A pole piece deviation correction control method for an automatic cutting lithium battery winding machine

Abstract

The present invention provides a pole piece correction control method for an automatic cutting lithium battery winding machine, the method comprising: installing a width sensor on the winding machine, acquiring width data of the currently wound pole piece in real time, transmitting the acquired width data to a correction control system as an input parameter for control decision-making; while adjusting the control strategy, analyzing historical production data through a machine learning algorithm to obtain an optimal control parameter combination under different pole piece widths, forming a self-learning parameter optimization model for guiding correction control decisions for subsequent batches of products; during the winding process, continuously monitoring the offset and motion state of the pole piece, filtering the collected data through a Kalman filter algorithm, eliminating interference noise, improving the accuracy of offset detection, and providing a reliable feedback signal for a correction actuator.

Description

Technical Field

[0001] The present invention relates to the field of information technology, in particular to the field of lithium battery winding machines, and specifically to a pole piece deviation correction control method of an automatic cutting lithium battery winding machine. Background Art

[0002] During the production of lithium-ion battery winding machines, fluctuations in electrode width pose a significant challenge to corrective control. The significant variations in electrode width between batches lead to variations in the electrode's motion characteristics during winding. Narrow electrode widths are prone to frequent deviation, while wide electrode widths are more susceptible to wrinkling during high-speed winding. Width fluctuations not only affect the control accuracy of the corrective actuator but also complicate the design of the control algorithm.

[0003] Traditional deviation correction control methods are mostly based on the motion model of fixed-width pole pieces and cannot adapt to width fluctuations. When the actual pole piece width does not match the preset value, it is difficult for the control system to make timely and effective adjustments, resulting in the pole piece offset exceeding the allowable range. If an overly conservative control strategy is adopted, although pole piece offset can be avoided, the production efficiency of the winder will be significantly reduced. In addition, due to the rapid iteration of lithium battery production processes, the width specifications of new pole pieces change more frequently. Whenever a new pole piece model is introduced, the deviation correction control program needs to be revised, which increases the workload of system maintenance. For some small-batch customized products, it is even necessary to write control code separately for each batch of pole pieces, which undoubtedly reduces the flexibility and versatility of the winder.

[0004] Therefore, an adaptive deviation-correction control method is urgently needed that can automatically adjust control strategies and parameters based on real-time changes in pole piece width. This method requires establishing a dynamic mapping relationship between pole piece width and motion characteristics, and designing corresponding control algorithms for different width ranges. At the same time, it is also necessary to consider how to maximize winding efficiency while maintaining control accuracy. This places higher demands on the real-time performance, stability, and robustness of the control system. Summary of the Invention

[0005] The present invention provides a pole piece deviation correction control method for an automatic cutting lithium battery winding machine, characterized in that the method comprises:

[0006] S101, installing a width sensor on the winding machine to obtain the width data of the currently wound pole piece in real time, and transmitting the obtained width data to the correction control system as an input parameter for control decision;

[0007] S102: Based on the acquired pole piece width data, access a pre-established mapping table of pole piece width and motion characteristics to obtain theoretical motion parameters of the pole piece at the current width, including a control accuracy threshold of a deviation correction actuator and an upper limit of a winding speed. The theoretical motion parameters specifically include an expected tension, an expected linear speed, and an expected offset range of the pole piece during the winding process.

[0008] S103, using an adaptive control algorithm to compare the pole piece width and motion parameters with the actual control accuracy of the current correction actuator. If the actual control accuracy is lower than the theoretical threshold requirement, the control strategy is dynamically adjusted to reduce the winding speed or increase the correction execution force until the accuracy requirement is met;

[0009] S104: While adjusting the control strategy, historical production data is analyzed using a machine learning algorithm to obtain the optimal control parameter combination for different electrode widths, forming a self-learning parameter optimization model to guide the correction control decision-making for subsequent batches of products;

[0010] S105, for small batches of customized products, obtains the actual width, material properties, and production process parameters of the batch of electrodes through rapid testing, and performs multi-dimensional matching with the historical data in the parameter optimization model. The control strategy under the most similar conditions is selected as the initial strategy, and then fine-tuned and optimized to shorten the adaptation time of the control system;

[0011] S106, during the winding process, continuously monitor the offset and motion state of the pole piece, filter the collected data using the Kalman filter algorithm to remove interference noise, improve the accuracy of offset detection, and provide a reliable feedback signal for the correction actuator;

[0012] S107 adopts an intelligent decision-making algorithm based on fuzzy logic, takes the obtained offset detection results as the main input, and comprehensively considers the influencing factors of pole piece width and winding speed, dynamically adjusts the control force and response time of the correction actuator, adjusts the displacement of the actuator according to the offset size, and adjusts the response speed of the actuator according to the offset speed. While ensuring the correction accuracy, it improves the winding efficiency and realizes adaptive correction control under width fluctuations.

[0013] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0014] The present invention discloses an adaptive deviation correction control method suitable for a winding machine. The method obtains pole piece width data in real time, combines a pre-established width and motion characteristic mapping table, determines theoretical motion parameters, adopts an adaptive control algorithm, and dynamically adjusts the control strategy to meet the accuracy requirements. At the same time, machine learning is used to analyze historical data to form a self-learning parameter optimization model to guide subsequent batch production. For small-batch customized products, the control strategy under similar conditions is quickly matched and fine-tuned. During the winding process, the Kalman filter algorithm is used to improve the accuracy of offset detection, and the intelligent decision-making algorithm based on fuzzy logic is used to dynamically adjust the deviation correction execution by comprehensively considering multiple factors. The present invention realizes precise deviation correction control under the condition of pole piece width fluctuation, improves winding efficiency and product quality, has strong adaptability, and is particularly suitable for flexible production needs of multiple varieties and small batches. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a flow chart of a pole piece deviation correction control method of an automatic cutting lithium battery winding machine. DETAILED DESCRIPTION

[0016] To further understand the content of the present invention, the present invention is described in detail with reference to the accompanying drawings and examples. The present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are intended only to illustrate the relevant invention and are not intended to limit the invention. It should also be noted that, for ease of description, only portions relevant to the invention are shown in the accompanying drawings.

[0017] Reference Figure 1 In this embodiment, a pole piece deviation correction control method of an automatic cutting lithium battery winding machine may specifically include:

[0018] S101. Install a width sensor on the winding machine to obtain the width data of the currently wound pole piece in real time, and transmit the obtained width data to the correction control system as an input parameter for control decision-making.

[0019] By installing a width sensor on the winding machine, the width data of the wound pole piece is acquired in real time, and the acquired width data is transmitted to the correction control system as an input parameter for the correction control system to make control decisions. Based on the winding pole piece width data received by the correction control system, it is determined whether the width of the wound pole piece exceeds the preset width threshold range. If it exceeds the preset width threshold range, the correction control system is triggered to perform correction control. The correction control system determines the correction control parameters for the current winding pole piece width based on the pre-established correspondence between the winding pole piece width and the correction control parameters, and transmits the correction control parameters to the winding machine's actuator. The winding machine's actuator adjusts the winding machine's operating state based on the received correction control parameters and controls the width of the wound pole piece to maintain the width of the wound pole piece within the preset width threshold range. During the process of correcting the width of the wound pole piece, the width data of the wound pole piece is continuously acquired and transmitted to the correction control system in real time, forming a closed-loop control. The correction control system uses a PID control algorithm. Based on the deviation between the acquired pole piece width data and the preset width threshold, the PID control algorithm parameters are dynamically adjusted to achieve precise control of the pole piece width. The width sensor installed on the winding machine continuously obtains the pole piece width data and transmits the width data to the host computer. The changing trend of the pole piece width is analyzed to predict possible abnormalities in the pole piece width and take preventive measures in advance.

[0020] Specifically, the width sensor on the winder acts as the eyes of the production line, monitoring the width of the electrode in real time. A laser displacement sensor can be used, emitting a laser beam onto the electrode surface and calculating the distance based on the reflected light, thereby determining the width. Assuming the standard electrode width is 100 mm, with an allowable error range of ±1 mm, the width threshold range is 99 mm to 101 mm. The sensor collects width data 10 times per second and immediately transmits it to the web-correction control system. The web-correction control system acts as the brain of the production line, receiving the width data and making decisions. For example, if the sensor measures a width of 102 mm at a given moment, exceeding the preset width threshold, the web-correction control system initiates the web-correction process. The web-correction control system has a pre-established mapping table between width and web-correction parameters. For example, a width deviation of +2 mm corresponds to a 1-degree adjustment of the guide roller angle. This table can be adjusted and optimized based on actual conditions, for example, by fitting experimental data or setting based on experience. This ensures accurate and efficient web-correction. The web-correction parameters act like instructions to the production line, telling the actuators how to operate. The winding machine's actuator, such as the guide roller motor, will adjust the angle of the guide roller after receiving the correction parameters, thereby changing the running trajectory of the pole piece and ultimately controlling the width of the pole piece.

[0021] In this example, the actuator adjusts the guide roller angle by 1 degree. The entire correction process is a closed-loop control process, like a continuous cycle. The width sensor continuously collects data, the correction control system continuously adjusts parameters, and the actuator continuously executes actions, ultimately maintaining the electrode width within the set range. This closed-loop control approach effectively addresses various interference factors and ensures stable electrode width. The PID control algorithm is the core of the correction control system, acting like a precision regulator. It dynamically adjusts control parameters based on width deviation, deviation change rate, and accumulated deviation to achieve precise control. For example, when the width deviation is large, the PID algorithm increases control effort to quickly correct the deviation; when the width deviation is small, the PID algorithm reduces control effort to avoid overcorrection. The host computer acts like the command center of the production line, receiving width data and performing analysis and predictions. For example, by analyzing width data trends, it can predict potential electrode width anomalies, such as a continuous increase or decrease in width. The host computer can issue an early alert, prompting operators to take preventive measures, such as adjusting equipment parameters or replacing materials, to avoid quality issues. The advantage of doing this is that problems can be discovered in advance, avoiding greater losses and improving production efficiency.

[0022] S102: Based on the acquired pole piece width data, access a pre-established mapping table of pole piece width and motion characteristics to obtain theoretical motion parameters of the pole piece at the current width, including the control accuracy threshold of the deviation correction actuator and the upper limit of the winding speed. The theoretical motion parameters specifically include the expected tension, expected linear speed, and expected deviation range of the pole piece during the winding process.

[0023] Obtain the current pole piece width data and use it as the query condition. Access the system's pre-established mapping table of pole piece width and motion characteristics, and query the corresponding theoretical motion parameter records based on the current pole piece width. Extract the control accuracy threshold and winding speed upper limit data of the correction actuator from the query results for subsequent correction control and winding speed control. Extract the expected tension, expected linear speed, and expected offset range data of the pole piece during the winding process from the query results as a reference for subsequent winding process control. Compare the extracted correction actuator control accuracy threshold with the actual correction actuator control accuracy. If the actual control accuracy exceeds the threshold range, the correction control strategy adjustment is triggered. Compare the extracted winding speed upper limit with the current winding speed. If the current winding speed exceeds the upper limit, the winding deceleration control is triggered. During the winding process, the pole piece tension, linear speed, and offset data are collected in real time and compared with the expected tension, expected linear speed, and expected offset range, respectively. Dynamically adjust the winding control parameters based on the deviation to ensure a smooth winding process.

[0024] Specifically, the purpose of establishing a mapping table between pole piece width and kinematic characteristics is to correlate the pole piece width with its optimal kinematic parameters during the winding process, thereby enabling precise control. This table can be derived from extensive experimental data and theoretical analysis. For example, for a pole piece with a width of 100 mm, the optimal tension is 20 Newtons, the optimal linear speed is 5 meters per second, the optimal offset range is ±0.5 mm, the control accuracy threshold of the deviation correction actuator is 0.1 mm, and the upper limit of the winding speed is 6 meters per second. The determination of these parameters requires consideration of various factors, including the pole piece's material properties, thickness, and the mechanical characteristics of the winding machine. Current pole piece width data can be obtained using a width sensor installed on the winding machine. The width sensor measures the pole piece width in real time and transmits the data to the control system. For example, if the current measured pole piece width is 98 mm, the corresponding theoretical kinematic parameters are queried from the pole piece width and kinematic characteristics mapping table based on the acquired width data (e.g., 98 mm). Since the mapping table may not contain a record that exactly corresponds to a width of 98 mm, interpolation can be used to obtain the value. Assuming that there are records of 95 mm and 100 mm in the table, the theoretical motion parameters corresponding to 98 mm can be calculated through linear interpolation based on the data of these two records.

[0025] Extract the required parameters from the query results. For example, after table lookup or interpolation calculation, the control accuracy threshold of the correcting actuator corresponding to a width of 98 mm is 0.09 mm, the upper limit of the winding speed is 5.8 m / s, the expected tension is 19 N, the expected linear velocity is 4.8 m / s, and the expected offset range is plus or minus 0.4 mm. The extracted control accuracy threshold is compared with the actual control accuracy of the correcting actuator. Assuming the actual control accuracy is 0.11 mm, it exceeds the threshold of 0.09 mm, indicating that the current correcting control accuracy is insufficient and the correcting control strategy needs to be adjusted. For example, the PID parameters of the correcting control system can be increased to improve the control system's response speed and accuracy. Compare the extracted winding speed upper limit with the current winding speed. Assuming the current winding speed is 6 m / s, it exceeds the upper limit of 5.8 m / s. In this case, the winding deceleration control needs to be triggered to reduce the winding speed to prevent problems such as wrinkling or breakage of the pole piece due to excessive speed. During the winding process, real-time data on the pole piece's tension, linear velocity, and offset is collected. For example, the collected tension is 18 Newtons, the linear speed is 4.7 meters per second, and the offset is 0.3 millimeters. The collected data is compared with the expected value. For example, the tension deviation is negative 1 Newton, the linear speed deviation is negative 0.1 meters per second, and the offset is within the expected range. The winding control parameters are dynamically adjusted according to the deviation. For example, because the tension is lower than the expected value, the output of the winding tension control system can be appropriately increased to bring the tension of the pole piece back to the expected range. Because the linear speed is lower than the expected value, the speed of the winding motor can be appropriately increased to bring the linear speed of the pole piece to the expected value. Through continuous monitoring and adjustment, the winding process is ensured to proceed smoothly and the quality of the final wound product is guaranteed. This can effectively control the tension, speed and position of the pole piece during the winding process, prevent quality defects, and improve the product yield.

[0026] S103. Adopt an adaptive control algorithm to compare the pole piece width and motion parameters with the actual control accuracy of the current deviation correction actuator. If the actual control accuracy is lower than the theoretical threshold requirement, dynamically adjust the control strategy to reduce the winding speed or increase the deviation correction execution force until the accuracy requirement is met.

[0027] Obtain production data such as pole piece width and motion parameters, and input them into the adaptive control algorithm model. The actual control accuracy data of the correction actuator is collected in real time through sensors and compared with the preset theoretical threshold. If the actual control accuracy is lower than the theoretical threshold, the adaptive control algorithm is triggered to dynamically adjust the control strategy according to the degree of deviation. Through the machine learning algorithm, combined with historical data and real-time feedback, it is determined whether it is more conducive to improving accuracy to reduce the winding speed or increase the correction execution force. According to the optimization suggestions output by the algorithm, the control system automatically adjusts the speed of the winding motor or the driving force of the correction actuator. Continuously monitor the actual control accuracy of the correction actuator and compare it with the dynamically updated theoretical threshold to form a closed-loop feedback control. When the actual control accuracy continues to meet the accuracy requirements, maintain the current control strategy; if deviation occurs again, repeat steps 3-6 until it stabilizes and meets the standards.

[0028] S104. While adjusting the control strategy, historical production data is analyzed through machine learning algorithms to obtain the optimal control parameter combination under different electrode widths, forming a self-learning parameter optimization model to guide the correction control decisions of subsequent batches of products.

[0029] Historical production data, including process parameters and quality data for different electrode widths, is acquired. This data is de-noised and normalized according to pre-set data cleaning rules to generate a production dataset suitable for analysis. A clustering algorithm is used to analyze the production dataset, dividing the data into clusters based on electrode width. An optimization model is trained for each cluster to determine the optimal process parameter combination for that cluster. In subsequent production batches, the current electrode width is obtained, its cluster is determined, and the optimal process parameter combination is obtained from the corresponding optimization model. This combination is used as the initial parameters for production control. Process parameters and quality data are collected in real time during the production process. If quality deviations exceeding pre-set thresholds are detected, a corrective control process is triggered. This corrective control process selects the optimal parameter combination under similar operating conditions from the optimization model for the current cluster based on the direction and magnitude of the deviation. The control strategy is then adjusted to ensure that the process parameters converge towards the optimal combination. The corrected process parameters and quality data are fed back to the optimization model, which is updated using an incremental learning algorithm to adapt to the latest production conditions, enabling self-learning and continuous optimization. The generalization performance of the optimization model is evaluated regularly. When the prediction error of the model on a new production batch exceeds the preset threshold, the clusters are re-divided and the optimization model is trained to adapt to changes in production conditions and ensure the effectiveness of the control strategy.

[0030] Specifically, historical production data is obtained, such as process parameters such as winding tension, winding speed, and corrective force for different electrode widths (e.g., 120mm, 125mm, 130mm, etc.), as well as quality data such as thickness uniformity and surface flatness of the final product. This data may come from a production database, an MES system, or other data recording platform. To ensure data quality, data cleaning is performed according to preset rules, such as removing outliers and filling missing values. For example, the winding tension range is set between 10-20N. Data outside this range is considered an outlier and requires removal or special processing. At the same time, data of different dimensions is normalized, such as scaling thickness uniformity values ​​to a range of 0-1, to facilitate subsequent analysis and modeling. The cleaned data is then partitioned according to electrode width using a clustering algorithm, such as the K-Means algorithm. Assume that the data is divided into three categories based on electrode width: narrow-width electrodes (e.g., less than 123mm), medium-width electrodes (e.g., between 123mm and 127mm), and wide-width electrodes (e.g., greater than 127mm). For each type of pole piece, an optimization model, such as a support vector machine or neural network model, is trained. The model inputs are process parameters, and the outputs are quality indicators. Through training, the optimal process parameter combination for each cluster is obtained. For example, the optimal process parameter combination for narrow-width pole pieces is: winding tension 12N, winding speed 10m / min, and guide force 5N. The benefit of this is that more precise control strategies can be developed for pole pieces of different widths.

[0031] When producing a new batch of pole pieces, first measure the width of the current pole piece, for example, 125mm. Determine the cluster to which it belongs based on the width, for example, it belongs to a medium-width pole piece. Then obtain the optimal process parameter combination from the corresponding optimization model (that is, the optimization model corresponding to the medium-width pole piece), such as a winding tension of 15N, a winding speed of 12m / min, and a correcting force of 8N, and apply it as the initial parameters to production control. This can effectively reduce parameter debugging time and improve production efficiency. During the production process, various process parameters and quality data are collected in real time. For example, data such as winding tension, winding speed, correcting force, and thickness uniformity are collected every 1 second. If it is detected that the quality deviation exceeds the preset threshold, for example, the thickness uniformity is lower than 95, the correction control process is triggered. This can timely detect and correct deviations in the production process to ensure product quality. In the correction control process, the optimal parameter combination under similar working conditions is selected from the optimization model of the current cluster based on the direction and magnitude of the deviation. For example, if thickness uniformity is low and the deviation is large, a parameter combination with lower winding tension and slower winding speed is selected from the optimization model of the current cluster (medium-width pole piece), such as a winding tension of 13N and a winding speed of 11m / min. By adjusting the control strategy, the process parameters converge to the optimal combination, thereby improving product quality. This allows for faster and more accurate deviation correction control.

[0032] The corrected process parameters and quality data are fed back to the optimization model, and the model is updated using an incremental learning algorithm. For example, new data is added to the training set and the model is retrained to adapt it to the latest production conditions. This allows the model to continuously learn new data, improving the model's prediction accuracy and adaptability. The generalization performance of the optimization model is regularly evaluated. For example, after producing every 100 batches of products, the model's prediction error on the new production batches is evaluated. If the model's prediction error exceeds a preset threshold, such as more than 5%, the clusters are re-divided and the optimization model is trained to adapt to changes in production conditions. For example, as the production process improves, the original cluster division may no longer be applicable, and the clusters need to be re-divided, and the optimization model needs to be trained for the new clusters. This ensures the effectiveness of the control strategy and continuously improves product quality.

[0033] S105. For small batch customized products, the actual width, material properties and production process parameters of the electrode batch are obtained through rapid testing, and multi-dimensional matching is performed with the historical data in the parameter optimization model. The control strategy under the most similar conditions is selected as the initial strategy, and then fine-tuning and optimization are performed to shorten the adaptation time of the control system.

[0034] The actual width, material properties and production process parameters of small-batch customized electrodes are obtained and stored in a database to form structured data; a pre-trained machine learning model is used to predict the optimal initial control strategy of the electrode based on the structured data; if the confidence of the optimal initial control strategy is lower than a preset threshold, the K-nearest neighbor algorithm is used to search the historical database for K samples most similar to the current electrode parameters, and the weighted average is calculated according to the control strategies corresponding to the K samples to obtain a new initial control strategy; the optimal initial control strategy is applied to the production process, and the key indicators of the production process are monitored in real time, and the key indicator data is stored in the database; the Bayesian optimization algorithm is used to fine-tune and optimize the optimal initial control strategy with the key indicators as the optimization target to obtain the optimized control strategy parameters; if the difference between the optimized control strategy parameters and the optimal initial control strategy parameters is less than a preset threshold, it is judged that the control system has reached a stable state; the optimized control strategy parameters and the corresponding key indicator data of the production process are stored in the historical database for updating the machine learning model and the sample library of the K-nearest neighbor algorithm.

[0035] Specifically, the actual width, material properties, and production process parameters of small-batch custom electrodes are obtained and stored in a database to form structured data. For example, for a custom electrode order with a width of 100 mm, using a specific graphene material formulation and a coating speed of 50 m / min, the production process parameters, such as coating roller pressure and drying temperature, will be recorded and stored in the database. This data will be stored in a structured format, such as a relational database, to facilitate subsequent query and analysis. This approach facilitates the retrieval and analysis of historical data, providing data support for subsequent model training and optimization. A pre-trained machine learning model (such as a support vector machine) is used to predict the optimal initial control strategy for the electrode based on historical data. The model's input features include electrode width, material properties, and production process parameters. The model's output is the predicted initial control strategy parameters. For example, for an electrode with a width of 95 mm and using the same graphene material formulation as in the example above, the model can predict the optimal initial control parameters, such as a coating roller pressure of 2 MPa and a drying temperature of 80°C, based on historical data. Support vector machines can effectively process high-dimensional data and have good generalization capabilities. They can predict the optimal initial control strategy for a new electrode based on historical data. If the confidence level of the predicted initial control strategy parameters falls below a preset threshold, the K-nearest neighbor algorithm is used to search the historical database for the K samples most similar to the current electrode parameters. A weighted average of the control strategies corresponding to these K samples is calculated to obtain a new initial control strategy. For example, if the confidence level of the model's predicted initial control strategy is less than 80%, the K-nearest neighbor algorithm can be used to search the historical database for the five samples most similar to the current electrode parameters. Assuming the coating roller pressures of these five samples are 1, 2, 3, 2, and 1 MPa, respectively, the weighted average is calculated to obtain the new coating roller pressure, which serves as the new initial control strategy parameter. The K-nearest neighbor algorithm effectively utilizes historical data to improve the reliability of the control strategy, especially when the prediction confidence level is low. The obtained initial control strategy is applied to the production process, and key process indicators such as temperature, pressure, and speed are monitored in real time. These key indicator data are stored in a database for subsequent model optimization. For example, the predicted coating roller pressure of 2 MPa and drying temperature of 80°C were applied to actual production, and key indicators such as temperature, pressure, and speed were monitored in real time. This data was stored in a database for subsequent model optimization. This enabled real-time monitoring of the production process and provided data support for subsequent model optimization. The initial control strategy was fine-tuned using a Bayesian optimization algorithm. The input of the Bayesian optimization algorithm was the control strategy parameters, and the output was the key indicators of the production process. The optimization goal was to maximize or minimize a key indicator, such as minimizing production time or maximizing product quality.For example, with the goal of minimizing production time, a Bayesian optimization algorithm can fine-tune the coating roller pressure and drying temperature based on the initial control strategy to find the optimal parameter combination that minimizes production time. Bayesian optimization algorithms can effectively search the parameter space to find the optimal control strategy.

[0036] Based on the output of the Bayesian optimization algorithm, the control strategy parameters are adjusted. If the difference between the adjusted control strategy parameters and the initial control strategy parameters is less than a preset threshold, the control system is considered to have reached a stable state. For example, the Bayesian optimization algorithm finds the optimal coating roller pressure of 25 MPa and the drying temperature of 82 degrees Celsius. If the difference between these two parameters and the initial control strategy parameters is less than a preset threshold, such as 5%, the control system is considered to have reached a stable state. This prevents over-adjustment of the control system and ensures the stability of the production process. The optimized control strategy parameters and the corresponding key production process indicators are stored in a historical database to update the sample library of the machine learning model and the K-nearest neighbor algorithm, enabling continuous model learning and optimization. For example, the optimized coating roller pressure of 25 MPa, the drying temperature of 82 degrees Celsius, and the corresponding key production process indicators are stored in a historical database. This data is used to update the sample library of the support vector machine model and the K-nearest neighbor algorithm, enabling the model to continuously learn from new production conditions, improving prediction accuracy and control effectiveness. This ensures continuous model learning and optimization, making the control strategy more adaptable to changes in the production environment.

[0037] S106. During the winding process, the offset and movement state of the pole piece are continuously monitored, and the collected data is filtered using the Kalman filter algorithm to eliminate interference noise, improve the accuracy of offset detection, and provide a reliable feedback signal for the correction actuator.

[0038] Based on the motion of the pole piece during the winding process, sensors continuously monitor the pole piece offset to obtain raw data reflecting the pole piece offset state. The obtained raw offset data is filtered using a Kalman filter algorithm. By establishing state and observation equations, the optimal pole piece offset is estimated. During the Kalman filter process, appropriate process noise covariance matrices and observation noise covariance matrices are set to effectively eliminate interfering noise and improve the accuracy of the offset estimation. Based on the optimal estimate of the pole piece offset obtained by the Kalman filter, it is determined whether the offset exceeds the preset allowable range. If so, a correction actuator is triggered to make adjustments. Based on the magnitude and direction of the offset, the correction actuator controls the winding equipment's motion parameters, such as winding speed and tension, to achieve real-time correction of the pole piece position. During the correction process, changes in pole piece offset are continuously monitored, and the control parameters of the correction actuator are dynamically adjusted through feedback control to ensure that the pole piece offset remains within the allowable range. The pole piece offset data and correction execution parameters are recorded during the correction process to form a data log for subsequent process optimization and equipment maintenance.

[0039] S107. Using the offset detection results obtained in the previous step as the primary input, an intelligent decision-making algorithm based on fuzzy logic is employed. Taking into account multiple factors, including pole piece width and winding speed, the algorithm dynamically adjusts the control force and response time of the corrective actuator. Specific adjustments include adjusting the actuator's displacement based on the offset and its response speed based on the offset speed. This ensures corrective accuracy while improving winding efficiency, achieving adaptive corrective control under width fluctuations.

[0040] Sensors collect parameters such as pole piece width and winding speed in real time to obtain data on influencing factors under the current production status. This data is preprocessed to remove outliers and smooth data fluctuations, providing high-quality input for subsequent analysis. This preprocessed data is then fed into an intelligent decision-making model based on fuzzy logic, where the importance of each influencing factor is determined through fuzzy reasoning. Based on the offset detection results and the importance of the influencing factors, a fuzzy rule base is used to dynamically generate control strategies for the correcting actuator. If the offset is large, a larger actuator displacement control strategy is generated; if the offset speed is fast, a faster actuator response speed control strategy is generated. The generated control strategies are converted into specific actuator control instructions, driving the correcting actuator to make real-time adjustments. The correction process is continuously monitored to obtain offset changes before and after correction, evaluate the correction effect, and feed the evaluation results back to the intelligent decision-making model for adaptive optimization.

[0041] Specifically, sensors collect parameters such as pole piece width and winding speed in real time, providing a data foundation for subsequent analysis. For example, a laser displacement sensor measures pole piece width, and an encoder measures winding speed. These sensor data constitute factors influencing the current production status. For example, the measured pole piece width is 60 mm and the winding speed is 10 m / min. This data reflects the current production status and serves as the basis for subsequent analysis and control. After data acquisition, preprocessing is required to remove outliers and data fluctuations. For example, the collected pole piece width data may exhibit spikes due to transient sensor interference. These are outliers and need to be removed. Simple mean filtering or median filtering can be used to smooth the data. For example, the continuously collected width data is 59 mm, 61 mm, 65 mm, 58 mm, and 62 mm. 65 mm is clearly an outlier and needs to be removed. After removal, mean filtering is performed to smooth the width data to approximately 60 mm. This prevents outlier data from interfering with subsequent analysis and control, improving system stability. The preprocessed data is input into an intelligent decision-making model based on fuzzy logic. Fuzzy logic is used to process fuzzy concepts such as "large," "small," "fast," and "slow." By fuzzifying influencing factors, for example, a pole piece width of 60 mm can be fuzzified into "normal width" and a winding speed of 10 m / min can be fuzzified into "medium speed." The fuzzy reasoning process is based on pre-defined fuzzy rules. For example, a rule could be: If the pole piece width is "normal width" and the winding speed is "medium speed," then the offset impact is "medium."

[0042] Through fuzzy reasoning, the importance of each influencing factor can be determined. For example, the importance of pole piece width is 8, and the importance of winding speed is 5. This approach has the advantage of simulating human experience and knowledge, handling factors that are difficult to precisely quantify, and making the control strategy more flexible. Based on the offset detection results and the importance of the influencing factors, a fuzzy rule base is used to dynamically generate a control strategy for the corrective actuator. For example, if the offset after Kalman filtering is 1mm, it is fuzzified as "small offset." Combining the importance of the influencing factors and the preset fuzzy rules, for example, if the offset is "small" and the pole piece width has a "high" impact, then the actuator displacement is "small." This ultimately generates a specific actuator control instruction, such as "move the actuator 5mm to the left." This allows dynamic adjustment of the control strategy based on actual conditions, achieving more precise control. For example, if the offset is large, such as 5mm, it is fuzzified as "large offset," and the offset speed is fast, fuzzified as "fast offset." According to the fuzzy rules, for example, if the offset is "large offset" and the offset speed is "fast offset," then the actuator displacement is "large" and the actuator response speed is "fast." The generated control strategy might be: quickly move the actuator 2mm to the right. This strategy can quickly and effectively correct large offsets and ensure product quality. After the control strategy is converted into specific actuator control instructions, the correction actuator is driven to make real-time adjustments. For example, the control instruction "move the actuator 2mm to the right" will be converted into an electrical signal or other type of control signal, driving the actuator motor to rotate the corresponding angle or distance to achieve precise displacement control. The correction process is continuously monitored. For example, the offset before correction is 5mm, and the offset after correction is 1mm. If the evaluation results show that the correction effect is good, this result is fed back to the intelligent decision-making model. If the correction effect is not good, the fuzzy rules or other parameters are adjusted to achieve adaptive optimization. This feedback mechanism can continuously improve the control strategy and enhance the adaptability and robustness of the system.

[0043] Through fuzzy logic, intelligent decision-making is performed on the pre-processed influencing factor data. The importance of each factor is determined based on fuzzy reasoning. Combined with the offset changes obtained by correction monitoring, the correction effect is evaluated to achieve adaptive optimization of the decision-making model.

[0044] Based on the influencing factor data, a fuzzy clustering algorithm is used to group and classify the data, generating data sets of different categories of influencing factors. For each category of influencing factor data set, fuzzy logic reasoning is used to determine the importance weights of each influencing factor within that category. The deviation change data for each influencing factor, output by the correction monitoring module, is obtained and weighted with the factor importance weights derived from fuzzy reasoning to obtain a comprehensive deviation change value. Based on this comprehensive deviation change value, a fuzzy evaluation method is used to determine the current deviation correction effect. If the deviation correction effect does not reach a preset threshold, the adaptive optimization mechanism of the decision model is triggered. An incremental learning algorithm is used to dynamically update the fuzzy inference rule base and membership function, adjusting the parameters of the fuzzy logic intelligent decision model to achieve adaptive optimization of the decision model. The optimized fuzzy logic intelligent decision model is applied to the influencing factor data processing, outputting optimized factor importance weights to guide the deviation correction process. These steps are continuously iterated, continuously improving the deviation correction effect through adaptive optimization of the decision model and feedback from correction monitoring, achieving accurate assessment of influencing factors and intelligent decision-making.

[0045] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A pole piece deviation correction control method for an automatic cutting lithium battery winding machine, characterized in that: The method comprises: S101, installing a width sensor on the winding machine to obtain the width data of the currently wound pole piece in real time, and transmitting the obtained width data to the correction control system as an input parameter for control decision; S102: Based on the acquired pole piece width data, access a pre-established mapping table of pole piece width and motion characteristics to obtain theoretical motion parameters of the pole piece at the current width, including a control accuracy threshold of a deviation correction actuator and an upper limit of a winding speed. The theoretical motion parameters specifically include an expected tension, an expected linear speed, and an expected offset range of the pole piece during the winding process. S103: Using an adaptive control algorithm, the pole piece width and motion parameters are compared with the actual control accuracy of the current deviation correction actuator. If the actual control accuracy is less than the theoretical threshold requirement, the control strategy is dynamically adjusted to reduce the winding speed or increase the deviation correction execution force until the accuracy requirement is met. S104: While adjusting the control strategy, historical production data is analyzed using a machine learning algorithm to obtain the optimal control parameter combination for different electrode widths, forming a self-learning parameter optimization model to guide the correction control decision-making for subsequent batches of products; S105, for small-batch customized products, obtains the actual width, material properties, and production process parameters of the pole piece through rapid testing, and performs multi-dimensional matching with the historical data in the parameter optimization model. The control strategy under the most similar conditions is selected as the initial strategy, and then fine-tuned and optimized to shorten the adaptation time of the control system; S106, during the winding process, continuously monitor the offset and motion state of the pole piece, filter the collected data using the Kalman filter algorithm to remove interference noise, improve the accuracy of offset detection, and provide a reliable feedback signal for the correction actuator; S107 adopts an intelligent decision-making algorithm based on fuzzy logic, takes the obtained offset detection results as the main input, and comprehensively considers the influencing factors of pole piece width and winding speed, dynamically adjusts the control force and response time of the correction actuator, adjusts the displacement of the actuator according to the offset size, and adjusts the response speed of the actuator according to the offset speed. While ensuring the correction accuracy, it improves the winding efficiency and realizes adaptive correction control under width fluctuations.

2. The method according to claim 1, characterized in that The step S101 includes: installing a width sensor on the winding machine to obtain width data of the wound pole piece in real time, and transmitting the obtained width data to the correction control system as an input parameter for the correction control system to make control decisions; According to the winding pole piece width data received by the correction control system, it is determined whether the width of the winding pole piece exceeds the preset width threshold range. If it exceeds the preset width threshold range, the correction control system is triggered to perform correction control; The correction control system determines the correction control parameters for the current winding pole piece width according to the pre-established correspondence between the winding pole piece width and the correction control parameters, and transmits the correction control parameters to the winding machine's actuator; The actuator of the winding machine adjusts the operating state of the winding machine according to the received correction control parameters, controls the width of the wound pole piece, and maintains the width of the wound pole piece within a preset width threshold range; During the process of correcting the width of the wound pole piece, the width data of the wound pole piece is continuously acquired and transmitted to the correction control system in real time to form a closed-loop control; The correction control system adopts PID control algorithm. According to the deviation between the acquired winding pole piece width data and the preset width threshold, the parameters of the PID control algorithm are dynamically adjusted to achieve precise control of the winding pole piece width. The width sensor installed on the winding machine continuously obtains the width data of the wound pole piece, and transmits the width data to the host computer to analyze the changing trend of the wound pole piece width, predict the abnormal situation of the wound pole piece width, and take preventive measures in advance.

3. The method according to claim 1, characterized in that The S102 includes: obtaining width data of the current electrode and using it as a query condition; Access the system's pre-established mapping table of pole piece width and motion characteristics, and query the corresponding theoretical motion parameter records based on the current pole piece width; Extract the control accuracy threshold of the deviation correction actuator and the upper limit data of the winding speed from the query results for subsequent deviation correction control and winding speed control; Extract the expected tension, expected linear speed and expected offset range of the pole piece during the winding process from the query results as the basis for subsequent winding process control; The extracted control accuracy threshold of the correction actuator is compared with the actual control accuracy of the correction actuator. If the actual control accuracy exceeds the threshold range, the correction control strategy adjustment is triggered; The extracted winding speed upper limit is compared with the current winding speed. If the current winding speed exceeds the upper limit, the winding deceleration control is triggered; During the winding process, the pole piece tension, line speed and offset data are collected in real time and compared with the expected tension, expected line speed and expected offset range respectively. The winding control parameters are dynamically adjusted according to the deviation to ensure a smooth winding process.

4. The method according to claim 1, wherein The S103 includes: obtaining pole piece width and motion parameter production data, and inputting the data into the adaptive control algorithm model; The actual control accuracy data of the correction actuator is collected in real time through sensors and compared with the preset theoretical threshold value; If the actual control accuracy is less than the theoretical threshold, the adaptive control algorithm is triggered to dynamically adjust the control strategy according to the degree of deviation; Through machine learning algorithms, combined with historical data and real-time feedback, it is determined whether reducing the winding speed or increasing the correction force is more conducive to improving accuracy; Based on the optimization suggestions output by the algorithm, the control system automatically adjusts the speed of the winding motor or the driving force of the deviation correction actuator; Continuously monitor the actual control accuracy of the correction actuator and compare it with the dynamically updated theoretical threshold to form a closed-loop feedback control; When the actual control accuracy continues to meet the accuracy requirements, the current control strategy is maintained.

5. The method according to claim 1, wherein The S104 includes: acquiring historical production data, including process parameters and quality data under different electrode widths, and performing denoising and standardization processing on the data according to preset data cleaning rules to obtain a production data set that can be used for analysis; A clustering algorithm is used to analyze the production data set, dividing the data into several clusters according to the electrode width. An optimization model is trained for each cluster to obtain the optimal process parameter combination under the cluster. In subsequent production batches, the current electrode width is obtained, the cluster to which it belongs is determined, and the optimal process parameter combination is obtained from the corresponding optimization model and applied as the initial parameter to production control; During the production process, various process parameters and quality data are collected in real time. If quality deviations exceeding the preset threshold are detected, the deviation correction control process is triggered. In the deviation correction control process, according to the deviation direction and magnitude, the optimal parameter combination under similar working conditions is selected from the current clustered optimization model, and the control strategy is adjusted to make the process parameters converge to the optimal combination; The corrected process parameters and quality data are fed back to the optimization model, and the model is updated using an incremental learning algorithm to adapt it to the latest production conditions, enabling self-learning and continuous optimization of the model. The generalization performance of the optimization model is evaluated regularly. When the prediction error of the model on a new production batch exceeds the preset threshold, the clusters are re-divided and the optimization model is trained to adapt to changes in production conditions and ensure the effectiveness of the control strategy.

6. The method according to claim 1, characterized in that The S105 includes: obtaining the actual width, material properties and production process parameters of the small batch customized electrode, storing them in a database, and forming structured data; Use pre-trained machine learning models to predict the optimal initial control strategy for the pole piece based on structured data; If the confidence of the optimal initial control strategy is less than the preset threshold, the K nearest neighbor algorithm is used to search the historical database for the K samples most similar to the current pole piece parameters. The weighted average value is calculated based on the control strategies corresponding to the K samples to obtain a new initial control strategy. Apply the optimal initial control strategy to the production process, monitor key indicators in the production process in real time, and store key indicator data in the database; Using the Bayesian optimization algorithm and taking key indicators as optimization targets, the optimal initial control strategy is fine-tuned and optimized to obtain the optimized control strategy parameters; If the difference between the optimized control strategy parameters and the optimal initial control strategy parameters is less than a preset threshold, it is judged that the control system has reached a stable state; The optimized control strategy parameters and the corresponding key indicator data of the production process are stored in the historical database for updating the sample library of the machine learning model and the K-nearest neighbor algorithm.

7. The method according to claim 1, characterized in that The step S106 includes: continuously monitoring the offset of the pole piece through a sensor according to the motion state of the pole piece during the winding process, and obtaining raw data reflecting the offset state of the pole piece; The Kalman filter algorithm is used to filter the acquired raw offset data, and the optimal value of the pole piece offset is estimated by establishing the state equation and observation equation. In the Kalman filtering process, by setting appropriate process noise covariance matrix and observation noise covariance matrix, the interference noise can be effectively eliminated and the accuracy of offset estimation can be improved; Based on the optimal estimated value of the pole piece offset obtained by Kalman filtering, it is determined whether the offset exceeds the preset allowable range. If it exceeds the range, the correction actuator is triggered to make adjustments; The deviation correction actuator adjusts the winding speed and tension according to the size and direction of the offset by controlling the motion parameters of the winding equipment, thereby achieving real-time correction of the pole piece position. During the correction process, the pole piece offset changes are continuously monitored, and the control parameters of the correction actuator are dynamically adjusted through feedback control to ensure that the pole piece offset is always within the allowable range. The pole piece offset data and correction execution parameters during the correction process are recorded to form a data log for subsequent process optimization and equipment maintenance.

8. The method according to claim 1, characterized in that The S107 includes: collecting the pole piece width and winding speed parameters in real time through the sensor to obtain the influencing factor data under the current production state; Preprocess the acquired influencing factor data to remove outliers and smooth data fluctuations to provide high-quality input for subsequent analysis; The pre-processed influencing factor data is input into the intelligent decision-making model based on fuzzy logic, and the importance of each influencing factor is determined through fuzzy reasoning; According to the deviation detection results and the importance of the influencing factors, the control strategy of the deviation correction actuator is dynamically generated using the fuzzy rule base; If the offset is greater than the preset threshold, a control strategy greater than the actuator displacement is generated; If the offset speed is greater than the preset threshold, a control strategy is generated that is greater than the actuator response speed; Convert the generated control strategy into specific actuator control instructions to drive the correction actuator to make real-time adjustments; Continuously monitor the correction process, obtain the offset changes before and after correction, evaluate the correction effect, and feed the evaluation results back to the intelligent decision-making model to achieve adaptive optimization.

Images