Wire rod surface glossiness control method and system

Abstract

The invention provides a wire rod surface glossiness control method and system, and relates to the technical field of metal wire rod surface glossiness control, and the technical scheme is characterized in that the initial glossiness and the real-time glossiness of the wire rod surface are obtained; inputting the initial glossiness and the real-time glossiness into a trained long-short-term memory network model, and generating predicted glossiness in future 2s; and calculating a first deviation between the predicted glossiness and the real-time glossiness at the same time node and a second deviation between the real-time glossiness at different time nodes so as to judge a glossiness change trend. The wire surface glossiness control method and system provided by the invention have the advantages that advanced prediction and subsequent accurate compensation of the wire surface glossiness in the thick wire fine-drawing process are realized, and the uniformity and coordination of the wire diameter surface glossiness are ensured.

Description

Technical Field

[0001] This application relates to the field of surface gloss control technology for metal wires, and more specifically, to a method and system for controlling the surface gloss of wires. Background Technology

[0002] In metal wire processing, the uniformity of surface gloss after drawing thick wire into thinner strands directly affects the quality of subsequent coating and product performance. Existing technologies often employ a single gloss sensor for fixed-point detection, which suffers from problems such as large blind spots, only providing post-processing alarms, and the inability to correlate with process parameters for tracing the root cause. This leads to delayed manual adjustments, large gloss fluctuations (typically ±5 GU), and a product pass rate below 85%. Furthermore, the detection data is not linked to the drawing equipment, hindering automatic control. Therefore, a technical solution that combines real-time performance, correlation capabilities, and closed-loop control is urgently needed.

[0003] There is currently no effective technical solution to the above problems. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for controlling the surface gloss of wire, which has the advantages of predicting the surface gloss of wire in advance and making accurate compensation in the subsequent process of drawing thick wire into thinner wire, thus ensuring uniform and coordinated surface gloss of wire diameter.

[0005] In a first aspect, this application provides a method for controlling the gloss of a wire surface, the method comprising the following steps: Obtain the initial gloss and real-time gloss of the wire surface; The initial gloss and real-time gloss are input into a trained long short-term memory network model to generate the predicted gloss within the next 2 seconds. Calculate the first deviation between the predicted gloss and the real-time gloss at the same time point, the second deviation between the real-time gloss at different time points, and adjust the parameters to determine the gloss change trend. Based on the absolute value of the first deviation and the gloss change trend represented by the second deviation, at least one of the following parameters is adjusted in stages: drawing speed, die pressure, lubricant flow rate, and temperature. The gloss of the adjusted wire surface is detected and compensation information is generated. This compensation information is used to compensate for the gloss in the fine wire painting process.

[0006] Furthermore, in this application, the steps of obtaining the initial gloss and real-time gloss of the wire surface also include: Pre-processing information is generated based on the initial gloss of the wire surface. The pre-processing information includes at least pre-processing information and corresponding equipment parameters. Perform preprocessing operations based on the preprocessing information; Update the initial gloss and real-time gloss after the job is acquired.

[0007] Furthermore, in this application, the step of adjusting at least one parameter among drawing speed, die pressure, lubricant flow rate, and temperature in stages according to the absolute value of the first deviation and the gloss change trend represented by the second deviation includes: When the absolute value of the first deviation is ≤1GU and the second deviation does not show an increasing trend, the parameters are automatically fine-tuned. When the absolute value of the first deviation is 1 to 1.5 GU or the second deviation shows an intermittent increasing trend, a first-level warning is triggered and the parameters are automatically adjusted. When the absolute value of the first deviation is greater than 1.5 GU or the second deviation shows a continuous increasing trend, a level two warning is triggered and the process is suspended.

[0008] Furthermore, in this application, the step of triggering a secondary warning and suspending the process when the absolute value of the first deviation is >1.5 GU or the second deviation shows a continuous increasing trend further includes: The equipment is tested for malfunctions, and the prediction deviation and standard deviation set according to the target gloss of the wire are verified when generating the predicted gloss. When the equipment is fault-free, set new prediction deviation and standard deviation values ​​according to the actual scenario and restart the equipment; the actual scenario is defined by the wire material, target wire diameter, target gloss and current processing stability. When the equipment malfunctions, repair the equipment and then restart it. Save all current data and associate it with the cable's unique code for traceability and analysis.

[0009] Furthermore, in this application, the step of setting new prediction bias and standard deviation values ​​according to the actual scenario and restarting the device when the device is fault-free also includes: When data interruption occurs, missing data is supplemented and adjustments are made based on valid historical data prior to the interruption, using linear extrapolation or a prediction model based on process parameter correlation. The actual scenario was re-planned based on the supplemented and adjusted data; Set new prediction bias and standard deviation based on the redesigned actual scenario and restart the device.

[0010] Furthermore, in this application, when data interruption occurs, the step of supplementing missing data and making adjustments based on valid historical data prior to the interruption using linear extrapolation or a prediction model based on process parameter correlation also includes: If the data interruption duration is ≤100ms, the process parameters of drawing speed and die pressure are completed by linear extrapolation based on the data change trend before the interruption. If the data interruption duration is greater than 100ms, for real-time gloss data, a mapping relationship model between gloss and drawing speed, die pressure, and lubricant parameters is established to back-calculate and complete the data.

[0011] Furthermore, in this application, the step of setting new prediction bias and standard deviation values ​​according to the actual scenario and restarting the device when the device is fault-free also includes: When there is no data interruption, new prediction bias and standard deviation are generated based on the complete processing data of the current wire. Based on historical processing data, re-verify whether the new prediction deviation and standard deviation are within the effective parameter range of the same historical scenario, whether they meet the equipment control capability limits, and whether they match the current processing trend; Restart the device after verification.

[0012] Furthermore, in this application, the step of detecting the gloss of the adjusted wire surface and generating compensation information further includes: Calculate the actual deviation between the adjusted gloss and the target gloss; When the actual deviation is not greater than the compensation error value preset based on the wire type, a compensation instruction containing the actual deviation information is generated so that the painting mechanism can adjust the paint layer thickness and curing temperature. When the actual deviation is greater than the compensation error value, compensation information containing the cause analysis of the deviation is generated, and the drawing adjustment parameters are updated to trigger the reprocessing of the wire.

[0013] Furthermore, in this application, the step of detecting the gloss of the adjusted wire surface and generating compensation information further includes: Collect wire information, process parameters, gloss data and quality feedback from historical processing to establish a structured database; The Long Short-Term Memory (LSTM) network model is incrementally trained using historical data from the database to optimize its prediction accuracy.

[0014] Secondly, this application also provides a wire surface gloss control system, the system comprising: The first acquisition module is used to acquire the initial gloss and real-time gloss of the wire surface; The first control module is used to input the initial gloss and real-time gloss into a trained long short-term memory network model to generate the predicted gloss within the next 2 seconds. The second control module is used to calculate the first deviation between the predicted gloss and the real-time gloss at the same time point, and the second deviation between the real-time gloss at different time points, in order to determine the gloss change trend. The third control module is used to adjust at least one parameter among drawing speed, die pressure, lubricant flow rate and temperature in stages according to the absolute value of the first deviation and the gloss change trend represented by the second deviation. The fourth control module is used to detect the gloss of the adjusted wire surface and generate compensation information, which is used for gloss compensation in the fine wire painting process.

[0015] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing embodiments of this application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Beneficial effects

[0016] 1) Overcoming the lag of post-event control to achieve proactive prediction and intervention: By introducing a trained Long Short-Term Memory (LSTM) network model, the system accurately predicts the gloss level for the next two seconds using initial and real-time gloss data. Combined with real-time calculated first and second deviations, a hierarchical control mechanism based on dual deviations and trend judgment is constructed. This allows the system to proactively and hierarchically adjust core process parameters such as drawing speed and die pressure before the gloss level actually exceeds the standard, based on the trend and magnitude of the predicted deviations. This fundamentally solves the core pain point of existing technologies—"post-event alarms and delayed manual adjustments"—achieving a fundamental shift from passive response to proactive prevention.

[0017] 2) Constructing a closed-loop control system that integrates hardware and software to improve control accuracy and yield: The system not only automatically adjusts equipment parameters but also automatically diagnoses (hardware faults and software parameter verification) and intelligently recovers in the event of severe anomalies. Ultimately, it ensures quality through cross-process compensation or self-iterative optimization. This closed loop reduces the gloss fluctuation range and stabilizes it within ±1.5 GU, improving the product yield and solving the problems of "no linkage between inspection data and drawing equipment" and "inability to achieve automatic control."

[0018] 3) Enhance system robustness and adaptability, reducing maintenance costs and unnecessary downtime: When data interruption occurs, the system can intelligently complete the data through linear extrapolation or correlation models, ensuring control continuity; when serious deviations occur, the system can handle different situations (optimize parameters and restart if there is no fault, repair and restart if there is a fault), avoiding blind downtime; through full-process data traceability and incremental model training, the system has the ability to continuously learn and optimize. These designs reduce unnecessary downtime caused by misjudgment or single faults, reduce reliance on manual intervention and maintenance costs, and enable the system to adapt to the needs of large-scale production of different materials and working conditions.

[0019] 4) Achieving end-to-end quality collaboration and traceability to meet compliance requirements of high-end industries: This invention extends quality control from a single process to the entire process. By generating precise compensation instructions for the downstream painting process, cross-process collaborative quality assurance is achieved. Simultaneously, all key data is linked and stored with the unique code of the wire, forming a complete quality traceability chain. This not only solves the problem of "inability to trace the root cause by associating process parameters," but also makes production process data transparent and traceable, meeting the stringent compliance and quality consistency requirements of industries such as automotive and electronics for key metal components. Attached Figure Description

[0020] Figure 1 A flowchart of a method for controlling the surface gloss of wires provided in an embodiment of this application; Figure 2 This is a schematic diagram of a first structure of a wire surface gloss control system provided in an embodiment of this application.

[0021] Labeling explanation: 201, First acquisition module; 202, First control module; 203, Second control module; 204, Third control module; 205, Fourth control module. Detailed Implementation

[0022] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0023] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] The following discloses and provides many different implementation methods or examples to achieve the purpose of the present invention and to solve the problems existing in the prior art.

[0025] Please refer to Figure 1 As shown in the figure, this application provides a method for controlling the surface gloss of wires, which includes the following steps: S1. Obtain the initial gloss and real-time gloss of the wire surface.

[0026] In this embodiment, initial gloss refers to the measured gloss value of the wire surface before entering the current drawing process. It represents the initial state of processing and is used to assess the raw material condition and serve as a benchmark for subsequent predictions. Real-time gloss refers to the surface gloss measurement value of the wire continuously or at high frequency during the current drawing process. It reflects the effect of processing parameters in real time. Obtaining initial gloss is to establish a processing benchmark and eliminate interference from raw material differences; obtaining real-time gloss is to monitor the processing effect in real time. The combination of the two provides the necessary time-series data and state starting point for subsequent prediction and control.

[0027] S2. Input the initial gloss and real-time gloss into the trained long short-term memory network model to generate the predicted gloss within the next 2 seconds.

[0028] In this embodiment, LSTM is a special type of recurrent neural network (RNN) capable of learning and memorizing dependencies in long-term time series. Training means the model has completed its learning process using a large amount of historical processing data (including gloss sequences and corresponding process parameters), enabling it to predict future trends based on current and historical gloss data. This distinguishes it from simple linear extrapolation or statistical prediction models. The chosen prediction duration of 2 seconds balances the effectiveness (enough to allow for control measures in advance) and accuracy (within a credible timeframe for industrial process control). Step S2 leverages the advantages of the LSTM model in processing time-series data to proactively infer the gloss trend in the short term, rather than passively waiting for deviations to occur. This provides a basis for decision-making in subsequent preventative adjustments, unlike existing post-event correction methods.

[0029] S3. Calculate the first deviation between the predicted gloss and the real-time gloss at the same time point, and the second deviation between the real-time gloss at different time points to determine the gloss change trend.

[0030] In this embodiment, the first deviation is defined as the difference between the predicted gloss and the real-time gloss at the same specific moment (time node). It directly reflects the immediate accuracy of the prediction model and the static difference between the current process state and the expectation. The second deviation is defined as the difference between the real-time gloss at different time nodes (usually adjacent sampling moments). It is used to characterize the rate or direction of change of the real-time gloss itself over time. Judging the trend of gloss change clarifies the purpose of calculating the second deviation. By analyzing continuous second deviations (such as calculating their moving average and sign), it is possible to qualitatively or quantitatively determine whether the gloss is increasing, decreasing, or remaining stable. Step S3 constructs a dual-deviation analysis system. The first deviation is used to evaluate and correct the gap between the prediction and reality; the second deviation is used to capture the dynamic characteristics of the process itself (such as whether it is out of control or drifting). The combination of the two provides a more comprehensive basis for control decisions: it focuses on both the current magnitude of the deviation and the possible direction of the deviation.

[0031] S4. Based on the absolute value of the first deviation and the gloss change trend represented by the second deviation, adjust at least one of the following parameters in stages: drawing speed, die pressure, lubricant flow rate, and temperature.

[0032] In this embodiment, the adjustment is based on two dimensions: the absolute value of the first deviation (the magnitude of the deviation) and the trend represented by the second deviation (the direction and severity of the deviation's change). "Grading" means that the adjustment is not a single, linear process, but rather different intensities or strategies (such as fine-tuning, strong adjustment, and emergency intervention) are adopted according to the magnitude of the deviation and the severity of the trend. Drawing speed, die pressure, lubricant flow rate, and temperature are key process parameters that directly affect the surface gloss of the wire. Drawing speed affects the friction and deformation rate; die pressure affects the surface finishing effect; lubrication conditions (flow rate and temperature) affect the coefficient of friction and surface quality. This clarifies the physical object of the adjustment, reflecting an understanding of the process mechanism. The presence of at least one parameter indicates that the adjustment can be performed on a single parameter or a combination of multiple parameters, ensuring flexibility. Step S4 maps the "magnitude of deviation" and "trend of change" information obtained in step S3 into specific adjustment actions for the key process parameters. The logic of "graded adjustment" embodies the essence of intelligent control: small deviations are gently adjusted to maintain stability; large deviations or deteriorating trends are decisively intervened to prevent quality accidents, thereby solving the drawbacks of lag or over-adjustment in traditional control.

[0033] S5. Detect the gloss of the adjusted wire surface and generate compensation information, which is used for gloss compensation in the fine wire painting process.

[0034] In this embodiment, the actual surface state of the wire after this round of adjustment ("adjusted gloss") is obtained through re-detection. Compensation information is generated based on the final comparison between the adjusted gloss and the target value, producing a data packet or instruction set. The purpose and flow of this compensation information are clearly defined for use in the fine wire coating process to compensate for gloss. It will be transmitted to the downstream fine wire coating process as input for adjusting its own process parameters (such as coating thickness, curing temperature, etc.) to ultimately correct any minor gloss deviations that may remain from the drawing process. Step S5 achieves cross-process quality collaborative control. It acknowledges that the control of the drawing process may have limits to precision. By generating compensation information, the quality status of the current process is communicated to the next process, enabling the coating process to perform targeted compensation, thereby ensuring that the gloss meets the standards at the final product level.

[0035] In some embodiments, the step of obtaining the initial gloss and real-time gloss of the wire surface further includes: S101. Generate pretreatment information based on the initial gloss of the wire surface. The pretreatment information includes at least the pretreatment type and the corresponding equipment parameters.

[0036] S102. Perform preprocessing operations based on the preprocessing information.

[0037] S103, Update the initial gloss and real-time gloss after the job is acquired.

[0038] In step S101, preprocessing information is generated based on the initial gloss of the wire surface. This means that the generated preprocessing information is dynamic and targeted, and its content depends on the specific state of the wire surface detected for the first time (such as the gloss value and uniformity), ensuring the surface state level (such as qualified, slightly contaminated, moderately oxidized, etc.) rather than a fixed preset instruction.

[0039] Pretreatment type refers to the category of physical or chemical treatment methods used to improve surface condition. Examples include: "high-pressure air curtain dust removal," "laser descaling," "chemical degreasing," and "ultrasonic cleaning." Different types of treatment target different surface defects (such as dust, oxide layers, and oil stains) and correspond to specific equipment.

[0040] The corresponding equipment parameters refer to the specific operating values ​​that the relevant preprocessing equipment needs to be set to in order to perform the above-mentioned "preprocessing types". For example: Dust removal type: air pressure (0.3~0.6MPa), spray angle, and action time; Deoxidation type: laser power (50~100W), scanning speed, irradiation mode; Degreasing type: chemical agent concentration, temperature, and soaking time.

[0041] For example, the equipment parameters for "high-pressure air curtain dust removal" may include "air pressure value (e.g., 0.5MPa)", "air curtain width", and "duration".

[0042] The phrase "at least includes" indicates that the "preprocessing information" is a structured data packet, in which "type" and "device parameters" are essential core elements, but it can also contain other auxiliary information, such as processing priority and target area coordinates, which leaves reasonable room for specific adjustments to the implementation plan.

[0043] Therefore, step S101 establishes a dynamic preprocessing planning mechanism based on the detection results. It realizes the transformation from passive detection to proactive intervention preparation, ensuring that the preprocessing operation is "targeted" and avoiding the problems of "overprocessing" (wasting resources) or "underprocessing" (leaving defects) that may be caused by uniform preprocessing in the prior art.

[0044] In step S102, the preprocessing operation is carried out according to the preprocessing information. This is based on the preprocessing information generated in step S101. The control system or operator parses the information into specific equipment control instructions and drives the corresponding preprocessing device (such as fan, laser, pump valve) to run according to the specified type and parameters, thereby converting the information instructions in step S101 into actual surface treatment actions.

[0045] The purpose of step S102 is to actively eliminate or reduce surface defects that exist in the wire before it enters the core drawing process, so as to create excellent initial surface conditions for obtaining high-quality and uniform gloss. This is a fundamental step in the entire gloss control chain to ensure the effect.

[0046] For example, a preprocessing job may specifically include, but is not limited to, the following: High-pressure air curtain dust removal: control the opening of the solenoid valve to regulate the air pressure, and control the start-up timing and angle of the nozzle array; Laser deoxidation: Controlling the power output of the laser generator and controlling the movement trajectory and speed of the scanning galvanometer; Chemical degreasing: The pump and valve system is controlled to regulate the flow rate of the chemical solution, and the heater is controlled to maintain the set temperature.

[0047] In step S103, the initial gloss after the operation specifically refers to the gloss of the wire surface that is re-detected after the pretreatment operation is completed. It replaces the original "initial gloss" and becomes the authoritative benchmark value representing the latest and most authentic surface state of the wire before it enters the drawing process.

[0048] Post-operation real-time gloss refers to the continuous gloss data collected immediately after the "initial gloss" benchmark has been re-established. The "real-time gloss" sequence collected at this time is dynamic data associated with the updated and accurate initial benchmark.

[0049] The "update" step in step 103 is used to replace the old data with the new data and to verify the effectiveness of the preprocessing operation. If the gloss level obtained after the update is still not up to standard, the system can determine that the preprocessing is invalid, thereby triggering an alarm or requiring reprocessing.

[0050] Therefore, step S103 is used to ensure that the pretreatment achieves the expected goals, providing qualified raw materials for the drawing process, and providing an accurate and reliable input benchmark for all subsequent calculations (especially the predictions of the LSTM model). If the prediction model is based on an unrealistic "initial gloss" that is severely affected by surface contaminants, its output will be meaningless. Step S103 fundamentally eliminates this risk of "garbage in, garbage out".

[0051] In some embodiments, the step of adjusting at least one parameter among drawing speed, die pressure, lubricant flow rate, and temperature in stages according to the absolute value of the first deviation and the gloss change trend characterized by the second deviation includes: S401. When the absolute value of the first deviation is ≤1GU and the second deviation does not show an increasing trend, the parameters are automatically fine-tuned. S402. When the absolute value of the first deviation is 1 to 1.5 GU or the second deviation shows an intermittent increasing trend, a first-level warning is triggered and the parameters are automatically adjusted. S403. When the absolute value of the first deviation is greater than 1.5GU or the second deviation shows a continuous increasing trend, a level 2 warning is triggered and the process is suspended.

[0052] In step S401, the absolute value of the first deviation is ≤1 GU. This is a quantitative threshold judgment. 1 GU (gloss unit) is a key performance indicator threshold set in this embodiment, which represents the normal process fluctuation range allowed by the system without emergency intervention. When the difference between the predicted value and the measured value is within this range, the processing status is considered to be basically under control.

[0053] The second deviation does not show an increasing trend. This is a dynamic trend judgment, which means that by calculating the difference between adjacent points or the moving average of continuously collected real-time gloss data, it is judged that the value does not show a continuous upward or downward trend, that is, it is in a "stable" or "random small fluctuation" state.

[0054] "And" means that two conditions must be met simultaneously, which means that even if the deviation is small, if the trend is deteriorating (such as a continued decline), more aggressive regulatory measures may be needed.

[0055] The scenario corresponding to step S401 is a slight, stable process fluctuation. The system determines that the current situation is good, but there is a minor deviation. The adjustment goal at this time is to perform fine-tuning to further reduce the deviation or maintain it at an excellent level, preventing its accumulation or development.

[0056] Fine-tuning means adjusting the range very small and the adjustment action is smooth. For example, the drawing speed may be adjusted by only 0.05~0.1m / s, or the lubricant flow rate may be corrected by 1~2%. This kind of adjustment is usually achieved by the proportional term of the PID controller or a very small fixed correction value, in order to avoid introducing new fluctuations due to over-adjustment.

[0057] In step S402, the absolute value of the first deviation is 1 to 1.5 GU, which is a warning threshold range. The deviation exceeds the normal range but has not yet reached a dangerous level. This indicates that the process status has deviated significantly and requires attention and strong intervention.

[0058] The second deviation indicates an intermittent increasing trend, meaning that the real-time gloss change trend shows an unstable and discontinuous increase or decrease, such as increased fluctuations, sometimes exceeding the standard and sometimes recovering. This may indicate unstable equipment status, changes in raw material batches, or interference from the external environment.

[0059] The "or" in step S402 means that the response at this level can be triggered if either of the two conditions is met. This broadens the coverage of the warning, focusing on both the "magnitude" of the deviation and the "change characteristics" of the deviation.

[0060] "Level 1 warning" is a warning that can be triggered by flashing prompts on the user interface, logging, or mild audio-visual signals. Its purpose is to inform operators that the system has detected a potential risk, requiring attention and preparation for possible intervention, but it does not force a production stoppage.

[0061] The "automatic parameter adjustment" is more powerful than the "fine-tuning." The system calculates and executes a more significant parameter correction based on the magnitude and direction of the deviation, aiming to quickly bring the process back on track. For example, it may adjust the mold pressure by 0.5~1.0MPa or make comprehensive adjustments to the lubrication system.

[0062] In step S403, the absolute value of the first deviation is >1.5GU, which is a serious exceedance threshold. This indicates that the prediction model has failed or there is a major anomaly in the process, which has far exceeded the allowable processing error, and the product quality can no longer be guaranteed.

[0063] The second deviation indicates a continuous increasing trend, meaning that the real-time gloss shows a clear and continuous unidirectional deterioration (such as a linear increase or decrease). This is a strong signal that the process is about to or has already gone out of control, and may be a direct manifestation of hardware problems such as severe mold wear or lubrication system failure.

[0064] The "or" in step S403 means that the highest level response is triggered when either of the two conditions is met, ensuring that any serious anomaly can be caught.

[0065] Level 2 warning is the highest level of emergency alert, usually accompanied by strong audible and visual alarms, system interface locking, or automatic notification to relevant personnel. Its purpose is to require immediate manual intervention for inspection.

[0066] Suspending the process is the ultimate safety measure. The system will issue an emergency stop or safety stop command to the drawing equipment to stop the wire feeding and processing. The purpose is to prevent production from continuing under unknown circumstances, which could lead to a large number of defective products and potentially further damage to the equipment.

[0067] It should be noted that the "parameters" in steps S401 to S403 refer to the drawing speed, die pressure, lubricant flow rate, and temperature.

[0068] Fine-tuning refers to small-amplitude, high-frequency parameter adjustments performed when the system is determined to be in the first-level (normal) operating condition (i.e., the first deviation ≤ 1 GU and the trend is stable), aimed at precise optimization. The adjustment amount is very low relative to the normal operating range or rated value of the parameter. For example, the adjustment range for drawing speed is typically within ±0.1 m / s (approximately 2~5% of the rated speed); the adjustment range for die pressure is typically within ±0.3 MPa; and the adjustment range for lubrication parameters such as flow rate or temperature typically does not exceed ±5% of its set value.

[0069] The purpose of fine-tuning is to eliminate or suppress small random fluctuations, so that the gloss more closely matches the target value or predicted trajectory, and to pursue the ultimate steady-state accuracy. It usually adopts proportional (P) control or proportional-integral (PI) control mode, with the controller gain set low to avoid causing system oscillation, smooth operation, and minimal impact on the production process.

[0070] In this context, "adjustment" refers to the parameter adjustment action of moderate to significant magnitude, aimed at correcting obvious deviations, performed when the system determines that it is in the second-level (warning) condition (i.e., the first deviation is 1~1.5GU or the trend is intermittently abnormal).

[0071] The adjustment range of "adjustment" is significantly greater than that of "fine adjustment". For example, the adjustment range of drawing speed reaches ±0.2~0.3m / s; the adjustment range of die pressure reaches ±0.5~1.0MPa; and the lubrication parameters are adjusted by 10~15% based on the set value or switched to the standby control mode.

[0072] The purpose of adjustment is to quickly curb the trend of significant deviation expansion that has already occurred, or to forcefully pull the process state back to an acceptable range (first-level operating condition). Its core objective is to correct deviations. It usually adopts proportional-integral-derivative (PID) control with high gain, or adopts model-based feedforward-feedback composite control to generate sufficiently strong corrective force. It may simultaneously adjust multiple parameters to form a synergistic effect.

[0073] In some embodiments, when the absolute value of the first deviation is >1.5 GU or the second deviation indicates a continuous increasing trend, the step of triggering a secondary warning and suspending the process further includes: S4031. Detect the fault status of the detection equipment and verify the prediction deviation and standard deviation set according to the target gloss of the wire when generating the predicted gloss. S4032. When the equipment is fault-free, set new prediction deviation and standard deviation values ​​according to the actual scenario and restart the equipment; the actual scenario is defined by the wire material, target wire diameter, target gloss and current processing stability. S4033. When the equipment malfunctions, repair the equipment and restart it; S4034. Save all current data and associate it with the wire's unique code for traceability analysis.

[0074] In step S4031, the equipment's fault condition is detected; this is an active diagnostic process. After the system pauses, it automatically or guides the operator to perform status checks on critical hardware components, which typically include: Sensor system: Check whether the signals of the gloss sensor, speed encoder, pressure sensor, etc. are normal, or whether they have drifted or failed; Actuator: Check whether the drawing motor, hydraulic system, lubrication pump valve, etc. are overloaded, stuck, or have abnormal response; Mechanical parts: Check whether the mold has excessive wear, cracks, or adhesive residue.

[0075] Verifying the prediction deviation used to generate the predicted gloss and the standard deviation set according to the target gloss of the wire is a review process of the software control logic and parameter settings, which the system or operator needs to verify. Prediction bias: refers to the inherent error range (e.g., ±0.5GU) allowed by the LSTM model during prediction. Verify whether its settings match the material and specifications of the wire being processed, and whether there are systematic prediction inaccuracies caused by the model not being updated in time or incorrect parameter settings.

[0076] Standard deviation: refers to the allowable processing error (e.g., ±1 GU) set manually based on the target gloss level of the wire (e.g., 50 GU for carbon steel). Verify whether it is set reasonably, whether it is too strict or too lenient, and whether it has not been updated in sync with changes in product specifications.

[0077] The purpose of step S4031 is to determine the primary dual-track diagnostic tasks after shutdown—hardware and software. Detecting equipment faults aims to locate physical, non-healable root causes (such as mold damage); verifying parameters aims to locate logical or setting-level, adjustable root causes (such as parameter mismatch). This avoids the drawbacks of blindly repairing based solely on experience after traditional shutdowns and provides a basis for precise handling in the next step.

[0078] In step S4032, "no equipment failure" means that through the diagnosis in the previous step S4031, it is confirmed that all key hardware (sensors, actuators, molds) are in normal working condition and there is no physical damage or functional failure.

[0079] "Real-world scenario" is a multi-dimensional processing condition description tag, defined in four key dimensions: The material of the wire, such as carbon steel, stainless steel, or copper alloy, determines its deformation and surface characteristics.

[0080] Target wire diameter: The final diameter to be machined, which affects the deformation rate and friction conditions.

[0081] Target gloss: The quality requirement value for this batch of products.

[0082] Current processing stability: This refers to whether, before the pause (excluding anomalies that caused the pause), the process parameters (speed, pressure, etc.) and gloss were in a relatively stable plateau period, or were already undergoing drastic adjustments or fluctuations. For example, the current processing stability can be quantified by calculating the standard deviation σ or coefficient of variation CV of real-time process parameters (speed, pressure, etc.) and gloss over a past period, such as 10 seconds. The standard deviation σ and coefficient of variation CV are statistical indicators that reflect the dispersion (magnitude of fluctuation) of the data.

[0083] Setting new prediction bias and standard deviation is a parameter optimization and adaptation process. Since the hardware is not faulty, the cause of severe deviation is likely that the control parameters (prediction bias and standard deviation) do not match the current "actual scenario". For example, after changing from stainless steel to carbon steel, the prediction bias was not adjusted accordingly; or when processing large-diameter wire, the standard deviation was not appropriately relaxed. In this case, the system needs to recalculate or select a more suitable set of parameters based on an accurate definition of the "actual scenario".

[0084] After the parameters are updated, the system is unsuspended and production resumes according to the new control parameters, i.e., the equipment is restarted.

[0085] Step S4032 provides a software solution when the hardware is in good working order. It embodies the adaptive capability of the control system—the ability to recognize changes in operating conditions and dynamically adjust its control “tolerance” boundaries (prediction deviation) and “target” boundaries (standard deviation), rather than rigidly using a fixed set of parameters, thereby enhancing the system’s adaptability and robustness to different product specifications and production conditions.

[0086] In step S4033, repairing the equipment refers to performing necessary repair, replacement, or calibration operations for specific hardware faults diagnosed (such as replacing worn molds, calibrating drifting sensors, or clearing blocked lubrication lines). This is a physical repair process that must be completed manually or by an automated maintenance system.

[0087] Restarting means restarting the production line after confirming that the fault has been eliminated and the equipment has returned to normal.

[0088] Step S4033 clarifies the standard handling path after a physical fault is discovered. It emphasizes that for hardware problems, substantive repairs must be carried out, and attempts should not be made to "cover up" or "bypass" them by adjusting software parameters.

[0089] In step S4034, all current data includes at least: the deviation data sequence before triggering the level 2 warning, the equipment diagnostic results, the fault description (if any), the parameter values ​​before and after verification, the set "actual scenario" label, and the measures taken (repair or parameter adjustment).

[0090] "Associated with the unique code of the wire": This means binding and storing the above data packet with the unique identification code of the wire being processed when the segment (or batch) caused the anomaly, so as to facilitate subsequent traceability analysis.

[0091] Step S4034 establishes a comprehensive quality traceability and knowledge accumulation mechanism: on the one hand, it is traceable, meaning that when a quality problem is subsequently discovered in a batch of products, the complete data at the time of production can be traced through the wire code, accurately reproducing the failure scenario and pinpointing whether it is due to equipment failure, parameter mismatch, or other reasons, facilitating quality zeroing and responsibility identification; on the other hand, it is analytical, meaning that the accumulated abnormal event data can form a case library for subsequent analysis of common problems, optimization of early warning thresholds, and even as training data to improve the LSTM model's ability to predict abnormal patterns, thereby achieving continuous improvement of the system.

[0092] This embodiment expands the "Level 2 Warning Pause" from a simple safety action into a complete anomaly management closed loop that includes "diagnosing hardware / software problems, decision-making (repair / optimization), execution (repair / setting), and recording (tracing)". This enables the system to not only stop losses in time when dealing with the most serious anomalies, but also to find out the cause, solve the problem and accumulate experience, thereby improving continuous optimization capabilities.

[0093] In some embodiments, when the device is fault-free, the steps of setting new prediction bias and standard deviation values ​​according to the actual scenario and restarting the device further include: S40321. When data interruption occurs, based on the valid historical data before the interruption, use linear extrapolation or a prediction model based on process parameter correlation to complete the missing data and make adjustments. S40322. Re-plan the actual scenario based on the completed and adjusted data; S40323. Set new prediction bias and standard deviation values ​​according to the replanned actual scenario and restart the device.

[0094] In step S40321, a data interruption refers to an event in which some or all of the critical data is lost on the data acquisition link of the control system within a certain period before the secondary warning is triggered and the system is suspended. Critical data includes at least real-time gloss, drawing speed, die pressure, lubricant flow rate, and temperature. The interruption may be caused by momentary sensor malfunctions, communication interference, or data acquisition card abnormalities.

[0095] Valid historical data prior to the interruption refers to a complete and verified data sequence that was normally collected and stored by the system before the data interruption event occurred. This is the only reliable basis for data recovery.

[0096] Using linear extrapolation or predictive models based on process parameter correlation to fill in missing data is the core technical means of data repair. The system must first reconstruct a complete and coherent data sequence in order to accurately analyze the "real-world scenario".

[0097] Among them, the linear extrapolation method is suitable for short-term interruptions or situations where data changes are stable. It assumes that the process parameters change linearly during the interruption according to the instantaneous rate of change (slope) before the interruption, thereby calculating the estimated value at the missing moment. This is a simple and fast completion method.

[0098] Predictive models based on process parameter correlations are suitable for situations involving prolonged interruptions or strong coupling between parameters. This model (which may differ from LSTM and focus more on static or short-term correlations) characterizes the mathematical relationship between gloss and parameters such as speed, pressure, and lubrication. When gloss data is missing, the corresponding gloss estimate can be derived from the known process parameter sequence using this model; and vice versa.

[0099] "And make adjustments" means that after the data is completed, the completed values ​​may need to be post-processed, such as checking for reasonableness (e.g., whether they exceed physical limits) and smoothing filtering, to ensure that the repaired data sequence can be used for reliable scene analysis.

[0100] The purpose of step S40321 is to solve the problem of incomplete information. After a data interruption occurs, the data foundation used to judge the "actual scenario" has gaps. The algorithm makes a reliable estimate based on historical patterns, thereby restoring the real process state during the interruption to the greatest extent possible, and providing the possibility for accurate planning of the actual scenario in the next step.

[0101] In step S40322, "replanning" here is to repeat the analysis logic of "according to the actual scenario" in step S4032 based on data repair, but the input data is the repaired complete sequence.

[0102] The system uses the completed data to recalculate a fourth dimension beyond wire material, target wire diameter, and target gloss: the current stable processing state. This includes assessing whether the fluctuations in process parameters and the gloss trend are truly stable before the interruption and during the interruption period inferred from the completed data. The completed data may reveal that there was a slow abnormal trend before the interruption, which is the root cause of the final serious deviation.

[0103] Step S40322 ensures that the judgment of the "actual scenario" is based on the most complete and accurate data possible, avoiding misjudgment of the scenario due to inaccurate data (for example, misjudging an unstable state as stable), which would lead to the setting of incorrect new parameters.

[0104] In step S40323, the system will select or generate a set of new prediction bias (model tolerance) and standard deviation (control target tolerance) that best match this corrected scenario understanding, based on the experience database or through algorithm calculation.

[0105] For example, if the replanned scenario is determined to be "material carbon steel - wire diameter 3mm - target gloss 52GU - slight fluctuation in processing condition", then the new standard deviation set may be slightly larger than that of the "stable" scenario, and the prediction bias may also be adjusted accordingly. After completing the parameter update, the system will restart.

[0106] Step S40323 completes the entire logical chain from data repair to scenario correction and parameter optimization. It ensures that after experiencing a specific anomaly such as data interruption, the control parameters used by the system to resume operation are adapted based on the most reasonable reconstruction of historical conditions, rather than based on incomplete or misleading information. This greatly improves the system's ability to recover robustly from complex anomalies.

[0107] In some embodiments, when data interruption occurs, the step of supplementing missing data and making adjustments based on valid historical data prior to the interruption using linear extrapolation or a prediction model based on process parameter correlation further includes: S403211. If the data interruption duration is ≤100ms, the process parameters of drawing speed and die pressure shall be completed by linear extrapolation based on the trend of data change before the interruption. S403212. If the data interruption duration is greater than 100ms, for real-time gloss data, a mapping relationship model between gloss and drawing speed, die pressure, and lubricant parameters is established to back-calculate and complete the data.

[0108] In step S403211, the data interruption duration is ≤100ms, which is a critical time threshold. 100ms corresponds to 10 data points collected at a sampling frequency of 100Hz, which is a very short time window, usually caused by transient communication interference or sampling jitter. It is reasonable to assume that the physical state of the process system will not change abruptly within this time.

[0109] "For drawing speed and die pressure process parameters": This clarifies the data types to which this rule applies. Drawing speed and die pressure are typical physical quantities with relatively large inertia and slow change. Within a short time (≤100ms), their changes can be approximated as linear.

[0110] The missing data is filled using a linear extrapolation method based on the trend of data changes before the interruption. Specifically, effective speed or pressure data for a short period of time before the interruption (such as the most recent 20 points, i.e., 200ms) is extracted, and its rate of change is calculated (e.g., the slope is obtained using linear regression). Then, it is assumed that the parameter continues to change linearly at this rate of change within 100ms of the interruption, thereby calculating the estimated value at each missing time point during the interruption.

[0111] Step S403211, for brief and smooth interruptions in process parameters, employs a computationally simple and fast-responding linear extrapolation method. This method can obtain sufficiently accurate supplementary values ​​with low computational cost, meeting the requirements for rapid recovery. It should be noted that this method is based on a reasonable simplification of the short-term behavior (inertia) of the physical process.

[0112] In step S403212, the data interruption duration is >100ms, which is another time threshold judgment. The interruption time is relatively long, exceeding the range that can be safely assumed to be a linear change, and the process state may have changed significantly during this period.

[0113] The section on "real-time gloss data" clarifies the core data type to which this rule applies. Gloss is the target variable to be controlled and the direct basis for scene planning and deviation calculation, making the integrity of the data crucial.

[0114] This intelligent completion method, based on multivariate correlation, uses a model to inversely complete the model by establishing a mapping relationship between gloss and drawing speed, die pressure, and lubricant parameters. Its technical essence is: The system learns and builds a mathematical model in advance or from historical data that can describe how gloss (output) is affected by key process parameters (inputs: speed, pressure, lubricant flow / temperature). This could be a multiple linear regression model, a simple neural network, or an empirical formula. Unlike LSTM prediction models, this model focuses more on static or quasi-static input-output relationships rather than complex time series predictions.

[0115] When gloss data is missing for an extended period, the system possesses complete (or supplemented by rule one) process parameter data from the period of interruption. Therefore, it inputs the known process parameter sequence into this mapping model, which calculates the estimated gloss value at the corresponding time point, thus supplementing the missing gloss data.

[0116] Step S403212 is a solution for complex situations involving long-term interruptions and missing data for critical targets. It recognizes that linear extrapolation is no longer reliable over longer timescales, and that gloss changes may be non-linear. By leveraging the inherent physical relationship between process parameters and gloss (which are precisely the causes driving gloss changes), the "result" can be deduced from the "cause," thus achieving more reliable data reconstruction.

[0117] In some embodiments, when the device is fault-free, the steps of setting new prediction bias and standard deviation values ​​according to the actual scenario and restarting the device further include: S40324. When there is no data interruption, generate new prediction deviation and standard deviation based on the complete processing data of the current wire. S40325. Based on historical processing data, re-verify whether the new prediction deviation and standard deviation are within the effective parameter range of the same historical scenario, whether they meet the equipment control capability limits, and whether they match the current processing trend. S40326. Restart the device after verification.

[0118] In step S40324, there is no data interruption. This is a clear prerequisite, meaning that from the start of this processing until the triggering of the Level 2 warning and suspension, the collection, transmission and recording of all key data are continuous, complete and without omissions.

[0119] The current complete processing data of the wire is the most ideal basis for analysis. It contains all timestamp-aligned, true initial gloss, real-time gloss, predicted gloss, drawing speed, die pressure, lubricant parameters, and all calculated deviation sequences of the wire segment (or batch) from the start of the drawing process to the time the abnormality occurred and the process was suspended. This is a complete dataset that requires no estimation or repair.

[0120] The system generates new prediction bias and standard deviation values ​​by directly calculating them based on this complete and authentic historical data using built-in optimization algorithms (e.g., statistical process control (SPC) methods, matching algorithms based on historical similar cases, or algorithms that use complete data to quickly re-evaluate the control model).

[0121] Because the data is complete, the system can accurately analyze the root causes of serious deviations. For example, it can accurately calculate the average prediction error of the LSTM model over a period of time (which directly relates to how large the new "prediction bias" should be), and it can also accurately assess the natural fluctuation range of gloss under the current operating conditions (which directly relates to how wide the new "standard deviation" should be). In other words, the generation of new parameters is data-driven and evidence-based.

[0122] In step S40325, historical processing data refers to the processing data of all previous successful batches and their corresponding optimal parameter sets stored in the database.

[0123] Regarding re-verification, this is a multi-dimensional and multi-constraint security and rationality verification process, specifically including: "Whether it falls within the range of valid parameters in the same historical scenario" (reasonableness verification): The system compares the newly generated parameters with the historical range of parameters that have been successfully used in the same or highly similar "actual scenarios" (such as the same material, similar wire diameter, and similar target gloss) in the database. The new parameters should fall within this empirical range, or at least not deviate significantly, to ensure that they are practically proven and safe and reliable.

[0124] "Does the equipment control capability limit need to be met?" (Feasibility Verification): The system checks whether the new parameters are physically feasible. For example, if the new standard deviation is set too small (e.g., ±0.2 GU), it may exceed the limits of the current equipment accuracy and sensor resolution, making it impossible to implement. This verification ensures that the parameter settings are engineering-feasible.

[0125] "Does it match current processing trends?" (Prospective / Adaptive Verification): The system analyzes the trend at the end of the complete data sequence. If the data shows a slow, upward "drift" trend in gloss, the newly set "Prediction Bias" may need to be adjusted to compensate for this trend. This verification ensures that the new parameters are not based solely on past static averages but are adapted to upcoming dynamic changes.

[0126] Step S40325 establishes a crucial safety barrier and optimization feedback loop, preventing the system from generating overly aggressive or unrealistic parameters based on data from a single anomalous event. By incorporating verification from three dimensions—historical experience, physical constraints, and trend prediction—it ensures that the generated new parameters are scientifically sound, safe, feasible, and possess a certain degree of foresight. This represents a key improvement from "data-generated parameters" to "intelligent parameter adoption."

[0127] In step S40326, "verification successful" means that all three verifications above have passed smoothly. If any verification fails (for example, the new parameter exceeds the historical range, exceeds the device's capabilities, or is seriously inconsistent with the trend), the system will not start directly, but may trigger an alarm, requiring manual intervention for review, or return to the previous step to regenerate the parameter.

[0128] After all verifications are passed, the system applies the newly generated, multi-verified prediction deviation and standard deviation values, and lifts the pause, resuming production by restarting the equipment.

[0129] Step S40326 is the final execution decision point, which emphasizes the prerequisite for restarting—it must be based on a set of rigorously validated, high-quality new control parameters. This reflects a prudent and robust recovery strategy, minimizing the risk of anomalies recurring immediately after restarting due to improper parameter settings.

[0130] In some embodiments, the step of detecting the gloss of the adjusted wire surface and generating compensation information further includes: S501. Calculate the actual deviation between the adjusted gloss and the target gloss. S502. When the actual deviation is not greater than the compensation error value preset based on the wire type, a compensation instruction containing the actual deviation information is generated so that the painting mechanism can adjust the paint layer thickness and curing temperature. S503. When the actual deviation is greater than the compensation error value, compensation information containing the cause analysis of the deviation is generated, and the drawing adjustment parameters are updated to trigger the reprocessing of the wire.

[0131] In step S501, "adjusted gloss": This is a final state measurement value, referring to the surface gloss value (denoted as G) detected by the gloss sensor at the final node leaving the drawing process after the wire has undergone all the aforementioned steps (which may include multiple rounds of adjustment, abnormal pauses and recoveries). final It represents the final output of the wire drawing process for surface finishing.

[0132] "Target gloss": This is the pre-set standard value for gloss (denoted as G) that is expected to be achieved for this batch of wires. target ( ), which is the absolute benchmark for quality judgment.

[0133] "Actual deviation": refers to the algebraic or absolute difference between the two, i.e., δ = |G| final -G target | or δ=G final -G target This is a key quantitative quality indicator that directly reflects the degree to which the processing accuracy of the drawing process conforms to the target value.

[0134] Step S501 is a data preparation step prior to quality assessment. It concretizes the action of "inspecting the adjusted gloss" into a specific, calculable difference. This "actual deviation" value δ is an objective, quantitative input that will be used for decision-making.

[0135] In step S502, a compensation error value is preset based on the wire type. This is a key quality tolerance threshold (denoted as ϵ). It is not fixed, but is preset according to the physical characteristics of different wires, application requirements, and the compensation capability of the coating process. For example: For carbon steel wire, the compensation error value ϵ can be set to ±0.8GU; For stainless steel wire (which has a brighter surface but may require higher consistency), the compensation error value ϵ can be set to ±0.7GU; For copper alloy wires, the compensation error value ϵ can be set to ±1.0GU.

[0136] The actual deviation not exceeding the compensation error value means that δ≤ϵ. This is a passability criterion, which means that although there is a deviation, it is within the allowable range and falls into the category of acceptable and repairable.

[0137] The compensation instruction is a structured data packet, the core of which transmits the magnitude and direction of the actual deviation δ (positive deviation represents a brighter appearance, negative deviation represents a darker appearance). Its purpose is to guide the downstream painting process in collaborative quality correction. That is, if the drawing process has already controlled the deviation within an acceptable compensable range, costly rework is unnecessary. After receiving this instruction, the painting process can adjust its own processes to mask or correct this minor deviation, for example: If the actual deviation δ is negative (too dark): the paint layer thickness can be appropriately increased (e.g., by 2μm) and / or the curing temperature can be increased (e.g., by 5℃) to make the paint film brighter in order to compensate. If the actual deviation δ is positive (too bright): the paint layer thickness can be appropriately reduced (e.g., reduced by 2μm) and / or the curing temperature can be increased (e.g., reduced by 5℃) to reduce the reflectivity of the paint film and compensate for it.

[0138] Step S502 embodies the ideas of "inter-process collaborative optimization" and "cost-effectiveness optimization". It avoids the potential decrease in production efficiency and increase in energy consumption that may result from pursuing the ultimate precision of the drawing process. Instead, it utilizes the process flexibility of downstream processes for economical and efficient final adjustments, thereby improving overall production efficiency while ensuring the quality of the final product.

[0139] In step S503, the actual deviation being greater than the compensation error value means that δ>ϵ. This is a non-compliance criterion, which means that the deviation has exceeded the limit that the downstream painting process can effectively compensate for. If painting is forced, the gloss of the final product will inevitably be unqualified.

[0140] The system generates compensation information that includes an analysis of the causes of deviations. This compensation information is essentially a quality anomaly report and a source tracing analysis report. The system integrates all data from the entire processing flow (initial state, adjustment records, abnormal events, and final deviations) to analyze possible causes of deviations (e.g., "insufficient mold pressure adjustment response in stage two," "abnormal lubricant temperature fluctuations," "large jump in prediction by the LSTM model at second X," etc.). This compensation information is primarily used for quality traceability, responsibility identification, and providing a basis for parameter optimization, rather than for paint compensation.

[0141] Updating the pull-out adjustment parameters refers to the automatic or manual correction of control parameters (such as PID gain, adjustment range, or predicted deviation / standard deviation set in the aforementioned steps) based on the analysis of the cause of the deviation, in order to avoid similar problems in the next processing.

[0142] Triggering wire reprocessing is a quality rejection and corrective action. The system marks the non-conforming section (or roll) of wire and guides it back or arranges for re-drawing processing to ensure that no non-conforming products flow into the next process and the final product.

[0143] Step S503 does not simply remove defective products, but rather analyzes the causes and updates the parameters, turning each failure into an opportunity to optimize the control system itself, demonstrating the system's ability to continuously improve itself.

[0144] In some embodiments, the step of detecting the gloss of the adjusted wire surface and generating compensation information further includes: S504. Collect wire information, process parameters, gloss data and quality feedback from historical processing to establish a structured database; S505. Using historical data from the database, incrementally train the long short-term memory network model to optimize its prediction accuracy.

[0145] In step S504, the historical processing process refers to the entire process of all completed processing batches, regardless of whether they were successful or not.

[0146] Collect wire information, process parameters, and gloss data from historical processing. Specifically, the data collected here should include at least: Wire information: unique code, material, target wire diameter, target gloss level; Process parameters: a full-cycle sequence of timestamp-aligned drawing speed, die pressure, lubricant flow rate, and temperature; Gloss data: Complete sequences of various gloss levels, including initial, real-time, predicted, and adjusted gloss levels; Quality feedback: final actual deviation value, whether reprocessing is required, compensation information, and abnormal event records.

[0147] "Building a structured database" is not simply about piling up data, but about cleaning, associating, and organizing the aforementioned heterogeneous data according to a unified format and relationships. For example, each record uses a "unique cable code" as the primary key, linking all relevant information (static attributes, dynamic time-series data, event tags) to form a high-quality knowledge base that is easy to query, analyze, and use for machine learning.

[0148] Step S504 addresses the common problems of data silos and data loss in industrial settings, transforming each production process into a data sample that can be analyzed. This structured database serves as the sole data source and factual basis for all subsequent model optimization, parameter tuning, and knowledge mining.

[0149] In step S505, "using historical data in the database" means selecting a dataset for model retraining from the structured database according to specific rules (such as selecting recent data, data of the same material, or data containing specific abnormal patterns).

[0150] "Incremental training" is the core technique for model optimization. It does not mean rebuilding the LSTM model from scratch, but rather using new historical data to "fine-tune" or "continue training" the model while retaining its original knowledge (weights and parameters).

[0151] The training process involves inputting new data into the model, calculating the error (loss) between its predicted output and the true value, and then updating the model's network weights with a small learning rate through the backpropagation algorithm. This process enables the model to adapt to new data patterns, correct existing prediction biases, and learn new working condition characteristics without forgetting the general rules it has learned.

[0152] "Optimizing its prediction accuracy" refers to making the LSTM model more accurate and robust in predicting future gloss levels by continuously absorbing the latest production experience. For example, the model can learn the subtle effects of a newly purchased batch of raw materials on gloss levels, or the new characteristics exhibited by equipment after maintenance.

[0153] Step S505 enables the continuous evolution of the entire control system, transforming the LSTM model, the core prediction engine, from a static model that remains unchanged once deployed into a dynamic agent capable of continuously learning from production practice. Continuous optimization of prediction accuracy leads to more precise parameter tuning, fewer abnormal failures, and a sustained increase in the final pass rate.

[0154] Please refer to Figure 2 This application provides a wire surface gloss control system, which corresponds to the method in the above embodiments. Specifically, the system includes: The first acquisition module 201 is used to acquire the initial gloss and real-time gloss of the wire surface; The first control module 202 is used to input the initial gloss and real-time gloss into a trained long short-term memory network model to generate the predicted gloss within the next 2 seconds. The second control module 203 is used to calculate the first deviation between the predicted gloss and the real-time gloss at the same time node, and the second deviation between the real-time gloss at different time nodes, in order to determine the gloss change trend. The third control module 204 is used to adjust at least one parameter among drawing speed, die pressure, lubricant flow rate and temperature in stages according to the absolute value of the first deviation and the gloss change trend represented by the second deviation. The fourth control module 205 is used to detect the gloss of the adjusted wire surface and generate compensation information. The compensation information is used for gloss compensation in the fine wire painting process.

[0155] As can be seen from the above, the wire surface gloss control method and system provided in this application have the following advantages: 1) Overcoming the lag of post-event control to achieve proactive prediction and intervention: By introducing a trained Long Short-Term Memory (LSTM) network model, the system accurately predicts the gloss level for the next two seconds using initial and real-time gloss data. Combined with real-time calculated first and second deviations, a hierarchical control mechanism based on dual deviations and trend judgment is constructed. This allows the system to proactively and hierarchically adjust core process parameters such as drawing speed and die pressure before the gloss level actually exceeds the standard, based on the trend and magnitude of the predicted deviations. This fundamentally solves the core pain point of existing technologies—"post-event alarms and delayed manual adjustments"—achieving a fundamental shift from passive response to proactive prevention.

[0156] 2) Constructing a closed-loop control system that integrates hardware and software to improve control accuracy and yield: The system not only automatically adjusts equipment parameters but also automatically diagnoses (hardware faults and software parameter verification) and intelligently recovers in the event of severe anomalies. Ultimately, it ensures quality through cross-process compensation or self-iterative optimization. This closed loop reduces the gloss fluctuation range and stabilizes it within ±1.5 GU, improving the product yield and solving the problems of "no linkage between inspection data and drawing equipment" and "inability to achieve automatic control."

[0157] 3) Enhance system robustness and adaptability, reducing maintenance costs and unnecessary downtime: When data interruption occurs, the system can intelligently complete the data through linear extrapolation or correlation models, ensuring control continuity; when serious deviations occur, the system can handle different situations (optimize parameters and restart if there is no fault, repair and restart if there is a fault), avoiding blind downtime; through full-process data traceability and incremental model training, the system has the ability to continuously learn and optimize. These designs reduce unnecessary downtime caused by misjudgment or single faults, reduce reliance on manual intervention and maintenance costs, and enable the system to adapt to the needs of large-scale production of different materials and working conditions.

[0158] 4) Achieving end-to-end quality collaboration and traceability to meet compliance requirements of high-end industries: This invention extends quality control from a single process to the entire process. By generating precise compensation instructions for the downstream painting process, cross-process collaborative quality assurance is achieved. Simultaneously, all key data is linked and stored with the unique code of the wire, forming a complete quality traceability chain. This not only solves the problem of "inability to trace the root cause by associating process parameters," but also makes production process data transparent and traceable, meeting the stringent compliance and quality consistency requirements of industries such as automotive and electronics for key metal components.

[0159] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for controlling the surface gloss of wire, characterized in that, The method includes the following steps: Obtain the initial gloss and real-time gloss of the wire surface; The initial gloss and real-time gloss are input into a trained long short-term memory network model to generate the predicted gloss within the next 2 seconds. Calculate the first deviation between the predicted gloss and the real-time gloss at the same time point, and the second deviation between the real-time gloss at different time points to determine the gloss change trend; Based on the absolute value of the first deviation and the gloss change trend represented by the second deviation, at least one of the following parameters is adjusted in stages: drawing speed, die pressure, lubricant flow rate, and temperature. The gloss of the adjusted wire surface is detected and compensation information is generated. This compensation information is used to compensate for the gloss in the fine wire painting process.

2. The method for controlling the surface gloss of wires according to claim 1, characterized in that, The steps for obtaining the initial gloss and real-time gloss of the wire surface also include: Pre-treatment information is generated based on the initial gloss of the wire surface, and the pre-treatment information includes at least the pre-treatment type and the corresponding equipment parameters; Perform preprocessing operations based on the preprocessing information; Update the initial gloss and real-time gloss after the job is acquired.

3. The method for controlling the surface gloss of wires according to claim 2, characterized in that, The step of adjusting at least one parameter among drawing speed, die pressure, lubricant flow rate, and temperature in stages, based on the absolute value of the first deviation and the gloss change trend represented by the second deviation, includes: When the absolute value of the first deviation is ≤1GU and the second deviation does not show an increasing trend, the parameters are automatically fine-tuned. When the absolute value of the first deviation is 1 to 1.5 GU or the second deviation shows an intermittent increasing trend, a first-level warning is triggered and the parameters are automatically adjusted. When the absolute value of the first deviation is greater than 1.5 GU or the second deviation shows a continuous increasing trend, a level two warning is triggered and the process is suspended.

4. The method for controlling the surface gloss of wires according to claim 3, characterized in that, When the absolute value of the first deviation is greater than 1.5 GU or the second deviation shows a continuous increasing trend, the steps to trigger a second-level warning and suspend the process further include: The equipment is tested for malfunctions, and the prediction deviation and standard deviation set according to the target gloss of the wire are verified when generating the predicted gloss. When the equipment is fault-free, set new prediction deviation and standard deviation values ​​according to the actual scenario and restart the equipment; the actual scenario is defined by the wire material, target wire diameter, target gloss and current processing stability. When the equipment malfunctions, repair the equipment and then restart it. Save all current data and associate it with the cable's unique code for traceability and analysis.

5. The method for controlling the surface gloss of wires according to claim 4, characterized in that, When the equipment is fault-free, the steps of setting new prediction bias and standard deviation values ​​according to the actual scenario and restarting the equipment also include: When data interruption occurs, missing data is supplemented and adjustments are made based on valid historical data prior to the interruption, using linear extrapolation or a prediction model based on process parameter correlation. The actual scenario was re-planned based on the supplemented and adjusted data; Set new prediction bias and standard deviation based on the redesigned actual scenario and restart the device.

6. The method for controlling the surface gloss of wires according to claim 5, characterized in that, When data interruption occurs, the steps of supplementing missing data and making adjustments based on valid historical data prior to the interruption, using linear extrapolation or a prediction model based on process parameter correlation, also include: If the data interruption duration is ≤100ms, the process parameters of drawing speed and die pressure are completed by linear extrapolation based on the data change trend before the interruption. If the data interruption duration is greater than 100ms, for real-time gloss data, a mapping relationship model between gloss and drawing speed, die pressure, and lubricant parameters is established to back-calculate and complete the data.

7. The method for controlling the surface gloss of wires according to claim 6, characterized in that, When the equipment is fault-free, the steps of setting new prediction bias and standard deviation values ​​according to the actual scenario and restarting the equipment also include: When there is no data interruption, new prediction bias and standard deviation are generated based on the complete processing data of the current wire. Based on historical processing data, re-verify whether the new prediction deviation and standard deviation are within the effective parameter range of the same historical scenario, whether they meet the equipment control capability limits, and whether they match the current processing trend; Restart the device after verification.

8. The method for controlling the surface gloss of wires according to claim 7, characterized in that, The steps of detecting the gloss of the adjusted wire surface and generating compensation information also include: Calculate the actual deviation between the adjusted gloss and the target gloss; When the actual deviation is not greater than the compensation error value preset based on the wire type, a compensation instruction containing the actual deviation information is generated so that the painting mechanism can adjust the paint layer thickness and curing temperature. When the actual deviation is greater than the compensation error value, compensation information containing the cause analysis of the deviation is generated, and the drawing adjustment parameters are updated to trigger the reprocessing of the wire.

9. The method for controlling the surface gloss of wires according to claim 8, characterized in that, The steps of detecting the gloss of the adjusted wire surface and generating compensation information also include: Collect wire information, process parameters, gloss data and quality feedback from historical processing to establish a structured database; The Long Short-Term Memory Network model is incrementally trained using historical data from the structured database to optimize its prediction accuracy.

10. A wire surface gloss control system, characterized in that, The system includes: The first acquisition module (201) is used to acquire the initial gloss and real-time gloss of the wire surface; The first control module (202) is used to input the initial gloss and real-time gloss into a trained long short-term memory network model to generate the predicted gloss within the next 2 seconds. The second control module (203) is used to calculate the first deviation between the predicted gloss and the real-time gloss at the same time node, and the second deviation between the real-time gloss at different time nodes to determine the gloss change trend. The third control module (204) is used to adjust at least one parameter among drawing speed, die pressure, lubricant flow rate and temperature in stages according to the absolute value of the first deviation and the gloss change trend represented by the second deviation. The fourth control module (205) is used to detect the gloss of the adjusted wire surface and generate compensation information, which is used for gloss compensation in the fine wire painting process.

Images