Method and system for automatic control of the manufacturing process of flexible joint copper strip foils

Abstract

The present application relates to the technical field of production automation control, and discloses a flexible connection red copper strip foil production manufacturing process automation control method and system, the method comprises collecting the thickness, shape data and annealing furnace temperature of the finishing outlet section, and integrating into real-time process data set. Extract the fluctuation characteristics and classify the anomaly, predict the conductivity deviation based on the abnormal data regression analysis. The deviation and the thickness fluctuation rate are spliced into a multi-dimensional feature vector, and the smooth conductivity deviation is obtained through smoothing filtering. The historical data is combined to calculate and optimize the annealing temperature and the temperature correction coefficient, and the heating power is adjusted to obtain a stable material structure. Finally, the actual conductivity is verified, and the next cycle monitoring reference is determined through data feedback. The method can realize high stability of the conductivity of the flexible connection red copper strip foil, and reduce the production scrap rate.

Description

Technical Field

[0001] This invention relates to the field of automated control technology in manufacturing, and in particular to an automated control method and system for the manufacturing process of flexible connecting copper strip foil. Background Technology

[0002] Currently, advanced non-ferrous metal materials are key foundational materials supporting the development of strategic emerging industries such as new energy and high-end equipment manufacturing. Flexible connecting copper strip foil, as a typical representative of high-precision advanced non-ferrous metal materials, is a core component for high-current transmission in new energy vehicle battery systems. Its conductivity and thickness uniformity directly determine the battery's energy transmission efficiency and operational safety. The annealing process determines the final conductivity by controlling the internal grain structure and stress state of the material; its temperature control precision is crucial for ensuring product quality.

[0003] In a current technology, the production of flexible copper strip foil employs an independent control mode for the finishing rolling and annealing processes, with each process relying solely on preset fixed parameters for single-variable closed-loop adjustment. However, thickness, strip shape, and annealing temperature are strongly coupled. Thickness fluctuations during finishing rolling alter the material's thermal conductivity, subsequently affecting the temperature field distribution during annealing. Uneven temperature field distribution leads to inconsistent recrystallization within the material, ultimately causing conductivity fluctuations. Current technology cannot detect this multivariate coupling effect. When a slight thickness deviation occurs during finishing rolling, the annealing temperature remains at the preset value, resulting in a mismatch between temperature and the actual material state. This leads to defects such as coarse grains and stress concentration, and the deviation is amplified and propagated along the process, only becoming apparent during finished product inspection.

[0004] In summary, the lack of multivariate coupling sensing and chain reaction prediction mechanisms in existing technologies leads to poor conductivity stability and high scrap rate in the produced flexible connecting copper foil strips. Summary of the Invention

[0005] This invention provides an automated control method and system for the production process of flexible connected copper strip foil, in order to solve the problems of poor conductivity stability and high scrap rate of flexible connected copper strip foil.

[0006] In a first aspect, to address the aforementioned technical problems, this invention provides an automated control method for the manufacturing process of flexible connected copper strip foil, comprising: collecting thickness and shape data from the finishing rolling exit section, obtaining the annealing temperature in the annealing furnace, and integrating them to obtain a real-time process dataset; extracting features from the process dataset to obtain fluctuation data, classifying the fluctuation data into anomalies to obtain abnormal fluctuation data; performing regression analysis on the shape data and the annealing temperature based on the abnormal fluctuation data to obtain a predicted conductivity deviation; and performing feature splicing between the predicted conductivity deviation and the thickness fluctuation rate. The process involves: obtaining a multi-dimensional feature vector; performing smoothing filtering on the multi-dimensional feature vector to obtain a smoothed conductivity deviation; acquiring historical data on annealing temperature and conductivity; performing regression calculation on the smoothed conductivity deviation based on the historical data to obtain an optimized annealing temperature; conducting potential chain reaction analysis on the optimized annealing temperature to obtain a temperature correction coefficient; adjusting the heating power of the annealing furnace based on the temperature correction coefficient; obtaining the material structure after annealing; verifying the conductivity based on the material structure to obtain the actual conductivity; and providing data feedback based on the actual conductivity to determine the monitoring benchmark for the next cycle.

[0007] Preferably, the process of collecting thickness and shape data from the finishing mill exit section, obtaining the annealing temperature inside the annealing furnace, and integrating them to obtain a real-time process dataset includes: performing laser scanning on the finishing mill exit section to obtain thickness and shape data; performing thermoelectric sensing processing on the inside of the annealing furnace to obtain the annealing temperature; and aligning the thickness data, shape data, and annealing temperature with timestamps to obtain a real-time process dataset.

[0008] Preferably, the step of extracting features from the process dataset to obtain fluctuation data, and classifying the fluctuation data to obtain abnormal fluctuation data, includes: extracting thickness data and shape data within a preset range from the process dataset; extracting feature vectors from the thickness data and the shape data to obtain fluctuation data, wherein the fluctuation data includes thickness fluctuation rate and shape deviation rate; and classifying the thickness fluctuation rate and shape deviation rate using a preset support vector machine model to obtain abnormal fluctuation data.

[0009] Preferably, the step of performing regression analysis on the plate shape data and the annealing temperature based on the abnormal fluctuation data to obtain the predicted conductivity deviation includes: comparing the abnormal fluctuation data with a preset fluctuation threshold; if the abnormal fluctuation data does not exceed the fluctuation threshold, maintaining the current system operation state; if the abnormal fluctuation data exceeds the fluctuation threshold, performing regression analysis on the plate shape data and the annealing temperature based on a pre-acquired historical process database to obtain the predicted conductivity deviation.

[0010] Preferably, the step of performing feature concatenation processing on the predicted conductivity deviation and the thickness fluctuation rate to obtain a multi-dimensional feature vector, and performing smoothing filtering processing on the multi-dimensional feature vector to obtain a smoothed conductivity deviation, includes: normalizing the predicted conductivity deviation and the thickness fluctuation rate to obtain a standard feature sequence; concatenating the standard feature sequence to obtain a multi-dimensional feature vector; and performing Kalman filtering processing on the multi-dimensional feature vector to obtain a smoothed conductivity deviation.

[0011] Preferably, the step of acquiring historical data on annealing temperature and conductivity, performing regression calculation on the smoothed conductivity deviation based on the historical data to obtain an optimized annealing temperature, and performing potential cascading reaction analysis on the optimized annealing temperature to obtain a temperature correction coefficient includes: acquiring historical data containing annealing temperature and conductivity; performing polynomial fitting processing on the smoothed conductivity deviation based on the historical data to obtain a conductivity change trend; optimizing the smoothed conductivity deviation based on the conductivity change trend to obtain an optimized annealing temperature; predicting the temperature gradient based on the optimized annealing temperature to obtain a temperature gradient distribution; conducting a potential risk assessment based on the temperature gradient distribution to obtain a temperature correction coefficient; calculating the stress distribution based on the temperature gradient distribution to obtain a material stress distribution; and performing a safety verification based on the material stress distribution to obtain a temperature correction coefficient.

[0012] Preferably, the step of adjusting the heating power of the annealing furnace according to the temperature correction coefficient to obtain the material structure after annealing includes: inputting the temperature correction coefficient into a preset adaptive controller for control gain conversion processing to obtain a power adjustment command; adjusting the annealing furnace according to the power adjustment command to obtain an updated heating power; and heating according to the updated heating power to obtain the material structure.

[0013] Preferably, the step of verifying the conductivity based on the material structure to obtain the actual conductivity, and providing data feedback based on the actual conductivity to determine the monitoring benchmark for the next cycle includes: collecting conductivity data on the material structure through a sensor at the outlet of the annealing furnace to obtain the actual conductivity; comparing the actual conductivity with a preset target conductivity to obtain the degree of conductivity improvement; and updating the support vector machine model based on the degree of conductivity improvement to obtain the monitoring benchmark data for the next cycle.

[0014] Secondly, this invention provides an automated control system for the production process of flexible connected copper strip foil, comprising: a data acquisition module for acquiring thickness and shape data at the finishing mill exit section, obtaining the annealing temperature in the annealing furnace, and integrating them to obtain a real-time process dataset; an anomaly classification module for extracting features from the process dataset to obtain fluctuation data, classifying the fluctuation data into anomalies to obtain abnormal fluctuation data; a trend prediction module for performing regression analysis on the shape data and the annealing temperature based on the abnormal fluctuation data to obtain a predicted conductivity deviation; and a fusion filtering module for performing feature splicing processing on the predicted conductivity deviation and the thickness fluctuation rate. A multidimensional feature vector is obtained, and the multidimensional feature vector is smoothed and filtered to obtain a smoothed conductivity deviation; a temperature optimization module is used to acquire historical data of annealing temperature and conductivity, and to perform regression calculation on the smoothed conductivity deviation based on the historical data to obtain an optimized annealing temperature. Potential cascading reaction analysis is performed on the optimized annealing temperature to obtain a temperature correction coefficient; a power adjustment module is used to adjust the heating power of the annealing furnace according to the temperature correction coefficient, and to obtain the material structure after annealing; a verification feedback module is used to verify the conductivity based on the material structure to obtain the actual conductivity, and to provide data feedback based on the actual conductivity to determine the monitoring benchmark for the next cycle.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention collects multi-dimensional process data on finishing thickness, plate shape and annealing temperature, uses support vector machine algorithm to classify abnormal fluctuation patterns, and locates the starting point of chain reaction through decision boundary analysis. This solves the defects of existing technology that cannot perceive the coupling effect of multiple variables and can only detect quality problems after the fact, and realizes the early prediction of abnormal risks, effectively blocking the transmission and amplification of deviations. (2) This invention integrates conductivity deviation and thickness data, uses Kalman filter to smooth multivariate coupling noise to obtain accurate deviation, combines temperature-conductivity quantitative model to calculate initial optimization temperature, and uses Bayesian network to predict stress change to obtain temperature correction coefficient. This solves the problem of easy over-adjustment of temperature adjustment in the prior art, which causes chain reaction, and improves the accuracy of annealing temperature control. (3) This invention forms a single closed-loop control by collecting process data of the finishing rolling end in real time and combining dynamic correction of the annealing end. Based on the feedback of online measured conductivity, the model parameters and monitoring benchmark are updated. This solves the problem of independent control of a single link and lack of continuous optimization capability in the existing technology, realizes dynamic adaptive adjustment of the production process, and significantly improves the stability and consistency of conductivity of advanced non-ferrous metal materials. Attached Figure Description

[0016] Figure 1This is a schematic diagram of the automated control method for the production process of flexible connecting copper strip foil provided in the first embodiment of the present invention; Figure 2 This is a comparison chart of the convergence trend of conductivity deviation during the annealing temperature optimization process provided in the embodiments of the present invention; Figure 3 This is a schematic diagram of the automated control system structure for the production process of flexible connected copper strip foil provided in the second embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Reference Figure 1 The first embodiment of the present invention provides an automated control method for the manufacturing process of flexible connecting copper strip foil, comprising the following steps: S11: Collect thickness and shape data of the finishing mill exit section, obtain the annealing temperature in the annealing furnace, and integrate them to obtain a real-time process dataset. S12, perform feature extraction on the process dataset to obtain fluctuation data, and perform anomaly classification on the fluctuation data to obtain abnormal fluctuation data; S13, Based on the abnormal fluctuation data, regression analysis is performed on the plate shape data and the annealing temperature to obtain the predicted conductivity deviation; S14, perform feature concatenation processing on the predicted conductivity deviation and the thickness fluctuation rate to obtain a multi-dimensional feature vector, and perform smoothing filtering processing on the multi-dimensional feature vector to obtain a smoothed conductivity deviation. S15, acquire historical data of annealing temperature and conductivity, perform regression calculation on the smoothed conductivity deviation based on the historical data to obtain the optimized annealing temperature, and perform potential chain reaction analysis on the optimized annealing temperature to obtain the temperature correction coefficient. S16, The heating power of the annealing furnace is adjusted according to the temperature correction coefficient, and the material structure is obtained after annealing; S17, Conduct conductivity verification based on the material structure to obtain the actual conductivity, and provide data feedback based on the actual conductivity to determine the monitoring benchmark for the next cycle.

[0019] In step S11, thickness and shape data of the finishing mill exit section are collected, annealing temperature inside the annealing furnace is obtained, and real-time process dataset is obtained. This includes: performing laser scanning on the finishing mill exit section to obtain thickness and shape data; performing thermoelectric sensing processing on the inside of the annealing furnace to obtain the annealing temperature; and aligning the thickness data, shape data, and annealing temperature with timestamps to obtain the real-time process dataset.

[0020] It should be noted that the laser sensor array comprises 10 high-precision laser sensors evenly distributed along the width of the rolled piece. Each laser sensor acquires thickness data at a frequency of 1000Hz. The thickness measurement range is 0.5mm to 5.0mm, with a measurement accuracy of ±0.01mm. The laser sensors simultaneously acquire flatness data, recording the waviness deviation per meter of the rolled piece surface. The preset threshold for waviness deviation is ±0.2mm. The acquired thickness and flatness data are transmitted in real-time to the central processing system via industrial Ethernet. Fourier transform algorithms are used to perform frequency domain analysis on thickness fluctuations and flatness deviations, extracting the main frequency components within the 0.1Hz to 10Hz range to assess the stability of the rolling process. If the peak value of the thickness fluctuation frequency exceeds 5Hz, an alarm is automatically triggered, indicating a potential roll wear problem and linking it to subsequent equipment maintenance plans.

[0021] It should be noted that the thermocouple array contains 20 thermocouples covering the upper, lower, and central areas of the furnace. The thermocouples measure temperatures from 200℃ to 600℃. The measurement accuracy is ±1℃. Annealing temperature data is collected every 5 seconds. The temperature gradient within the furnace is calculated using a time series analysis algorithm. Temperature data from two adjacent sampling times are extracted, and the ratio of the temperature difference to the time interval is calculated to obtain the rate of change over time. Combined with the spatial coordinates of the thermocouples, the ratio of the temperature difference between adjacent thermocouples to their spatial distance is calculated to obtain the spatial temperature gradient in different areas of the furnace. If the temperature gradient exceeds 10℃ / m, the heating power is automatically adjusted to optimize the temperature field. Multivariate correlation analysis is performed on the temperature data, thickness data, and plate shape data. Principal component analysis was used to extract key variable influencing factors. A covariance matrix containing three types of data—thickness, plate shape, and temperature—was constructed. Eigenvalue decomposition was performed on the covariance matrix, and the eigenvalues ​​were sorted from largest to smallest. Principal components with a cumulative variance contribution rate exceeding 85% were extracted. The correlation coefficient between thickness and temperature was calculated to be 0.85, indicating that temperature has a significant impact on thickness uniformity.

[0022] It should be noted that, based on the unified clock of the central processing system, a corresponding timestamp is added to each set of thickness data, shape data, and annealing temperature data. The three types of data are synchronized and matched according to their timestamps, and outlier data points with a time difference exceeding 100ms are removed. For data gaps resulting from this removal, linear interpolation is used to fill them, extracting the two closest valid data points before and after the gap location. The value at the gap location is calculated based on the time distance ratio to ensure the continuity of the data sequence. The synchronized thickness data, shape data, and annealing temperature data are integrated into a real-time process dataset, which is stored in a distributed database for subsequent process optimization models to access, forming a complete closed-loop logic from data acquisition to analysis and feedback adjustment.

[0023] In step S12, feature extraction is performed on the process dataset to obtain fluctuation data, and anomaly classification is performed on the fluctuation data to obtain abnormal fluctuation data. This includes: extracting thickness data and plate shape data within a preset range from the process dataset; extracting feature vectors from the thickness data and the plate shape data to obtain fluctuation data, wherein the fluctuation data includes thickness fluctuation rate and plate shape deviation rate; and classifying the thickness fluctuation rate and plate shape deviation rate using a preset support vector machine algorithm to obtain abnormal fluctuation data.

[0024] It should be noted that the operation of extracting real-time process datasets from the distributed database is automatically triggered after each acquisition cycle, with the acquisition cycle set to 1000Hz to match the sampling frequency of the laser sensor. The extracted thickness data ranges from 0.5mm to 5.0mm, covering the standard thickness specifications of all mainstream flexible copper strip foil products. Data outside this range is invalid due to sensor malfunction or rolling mill anomalies. The wave height deviation range of the extracted plate shape data is ±0.2mm, which is the acceptable threshold for plate flatness. Exceeding this range will lead to uneven heating of the material during subsequent annealing, resulting in decreased conductivity and internal stress concentration. These data extraction ranges are determined based on the measurement range of the laser sensor and the product quality requirements of the flexible copper strip foil. The real-time acquired data is batch-processed at a preset frequency, which is set according to the real-time response requirements of the industrial control system to ensure that abnormal fluctuations can be identified and trigger subsequent processing within 1 second. The data cleaning process is performed simultaneously during the extraction process. The 3σ principle is used to remove missing values, duplicate values, and outliers that exceed the sensor's range. The cleaning threshold is determined based on the statistical results of normal production data over the past 6 months. After multiple batch verifications, this cleaning method can effectively filter more than 99.9% of invalid data, ensuring the validity of the data input into subsequent steps.

[0025] It is worth noting that time-series difference calculations were performed on the extracted thickness data, and the thickness fluctuation rate was calculated using the sliding window method. The sliding window size was set to 100 sampling points, corresponding to a time window of 0.1 seconds. This window size was verified through multiple batches of production experiments, as it can capture rapid fluctuations caused by roll vibration and tension fluctuations during production, while effectively filtering high-frequency noise generated by the sensors themselves. The difference between the maximum and minimum thickness values ​​within each window was calculated, and the difference was divided by the window time to obtain the thickness change per unit time. This change is the thickness fluctuation rate, with units of mm / s. The extracted plate shape data underwent length normalization. For the width direction plate shape data collected by 10 laser sensors at each sampling time, the root mean square error of the error compared to the standard straightness was calculated. The error was divided by the sampling length to obtain the wave height deviation per unit length. This deviation is the plate shape deviation rate, with units of mm / m. The thickness volatility and plate shape deviation rate are aligned in the time dimension. Each timestamp corresponds to a set of thickness volatility and plate shape deviation rate values. These values ​​are then concatenated to form a two-dimensional feature vector, which is the volatility data.

[0026] It is worth further explaining that a pre-trained Support Vector Machine (SVM) model is used to classify the two-dimensional feature vectors. The kernel function of this SVM model is set to a radial basis function (RBF), which is suitable for handling nonlinear classification problems in process data. The training dataset contains more than 1 million labeled production samples from the past 12 months, of which normal samples account for approximately 96.8% and abnormal samples account for approximately 3.2%. The sample annotation was completed by senior process engineers with more than 5 years of experience based on the final conductivity test results of the products. Process data corresponding to products with conductivity lower than the national standard requirements were labeled as abnormal samples. A 5-fold cross-validation method was used to optimize the model parameters. The dataset was randomly divided into 5 parts, with 4 parts used for training and 1 part used for validation in rotation. The final optimized penalty parameter C value was 10, and the kernel function parameter γ value was 0.1. The model performance was validated with multiple batches of production data, achieving a classification accuracy of over 95%, a precision of 94.2%, a recall of 93.7%, and an F1 score of 93.9%. The classification process divides the input feature vector into two categories: normal fluctuations and abnormal fluctuations. The threshold for judging abnormal fluctuations is thickness fluctuation rate ±0.05 mm / s or plate shape deviation rate ±0.1 mm / m, which is determined based on the 99.7th percentile of historical data. The output abnormal fluctuation data includes the timestamp of the abnormality, the type of abnormality, the magnitude of the abnormal value, and the corresponding original process data, providing accurate input basis for subsequent conductivity deviation prediction.

[0027] In step S13, regression analysis is performed on the plate shape data and the annealing temperature based on the abnormal fluctuation data to obtain the predicted conductivity deviation. This includes: comparing the abnormal fluctuation data with a preset fluctuation threshold; if the abnormal fluctuation data does not exceed the fluctuation threshold, maintaining the current system operation state; if the abnormal fluctuation data exceeds the fluctuation threshold, regression analysis is performed on the plate shape data and the annealing temperature based on a pre-acquired historical process database to obtain the predicted conductivity deviation.

[0028] It should be noted that the comparison operation between abnormal fluctuation data and the preset fluctuation threshold is performed synchronously with the generation operation of abnormal fluctuation data, and the comparison frequency is set to 5000 times per minute, consistent with the batch processing frequency. The preset fluctuation threshold includes two independent judgment conditions: thickness fluctuation rate ±0.07mm / s and plate shape deviation rate ±0.15mm / m. The two conditions are ORed, and if either condition is met, it is judged as exceeding the threshold. This threshold is determined based on the conductivity quality requirements of flexible connecting copper foil strips, combined with the statistical analysis results of production data over the past 12 months. After multiple batches of on-site verification, when the thickness fluctuation rate exceeds ±0.07mm / s or the plate shape deviation rate exceeds ±0.15mm / m, the product conductivity failure rate will increase from 0.3% to over 4.5%. Therefore, this value is set as the critical judgment condition to trigger the subsequent conductivity prediction process. During the comparison process, each set of abnormal fluctuation data is compared with two thresholds one by one, and the temporal changes of the comparison results are recorded synchronously. A continuous judgment logic is adopted. Only when the abnormal fluctuation data exceeds the threshold for three or more consecutive sampling periods is it judged as a valid abnormality and the subsequent processing is triggered, so as to avoid false triggering caused by instantaneous noise of the sensor.

[0029] It is worth noting that when abnormal fluctuation data does not exceed the preset fluctuation threshold, the current production process is determined to be in a stable and controllable state. The current system operation is maintained, and all process parameters for the finishing rolling and annealing processes remain unchanged. Simultaneously, the current abnormal fluctuation data, comparison results, and corresponding original process data are synchronously stored in the local historical database for updating the training dataset of the support vector machine classification model and dynamically calibrating the fluctuation threshold. While maintaining the current operating state, the complete process of process data acquisition, feature extraction, anomaly classification, and threshold comparison continues to be executed, ensuring that any abnormal fluctuations occurring during production can be identified in a timely manner.

[0030] It is worth further explaining that when abnormal fluctuation data exceeds the preset fluctuation threshold, the conductivity deviation prediction process is automatically triggered. Historical correlation data between plate shape data and annealing temperature is extracted from the historical process database, and a prediction model is constructed using a random forest regression algorithm. The training dataset for this model contains over 2 million labeled production samples from the past 18 months. Sample features include plate shape deviation rate, annealing temperature, temperature change rate, and corresponding time series features. The sample labels are the actual conductivity deviation values ​​of the corresponding products, and the label data is provided by a high-precision conductivity testing device in an offline laboratory. Model parameters were optimized and determined using a grid search combined with 5-fold cross-validation. The final settings included 100 decision trees, a maximum depth of 15, a minimum number of sample splits of 2, and a minimum number of leaf nodes of 1. The model performance was validated using multiple batches of production data, with prediction errors controlled within ±0.5%, a coefficient of determination R² of 0.85, and a mean absolute error of 0.12 MS / m. During the regression analysis, the current plate shape data and annealing temperature data are used to construct a feature matrix, which is then input into the trained random forest model to output the predicted conductivity deviation. For example, in a certain implementation scenario, the current plate shape deviation rate is 0.2 mm / m, the annealing temperature fluctuation is ±12℃, and the predicted conductivity deviation is calculated by substituting into the model as -0.8 MS / m. This value will serve as the core basis for subsequent process parameter adjustments.

[0031] It should be noted that the preset fluctuation thresholds in this step (thickness fluctuation rate ±0.07mm / s, plate shape deviation rate ±0.15mm / m) and the anomaly detection thresholds of the support vector machine model in step S12 (thickness fluctuation rate ±0.05mm / s, plate shape deviation rate ±0.1mm / m) have different functions: the former is used to trigger the conductivity deviation prediction process, with a wider range of values ​​to avoid oversensitivity leading to frequent invalid predictions; the latter is used to classify abnormal fluctuation data, with a narrower range of values ​​to ensure that early anomalies are identified in a timely manner. The two form a hierarchical early warning mechanism, together constituting a complete anomaly detection system.

[0032] In step S14, the predicted conductivity deviation and the thickness fluctuation rate are subjected to feature concatenation processing to obtain a multi-dimensional feature vector, and the multi-dimensional feature vector is subjected to smoothing filtering processing to obtain a smoothed conductivity deviation. This includes: normalizing the predicted conductivity deviation and the thickness fluctuation rate to obtain a standard feature sequence; concatenating the standard feature sequence to obtain a multi-dimensional feature vector; and performing Kalman filtering processing on the multi-dimensional feature vector to obtain a smoothed conductivity deviation.

[0033] It should be noted that min-max normalization was performed on the predicted conductivity deviation and thickness fluctuation rate to obtain a standard feature sequence. Normalization maps features of different dimensions to a unified range of 0 to 1, eliminating the influence of dimensional differences on subsequent filtering. The normalization range for the predicted conductivity deviation was set to -2MS / m to 2MS / m, based on 99.9% of the conductivity deviation distribution in historical production data. Statistically, the thickness fluctuation rate in normal production is typically within ±0.05mm / s; therefore, the normalization range for the thickness fluctuation rate was set to -0.1mm / s to 0.1mm / s, corresponding to the maximum possible fluctuation range of the thickness fluctuation rate. The normalization process was performed using a sliding window method, with the window size consistent with the feature extraction sliding window of 100 sampling points to ensure temporal consistency in data processing.

[0034] It is worth noting that the normalized predicted conductivity deviation standard sequence and the thickness fluctuation rate standard sequence are aligned in time dimension, and a feature concatenation operation is performed to obtain a two-dimensional multidimensional feature vector. Each timestamp corresponds to a set of normalized predicted conductivity deviation values ​​and thickness fluctuation rate values, which are concatenated column-wise to form a feature matrix of dimension 2×N, where N is the number of sampling points. The feature concatenation process strictly maintains the one-to-one correspondence of timestamps, eliminating mismatched data points with a time difference exceeding 100ms to ensure the temporal synchronization of the feature vector. The concatenated multidimensional feature vector simultaneously contains conductivity deviation information and thickness fluctuation information, comprehensively characterizing the combined impact of multivariate coupling on material properties.

[0035] Kalman filtering is performed on the multidimensional eigenvectors to obtain a smoothed conductivity deviation. The state vector of the Kalman filter is set to two dimensions, corresponding to the conductivity deviation state and the thickness fluctuation state, respectively. The state transition matrix A is a 2×2 identity matrix, representing that the state remains unchanged in the absence of external disturbances. The observation matrix H is a 2×2 identity matrix, indicating that both state components can be directly observed. The state noise covariance matrix Q is set as a diagonal matrix, with diagonal elements of 0.01 and 0.005, which are determined based on the statistical characteristics of sensor measurement noise. The measurement noise covariance matrix R is also set as a diagonal matrix, with diagonal elements of 0.05 and 0.02. It should be noted that the diagonal elements 0.01 and 0.005 of the state noise covariance matrix Q are obtained by statistically analyzing the variance of the differences between state variables at adjacent time points in historical normal production data. In the diagonal elements 0.05 and 0.02 of the measurement noise covariance matrix R, the observation noise of thickness fluctuation rate is determined by the nominal accuracy of the laser sensor, while the observation noise of predicted conductivity deviation is determined by the prediction residual variance of the regression model on the validation set. Compared to low-pass filtering, Kalman filtering can dynamically adjust the filter gain according to the statistical characteristics of signal and noise, maintaining a smoothing effect even when thickness fluctuations are severe. The filtering process sequentially executes a prediction step and an update step. In the prediction step, the prior state estimate at the current time is calculated based on the posterior state estimate and state transition matrix of the previous time step, and the prior error covariance matrix is ​​also calculated. In the update step, the Kalman gain is calculated based on the measurement value at the current time step, and then the prior state estimate is corrected using the Kalman gain to obtain the posterior state estimate, while the posterior error covariance matrix is ​​updated. After verification in multiple production batches, Kalman filtering can effectively decouple the effects of multivariate coupling, reducing the error of conductivity deviation estimation by more than 30%. For example, in a certain implementation scenario, the original predicted conductivity deviation was -1.0 MS / m, which was adjusted to -0.8 MS / m after Kalman filtering and smoothing. The weight of the influence of thickness fluctuation on conductivity was reduced from 0.75 to 0.2.

[0036] It should be noted that the first dimension extracted from the output state vector of the Kalman filter is the smoothed conductivity deviation. This smoothed conductivity deviation data will serve as the core input parameter for subsequent annealing temperature optimization. Intermediate state data generated during the filtering process will be synchronously stored in the historical process database for periodic optimization of the Kalman filter parameters. Every 24 hours, the system re-estimates the state noise covariance and measurement noise covariance using the maximum likelihood estimation method based on the filtering error data from the past 24 hours, ensuring the long-term operational stability of the filter.

[0037] In step S15, historical data on annealing temperature and conductivity are acquired. Regression calculations are performed on the smoothed conductivity deviation based on the historical data to obtain an optimized annealing temperature. Potential cascading reaction analysis is then performed on the optimized annealing temperature to obtain a temperature correction coefficient. This includes: acquiring historical data containing annealing temperature and conductivity; performing polynomial fitting on the smoothed conductivity deviation based on the historical data to obtain a conductivity change trend; optimizing the smoothed conductivity deviation based on the conductivity change trend to obtain an optimized annealing temperature; predicting the temperature gradient at the optimized annealing temperature to obtain a temperature gradient distribution; calculating the stress distribution based on the temperature gradient distribution to obtain a material stress distribution; and performing a safety check based on the material stress distribution to obtain a temperature correction coefficient.

[0038] It should be noted that the operation of obtaining historical data including annealing temperature and conductivity is automatically executed after triggering the temperature optimization process. Historical data is stored in a distributed process database, containing over 3 million labeled production samples from the past 24 months. Each sample includes annealing temperature, temperature change rate, plate shape deviation rate, thickness fluctuation rate, and the corresponding actual conductivity value. An outlier removal operation is performed concurrently during data extraction. Box plots are used to remove invalid samples with temperatures exceeding the safe range of 400℃ to 500℃ and conductivity exceeding the normal range of 55MS / m to 59MS / m, ensuring the validity of the fitted data. A third-order polynomial fitting method is used to establish a quantitative relationship between annealing temperature and conductivity, obtaining the conductivity change trend. The third-order polynomial can accurately characterize the nonlinear relationship between temperature and conductivity during the recrystallization process of copper, avoiding the problems of insufficient fitting by low-order polynomials or overfitting by high-order polynomials. Validated by multiple batches of production data, the coefficient of determination R² of this fitting relationship reaches 0.92, and the mean absolute error is 0.18MS / m, accurately reflecting the influence of annealing temperature changes on conductivity.

[0039] Subsequently, the smoothed conductivity deviation is optimized based on the conductivity variation trend. The objective function of the optimization is to minimize the smoothed conductivity deviation, with the constraint that the annealing temperature must be within the safe range of 400℃ to 500℃. The gradient descent method is used for optimization calculation, with the current actual annealing temperature as the initial value, an iteration step size of 0.5℃, and a convergence condition that the difference in conductivity deviation between two adjacent iterations is less than 0.01 MS / m. The current annealing temperature is substituted into a third-order polynomial to obtain the current predicted conductivity value, and its deviation from the target conductivity is calculated. Then, the derivative of the objective function with respect to temperature is calculated, and the annealing temperature is updated according to the direction of the derivative and the iteration step size. This process is repeated until the convergence condition is met. During the optimization process, the smoothed conductivity deviation is substituted into the conductivity variation trend to calculate the temperature adjustment required to bring the conductivity deviation back to the acceptable range, thus obtaining the initial optimized annealing temperature.

[0040] It is worth further explaining that a three-dimensional heat conduction model of the annealing furnace was established using the finite element method (FEM). The model input parameters included the initial optimized annealing temperature, heating power distribution, foil conveyor speed, and material thermophysical parameters. The heat conduction governing equation is: Where ρ is the density of copper (8960 kg / m³), c is the specific heat capacity of copper (385 J / (kg·K)), k is the thermal conductivity of copper (401 W / (m·K)), and q is the heating power density. Boundary conditions were set as follows: furnace wall convective heat transfer coefficient 15 W / (m²·K), ambient temperature 25℃, foil inlet temperature 25℃, and initial condition: the entire furnace body's initial temperature was room temperature (25℃). During the simulation, the annealing furnace was divided into 100 elements along its length, 20 elements along its width, and 5 elements along its thickness. The temperature change of each element over time was calculated, ultimately obtaining the overall temperature gradient distribution within the furnace. The predicted temperature gradient distribution data includes the temperature differences between the upper and lower, middle, and inlet / outlet areas of the furnace body, in units of ℃ / m. For example, in one implementation scenario, the initial optimized annealing temperature was 443℃, corresponding to a predicted temperature gradient of 13℃ / m, exceeding the safety threshold of 8℃ / m, indicating a risk of stress concentration due to uneven temperature distribution.

[0041] Next, a Bayesian network model was used for stress distribution calculation and safety verification. This model includes five nodes: annealing temperature, temperature gradient, stress distribution, grain size, and conductivity. The probabilistic relationships between the nodes were learned based on historical production data and experimental data on material mechanical properties. Laboratory control experimental data determined the physical constraints and influence boundaries of the causal relationships between variables, while industrial production data quantified the probability distribution characteristics of multivariate coupling in real-world scenarios. The model training dataset contains 500,000 labeled samples from the past 12 months. Each sample consists of process parameters collected in real-time from the production site and material performance data tested offline in the laboratory. All testing equipment was calibrated. During dataset construction, outlier samples were removed and a small amount of missing data was filled to ensure data quality. Maximum likelihood estimation was used for parameter learning. For all continuous nodes, it was assumed that they follow a Gaussian distribution under the condition of their parent node. The conditional probability distribution parameters of each node were solved by maximizing the likelihood function of the training data. After training, the model underwent 5-fold cross-validation, achieving a prediction accuracy of 91.5%. During the evaluation process, the predicted temperature gradient distribution is input into the Bayesian network model to calculate the changes in peak stress and grain size uniformity caused by temperature adjustment, thereby determining the temperature correction coefficient. Specifically, the temperature gradient values ​​(in °C / m) calculated by finite element method are discretized into three levels: low (<5 °C / m), medium (5-10 °C / m), and high (>10 °C / m). The stress distribution nodes in the Bayesian network are discretized into three states: safe, critical, and dangerous. The conditional probability table P(stress|temperature gradient) between nodes is determined statistically based on historical finite element simulation data, as exemplarily as follows: when the temperature gradient is low, the probability of safe stress is 0.9, the probability of critical stress is 0.1, and the probability of dangerous stress is 0; when the temperature gradient is medium, the probability of safe stress is 0.3, the probability of critical stress is 0.6, and the probability of dangerous stress is 0.1; when the temperature gradient is high, the probability of safe stress is 0, the probability of critical stress is 0.3, and the probability of dangerous stress is 0.7. During safety verification, if the posterior probability of a dangerous stress distribution exceeds 0.5, it is deemed unsafe; if the probability of safety is greater than 0.8, it is deemed safe; all other cases are considered critical. The temperature correction coefficient is determined based on the safety verification results: if safe, the correction coefficient = 1; if critical, the correction coefficient = 0.95; if dangerous, the correction coefficient = 1 - 0.3 × P(dangerous), where P(dangerous) is obtained through Bayesian network inference. The temperature correction coefficient ranges from 0.85 to 1.15. When the predicted peak stress exceeds the safety threshold, the correction coefficient is less than 1, reducing the temperature adjustment range; when the predicted temperature gradient is lower than the safety threshold, the correction coefficient is greater than 1, appropriately increasing the adjustment range.For example, in the above implementation scenario, the actual annealing temperature is 450℃, the initial optimized temperature is 443℃, the initial temperature adjustment is -7℃, and the predicted peak stress is 85MPa, exceeding the safety threshold of 80MPa. Therefore, the calculated temperature correction coefficient is 0.92, and the actual temperature adjustment is 0.92 times the initial temperature adjustment, which equals -6.44℃, i.e., adjusting from 450℃ to 443.6℃. The corresponding predicted temperature gradient decreases to 8℃ / m, and the peak stress decreases to 79.4MPa, meeting the safety requirements. For example, as shown... Figure 2 As shown, after iteration using the third-order polynomial provided in this embodiment, it can be clearly seen that the conductivity deviation convergence effect is better.

[0042] In step S16, the heating power of the annealing furnace is adjusted according to the temperature correction coefficient to obtain the material structure after annealing. This includes: inputting the temperature correction coefficient into a preset adaptive controller for control gain conversion to obtain a power adjustment command; adjusting the annealing furnace according to the power adjustment command to obtain an updated heating power; and heating according to the updated heating power to obtain the material structure.

[0043] It should be noted that the initial parameters of the adaptive proportional-integral-derivative (PID) controller were calibrated based on the thermal inertia characteristics of the annealing furnace equipment and historical production data. Calibration was completed using a trial-and-error method combined with multiple batches of production tests. First, the integral and derivative parameters were set to 0. The proportional parameter was gradually increased, and the temperature control response was observed to determine the range of proportional parameters that ensured system stability and a fast response. Then, the integral parameter was introduced to eliminate steady-state error. Finally, the derivative parameter was introduced to suppress temperature overshoot. After more than 50 production tests under different operating conditions, the initial proportional parameters were finally determined. The integral parameter is 2.5. The differential parameter is 0.3. The value is 0.1. The control gain conversion process follows a linear mapping relationship, and the power adjustment amount... The calculation formula is: ;in This represents the difference between the final target temperature and the current actual temperature. A temperature correction factor is used to calculate the final target temperature, which is the initial optimized annealing temperature multiplied by the temperature correction factor. After verification in multiple production batches, the temperature control accuracy of this adaptive controller can be stably maintained within ±1℃, with a temperature overshoot of less than 2℃, effectively avoiding the impact of drastic temperature fluctuations on the internal structure of the material.

[0044] It is worth noting that the power adjustment command is sent to the zoned power adjustment module of the annealing furnace to execute the heating power adjustment operation and obtain the updated heating power. The annealing furnace adopts a design with 5 independent heating zones along its length, each equipped with an independent thyristor power adjustment module and temperature feedback sensor. The power adjustment command contains the target power value for each heating zone. The adjustment process adopts a closed-loop control method, collecting the actual temperature data of each heating zone every 2 seconds and comparing it with the target temperature of the corresponding heating zone to dynamically adjust the power output. The power adjustment range is limited to ±15% of the initial power to avoid sudden power changes that would cause violent fluctuations in the furnace temperature field. During the adjustment process, the temperature gradient change in the furnace is monitored simultaneously. When the temperature gradient in any area exceeds the safety threshold of 12℃ / m, the power adjustment step size is automatically halved to ensure a smooth temperature transition.

[0045] Subsequently, the copper strip foil in the annealing furnace was continuously heated according to the updated heating power to obtain a stable material structure. During the heating process, the copper strip foil passed through the annealing furnace at a constant speed, completing the recrystallization annealing process within a temperature range of 400℃ to 500℃. The system, through its built-in material structure analysis model, simulated the changes in heat distribution and the evolution of the internal structure of the material in real time based on the finite element method, predicting the grain size distribution and residual stress level of the material. When the temperature of all heating zones in the furnace stabilized within ±1℃ of the final target temperature, and the overall temperature gradient was less than 8℃ / m and lasted for more than 30 seconds, the material structure was determined to have reached a stable state. The stable material structure was characterized by uniform grain size, low and uniform residual stress distribution. According to statistics from multiple batches of production data, after adjusting the heating power using this method, the grain size uniformity of the material improved by 3.4%, and the peak residual stress decreased by 5.6MPa, effectively avoiding defects such as brittleness and cracking caused by uneven temperature. After heating is completed, the material structure data will be synchronously stored in the historical process database to update the parameters of the adaptive controller and the material structure analysis model, forming a closed-loop feedback mechanism for power adjustment and structural output.

[0046] In step S17, conductivity verification is performed based on the material structure to obtain the actual conductivity. Data feedback is then performed based on the actual conductivity to determine the monitoring benchmark for the next cycle. This includes: collecting conductivity data of the material structure through a sensor at the annealing furnace outlet to obtain the actual conductivity; comparing the actual conductivity with a preset target conductivity to obtain the degree of conductivity improvement; and updating the support vector machine algorithm based on the degree of conductivity improvement to obtain the monitoring benchmark data for the next cycle.

[0047] It should be noted that a high-precision contact conductivity sensor is deployed 2 meters from the furnace outlet to continuously collect conductivity data on the surface of the annealed copper strip foil, obtaining the actual conductivity. The sensor's measurement range is 50 MS / m to 65 MS / m, with a measurement accuracy of ±0.1 MS / m and a sampling frequency of 0.2 Hz, synchronized with the strip foil's running speed to ensure at least one conductivity data point is collected per meter of strip foil. During the acquisition process, a sliding window averaging filter method is used to preprocess the raw data, with the window size set to 5 sampling points to remove outliers exceeding the 3σ range and filter out measurement errors caused by sensor contact noise and the oxide layer on the strip foil surface. The preprocessed actual conductivity data is correlated with the corresponding process parameter data through timestamps to ensure that each conductivity data point can be traced back to the corresponding finishing thickness, shape, and annealing temperature parameters.

[0048] It should be further explained that the actual conductivity after preprocessing is compared point-by-point with the preset target conductivity, and simultaneously compared with the predicted conductivity to calculate the degree of conductivity improvement. The target conductivity is set according to the national standard GB / T2059-2017 for flexible connecting copper strip foil and customer-customized requirements. In this solution, the target conductivity is 58.0 MS / m, with an allowable deviation range of ±0.5 MS / m. The conductivity improvement is calculated by dividing the difference between the actual conductivity and the predicted conductivity by the difference between the target conductivity and the predicted conductivity. When the conductivity improvement is greater than or equal to 0.8, the temperature adjustment is considered effective; when the conductivity improvement is greater than or equal to 0.5 and less than 0.8, the adjustment effect is considered average; when the conductivity improvement is less than 0.5, the adjustment is considered ineffective, triggering an emergency update process for model parameters. For example, in a certain implementation scenario, the predicted conductivity is 56.7 MS / m, the actual conductivity is 57.8 MS / m, and the target conductivity is 58.0 MS / m. The calculated conductivity improvement is 0.846, indicating that the temperature adjustment effect is good.

[0049] It is worth noting that the support vector machine (SVM) classification algorithm is incrementally updated based on the improvement in conductivity to obtain the monitoring baseline data for the next period. The incremental update employs an online learning approach, adding only samples from the current production batch with a conductivity improvement of less than 0.8 to the training dataset, thus avoiding the computational resource consumption of a full update. During the update process, the radial basis function kernel of the SVM remains unchanged, and the penalty parameter and kernel function parameters are re-optimized using a grid search method. The parameter search range is between 1 and 20 for the penalty parameter and between 0.01 and 1 for the kernel function parameter, with the optimization objective being to maximize the model's F1 score. The model update frequency is once every 10 production batches; when three consecutive batches show an improvement of less than 0.5, a full update is immediately performed. Based on the updated SVM model, the anomaly detection thresholds for thickness fluctuation rate and plate shape deviation rate are recalculated to obtain the monitoring baseline data for the next period. For example, in the above implementation scenario, the updated penalty parameter was adjusted from 10 to 20, the kernel function parameter from 0.1 to 0.5, and the model classification accuracy improved from 95.2% to 95.7%. The corresponding thickness fluctuation monitoring benchmark was adjusted to ±0.068 mm / s, and the plate shape deviation monitoring benchmark was adjusted to ±0.145 mm / m. This significant parameter adjustment is due to the emergence of anomalous patterns in the newly added samples that are significantly different from historical support vectors, leading to a large shift in the decision boundary. The specific shift can be verified by solving the SVM dual problem. The updated monitoring benchmark data is synchronously distributed to the process monitoring module for anomaly fluctuation detection in the next production cycle.

[0050] Finally, all collected actual conductivity data, improvement calculation results, and model update parameters will be synchronously stored in the distributed historical process database, forming a complete process optimization archive to provide data support for subsequent model iterations and process parameter optimization. The system will automatically generate a quality analysis report for this production cycle, including key indicators such as conductivity pass rate, temperature adjustment effectiveness, and model performance changes, for process engineers' reference.

[0051] In summary, this invention discloses an automated control method for the production process of flexible copper strip foil, comprising: collecting thickness and shape data from the finishing mill exit section; obtaining the annealing temperature in the annealing furnace; integrating these data to obtain a real-time process dataset; extracting features from the process dataset to obtain fluctuation data; classifying the fluctuation data into anomalies to obtain abnormal fluctuation data; performing regression analysis on the shape data and the annealing temperature based on the abnormal fluctuation data to obtain a predicted conductivity deviation; performing feature concatenation on the predicted conductivity deviation and the thickness fluctuation rate to obtain a multi-dimensional feature vector; and performing smoothing filtering on the multi-dimensional feature vector to obtain a smoothed conductivity deviation; acquiring historical data on annealing temperature and conductivity; performing regression calculation on the smoothed conductivity deviation based on the historical data to obtain an optimized annealing temperature; performing potential chain reaction analysis on the optimized annealing temperature to obtain a temperature correction coefficient; adjusting the heating power of the annealing furnace based on the temperature correction coefficient; obtaining the material structure after annealing; verifying the conductivity based on the material structure to obtain the actual conductivity; and providing data feedback based on the actual conductivity to determine the monitoring benchmark for the next cycle. This invention collects process data through multiple sensors, combines multiple algorithms to achieve anomaly prediction and precise temperature control, and constructs a dual closed-loop feedback mechanism to achieve high conductivity stability of flexible copper foil strips and reduce production scrap rate.

[0052] Reference Figure 3 The second embodiment of the present invention provides an automated control system for the production process of flexible copper strip foil, comprising: a data acquisition module for acquiring thickness and shape data at the exit of the finishing rolling section, obtaining the annealing temperature in the annealing furnace, and integrating them to obtain a real-time process dataset; an anomaly classification module for extracting features from the process dataset to obtain fluctuation data, classifying the fluctuation data into anomalies to obtain abnormal fluctuation data; a trend prediction module for performing regression analysis on the shape data and the annealing temperature based on the abnormal fluctuation data to obtain a predicted conductivity deviation; and a fusion filtering module for performing feature splicing processing on the predicted conductivity deviation and the thickness fluctuation rate. The process involves several steps: First, a multi-dimensional feature vector is obtained, and this vector is smoothed and filtered to obtain a smoothed conductivity deviation. Second, a temperature optimization module acquires historical data on annealing temperature and conductivity. Based on this historical data, regression calculations are performed on the smoothed conductivity deviation to obtain an optimized annealing temperature. Potential cascading reaction analysis is then performed on the optimized annealing temperature to obtain a temperature correction coefficient. Third, a power adjustment module adjusts the heating power of the annealing furnace according to the temperature correction coefficient, resulting in a material structure after annealing. Fourth, a verification feedback module verifies the conductivity based on the material structure to obtain the actual conductivity. Based on the actual conductivity, data feedback is provided to determine the monitoring benchmark for the next cycle.

[0053] It should be noted that the automated control system for the production process of flexible connected copper strip foil provided in this embodiment of the invention is used to execute all the process steps of the automated control method for the production process of flexible connected copper strip foil in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0054] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0055] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for the automatic control of the manufacturing process of flexible jointed copper strip foils, characterized in that, include: The thickness and shape data of the finishing mill exit section are collected, the annealing temperature in the annealing furnace is obtained, and the data are integrated to obtain a real-time process dataset. Feature extraction is performed on the process dataset to obtain fluctuation data. Anomaly classification is then performed on the fluctuation data to obtain abnormal fluctuation data. The fluctuation data includes thickness fluctuation rate and plate shape deviation rate. Based on the abnormal fluctuation data, regression analysis is performed on the plate shape data and the annealing temperature to obtain predicted conductivity deviation. Feature concatenation is performed on the predicted conductivity deviation and the thickness fluctuation rate to obtain a multi-dimensional feature vector. This multi-dimensional feature vector is then smoothed and filtered to obtain a smoothed conductivity deviation. Historical data on annealing temperature and conductivity are acquired. Regression calculation is performed on the smoothed conductivity deviation based on the historical data to obtain an optimized annealing temperature. Potential cascading reaction analysis is performed on the optimized annealing temperature to obtain a temperature correction coefficient. Based on the temperature correction coefficient, the heating power of the annealing furnace is adjusted, resulting in a material structure after annealing. Conductivity verification is performed based on the material structure to obtain the actual conductivity. Data feedback is then performed based on the actual conductivity to determine the monitoring benchmark for the next cycle.

2. The automated control method for the production process of flexible connected copper strip foil according to claim 1, characterized in that, The process of collecting thickness and shape data from the finishing mill exit section, obtaining the annealing temperature inside the annealing furnace, and integrating them to obtain a real-time process dataset includes: performing laser scanning on the finishing mill exit section to obtain thickness and shape data; performing thermoelectric sensing processing on the inside of the annealing furnace to obtain the annealing temperature; and aligning the thickness data, shape data, and annealing temperature with timestamps to obtain a real-time process dataset.

3. The automated control method for the production process of flexible connected copper strip foil according to claim 1, characterized in that, The step of extracting features from the process dataset to obtain fluctuation data, and classifying the fluctuation data to obtain abnormal fluctuation data, includes: extracting thickness data and plate shape data within a preset range from the process dataset; extracting feature vectors from the thickness data and the plate shape data to obtain fluctuation data, wherein the fluctuation data includes thickness fluctuation rate and plate shape deviation rate; and classifying the thickness fluctuation rate and plate shape deviation rate using a preset support vector machine model to obtain abnormal fluctuation data.

4. The automated control method for the production process of flexible connected copper strip foil according to claim 1, characterized in that, The step of performing regression analysis on the plate shape data and the annealing temperature based on the abnormal fluctuation data to obtain the predicted conductivity deviation includes: comparing the abnormal fluctuation data with a preset fluctuation threshold; if the abnormal fluctuation data does not exceed the fluctuation threshold, maintaining the current system operation state; if the abnormal fluctuation data exceeds the fluctuation threshold, performing regression analysis on the plate shape data and the annealing temperature based on a pre-acquired historical process database to obtain the predicted conductivity deviation.

5. The automated control method for the production process of flexible connected copper strip foil according to claim 1, characterized in that, The step of performing feature concatenation processing on the predicted conductivity deviation and the thickness fluctuation rate to obtain a multi-dimensional feature vector, and then performing smoothing filtering processing on the multi-dimensional feature vector to obtain a smoothed conductivity deviation, includes: normalizing the predicted conductivity deviation and the thickness fluctuation rate to obtain a standard feature sequence; concatenating the standard feature sequence to obtain a multi-dimensional feature vector; and performing Kalman filtering processing on the multi-dimensional feature vector to obtain a smoothed conductivity deviation.

6. The automated control method for the production process of flexible connected copper strip foil according to claim 1, characterized in that, The process of acquiring historical data on annealing temperature and conductivity, performing regression calculations on the smoothed conductivity deviation based on the historical data to obtain an optimized annealing temperature, and conducting potential cascading reaction analysis on the optimized annealing temperature to obtain a temperature correction coefficient includes: acquiring historical data containing annealing temperature and conductivity; performing polynomial fitting processing on the smoothed conductivity deviation based on the historical data to obtain a conductivity change trend; optimizing the smoothed conductivity deviation based on the conductivity change trend to obtain an optimized annealing temperature; predicting the temperature gradient at the optimized annealing temperature to obtain a temperature gradient distribution; calculating the stress distribution based on the temperature gradient distribution to obtain a material stress distribution; and performing a safety verification based on the material stress distribution to obtain a temperature correction coefficient.

7. The automated control method for the production process of flexible connected copper strip foil according to claim 1, characterized in that, The step of adjusting the heating power of the annealing furnace according to the temperature correction coefficient to obtain the material structure after annealing includes: inputting the temperature correction coefficient into a preset adaptive controller for control gain conversion processing to obtain a power adjustment command; adjusting the annealing furnace according to the power adjustment command to obtain an updated heating power; and heating according to the updated heating power to obtain the material structure.

8. The automated control method for the production process of flexible connected copper strip foil according to claim 3, characterized in that, The step of verifying the conductivity based on the material structure to obtain the actual conductivity, and then providing data feedback based on the actual conductivity to determine the monitoring benchmark for the next cycle includes: collecting conductivity data of the material structure through a sensor at the outlet of the annealing furnace to obtain the actual conductivity; comparing the actual conductivity with a preset target conductivity to obtain the degree of conductivity improvement; and updating the support vector machine model based on the degree of conductivity improvement to obtain the monitoring benchmark data for the next cycle.

9. An automated control system for the production process of flexible connected copper strip foil, characterized in that, include: The data acquisition module is used to collect thickness and shape data of the finishing mill exit section, obtain the annealing temperature in the annealing furnace, and integrate them to obtain real-time process datasets. abnormal The classification module is used to extract features from the process dataset to obtain fluctuation data, and to classify the fluctuation data into anomalies to obtain abnormal fluctuation data. The trend prediction module is used to perform regression analysis on the plate shape data and the annealing temperature based on the abnormal fluctuation data to obtain the predicted conductivity deviation. The system comprises the following modules: a fusion filtering module, which performs feature concatenation processing on the predicted conductivity deviation and the thickness fluctuation rate to obtain a multi-dimensional feature vector, and then performs smoothing filtering on the multi-dimensional feature vector to obtain a smoothed conductivity deviation; a temperature optimization module, which acquires historical data on annealing temperature and conductivity, performs regression calculation on the smoothed conductivity deviation based on the historical data to obtain an optimized annealing temperature, and performs potential chain reaction analysis on the optimized annealing temperature to obtain a temperature correction coefficient; a power adjustment module, which adjusts the heating power of the annealing furnace according to the temperature correction coefficient, and obtains the material structure after annealing; and a verification feedback module, which verifies the conductivity based on the material structure to obtain the actual conductivity, and provides data feedback based on the actual conductivity to determine the monitoring benchmark for the next cycle.

Images