instantaneous viscosity characterization parameter within the sliding window and the delay influence coefficient. The event feature extraction module is used to respond to the viscosity drift event and extract the melt temperature change rate within the first time window before the event occurs, as well as the melt pressure fluctuation amplitude and the change in cooling water inlet and outlet temperature difference within the second time window after the event occurs. The model prediction correction module is used to input the melt temperature change rate, melt pressure fluctuation amplitude and cooling water inlet and outlet temperature difference into the pre-trained coupled compensation model, and output the feedforward correction amount for the current temperature control setpoint. The collaborative execution control module is used to perform layered correction of the temperature setpoint of each heating zone during the extrusion process of the cable insulation layer according to the feedforward correction amount, and to synchronously adjust the cooling water flow rate to obtain a stable melt temperature during the viscosity drift process.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application proposes a method and system for temperature control of cable insulation extrusion based on multi-parameter fusion. Through steps such as constructing instantaneous viscosity characterization parameters, identifying viscosity drift events, extracting multi-dimensional features before and after drift, reasoning of coupled compensation models, stratification correction of heating zones, coordinated adjustment of cooling water flow rate, and closed-loop determination of melt temperature stability, the method achieves accurate feedforward control and rapid stable recovery of melt temperature under viscosity drift disturbance during cable insulation extrusion. First, instantaneous viscosity characterization parameters are constructed based on melt pressure and temperature data. Sliding window fitting and hysteresis compensation are used to identify viscosity drift events and determine their occurrence time and drift amplitude. Then, a time window is extracted based on the event time to extract the melt temperature change rate, melt pressure fluctuation amplitude, and cooling water inlet and outlet temperature difference. Next, the multi-dimensional features are input into a pre-trained coupled compensation model to output a feedforward correction amount adapted to the current operating conditions. Subsequently, the correction amount is analyzed and the temperature of each heating zone is corrected in layers, and the cooling water flow rate is adjusted synchronously to achieve coordinated control. Finally, the deviation tolerance threshold is calculated based on historical stable data and the delay influence coefficient. The temperature recovery deviation is compared in real time to complete the determination of the melt temperature stability state and form a temperature control closed loop.

[0008] The technical solution of this application solves the problems of lag in temperature control, slow response, poor coordination between heating and cooling, and difficulty in suppressing viscosity drift in traditional cable insulation extrusion temperature control by multi-parameter feature fusion and intelligent feedforward compensation. It avoids uneven extrusion, thickness deviation and product quality defects caused by viscosity fluctuation. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A schematic flowchart of a cable insulation extrusion temperature control method based on multi-parameter fusion provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a cable insulation extrusion temperature control system based on multi-parameter fusion, provided in an embodiment of this application.

[0011] The components represented by each number in the attached diagram are explained below: Multi-source data acquisition module 01, feature parameter calculation module 02, viscosity event discrimination module 03, event feature extraction module 04, model prediction and correction module 05, and collaborative execution control module 06. Detailed Implementation

[0012] This application provides a method and system for temperature control of cable insulation extrusion based on multi-parameter fusion, which solves the technical problems in the prior art where the temperature control system only passively adjusts after the temperature deviates due to melt viscosity fluctuations, resulting in temperature fluctuation overshoot, unstable extrusion flow, and consequently uneven insulation layer size and thermal degradation of the material.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] Example 1, as shown in the appendix Figure 1 As shown, this application provides a method for temperature control of cable insulation extrusion based on multi-parameter fusion, the method comprising the following steps: S110: Real-time acquisition of multi-source time-series data during the extrusion process of cable insulation layer, wherein the multi-source time-series data includes at least melt pressure, screw speed, cooling water inlet and outlet temperature difference, and melt temperature; In this embodiment of the application, in the actual production scenario of cable insulation layer extrusion molding, in order to achieve real-time perception of the process parameters of the extrusion process and capture the dynamic changes of melt state and equipment operation, it is necessary to synchronize and collect high-frequency multi-source time-series data of the core process parameters in the extrusion process, so as to build a parameter perception system covering the key links of extrusion and ensure the scientificity and timeliness of subsequent temperature control decisions.

[0015] Among them, multi-source time-series data refers to time-series data collected at a fixed sampling frequency during the continuous production process of cable insulation extrusion, reflecting different process dimensions. Melt pressure, screw speed, cooling water inlet and outlet temperature difference, and melt temperature are the core parameters that determine the melt forming state. The real-time performance and accuracy of the data directly affect the effects of subsequent viscosity characterization calculations, drift event identification, and temperature control correction. Through the integrated acquisition of this type of data, it is possible to achieve comprehensive monitoring of the melt rheological characteristics and equipment operating status during the extrusion process.

[0016] In addition, various types of multi-source time-series data are collected at key monitoring points of the extrusion equipment using corresponding professional sensing and detection equipment. The acquisition equipment with different parameters is matched with the installation points to ensure that the collected data can truly reflect the actual state of the corresponding process.

[0017] In the method provided in this application embodiment, the melt pressure is obtained in real time by a melt pressure sensor installed at the extruder head, the screw speed is obtained in real time by an encoder of the extruder main drive motor, the temperature difference between the cooling water inlet and outlet is obtained by real-time acquisition and calculation of the difference by a first temperature sensor installed at the inlet of the cooling water circuit and a second temperature sensor installed at the outlet of the cooling water circuit, and the melt temperature is obtained in real time by insertion thermocouple at the melt flow channel of the extruder head.

[0018] Specifically, the acquisition of melt pressure relies on a melt pressure sensor, which is directly installed at the extruder head. This location is a critical node for melt extrusion molding, where the melt completes the final flow channel formation. The sensor installed at this location can directly capture the pressure changes during melt extrusion, avoiding pressure data distortion caused by installation point deviation.

[0019] In addition, the screw speed, as a core operating parameter of the extruder, is collected by the encoder of the main drive motor of the extruder. The encoder is linked with the main drive motor and can capture the number of rotations and speed changes of the motor in real time, and convert them into the actual operating speed data of the screw.

[0020] Secondly, the acquisition of the temperature difference between the inlet and outlet of the cooling water needs to be completed with two sets of temperature sensors. The first temperature sensor is installed at the inlet of the cooling water circuit, and the second temperature sensor is installed at the outlet of the cooling water circuit. The inlet and outlet real-time temperatures of the cooling water are collected by the two sets of sensors respectively, and the difference between the two is calculated to obtain the temperature difference data of the inlet and outlet of the cooling water, so as to intuitively reflect the temperature change of the cooling water during the heat exchange process.

[0021] Finally, the melt temperature is collected by an insertion thermocouple, which is directly inserted into the melt flow channel of the extruder head and in direct contact with the melt. This allows for real-time measurement of the actual temperature of the melt within the flow channel, thus avoiding temperature errors caused by non-contact measurement.

[0022] In the actual data acquisition process, the installation and debugging of all sensing and detection equipment must first be completed to ensure that the installation points of the melt pressure sensor, encoder, two sets of temperature sensors and insertion thermocouple are accurate and firmly fixed, and that there is no mechanical interference with the operating parts of the extrusion equipment. At the same time, the calibration of each device must be completed to eliminate the measurement errors of the device itself.

[0023] Secondly, all data acquisition devices need to be configured with a uniform sampling frequency. The acquisition frequencies of melt pressure, screw speed, cooling water inlet and outlet temperature difference, and melt temperature should be set to the same value to ensure the synchronization of various data in the time dimension, avoid time series data misalignment caused by different sampling frequencies, and ensure data matching in subsequent multi-parameter fusion analysis.

[0024] In addition, a stable data transmission channel needs to be configured for each acquisition device to synchronously transmit the acquired real-time data to the data processing unit, so as to realize the real-time reception and storage of data and prepare data for subsequent parameter calculation and event discrimination.

[0025] For example, in an extrusion production line for cross-linked polyethylene cable insulation, a melt pressure sensor is installed at the melt flow channel outlet of the extruder head, an encoder is installed on the shaft of the main drive motor of the extruder, and a first temperature sensor and a second temperature sensor are installed on the inlet and outlet pipes of the cooling water circuit, respectively. Simultaneously, an insertion thermocouple is inserted into the melt flow channel of the extruder head to directly contact the melt. The sampling frequency of all acquisition devices is uniformly set to 10Hz. The real-time data collected by each device is transmitted to the PLC data processing unit on-site via an industrial bus. This unit receives and stores time-series data of melt pressure, screw speed, cooling water inlet and outlet temperatures, and melt temperature in real time. By calculating the difference between the cooling water inlet and outlet temperatures, the cooling water inlet and outlet temperature difference is obtained, ultimately forming a multi-source time-series dataset containing four core parameters. This provides continuous and synchronous raw data support for subsequent calculation of instantaneous viscosity characterization parameters, viscosity drift event discrimination, and temperature control adjustment.

[0026] S120: Calculate the instantaneous viscosity characterization parameters based on the melt pressure and screw speed, and determine the delay influence coefficient based on the influence transmission relationship of the temperature difference between the inlet and outlet of the cooling water on the melt pressure; In this embodiment of the application, in order to accurately characterize the real-time rheological state of the melt and quantify the hysteretic effect of cooling water temperature changes on melt pressure, it is necessary to calculate the instantaneous viscosity characterization parameters by coupling process parameters and determine the delay influence coefficient by combining historical data mining, so as to support the subsequent identification of viscosity drift events.

[0027] First, based on synchronously acquired melt pressure and screw speed data, instantaneous viscosity characterization parameters are obtained through ratio calculation, outlier removal, and multi-stage smoothing. Specifically, the ratio of melt pressure to screw speed is used as the basic calculated value. Data is stored in a circular buffer queue. After outlier removal through normal distribution testing, an equal-weighted moving average filter is applied to obtain a coarse smoothed value. Then, three consecutive coarse smoothed values ​​are assigned different weights according to their time sequence and subjected to a weighted secondary smoothing process, with the middle value having the highest weight. The result is the final instantaneous viscosity characterization parameter.

[0028] Simultaneously, the moments of abrupt temperature changes at the inlet and outlet of cooling water and the moments of melt pressure response are extracted from historical operating data. After time difference calculation and elimination of extreme samples, the delay influence coefficient is determined. Specifically, the moments of temperature change and corresponding pressure response within multiple time intervals are extracted, and the time difference between the two is calculated as a delay time sample. After sorting the samples, extreme samples at both ends with a predetermined proportion are eliminated, and the arithmetic mean of the remaining intermediate samples is taken as the delay influence coefficient.

[0029] This step, by determining the instantaneous viscosity characterization parameters and the delay effect coefficient, not only effectively captures the viscosity state of the melt but also quantifies the hysteretic effect of cooling water temperature on melt pressure, providing a quantitative basis for the subsequent determination of the timing and amplitude of viscosity drift events.

[0030] Step S120 in the method provided in this application embodiment includes: At each sampling moment, the real-time values ​​of the melt pressure and the screw speed are read synchronously, and the ratio of the melt pressure to the screw speed is calculated as the instantaneous calculated value of the instantaneous viscosity characterization parameter. Instantaneous calculation values ​​at multiple consecutive sampling times are stored in a circular buffer queue, wherein the circular buffer queue is a first-in-first-out storage area with a fixed length, and the earliest stored instantaneous calculation value is automatically overwritten whenever a new instantaneous calculation value is stored. Perform a normal distribution test on all instantaneous calculated values ​​in the circular cache queue, and identify and remove abnormal instantaneous calculated values ​​that deviate from the mean by more than three times the standard deviation; After removing the abnormal instantaneous calculated values, the remaining instantaneous calculated values ​​are subjected to equal-weighted moving average filtering to obtain coarse smooth values ​​of the instantaneous viscosity characterization parameters; The three consecutively acquired coarse smoothing values ​​are subjected to weighted double smoothing, and the result of the weighted double smoothing is used as the final output instantaneous viscosity characterization parameter. The three coarse smoothing values ​​are assigned a first weight, a second weight, and a third weight in the order of their acquisition time, with the second weight being greater than the first weight and the third weight.

[0031] Extract the moments of temperature difference changes where the temperature difference between the inlet and outlet of the cooling water changes abruptly within multiple time intervals from historical operating data, and extract the moments of pressure response where the melt pressure begins to respond after the moments of temperature difference changes within each time interval. Calculate the time difference between the pressure response time and the temperature difference change time in each time interval, and use the time difference as a single delay time sample. The multiple single-time delay samples are sorted, and the extreme samples that are located at the two ends of the preset ratio range after sorting are removed. The arithmetic mean of the remaining intermediate samples is calculated, and the arithmetic mean is determined as the delay influence coefficient.

[0032] Specifically, the real-time values ​​of melt pressure and screw speed are first read synchronously at each sampling moment. This is a prerequisite for ensuring that the ratio between the two has practical technological significance and can avoid calculation deviations caused by asynchronous data acquisition. Based on this, the ratio of melt pressure to screw speed is directly calculated, and this ratio is used as the instantaneous calculated value of the instantaneous viscosity characterization parameter.

[0033] Among them, the instantaneous calculated value of the instantaneous viscosity characterization parameter is the original basic data reflecting the viscosity state of the melt, which can intuitively reflect the rheological characteristics of the melt under the coupling of melt pressure and screw speed. For example, if the melt pressure is read as 15 MPa and the screw speed is 30 r / min at a certain sampling moment, the ratio of the two, 0.5 MPa·min / r, is used as the instantaneous calculated value of the instantaneous viscosity characterization parameter at that moment, thus completing the basic data calculation for a single moment.

[0034] Furthermore, the instantaneous calculated values ​​from multiple consecutive sampling moments are stored in a circular buffer queue. This queue is set as a fixed-length first-in-first-out storage area, the length of which is set according to the sampling frequency and data processing requirements of the extrusion process, ensuring that the latest sets of instantaneous calculated values ​​are always stored in the queue.

[0035] For example, when using a high-speed extrusion process, the sampling frequency is high and the data updates quickly, so the queue length is set to 50 to ensure sufficient data samples and reliable statistical tests; when using a low-speed precision extrusion process, the data changes gradually, so the queue length is set to 30 to adapt to the pace of process changes and improve data processing efficiency.

[0036] Whenever a new instantaneous calculated value is stored, the queue automatically overwrites the oldest stored instantaneous calculated value, achieving dynamic data updates and retention without manual intervention. For example, if the length of the circular cache queue is set to 50, when the 51st instantaneous calculated value is stored, the queue will automatically remove the first stored value, always maintaining 50 sets of the latest data in the queue, providing a stable sample size for subsequent testing.

[0037] Furthermore, a normal distribution test is performed on all instantaneous calculated values ​​in the circular buffer queue. The purpose of the test is to identify and eliminate abnormal instantaneous calculated values ​​caused by factors such as sensor noise, equipment mechanical vibration, and instantaneous process fluctuations during the acquisition process.

[0038] Specifically, outliers are identified by deviating from the mean by more than three standard deviations. Three standard deviations correspond to a 99.7% confidence interval for a normal distribution. Instantaneous calculated values ​​exceeding this range are statistically low-probability events and lack practical reference value in the process; therefore, they are discarded. This step effectively filters data noise, ensuring the accuracy of subsequent data processing. For example, if the mean of instantaneous calculated values ​​in a circular buffer queue is 0.48 MPa·min / r and the standard deviation is 0.02 MPa·min / r, then values ​​exceeding the range of 0.42 MPa·min / r to 0.54 MPa·min / r will be identified as outliers and discarded.

[0039] After outlier removal, the remaining instantaneous calculated values ​​are subjected to equal-weighted moving average filtering. This equal-weighted moving average filtering assigns the same weight to all instantaneous calculated values, further smoothing out minor fluctuations in the data by averaging them. This reduces the impact of random interference and makes the calculation results more closely reflect the actual trend of melt viscosity variation. The value obtained after this step is the coarse-smoothed value of the instantaneous viscosity characterization parameter. This value is more stable than the instantaneous calculated values ​​and forms the basis for subsequent secondary smoothing processing.

[0040] Furthermore, a weighted secondary smoothing process is performed on the three consecutively acquired coarse smoothing values. This step is crucial for improving the accuracy of the viscosity characterization parameters, ensuring that the final parameter values ​​more closely reflect the real-time changes in melt viscosity. Specifically, the three coarse smoothing values ​​are assigned a first weight, a second weight, and a third weight sequentially according to the order of acquisition time. The second weight is greater than both the first and third weights, and the sum of all three is 1. This weighting method ensures that the latest coarse smoothing value has the highest weight, balancing the reference value of historical data with the validity of real-time data, and avoiding parameter response lag caused by single averaging.

[0041] Next, the result after weighted secondary smoothing is used as the final output instantaneous viscosity characterization parameter, which can stably reflect the real-time rheological properties of the melt during extrusion. For example, if the first weight of three consecutive coarse smoothing values ​​is set to 0.2, the second weight to 0.6, and the third weight to 0.2, and the three coarse smoothing values ​​are 0.47 MPa·min / r, 0.49 MPa·min / r, and 0.48 MPa·min / r respectively, the final instantaneous viscosity characterization parameter of 0.484 MPa·min / r obtained after weighted calculation is the final instantaneous viscosity characterization parameter at that moment.

[0042] While calculating the instantaneous viscosity characterization parameters, characteristic time points were extracted from the historical operating data of the cable insulation extrusion process to determine the delay effect coefficient. Specifically, the time of temperature difference change when the temperature difference between the inlet and outlet of the cooling water changes abruptly within multiple time intervals was first extracted, along with the pressure response time when the melt pressure begins to respond after the time of temperature difference change within each time interval.

[0043] Among them, the step change refers to the obvious sudden change in the temperature difference between the inlet and outlet of the cooling water in a short period of time, and the pressure response moment is the starting moment when the melt pressure begins to fluctuate regularly with the temperature difference. Extracting these two characteristic moments is the basis for calculating the lag time of the cooling water temperature on the melt pressure, and it is necessary to ensure that the extracted time interval covers different combinations of extrusion process parameters to make the subsequent calculation results more universal.

[0044] For example, within a certain historical time interval, the temperature difference between the inlet and outlet of the cooling water increases dramatically at 10:05, and the melt pressure begins to respond accordingly at 10:08. In this case, 10:05 is the moment of temperature difference change, and 10:08 is the moment of pressure response.

[0045] Next, the time difference between the pressure response moment and the temperature difference change moment within each time interval is calculated, and this time difference is used as a single delay time sample. This single delay time sample reflects the lag time of the effect of the cooling water inlet and outlet temperature difference change on the melt pressure within a single time interval. Multiple time intervals yield multiple sets of single delay time samples, providing a data foundation for subsequent statistical analysis. For example, if the time difference within the above time interval is 3 minutes, this value constitutes a single delay time sample.

[0046] Then, all acquired single-time delay time samples are sorted, arranged in ascending order, and extreme samples at both ends of the sorted range are removed. The preset range is determined using the interquartile range method, specifically removing extreme samples less than the first quartile minus 1.5 times the interquartile range and greater than the third quartile plus 1.5 times the interquartile range. This method can objectively identify outliers without assuming a data distribution pattern, effectively eliminating extreme delay time samples caused by historical process anomalies, data acquisition failures, etc., while preserving the statistical representativeness of the main samples, making the subsequently calculated average value more consistent with the actual process delay patterns.

[0047] Finally, the arithmetic mean of the remaining intermediate samples after removing extreme samples is calculated, and this arithmetic mean is determined as the delay effect coefficient. This delay effect coefficient is a quantitative representation of the hysteresis effect of the temperature difference between the inlet and outlet of cooling water on the melt pressure. It can uniformly reflect the hysteresis time characteristics under different extrusion process conditions, making the discrimination results of viscosity drift events more consistent with the actual parameter variation law of the extrusion process.

[0048] S130: By combining the rate of change of the instantaneous viscosity characterization parameter within the sliding window and the delay influence coefficient, determine the occurrence time and drift amplitude of the viscosity drift event; In this embodiment of the application, in order to identify the abnormal drift state of melt viscosity and eliminate the interference of the hysteresis of cooling water temperature on the judgment result, it is necessary to analyze the variation law of viscosity parameters through a sliding window and perform hysteresis compensation in combination with the delay influence coefficient, so as to accurately determine the occurrence time and drift amplitude of viscosity drift event.

[0049] First, a fixed-length sliding window is truncated forward from the current sampling time. Instantaneous viscosity characterization parameters of all sampling times within the window are extracted. After linear fitting of these parameters, the slope of the fitted line is used as the rate of change of the instantaneous viscosity characterization parameters within the sliding window. This rate of change intuitively reflects the trend and rate of change of melt viscosity within a specific time period.

[0050] Specifically, the delay influence coefficient at the current moment is read, and the rate of change of the viscosity characterization parameter is multiplied by the delay influence coefficient to obtain the equivalent instantaneous change after hysteresis compensation, thereby eliminating the discrimination bias caused by the hysteresis effect of the temperature difference between the inlet and outlet of the cooling water on the melt pressure.

[0051] Furthermore, the equivalent instantaneous change after hysteresis compensation is compared with the calibrated change threshold. When the equivalent instantaneous change exceeds the calibrated change threshold, it is determined that a viscosity drift event has occurred at the current moment, and the current moment is accurately recorded as the moment when the viscosity drift event occurred.

[0052] Simultaneously, the absolute value of the difference between the mean instantaneous viscosity characterization parameter within the sliding window and the instantaneous viscosity characterization parameter at the current moment is calculated. This absolute value of the difference is used as the drift amplitude of the viscosity drift event, thereby quantifying the degree of abnormal drift in melt viscosity.

[0053] This step captures the abnormal changes in melt viscosity by determining the timing and quantifying the magnitude of viscosity drift events. This provides clear time points and quantitative basis for subsequent extraction of drift event-related characteristic parameters and temperature control feedforward correction, ensuring that temperature control can respond specifically to viscosity drift issues.

[0054] Step S130 in the method provided in this application embodiment includes: Using the current sampling time as a reference, a sliding window of a fixed length is truncated forward, and the instantaneous viscosity characterization parameters of all sampling times within the sliding window are extracted; Linear fitting is performed on the instantaneous viscosity characterization parameter within the sliding window, and the slope of the fitted line is taken as the rate of change of the instantaneous viscosity characterization parameter within the sliding window; Read the delay impact coefficient at the current moment, multiply the rate of change by the delay impact coefficient, and obtain the equivalent instantaneous change after hysteresis compensation; The equivalent instantaneous change is compared with the calibrated change threshold. When the equivalent instantaneous change exceeds the calibrated change threshold, it is determined that a viscosity drift event has occurred at the current moment, and the current moment is recorded as the moment when the viscosity drift event occurred. Calculate the arithmetic mean of the instantaneous viscosity characterization parameters at all sampling times within the sliding window, and take the absolute value of the difference between the arithmetic mean and the instantaneous viscosity characterization parameters at the current time as the drift amplitude of the viscosity drift event.

[0055] Specifically, a fixed-length sliding window is first extracted based on the current sampling time. The length of this sliding window is set according to the parameter variation characteristics of the extrusion process and the sampling frequency, so as to cover the shortest time interval in which the melt viscosity changes effectively, ensuring that the extracted parameter data has statistical significance.

[0056] Subsequently, the instantaneous viscosity characterization parameters of all sampling times within the window are extracted. These continuous parameter data can fully reflect the continuous change process of melt viscosity within this time period, providing a complete data source for subsequent calculation of the rate of change.

[0057] For example, if the sampling frequency of the extrusion process is 10Hz, the sliding window length is set to 30, that is, 30 instantaneous viscosity characterization parameters are captured within 3 seconds, so as to achieve complete capture of the viscosity change trend in a short period of time.

[0058] Next, linear fitting is performed on all instantaneous viscosity characterization parameters within the sliding window. Linear fitting can transform discrete parameter data into continuous trend curves by constructing a linear function model, which intuitively reflects the overall direction and rate of change of melt viscosity within the sliding window. The fitting process uses the sampling time as the independent variable and the instantaneous viscosity characterization parameters as the dependent variable, and the least squares method is used to complete the fitting calculation, so that the fitted line is as close as possible to all discrete parameter data points to reduce fitting error.

[0059] Finally, the slope of the fitted straight line is used as the rate of change of the instantaneous viscosity characterization parameter within the sliding window. The sign of the slope reflects the direction of viscosity change, and the magnitude of the absolute value of the slope reflects the rate of viscosity change.

[0060] For example, the slope of the fitted line obtained by linear fitting is 0.02 MPa·min / r·s, indicating that the melt viscosity increases at a rate of 0.02 MPa·min / r per second within the sliding window, which intuitively reflects the trend of viscosity change.

[0061] Then, the delay influence coefficient at the current moment is read. This delay influence coefficient is a quantitative indicator determined in the early stage based on the influence transmission relationship of the temperature difference between the inlet and outlet of the cooling water on the melt pressure. It can accurately reflect the lag time of the influence of the cooling water temperature change on the melt pressure.

[0062] Furthermore, the previously calculated rate of change of instantaneous viscosity characterization parameter is multiplied by the delay effect coefficient to complete the hysteresis compensation calculation of viscosity change rate, and the equivalent instantaneous change after hysteresis compensation is obtained. This step can effectively eliminate the parameter deviation caused by the hysteresis effect of cooling water temperature, and make the calculated change more consistent with the actual instantaneous change state of melt viscosity, avoiding misjudgment or omission of drift events caused by hysteresis.

[0063] For example, if the viscosity change rate within the sliding window is 0.02 MPa·min / r·s and the delay effect coefficient at the current moment is 2 s, the calculated equivalent instantaneous change is 0.04 MPa·min / r, thus achieving effective compensation for the hysteresis effect.

[0064] Subsequently, the equivalent instantaneous change after hysteresis compensation is compared with the calibrated change threshold to determine whether the melt viscosity has abnormally drifted and to identify whether a viscosity drift event has occurred.

[0065] Among them, the calibration change threshold is the core criterion for judging whether the melt viscosity has undergone abnormal drift. It is determined by statistically analyzing the equivalent instantaneous change data under historical normal operating conditions, calculating the mean of the data set and adding three times the standard deviation. According to the characteristics of normal distribution, the viscosity change under normal process conditions has a 99.7% probability of falling within this range. The equivalent instantaneous change exceeding this range is regarded as a viscosity drift signal caused by process abnormality.

[0066] For example, if the average value of the equivalent instantaneous change under normal historical operating conditions is 0.01 MPa·min / r and the standard deviation is 0.005 MPa·min / r, then the calibrated change threshold is 0.025 MPa·min / r. When the compensated equivalent instantaneous change exceeds this value, it can be determined as an abnormal viscosity drift.

[0067] In addition, when the equivalent instantaneous change exceeds the calibrated change threshold, the viscosity drift event is directly determined to have occurred at the current moment, and the current sampling time is accurately recorded as the occurrence time of the viscosity drift event, thus defining a clear time node for subsequent targeted temperature control correction.

[0068] Finally, the drift amplitude of the viscosity drift event is calculated to quantify the degree of abnormal drift in melt viscosity. Specifically, the arithmetic mean of the instantaneous viscosity characterization parameters at all sampling times within the sliding window is used as a benchmark. This arithmetic mean represents the stable benchmark state of melt viscosity before the drift event occurs. The absolute value of the difference between this benchmark value and the instantaneous viscosity characterization parameter at the current time is then calculated. This absolute value of the difference is taken as the drift amplitude of the viscosity drift event. The magnitude of this value directly reflects the degree to which the melt viscosity deviates from the normal stable state, providing important quantitative characteristics for the feedforward correction amount of the subsequent coupled compensation model output adaptation.

[0069] For example, the average instantaneous viscosity characterization parameter within the sliding window is 0.5 MPa·min / r, and the instantaneous viscosity characterization parameter at the current moment is 0.65 MPa·min / r. The absolute value of the difference between the two, 0.15 MPa·min / r, is the drift amplitude of this viscosity drift event, which clearly quantifies the degree of abnormal deviation in viscosity.

[0070] S140: In response to the viscosity drift event, extract the melt temperature change rate within the first time window before the event occurs, and the melt pressure fluctuation amplitude and the change in cooling water inlet and outlet temperature difference within the second time window after the event occurs. In this embodiment of the application, in order to mine the associated process parameter features of viscosity drift events and obtain the input data of the coupled compensation model, it is necessary to extract the relevant parameter change features by capturing a time window around the moment when the drift event occurs, so as to support the accurate prediction of the subsequent temperature control feedforward correction.

[0071] First, a first time window is extracted backward from the moment of occurrence of the viscosity drift event. The melt temperature at all sampling moments within the window is extracted and linearly fitted. The slope of the fitted line is taken as the melt temperature change rate. This rate reflects the trend of melt temperature change before the drift occurs and is the core early feature of the associated drift event.

[0072] Specifically, a second time window is extracted based on the occurrence time of the viscosity drift event. The melt pressure at all sampling times within the window is extracted, and the difference between the maximum and minimum melt pressure is calculated. This difference is used as the melt pressure fluctuation amplitude to intuitively quantify the degree of melt pressure fluctuation after the drift occurs.

[0073] Furthermore, the temperature difference between the inlet and outlet of the cooling water at the beginning and end times within the second time window is extracted, and the difference between the two time points is calculated. This difference is used as the change in the temperature difference between the inlet and outlet of the cooling water to reflect the change characteristics of the cooling water heat exchange state after the drift occurs.

[0074] The lengths of the first and second time windows are determined based on the delay influence coefficient. The length of the first time window is a multiple of the delay influence coefficient, and the length of the second time window is a multiple of the delay influence coefficient. The first multiple is less than the second multiple, which adapts to the response rhythm of changes in process parameters.

[0075] This step extracts key process parameter change features before and after the viscosity drift event, providing a targeted input feature set for the coupled compensation model. This allows the model to output appropriate feedforward corrections based on the actual process parameter change patterns, ensuring the accuracy and adaptability of subsequent temperature control adjustments.

[0076] Step S140 in the method provided in this application embodiment includes: Based on the occurrence time of the viscosity drift event, a first time window is extracted backward, and the melt temperature at all sampling times within the first time window is extracted. The melt temperature within the first time window is linearly fitted, and the slope of the fitted line is taken as the rate of change of the melt temperature. Based on the occurrence time of the viscosity drift event, a second time window is extracted, and the melt pressure at all sampling times within the second time window is extracted. Calculate the difference between the maximum and minimum values ​​of the melt pressure within the second time window, and use the difference as the melt pressure fluctuation amplitude; Extract the temperature difference between the inlet and outlet of the cooling water at the start time and the temperature difference between the inlet and outlet of the cooling water at the end time within the second time window, calculate the difference between the temperature difference at the end time and the temperature difference at the start time, and use the difference as the change in the temperature difference between the inlet and outlet of the cooling water. The lengths of the first time window and the second time window are determined based on the delay impact coefficient. The length of the first time window is set to a first multiple of the delay impact coefficient, and the length of the second time window is set to a second multiple of the delay impact coefficient. The first multiple is less than the second multiple.

[0077] Specifically, a first time window is extracted based on the moment when the viscosity drift event occurs. This first time window is the range of process parameter characteristics collected before the drift event occurs, used to capture the continuous change trend of melt temperature before the drift occurs, providing data support for analyzing the early causes of the drift event.

[0078] Subsequently, melt temperature data from all sampling moments within the first time window are extracted. The obtained continuous temperature data completely records the dynamic change process of the melt temperature before the drift occurs, serving as the fundamental data source for calculating the rate of temperature change. For example, if the viscosity drift event occurs at the 100th sampling moment, and the first time window covers sampling moments 70 to 99, then all melt temperature data corresponding to these 30 sampling moments are extracted, achieving a complete capture of the temperature change process before the drift.

[0079] Next, linear fitting is performed on all melt temperature data extracted within the first time window to fit the trend curve of temperature change over time and calculate the melt temperature change rate.

[0080] Specifically, the fitting process uses the sampling time as the independent variable and the melt temperature as the dependent variable, and uses the least squares method to complete the fitting calculation, so that the fitted line is as close as possible to all discrete temperature data points, minimizing the fitting error. Finally, the slope of the fitted line is used as the rate of change of the melt temperature.

[0081] The positive or negative value of the slope reflects the rising or falling trend of the melt temperature before the drift occurs, while the absolute value of the slope reflects the rate of temperature change. This rate is a core pre-event parameter characteristic associated with viscosity drift events and can reflect the intrinsic process correlation between temperature change and viscosity drift.

[0082] For example, the slope obtained by linearly fitting the melt temperature data within the first time window is 0.8℃ / s, indicating that the melt temperature continued to rise at a rate of 0.8℃ per second before the drift occurred, clearly showing the trend of temperature change.

[0083] Subsequently, a second time window is extracted based on the occurrence time of the viscosity drift event. This second time window is the data acquisition interval for process parameter response characteristics after the drift event, used to observe the complete dynamic response process of melt pressure and cooling water inlet and outlet temperature difference after the drift occurs, providing data support for quantifying the process parameter fluctuations caused by the drift event. The second time window has a longer time span than the first time window, fully covering the fluctuation and adjustment phases of parameters after the drift.

[0084] Next, melt pressure data at all sampling times within the second time window are extracted. These data completely record the fluctuation and change process of melt pressure after the drift occurs, and serve as the basis for calculating the amplitude of melt pressure fluctuation.

[0085] For example, if the viscosity drift event occurs at the 100th sampling time, the second time window covers the 100th to 159th sampling times, and all melt pressure data corresponding to these 60 sampling times are extracted to achieve complete capture of the pressure fluctuation process after drift.

[0086] After extracting all melt pressure data within the second time window, extreme value analysis was performed on this data set to identify the maximum and minimum values. The difference between these values ​​was then calculated and used as the melt pressure fluctuation amplitude. This value directly quantifies the degree of melt pressure fluctuation after a drift event; a larger value indicates a more significant impact of the drift event on the melt pressure. For example, if the maximum melt pressure within the second time window is 18 MPa and the minimum is 12 MPa, the difference of 6 MPa represents the melt pressure fluctuation amplitude caused by this viscosity drift event, clearly quantifying the degree of pressure fluctuation.

[0087] Simultaneously, the cooling water inlet and outlet temperature difference data at the start and end times within the second time window are extracted. The temperature difference at the start time represents the cooling water heat exchange state at the initial stage of the drift event, while the temperature difference at the end time represents the cooling water heat exchange state after a certain period of adjustment following the drift event. By calculating the difference between the two temperature differences, this difference is used as the change in the cooling water inlet and outlet temperature difference. This value can reflect the trend and degree of change in the cooling water heat exchange state after the drift event, and embodies the dynamic response of the cooling water system to the viscosity drift event. It is an important cooling system characteristic parameter required by the coupled compensation model.

[0088] For example, the temperature difference between the inlet and outlet of the cooling water at the beginning of the second time window is 8°C and the temperature difference at the end of the second time window is 12°C. The difference of 4°C is the change in the temperature difference between the inlet and outlet of the cooling water, which intuitively reflects the change in the heat exchange temperature difference of the cooling water after drifting.

[0089] The lengths of the first and second time windows are not fixed values, but are dynamically determined based on the delay impact coefficient calculated in the previous period. The length of the first time window is set as a multiple of the delay impact coefficient, and the length of the second time window is set as a multiple of the delay impact coefficient, with the first multiple always being less than the second multiple.

[0090] The first multiplier is set to 1 to 1.5 times. This value is chosen because the first time window is used to capture the trend of melt temperature change before the event occurs. If the window is too large, historical temperature data unrelated to the drift event will be mixed in, resulting in distortion of the temperature change rate calculation. If the window is too small, it will be difficult to accumulate enough temperature data and stably fit an accurate change slope. Therefore, a value slightly larger than one delay period is chosen to ensure both the correlation of the data and the stability of the fitting.

[0091] In addition, the second multiple is set to 2 to 3 times. This value is because the second time window is used to observe the complete response process of melt pressure and cooling water temperature after the event occurs. The window needs to fully cover the dynamic adjustment stage after the disturbance. Taking two to three delay periods can ensure that the peak value of parameter fluctuation and the initial steady-state recovery process are captured, and fully reflect the parameter response characteristics after drift.

[0092] For example, if the delay effect coefficient is 10 seconds, the first multiple is 1.2 and the second multiple is 2.5, then the length of the first time window is 12 seconds and the length of the second time window is 25 seconds, thereby achieving accurate acquisition of different parameter characteristics before and after drift.

[0093] S150: Input the melt temperature change rate, melt pressure fluctuation amplitude and cooling water inlet and outlet temperature difference into the pre-trained coupled compensation model, and output the feedforward correction amount for the current temperature control setpoint; In this embodiment of the application, in order to quickly output a temperature control feedforward correction amount that adapts to the current operating conditions based on the associated process parameter characteristics of viscosity drift events, and to avoid the lag problem of traditional passive adjustment, the extracted multi-dimensional parameter features need to be input into the pre-trained coupled compensation model. The model outputs an accurate feedforward correction amount through intelligent calculation, so as to realize the active predictive adjustment of melt temperature and quickly suppress the process parameter fluctuations caused by viscosity drift.

[0094] In the method provided in this application embodiment, the pre-training step of the coupling compensation model includes: Multiple viscosity drift events are extracted from historical operating data. For each viscosity drift event, the corresponding melt temperature change rate, melt pressure fluctuation amplitude, and cooling water inlet and outlet temperature difference change are obtained as the sample input feature set. Each viscosity drift event is labeled with the optimal feedforward correction amount determined after process testing, which is used as the sample output label set. An initial coupling compensation model is constructed based on machine learning. The number of input layer nodes of the initial coupling compensation model corresponds to the number of features in the sample input feature set, and the number of output layer nodes corresponds to the output dimension of the feedforward correction. The sample input feature set is used as the model input, and the optimal feedforward correction value corresponding to the sample output label set is used as the supervision output. The training objective is to minimize the error between the model's predicted feedforward correction value and the supervision output. The model parameters are iteratively optimized, and training stops when the prediction error on the validation set converges to the target range, thus obtaining the trained coupling compensation model.

[0095] Specifically, the first step is to extract multiple actual viscosity drift events from historical operational data of cable insulation extrusion production. This historical operational data must cover different extrusion process types, equipment operating states, raw material characteristics, and production environments to ensure the comprehensiveness and representativeness of the extracted viscosity drift events. This includes both minor, slight viscosity drifts and significant, severe viscosity drifts, while also covering drift scenarios under different process conditions such as high-speed extrusion and low-speed precision extrusion.

[0096] After extracting each viscosity drift event, the melt temperature change rate, melt pressure fluctuation amplitude, and cooling water inlet and outlet temperature difference change corresponding to that drift event are retrieved from historical data. These three types of parameters are core process features directly related to viscosity drift and can fully characterize the occurrence state and impact of the viscosity drift. The three types of parameters are integrated into a set of data as the sample input feature set corresponding to the drift event. Multiple drift events form multiple sets of sample input feature sets, providing a sufficient feature data foundation for model training.

[0097] For example, a viscosity drift event during the extrusion process of a cross-linked polyethylene cable insulation layer is extracted from historical data. The melt temperature change rate of this event is 0.6℃ / s, the melt pressure fluctuation amplitude is 5MPa, and the temperature difference between the inlet and outlet of the cooling water is 4℃. These three sets of values ​​are then integrated into the sample input feature set of this event.

[0098] Furthermore, a corresponding sample output label set is assigned to each extracted viscosity drift event. This label set represents the optimal feedforward correction amount determined after process testing. The process testing must be conducted under the exact same extrusion process parameters, equipment status, and raw material characteristics as the specific drift event. By simulating the occurrence of the viscosity drift event, the values ​​of core temperature control parameters such as the heating power correction ratio and cooling water flow rate correction ratio are adjusted one by one. The recovery speed and stability of melt viscosity under different correction amounts are observed. Finally, the temperature control parameter correction combination that can quickly suppress viscosity drift and restore melt temperature and pressure to a stable process range is determined. This combination is the optimal feedforward correction amount corresponding to the drift event, and it is used as the sample output label set, corresponding one-to-one with the sample input feature set.

[0099] For example, for the drift events of the melt temperature change rate of 0.6℃ / s, melt pressure fluctuation amplitude of 5MPa, and cooling water inlet and outlet temperature difference change of 4℃, the optimal feedforward correction amount was determined by process test to be a 13% reduction in heating power and a 19% increase in cooling water flow rate. This combination of correction ratios was then labeled as the output label corresponding to the sample input feature set.

[0100] After organizing the sample input feature set and output label set, an initial coupling compensation model is built based on machine learning algorithms. The model architecture needs to match the dimensional features of the sample features and the output labels.

[0101] The number of nodes in the input layer of the initial coupling compensation model is consistent with the number of features in the sample input feature set. Since the sample input feature set includes three types of features: melt temperature change rate, melt pressure fluctuation amplitude, and cooling water inlet and outlet temperature difference change, the input layer is set with 3 nodes, which respectively receive the three types of feature parameters.

[0102] In addition, the number of output layer nodes is consistent with the output dimension of the feedforward correction. If the feedforward correction includes two output dimensions, namely the heating power correction ratio and the cooling water flow rate correction ratio, the output layer is set with two nodes, which correspond to the output of the two types of correction values ​​respectively.

[0103] Meanwhile, based on the complexity of the extrusion process data, hidden layers are appropriately set between the input and output layers. The number of hidden layers and nodes is adjusted according to the scale of the sample data. If the sample data is large and the drift scenario is complex, 2-3 hidden layers can be set, with 16-32 nodes per layer; if the sample data is small, it can be simplified to 1 hidden layer with 8-16 nodes. Through reasonable architecture construction, the model can fully learn the correlation between features and labels in the samples, while avoiding overfitting or slow convergence caused by an overly complex architecture.

[0104] For example, for a scenario that includes more than a thousand drift samples and two types of corrections, namely output heating power and cooling water flow rate, the initial coupling compensation model is set with 3 input layer nodes, 2 output layer nodes, and 2 hidden layers in the middle, with 24 nodes in each layer, forming a model network architecture adapted to the scenario.

[0105] Subsequently, the compiled sample input feature set and sample output label set are divided into training set, validation set, and test set according to a preset ratio. Specifically, 70% of the samples are typically allocated to the training set for iterative optimization of model parameters, 20% to the validation set for monitoring model performance during training, and 10% to the test set for final verification of the model's generalization ability. During the partitioning process, it is ensured that various drift scenarios are evenly distributed across different sets to avoid model training bias caused by uneven sample distribution.

[0106] Furthermore, the sample input feature set of the training set is used as the model input and fed into the constructed initial coupling compensation model. The model performs calculations based on the initial parameters and outputs the predicted feedforward correction. The model's predicted feedforward correction is then compared with the optimal feedforward correction corresponding to the sample output label set. With minimizing the error between the two as the core training objective, the gradient descent algorithm is used to iteratively update and optimize the model's internal parameters such as weights and biases. By continuously adjusting the parameters, the model's prediction error is reduced, allowing the model's output to gradually approach the actual optimal feedforward correction.

[0107] Meanwhile, during the parameter iterative optimization process, after each round of training, the model is tested using samples from the validation set, the error between the model's predicted value and the true label value on the validation set is calculated, and the changing trend of the validation set error is monitored in real time.

[0108] When the prediction error on the validation set no longer decreases significantly after several rounds and converges to the preset target range, and the prediction error on the test set also meets the process requirements, the iterative optimization of the model parameters is stopped. At this point, the model has fully learned the intrinsic process correlation between different viscosity drift characteristic parameters and the optimal feedforward correction amount, and has the ability to accurately output the appropriate feedforward correction amount based on the input characteristic parameters. Thus, the coupled compensation model that has been trained is obtained.

[0109] For example, the target range of the prediction error on the preset model validation set is within 5%. After multiple rounds of training, the prediction error on the validation set stabilizes at 4.2%, and there is no significant decrease after 10 consecutive rounds of training. The prediction error on the test set is 4.5%, which meets the process requirements. At this point, training is stopped, and the pre-training of the coupling compensation model is completed.

[0110] Throughout the model training process, it is necessary to periodically verify the training effect and adjust the parameters. If the prediction error on the validation set shows an upward trend, it indicates that the model is overfitting and the training parameters need to be adjusted in time, such as reducing the model's learning rate, increasing the training batch size, or augmenting the sample data. If the prediction error of the model is consistently high in a certain type of drift scenario, it indicates that the sample data for that type of scenario is insufficient. It is necessary to supplement the historical sample data for that type of drift scenario and reintegrate it into the training set for training to ensure that the model can maintain stable prediction accuracy under various extrusion process conditions and various viscosity drift scenarios.

[0111] For example, during training, if the validation set prediction error is found to have increased from 3.8% to 6.5%, the model learning rate is promptly adjusted from 0.001 to 0.0005, and the training batch size is adjusted from 16 batches / batch to 32 batches / batch. After the adjustment, the model's validation set prediction error gradually decreases and tends to stabilize. If the model prediction error for viscosity drift events under high-speed extrusion processes is found to be consistently above 7%, 50 additional drift sample data under high-speed extrusion processes are added to the training set. After retraining, the prediction error for this type of scenario is reduced to below 4%, ensuring the model's generalization ability and prediction stability.

[0112] Finally, the pre-trained coupled compensation model, completed through the above steps, can capture the process correlation between viscosity drift correlation characteristics and the optimal feedforward correction amount. In actual production, after receiving characteristic parameters such as melt temperature change rate, melt pressure fluctuation amplitude, and cooling water inlet and outlet temperature difference change amount, it can quickly calculate and output the appropriate feedforward correction amount, ensuring the timeliness and accuracy of temperature control feedforward correction.

[0113] S160: Based on the feedforward correction amount, the temperature setpoint of each heating zone during the extrusion process of the cable insulation layer is corrected in layers, and the cooling water flow rate is adjusted synchronously to obtain a stable melt temperature during the viscosity drift process.

[0114] In this embodiment of the application, in order to quickly eliminate the melt state fluctuations caused by viscosity drift and restore the melt temperature to the process stability range, it is necessary to coordinately adjust the temperature of each heating zone and the flow rate of cooling water according to the feedforward correction amount, so as to achieve rapid stabilization of melt temperature and ensure the continuity and stability of the extrusion molding process.

[0115] First, the feedforward correction value of the output is analyzed, and the temperature correction component corresponding to each heating zone and the flow regulation component of the cooling water circuit are extracted from it. These components are the quantitative adjustment basis to adapt to the current viscosity drift condition, and directly guide the stratified temperature adjustment of the temperature control system and the flow adjustment of the cooling system.

[0116] Specifically, for each heating zone in the extrusion process, the current temperature setpoint of each heating zone is first read, and the temperature setpoint is calculated with the corresponding temperature correction component to obtain the updated temperature setpoint of the corresponding heating zone. The updated temperature setpoint is then sent to the temperature controller of the corresponding heating zone in real time for execution, so as to realize the layered temperature control correction of each heating zone.

[0117] Furthermore, the current real-time flow value of the cooling water circuit is read, and the real-time flow value is calculated with the extracted cooling water flow adjustment component to obtain the updated cooling water flow setting value. The setting value is then sent to the flow controller of the cooling water circuit for execution, and the flow adjustment of the cooling system is completed synchronously, forming a coordinated control with the temperature correction of the heating zone.

[0118] Meanwhile, the real-time melt temperature at the extruder head is continuously collected, and this real-time melt temperature is compared with the average melt temperature during historical stable periods. The temperature recovery deviation between the two is calculated, and this deviation directly reflects the degree to which the melt temperature recovers to a stable state.

[0119] Based on the standard deviation of melt temperature fluctuations during historical stable periods and the previously determined delay effect coefficient, a deviation tolerance threshold is calculated. This deviation tolerance threshold is the core criterion for determining whether the melt temperature has returned to stability, adapting to the temperature fluctuation characteristics and hysteresis effects of the process.

[0120] When the calculated temperature recovery deviation is less than the deviation tolerance threshold, it is determined that the influence of viscosity drift has been effectively eliminated, and a stable melt temperature is obtained during the viscosity drift process.

[0121] This step achieves coordinated temperature control of heating and cooling by stratified correction of the temperature of each heating zone and synchronous adjustment of the cooling water flow rate. It suppresses melt temperature fluctuations caused by viscosity drift and allows the melt temperature to be accurately restored to the stable range of the process, providing temperature assurance for the continuous and stable extrusion molding of cable insulation layers.

[0122] Step S160 in the method provided in this application embodiment includes: The feedforward correction is analyzed to extract the temperature correction component and cooling water flow rate adjustment component corresponding to each heating zone. For each heating zone, the current temperature setting value is read, and the temperature setting value is added to the corresponding temperature correction component to obtain the updated temperature setting value for the corresponding heating zone. The updated temperature setpoint is sent to the temperature controller of the corresponding heating zone for execution. Read the real-time flow value of the current cooling water circuit, add the real-time flow value to the cooling water flow adjustment component, and obtain the updated cooling water flow setting value; The updated cooling water flow rate setting is sent to the cooling water circuit flow controller for execution. The real-time melt temperature at the extruder head is continuously collected, and the real-time melt temperature is compared with the average melt temperature during the historical stable period to calculate the temperature recovery deviation. Based on the standard deviation of melt temperature fluctuation during the historical stable period and the delay effect coefficient, the deviation tolerance threshold is calculated. When the temperature recovery deviation is less than the deviation tolerance threshold, it is determined that the stable melt temperature in the viscosity drift process has been reached.

[0123] First, the feedforward correction output of the coupled compensation model is structured and analyzed to extract the temperature correction component corresponding to each heating zone and the flow regulation component of the cooling water circuit, thereby providing a quantitative basis for the temperature stratification correction of each heating zone and the synchronous regulation of cooling water flow.

[0124] Specifically, the parsing process first reads the preset data structure of the feedforward correction, which includes dedicated temperature correction and cooling water flow correction fields, each storing corresponding adjustment parameter information. Then, it parses the temperature correction sequence from the temperature correction field, arranged according to the physical location of the heating zones. Each element in the temperature correction sequence uniquely corresponds to a temperature correction component for a specific heating zone.

[0125] Subsequently, following the ascending order of heating zone numbers, the temperature correction components for the first heating zone, the second heating zone, and so on up to the Nth heating zone are extracted sequentially from the temperature correction sequence. Here, N represents the total number of heating zones in the extruder. Finally, the cooling water flow rate adjustment component is directly extracted from the cooling water flow rate correction field, completing the full analysis of the feedforward correction.

[0126] For example, an extruder has three heating zones. When analyzing the feedforward correction, the temperature correction sequence ordered by zones 1, 2, and 3 is extracted from the temperature correction field as -2℃, -1.5℃, and -1℃, and the flow rate adjustment component is extracted from the cooling water flow rate correction field as 5L / min. This completes the analysis of the correction.

[0127] Furthermore, the temperature setpoints for each heating zone of the extruder are modified in layers to achieve temperature control adjustment for each heating zone and adapt to the process temperature requirements under viscosity drift.

[0128] Specifically, firstly, the parameter reading unit of the temperature control device retrieves the current process temperature setting value of each heating zone one by one, and then the current temperature setting value of each heating zone is summed with the corresponding temperature correction component. The result of the calculation is the updated temperature setting value of the heating zone.

[0129] The temperature setpoint is a precise temperature control value adapted to the current viscosity drift conditions. Subsequently, the updated temperature setpoints for each heating zone are sent to the temperature controllers of the corresponding heating zones. The temperature controllers then perform heating regulation operations according to the new setpoints, achieving stratified and precise temperature correction for each heating zone, so that the temperature adjustment of different heating zones is adapted to the influence of viscosity drift.

[0130] For example, the current setting of the first heating zone of the extruder is 180°C, and the corresponding temperature correction component is -2°C. After calculation and updating, the setting is 178°C. After this value is sent to the temperature controller of the first heating zone, the controller immediately performs temperature control adjustment according to 178°C.

[0131] While completing the correction and distribution of temperature setpoints for each heating zone, the flow rate of the cooling water circuit is simultaneously adjusted to achieve coordinated temperature control of heating and cooling, and quickly eliminate melt temperature fluctuations caused by viscosity drift.

[0132] Specifically, the real-time flow rate of the cooling water circuit is first read by a flow sensor, and then this real-time flow rate is summed with the analyzed cooling water flow rate adjustment component to obtain an updated cooling water flow rate setpoint. This cooling water flow rate setpoint, together with the heating zone temperature setpoint, forms a coordinated control strategy that can help mitigate temperature fluctuations caused by melt viscosity drift by adjusting the cooling water heat exchange efficiency.

[0133] Subsequently, the updated cooling water flow rate setpoint is sent to the flow controller in the cooling water circuit. The flow controller then performs flow regulation by adjusting the speed of the variable frequency water pump, thereby achieving coordinated temperature control for heating and cooling. For example, if the current real-time flow rate of the cooling water circuit is 30 L / min, the corresponding flow regulation component is 5 L / min, and the calculated updated setpoint is 35 L / min, the flow controller will immediately adjust the cooling water flow rate to 35 L / min upon receiving the instruction.

[0134] Furthermore, after issuing the adjustment commands for the heating zone temperature and cooling water flow rate, the real-time melt temperature at the extruder head is continuously collected. The extruder head is a key position for melt extrusion molding, and the melt temperature here directly reflects the actual effect of temperature control. The collection process maintains the same sampling frequency as before to ensure the continuity and real-time nature of the data.

[0135] Simultaneously, the average melt temperature during the historical stable period when no viscosity drift event occurred is retrieved from the historical operating data of the extrusion process. This average melt temperature serves as the process stability benchmark for the melt temperature. The difference between the real-time collected melt temperature and this historical stable average temperature is calculated, and the result is the temperature recovery deviation. The absolute value of this temperature recovery deviation directly reflects the degree to which the melt temperature recovers to a stable state; the smaller the deviation, the closer the melt temperature is to the process stability range.

[0136] For example, the average melt temperature during the historical stable period is 175℃, and the real-time melt temperature of the die head collected at a certain moment is 176.2℃. The calculated temperature recovery deviation is 1.2℃.

[0137] Subsequently, based on the standard deviation of melt temperature fluctuations during historical stable periods and the previously determined delay influence coefficient, the deviation tolerance threshold for determining whether the temperature is stable is calculated, thereby establishing a temperature stability judgment standard that fits the actual process and accurately determining whether the melt temperature has recovered to the stable state during the viscosity drift process.

[0138] The method provided in this application embodiment calculates the deviation tolerance threshold based on the standard deviation of melt temperature fluctuation during the historical stable period and the delay influence coefficient, including: Extract stable running segments from historical operating data where multiple viscosity-free drift events occur, and set the duration of the stable running segments to the third multiple of the delay influence coefficient; For each stable operating period, the melt temperature at all sampling times within the corresponding period is collected and the arithmetic mean is calculated as the historical stable temperature average for the corresponding period. For each stable operating period, the standard deviation of melt temperature fluctuation for the corresponding period is calculated based on the difference between the melt temperature at all sampling times within the corresponding period and the average historical stable temperature. The average standard deviation of melt temperature fluctuation is obtained by arithmetically averaging the standard deviations of the temperature fluctuations over multiple stable operating periods. The deviation tolerance threshold is obtained by multiplying the average standard deviation of fluctuation by the value of the delay effect coefficient.

[0139] Specifically, we first screened and extracted several stable operating periods from historical operating data where no viscosity drift events occurred. These periods correspond to the normal production stage where the cable insulation extrusion process is stable, the melt state is uniform, and there are no abnormal disturbances. They can truly reflect the temperature fluctuation characteristics of the equipment under ideal operating conditions.

[0140] During the selection process, the duration of each stable operating period was uniformly set to the third multiple of the delay influence coefficient to ensure that the time period was long enough to cover the complete temperature response cycle and to avoid statistical bias due to excessively short time periods. For example, if the delay influence coefficient is 10 seconds and the third multiple is 2, then the duration of the stable operating period is set to 20 seconds to ensure that the temperature data has sufficient statistical representativeness.

[0141] Next, for each stable operating period, melt temperature data at all sampling times within that period are collected. These data continuously and completely record the real-time changes in melt temperature under normal process conditions.

[0142] Furthermore, the collected melt temperature data are averaged arithmetically, and the result is used as the historical stable temperature average for that period. This average represents the ideal benchmark value of melt temperature when the process is stable, effectively eliminating the impact of random fluctuations. For example, if 20 melt temperature sampling points are collected during a stable operating period, and their arithmetic mean is 175℃, this value is the historical stable temperature average for that period.

[0143] After obtaining the historical average stable temperature, the standard deviation of melt temperature fluctuation during that period is further calculated. Specifically, the difference between the melt temperature and the historical average stable temperature is calculated for each sampling moment. The squares of each difference are then summed, divided by the number of sampling points in that period, and finally the square root is taken to obtain the standard deviation of melt temperature fluctuation for that period. The standard deviation of fluctuation quantifies the natural dispersion of melt temperature during stable operation; a smaller value indicates more stable temperature control.

[0144] In addition, since the standard deviation of fluctuations in a single stable period is random, it is necessary to perform an arithmetic mean of the standard deviations of melt temperature fluctuations calculated for multiple stable operating periods to obtain the average standard deviation of fluctuations.

[0145] The average standard deviation of fluctuation integrates multiple sets of normal operating condition data, which can more comprehensively reflect the inherent temperature fluctuation level of the extrusion system under drift-free conditions, thus avoiding interference from individual abnormal data with the threshold. For example, if five stable operating periods are selected, their fluctuation standard deviations are 0.8℃, 0.7℃, 0.9℃, 0.8℃, and 0.8℃, respectively, and the arithmetic mean yields an average fluctuation standard deviation of 0.8℃.

[0146] Finally, the average standard deviation of the fluctuation is multiplied by the value of the delay effect coefficient to obtain the final deviation tolerance threshold. The delay effect coefficient reflects the hysteresis characteristic (in seconds) of the effect of cooling water temperature changes on melt pressure. Its value is used here as a dimensionless gain factor in the deviation tolerance threshold calculation, allowing the judgment standard to adapt to the process hysteresis characteristics, thus avoiding misjudgment of stability before the temperature has fully recovered due to the hysteresis effect. For example, if the average standard deviation of the fluctuation is 0.8℃ and the delay effect coefficient is 2 seconds, then its value of 2 is used, resulting in a deviation tolerance threshold of 1.6℃.

[0147] Furthermore, after collecting the melt temperature at the extruder head position in real time and calculating the temperature recovery deviation, the temperature recovery deviation is compared with the calculated deviation tolerance threshold in real time.

[0148] Specifically, when the temperature recovery deviation is less than the deviation tolerance threshold, it indicates that the current melt temperature has returned to the stable fluctuation range allowed by the process, and the temperature disturbance caused by viscosity drift has been effectively eliminated by the feedforward correction measures. At this point, it is determined that the system has obtained a stable melt temperature during the viscosity drift process. For example, if the temperature recovery deviation is 1.2℃ and the deviation tolerance threshold is 1.6℃, the former is less than the latter, so it can be determined that the melt temperature has stabilized, and the extrusion process can continue to operate smoothly.

[0149] Conversely, if the temperature recovery deviation is greater than or equal to the deviation tolerance threshold, it indicates that the current melt temperature is still outside the stable fluctuation range allowed by the process, the temperature disturbance caused by viscosity drift has not been completely eliminated, and the control effect of the feedforward correction measures has not yet achieved the expected results. In this case, the melt temperature is not considered stable, and the current heating zone temperature correction and cooling water flow regulation strategy must be maintained. The melt temperature at the extruder head must continue to be collected in real time, the temperature recovery deviation must be calculated, and the deviation must be repeatedly compared with the threshold until the temperature recovery deviation is less than the deviation tolerance threshold. For example, if the temperature recovery deviation is 1.8℃ and the deviation tolerance threshold is 1.6℃, the former is greater than the latter, and the existing control strategy must be maintained, with continuous monitoring and adjustment until the temperature recovery deviation drops below 1.6℃.

[0150] Ultimately, through the aforementioned closed-loop determination process, the stable point of the melt temperature can be accurately captured, and the effectiveness of the feedforward correction can be confirmed in a timely manner. This avoids process fluctuations caused by premature determination of stability, as well as energy waste and product quality deviations caused by excessive control. It provides a guarantee for the stable operation of the cable insulation extrusion process, ensuring that the extruded cable insulation layer has uniform thickness and good physical properties.

[0151] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects: This application proposes a multi-parameter fusion-based temperature control method for cable insulation extrusion. First, it collects real-time multi-source time-series data during the extrusion process, including melt pressure, screw speed, cooling water inlet and outlet temperature difference, and melt temperature. Instantaneous viscosity characterization parameters are calculated based on melt pressure and screw speed, and a delay influence coefficient is determined based on the influence transmission relationship between cooling water heat transfer and melt pressure. Then, combining the rate of change of instantaneous viscosity characterization parameters within a sliding window with the delay influence coefficient, the occurrence time and amplitude of viscosity drift events are determined. Subsequently, in response to viscosity drift events, time windows before and after the event are extracted to obtain multi-dimensional features such as the melt temperature change rate, melt pressure fluctuation amplitude, and cooling water inlet and outlet temperature difference change. These features are then input into a pre-trained coupled compensation model, which infers and outputs a feedforward correction amount adapted to the current operating conditions. Finally, the feedforward correction amount is analyzed to perform layered corrections on the temperature setpoints of each heating zone, synchronously adjusting the cooling water flow rate. Based on historical stable data and the delay influence coefficient, a deviation tolerance threshold is calculated. By comparing the temperature recovery deviation in real time, a closed-loop determination of the melt temperature stability state is completed, achieving precise temperature control and rapid stabilization under viscosity drift disturbances.

[0152] The method provided in this application, through the technical solution of "multi-source data acquisition - viscosity parameter characterization - drift event identification - multi-dimensional feature extraction - intelligent model reasoning - hierarchical collaborative adjustment - closed-loop stability determination", solves the problems of lag in adjustment, slow response, weak coordination between heating and cooling, and difficulty in actively suppressing viscosity drift in the traditional temperature control of cable insulation extrusion. It improves the temperature control accuracy, process stability and consistency of insulation layer forming quality in the extrusion process.

[0153] Example 2, as shown in the appendix Figure 2 As shown, based on the inventive concept of the cable insulation extrusion temperature control method based on multi-parameter fusion provided in Embodiment 1, this application also provides a cable insulation extrusion temperature control system based on multi-parameter fusion, specifically including: The multi-source data acquisition module 01 is used to acquire multi-source time-series data in real time during the extrusion process of cable insulation layer. The multi-source time-series data includes at least melt pressure, screw speed, cooling water inlet and outlet temperature difference, and melt temperature. The feature parameter calculation module 02 is used to calculate the instantaneous viscosity characterization parameters based on the melt pressure and screw speed, and at the same time determine the delay influence coefficient based on the influence transmission relationship of the temperature difference between the inlet and outlet of the cooling water on the melt pressure. The viscosity event discrimination module 03 is used to determine the occurrence time and drift amplitude of the viscosity drift event by combining the rate of change of the instantaneous viscosity characterization parameter within the sliding window and the delay influence coefficient. Event feature extraction module 04 is used to respond to the viscosity drift event and extract the melt temperature change rate in the first time window before the event occurs, as well as the melt pressure fluctuation amplitude and the change in cooling water inlet and outlet temperature difference in the second time window after the event occurs. The model prediction correction module 05 is used to input the melt temperature change rate, melt pressure fluctuation amplitude and cooling water inlet and outlet temperature difference into the pre-trained coupled compensation model, and output the feedforward correction amount for the current temperature control setpoint. The collaborative execution control module 06 is used to perform layered correction of the temperature setpoint of each heating zone during the extrusion process of the cable insulation layer according to the feedforward correction amount, and to synchronously adjust the cooling water flow rate to obtain a stable melt temperature during the viscosity drift process.

[0154] In one embodiment, the multi-source data acquisition module 01 is further configured to: acquire the melt pressure in real time through a melt pressure sensor installed at the extruder head; acquire the screw speed in real time through an encoder of the main drive motor of the extruder; acquire and calculate the temperature difference between the inlet and outlet of the cooling water in real time through a first temperature sensor installed at the inlet of the cooling water circuit and a second temperature sensor installed at the outlet of the cooling water circuit; and acquire the melt temperature in real time through an insertion thermocouple at the melt flow channel of the extruder head.

[0155] In one embodiment, the feature parameter calculation module 02 is further configured to: synchronously read the real-time values ​​of the melt pressure and the screw speed at each sampling time, calculate the ratio of the melt pressure to the screw speed as the instantaneous calculated value of the instantaneous viscosity characterization parameter; store the instantaneous calculated values ​​of multiple consecutive sampling times into a circular buffer queue, wherein the circular buffer queue is a first-in-first-out storage area with a fixed length, and automatically overwrites the earliest stored instantaneous calculated value whenever a new instantaneous calculated value is stored; perform a normal distribution test on all instantaneous calculated values ​​in the circular buffer queue, identify and remove abnormal instantaneous calculated values ​​that deviate from the mean by more than three standard deviations; perform equal-weighted moving average filtering on the remaining instantaneous calculated values ​​after removing the abnormal instantaneous calculated values ​​to obtain a coarse smoothed value of the instantaneous viscosity characterization parameter; perform weighted secondary smoothing processing on the three consecutively acquired coarse smoothed values, and use the result of the weighted secondary smoothing processing as the final output instantaneous viscosity characterization parameter; wherein the three coarse smoothed values ​​are assigned a first weight, a second weight, and a third weight in the order of acquisition time, and the second weight is greater than the first weight and greater than the third weight.

[0156] Furthermore, the feature parameter calculation module 02 also includes: extracting the time of temperature difference change when the temperature difference between the inlet and outlet of the cooling water changes abruptly within multiple time intervals from historical operating data, and extracting the pressure response time when the melt pressure begins to respond after the time of temperature difference change within each time interval; calculating the time difference between the pressure response time and the time of temperature difference change within each time interval, and using the time difference as a single delay time sample; sorting the multiple single delay time samples, removing extreme samples that are located at both ends of each preset proportion range after sorting, calculating the arithmetic mean of the remaining intermediate samples, and determining the arithmetic mean as the delay influence coefficient.

[0157] In one embodiment, the viscosity event discrimination module 03 is further configured to: take the current sampling time as a reference, extract a fixed-length sliding window, and extract the instantaneous viscosity characterization parameters of all sampling times within the sliding window; perform linear fitting on the instantaneous viscosity characterization parameters within the sliding window, and use the slope of the fitted line as the rate of change of the instantaneous viscosity characterization parameters within the sliding window; read the delay influence coefficient at the current time, multiply the rate of change by the delay influence coefficient to obtain the equivalent instantaneous change after hysteresis compensation; compare the equivalent instantaneous change with a calibrated change threshold, and when the equivalent instantaneous change exceeds the calibrated change threshold, determine that a viscosity drift event has occurred at the current time, and record the current time as the occurrence time of the viscosity drift event; calculate the arithmetic mean of the instantaneous viscosity characterization parameters of all sampling times within the sliding window, and use the absolute value of the difference between the arithmetic mean and the instantaneous viscosity characterization parameters at the current time as the drift amplitude of the viscosity drift event.

[0158] In one embodiment, the event feature extraction module 04 is further configured to: take the occurrence time of the viscosity drift event as a reference, extract a first time window forward, and extract the melt temperature at all sampling times within the first time window; perform linear fitting on the melt temperature within the first time window, and use the slope of the fitted line as the melt temperature change rate; take the occurrence time of the viscosity drift event as a reference, extract a second time window backward, and extract the melt pressure at all sampling times within the second time window; calculate the difference between the maximum and minimum values ​​of the melt pressure within the second time window, and use the difference as the melt pressure fluctuation amplitude; extract the cooling water inlet and outlet temperature difference at the start time and the cooling water inlet and outlet temperature difference at the end time within the second time window, calculate the difference between the temperature difference at the end time and the temperature difference at the start time, and use the difference as the cooling water inlet and outlet temperature difference change; wherein, the lengths of the first time window and the second time window are determined according to the delay influence coefficient, the length of the first time window is set as a first multiple of the delay influence coefficient, the length of the second time window is set as a second multiple of the delay influence coefficient, and the first multiple is less than the second multiple.

[0159] In one embodiment, the model prediction correction module 05 is further configured to: extract multiple viscosity drift events from historical operating data; obtain the corresponding melt temperature change rate, melt pressure fluctuation amplitude, and cooling water inlet and outlet temperature difference change for each viscosity drift event as a sample input feature set; label each viscosity drift event with the optimal feedforward correction amount determined by subsequent process experiments as a sample output label set; construct an initial coupling compensation model based on machine learning; the number of input layer nodes of the initial coupling compensation model corresponds to the number of features in the sample input feature set, and the number of output layer nodes corresponds to the output dimension of the feedforward correction amount; use the sample input feature set as the model input, use the optimal feedforward correction amount corresponding to the sample output label set as the supervision output, take minimizing the error between the model-predicted feedforward correction amount and the supervision output as the training objective, iteratively optimize the model parameters, and stop training after the prediction error on the validation set converges to the target range to obtain the trained coupling compensation model.

[0160] In one embodiment, the collaborative execution control module 06 is further configured to: analyze the feedforward correction amount, extract the temperature correction component and cooling water flow rate adjustment component corresponding to each heating zone; for each heating zone, read the current temperature setpoint, add the temperature setpoint to the corresponding temperature correction component to obtain the updated temperature setpoint for the corresponding heating zone; send the updated temperature setpoint to the temperature controller of the corresponding heating zone for execution; read the real-time flow rate value of the current cooling water circuit, add the real-time flow rate value to the cooling water flow rate adjustment component to obtain the updated cooling water flow rate setpoint; send the updated cooling water flow rate setpoint to the cooling water circuit flow controller for execution; continuously collect the real-time melt temperature at the extruder head, compare the real-time melt temperature with the average melt temperature during the historical stable period, and calculate the temperature recovery deviation; calculate the deviation tolerance threshold based on the standard deviation of melt temperature fluctuation during the historical stable period and the delay influence coefficient; when the temperature recovery deviation is less than the deviation tolerance threshold, determine that the stable melt temperature in the viscosity drift process has been reached.

[0161] Furthermore, the collaborative execution control module 06 also includes: extracting multiple stable operating periods from historical operating data where no viscosity drift events occur, and setting the duration of the stable operating period to a third multiple of the delay influence coefficient; for each stable operating period, collecting the melt temperature at all sampling times within the corresponding time period and calculating the arithmetic mean as the historical stable temperature mean for the corresponding time period; for each stable operating period, calculating the melt temperature fluctuation standard deviation for the corresponding time period based on the difference between the melt temperature at all sampling times within the corresponding time period and the historical stable temperature mean; performing an arithmetic mean on the melt temperature fluctuation standard deviations of multiple stable operating periods to obtain the average fluctuation standard deviation; and multiplying the average fluctuation standard deviation by the value of the delay influence coefficient to obtain the deviation tolerance threshold.

Claims

1. A method for temperature control of cable insulation extrusion based on multi-parameter fusion, characterized in that, The method includes: Real-time acquisition of multi-source time-series data during the extrusion process of cable insulation layer, wherein the multi-source time-series data includes at least melt pressure, screw speed, cooling water inlet and outlet temperature difference, and melt temperature; Instantaneous viscosity characterization parameters are calculated based on the melt pressure and screw speed, and a delay influence coefficient is determined based on the transmission relationship of the influence of the cooling water inlet and outlet temperature difference on the melt pressure. By combining the rate of change of the instantaneous viscosity characterization parameter within the sliding window and the delay effect coefficient, the occurrence time and drift amplitude of the viscosity drift event can be determined. In response to the viscosity drift event, the melt temperature change rate within the first time window before the event occurs, and the melt pressure fluctuation amplitude and the change in cooling water inlet and outlet temperature difference within the second time window after the event occur are extracted. The melt temperature change rate, melt pressure fluctuation amplitude, and cooling water inlet and outlet temperature difference change are input into the pre-trained coupled compensation model, and the feedforward correction amount for the current temperature control setpoint is output. Based on the feedforward correction amount, the temperature setpoints of each heating zone during the extrusion process of the cable insulation layer are corrected in layers, and the cooling water flow rate is adjusted synchronously to obtain a stable melt temperature during the viscosity drift process.

2. The method for temperature control of cable insulation extrusion based on multi-parameter fusion according to claim 1, characterized in that, The melt pressure is obtained in real time by a melt pressure sensor installed at the extruder head. The screw speed is obtained in real time by an encoder of the main drive motor of the extruder. The temperature difference between the inlet and outlet of the cooling water is obtained by real-time acquisition and calculation of the difference by a first temperature sensor installed at the inlet of the cooling water circuit and a second temperature sensor installed at the outlet of the cooling water circuit. The melt temperature is obtained in real time by insertion thermocouple at the melt flow channel of the extruder head.

3. The method for temperature control of cable insulation extrusion based on multi-parameter fusion according to claim 1, characterized in that, The instantaneous viscosity characterization parameters are calculated based on the melt pressure and screw speed, including: At each sampling moment, the real-time values ​​of the melt pressure and the screw speed are read synchronously, and the ratio of the melt pressure to the screw speed is calculated as the instantaneous calculated value of the instantaneous viscosity characterization parameter. Instantaneous calculation values ​​at multiple consecutive sampling times are stored in a circular buffer queue, wherein the circular buffer queue is a first-in-first-out storage area with a fixed length, and the earliest stored instantaneous calculation value is automatically overwritten whenever a new instantaneous calculation value is stored. Perform a normal distribution test on all instantaneous calculated values ​​in the circular cache queue, and identify and remove abnormal instantaneous calculated values ​​that deviate from the mean by more than three times the standard deviation; After removing the abnormal instantaneous calculated values, the remaining instantaneous calculated values ​​are subjected to equal-weighted moving average filtering to obtain coarse smooth values ​​of the instantaneous viscosity characterization parameters; The three consecutively acquired coarse smoothing values ​​are subjected to weighted double smoothing, and the result of the weighted double smoothing is used as the final output instantaneous viscosity characterization parameter. The three coarse smoothing values ​​are assigned a first weight, a second weight, and a third weight in the order of their acquisition time, with the second weight being greater than the first weight and the third weight.

4. The method for temperature control of cable insulation extrusion based on multi-parameter fusion according to claim 1, characterized in that, The delay influence coefficient is determined based on the transmission relationship of the influence of the cooling water inlet and outlet temperature difference on the melt pressure, including: Extract the moments of temperature difference changes where the temperature difference between the inlet and outlet of the cooling water changes abruptly within multiple time intervals from historical operating data, and extract the moments of pressure response where the melt pressure begins to respond after the moments of temperature difference changes within each time interval. Calculate the time difference between the pressure response time and the temperature difference change time in each time interval, and use the time difference as a single delay time sample. The multiple single-time delay samples are sorted, and the extreme samples that are located at the two ends of the preset ratio range after sorting are removed. The arithmetic mean of the remaining intermediate samples is calculated, and the arithmetic mean is determined as the delay influence coefficient.

5. The method for temperature control of cable insulation extrusion based on multi-parameter fusion according to claim 1, characterized in that, By combining the rate of change of the instantaneous viscosity characterization parameter within the sliding window and the delay effect coefficient, the occurrence time and drift magnitude of the viscosity drift event are determined, including: Using the current sampling time as a reference, a sliding window of a fixed length is truncated forward, and the instantaneous viscosity characterization parameters of all sampling times within the sliding window are extracted; Linear fitting is performed on the instantaneous viscosity characterization parameter within the sliding window, and the slope of the fitted line is taken as the rate of change of the instantaneous viscosity characterization parameter within the sliding window; Read the delay impact coefficient at the current moment, multiply the rate of change by the delay impact coefficient, and obtain the equivalent instantaneous change after hysteresis compensation; The equivalent instantaneous change is compared with the calibrated change threshold. When the equivalent instantaneous change exceeds the calibrated change threshold, it is determined that a viscosity drift event has occurred at the current moment, and the current moment is recorded as the moment when the viscosity drift event occurred. Calculate the arithmetic mean of the instantaneous viscosity characterization parameters at all sampling times within the sliding window, and take the absolute value of the difference between the arithmetic mean and the instantaneous viscosity characterization parameters at the current time as the drift amplitude of the viscosity drift event.

6. The method for temperature control of cable insulation extrusion based on multi-parameter fusion according to claim 1, characterized in that, In response to the viscosity drift event, the melt temperature change rate within a first time window before the event occurs, and the melt pressure fluctuation amplitude and cooling water inlet / outlet temperature difference change within a second time window after the event occur, are extracted, including: Based on the occurrence time of the viscosity drift event, a first time window is extracted backward, and the melt temperature at all sampling times within the first time window is extracted. The melt temperature within the first time window is linearly fitted, and the slope of the fitted line is taken as the rate of change of the melt temperature. Based on the occurrence time of the viscosity drift event, a second time window is extracted, and the melt pressure at all sampling times within the second time window is extracted. Calculate the difference between the maximum and minimum values ​​of the melt pressure within the second time window, and use the difference as the melt pressure fluctuation amplitude; Extract the temperature difference between the inlet and outlet of the cooling water at the start time and the temperature difference between the inlet and outlet of the cooling water at the end time within the second time window, calculate the difference between the temperature difference at the end time and the temperature difference at the start time, and use the difference as the change in the temperature difference between the inlet and outlet of the cooling water. The lengths of the first time window and the second time window are determined based on the delay impact coefficient. The length of the first time window is set to a first multiple of the delay impact coefficient, and the length of the second time window is set to a second multiple of the delay impact coefficient. The first multiple is less than the second multiple.

7. The method for temperature control of cable insulation extrusion based on multi-parameter fusion according to claim 1, characterized in that, The pre-training steps of the coupling compensation model include: Multiple viscosity drift events are extracted from historical operating data. For each viscosity drift event, the corresponding melt temperature change rate, melt pressure fluctuation amplitude, and cooling water inlet and outlet temperature difference change are obtained as the sample input feature set. Each viscosity drift event is labeled with the optimal feedforward correction amount determined after process testing, which is used as the sample output label set. An initial coupling compensation model is constructed based on machine learning. The number of input layer nodes of the initial coupling compensation model corresponds to the number of features in the sample input feature set, and the number of output layer nodes corresponds to the output dimension of the feedforward correction. The sample input feature set is used as the model input, and the optimal feedforward correction value corresponding to the sample output label set is used as the supervision output. The training objective is to minimize the error between the model's predicted feedforward correction value and the supervision output. The model parameters are iteratively optimized, and training stops when the prediction error on the validation set converges to the target range, thus obtaining the trained coupling compensation model.

8. The method for temperature control of cable insulation extrusion based on multi-parameter fusion according to claim 1, characterized in that, Based on the feedforward correction amount, the temperature setpoints of each heating zone during the extrusion process of the cable insulation layer are corrected in layers, and the cooling water flow rate is adjusted synchronously to obtain a stable melt temperature during the viscosity drift process, including: The feedforward correction is analyzed to extract the temperature correction component and cooling water flow rate adjustment component corresponding to each heating zone. For each heating zone, the current temperature setting value is read, and the temperature setting value is added to the corresponding temperature correction component to obtain the updated temperature setting value for the corresponding heating zone. The updated temperature setpoint is sent to the temperature controller of the corresponding heating zone for execution. Read the real-time flow value of the current cooling water circuit, add the real-time flow value to the cooling water flow adjustment component, and obtain the updated cooling water flow setting value; The updated cooling water flow rate setting is sent to the cooling water circuit flow controller for execution. The real-time melt temperature at the extruder head is continuously collected, and the real-time melt temperature is compared with the average melt temperature during the historical stable period to calculate the temperature recovery deviation. Based on the standard deviation of melt temperature fluctuation during the historical stable period and the delay effect coefficient, the deviation tolerance threshold is calculated. When the temperature recovery deviation is less than the deviation tolerance threshold, it is determined that the stable melt temperature in the viscosity drift process has been reached.

9. The method for temperature control of cable insulation extrusion based on multi-parameter fusion according to claim 8, characterized in that, Based on the standard deviation of melt temperature fluctuations during the historical stable period and the delay effect coefficient, the deviation tolerance threshold is calculated, including: Extract stable running segments from historical operating data where multiple viscosity-free drift events occur, and set the duration of the stable running segments to the third multiple of the delay influence coefficient; For each stable operating period, the melt temperature at all sampling times within the corresponding period is collected and the arithmetic mean is calculated as the historical stable temperature average for the corresponding period. For each stable operating period, the standard deviation of melt temperature fluctuation for the corresponding period is calculated based on the difference between the melt temperature at all sampling times within the corresponding period and the average historical stable temperature. The average standard deviation of melt temperature fluctuation is obtained by arithmetically averaging the standard deviations of the temperature fluctuations over multiple stable operating periods. The deviation tolerance threshold is obtained by multiplying the average standard deviation of fluctuation by the value of the delay effect coefficient.

10. A cable insulation extrusion temperature control system based on multi-parameter fusion, characterized in that, The system is used to execute the cable insulation extrusion temperature control method based on multi-parameter fusion as described in any one of claims 1-9, the system comprising: A multi-source data acquisition module is used to acquire multi-source time-series data in real time during the extrusion process of cable insulation layer. The multi-source time-series data includes at least melt pressure, screw speed, cooling water inlet and outlet temperature difference, and melt temperature. The characteristic parameter calculation module is used to calculate the instantaneous viscosity characterization parameters based on the melt pressure and screw speed, and at the same time determine the delay influence coefficient based on the influence transmission relationship of the cooling water inlet and outlet temperature difference on the melt pressure; The viscosity event discrimination module is used to determine the occurrence time and drift amplitude of viscosity drift events by combining the rate of change of the instantaneous viscosity characterization parameter within the sliding window and the delay influence coefficient. The event feature extraction module is used to respond to the viscosity drift event and extract the melt temperature change rate within the first time window before the event occurs, as well as the melt pressure fluctuation amplitude and the change in cooling water inlet and outlet temperature difference within the second time window after the event occurs. The model prediction correction module is used to input the melt temperature change rate, melt pressure fluctuation amplitude and cooling water inlet and outlet temperature difference into the pre-trained coupled compensation model, and output the feedforward correction amount for the current temperature control setpoint. The collaborative execution control module is used to perform layered correction of the temperature setpoint of each heating zone during the extrusion process of the cable insulation layer according to the feedforward correction amount, and to synchronously adjust the cooling water flow rate to obtain a stable melt temperature during the viscosity drift process.

Images