Die cutting material pulling motion control method and system and product

Abstract

The invention discloses a die cutting material pulling motion control method and system and a product, and solves the problems that a roller-material interface slippage phenomenon always exists in a material pulling mechanism in the prior art, edge warping, detection errors and the like exist in a pole piece when the position of a pole lug is measured, and consequently a certain difference value exists between the length of the pole piece actually sent out by the material pulling mechanism and the set theoretical length; and along with the increase of the speed, the difference value is also increased, and the production yield is reduced due to the influence of the increase of the difference value. Detecting the size data of the pole piece through a size detection CCD, and counting the difference value between the target length and the actual measurement length of the manufactured pole piece; calculating and predicting a future difference value by using an autoregression model according to the difference value between the counted target length of the pole piece and the actually measured length; compensating a material pulling main drive according to the predicted value; manufacturing a new pole piece by adopting the compensated material pulling main drive; and reading the actual deviation value of the size detection CCD on the detection of the new pole piece, and carrying out prediction error judgment.

Description

Technical Field

[0001] The present application relates to the field of die-cutting and drawing technology, and in particular to a die-cutting and drawing motion control method, system and product. Background Art

[0002] With the rapid rise and continuous innovation of my country's new energy industry, the production process requirements for lithium batteries, a key part of new energy, are also constantly increasing. Among them, the die-cutting machine's sheeting accuracy is one of the key factors affecting lithium battery quality. The main process requirements are that the distance from the center of the tab to the edge (hereinafter referred to as the short margin) must meet the standard value ± tolerance, and the width of the entire electrode (hereinafter referred to as the electrode width) must meet the standard value ± tolerance. The tolerance is generally required to be within 0.25mm, and the production capacity must reach 0.3 seconds per piece. Therefore, the material drawing control speed must be very fast and the accuracy must be very high to meet the process requirements.

[0003] At present, the cutting control scheme of the die-cutting machine is: the position of the tab is measured at a fixed position before cutting (using CCD or high-speed probe function). Because the measurement position is fixed, the distance from the measurement position to the cutting point is fixed. The measured value and the fixed distance are compared to calculate the pulling compensation value. The compensation is used to improve the production accuracy during pulling, and the pole piece size is measured after cutting. The pole pieces that do not meet the process requirements will be discarded to ensure that the size of the pole pieces sent to the next process all meet the process requirements.

[0004] Since there is always roller-material interface slip in the pulling mechanism, there are warping and detection errors in the pole piece when measuring the position of the pole lug, resulting in a certain difference between the actual length of the pole piece delivered by the pulling mechanism and the set theoretical length. As the speed increases, the difference also increases, and the production yield is affected by the increase in the difference and reduced. Summary of the Invention

[0005] The present application provides a die-cutting drawing motion control method, system and product, which are intended to solve the problem that the drawing mechanism of the prior art always has the phenomenon of roller-material interface slippage, the pole piece has warping and detection errors when measuring the position of the pole ear, etc., resulting in a certain difference between the actual length of the pole piece delivered by the drawing mechanism and the set theoretical length, and as the speed increases, the difference also increases, and the production yield is affected by the increase in the difference and reduced.

[0006] In a first aspect, a method for controlling a die-cutting material drawing motion is provided, comprising:

[0007] The electrode size data is detected by the size detection CCD, and the difference between the target length and the actual measured length of the completed electrode is counted;

[0008] The autoregressive model is used to calculate and predict the future difference between the statistical target length of the pole piece and the actual measured length; the main drive of the pulling material is compensated according to the predicted value;

[0009] A new pole piece is made by using the compensated main drive for drawing material; the actual deviation value of the new pole piece detected by the size detection CCD is read, and a prediction error judgment is performed. If the prediction error is too large, the number of times the error exceeds the standard is increased by one. If the number of times the error exceeds the standard exceeds the set alarm number, the alarm is sounded and the machine is shut down, and a reminder is given as to whether the autoregressive model needs to be retrained; if the error is normal, the number of times the error exceeds the standard is cleared, and the current actual deviation value is input into the autoregressive model as new historical data for subsequent prediction calculation of future differences.

[0010] In the above solution, optionally, the calculating and predicting of the future difference based on the statistical difference between the target pole piece length and the actual measured length using the autoregressive model specifically includes:

[0011] The autoregressive model formula is

[0012]

[0013] Among them, y t represents the predicted value; φ i represents the autoregressive coefficient; c represents a constant, ∈ t represents noise; p represents the class of the autoregressive model, which is the preset value; i represents the lag index, which ranges from 1 to p, indicating that the i-th of the past p historical data points is used; t represents the position of the current time point in the time series, and y t-i Represents the historical value of the i-th step before time t;

[0014] Training the autoregressive model in an external tool based on historical data of the device includes: using the least squares method in the external tool to solve and obtain the coefficient φ i and the constant term c;

[0015] The coefficient φ obtained by training the model i and the constant term c are solidified into the PLC program;

[0016] Input the deviation value of the previous pole piece, call the PLC program to perform autoregressive model calculation, and read the predicted value output by the PLC program as the predicted result of the deviation value of the next pole piece.

[0017] In the above scheme, further optionally, after inputting the deviation value of the previous electrode, a step of judging the deviation value is also included: judging whether the deviation value is within the set process requirement range; if it is outside the range, the unreasonable data is eliminated and replaced with 0, wherein the process requirement range is determined according to different process standards and electrode types.

[0018] In the above solution, optionally, in the step of inputting the deviation value of the previous electrode, it also includes: if the electrode has just started to be produced, the deviation values ​​of the first n electrodes are defaulted to 0.

[0019] In the above solution, optionally, the order of the autoregressive model is determined according to the production capacity requirements of the die-cutting machine and the high-speed response requirements of the PLC to avoid overfitting or excessive calculation.

[0020] In the above solution, optionally, the predictive calculation of the future difference by the autoregressive model is completed in the PLC.

[0021] In a second aspect, a die-cutting material pulling compensation system is provided, comprising:

[0022] Data acquisition and processing module: used to detect the electrode size data through the size detection CCD, and to collect statistics on the difference between the target length and the actual measured length of the completed electrode;

[0023] Prediction and compensation module: used to calculate and predict the future difference based on the difference between the statistical pole piece target length and the actual measured length using the autoregressive model; compensate the main drive of the pulling material according to the predicted value;

[0024] Pole piece production and error monitoring feedback module: used to produce new pole pieces using the compensated pulling main drive; read the actual deviation value of the size detection CCD for the new pole piece detection, and make a prediction error judgment. If the prediction error is too large, the number of continuous error exceeding the standard will be increased by one. If the number of continuous error exceeding the standard exceeds the set alarm number, the alarm will be shut down and a reminder will be given whether the autoregressive model needs to be retrained; if the error is normal, the number of continuous error exceeding the standard will be cleared, and the current actual deviation value will be input into the autoregressive model as the new historical data for subsequent prediction calculation of future differences.

[0025] In a third aspect, a computer program product is provided, comprising a computer program / instruction, which implements the steps of the above method when executed by a processor.

[0026] Compared with the prior art, this application has at least the following beneficial effects:

[0027] This application is based on further analysis and research on the problems of the existing technology, and recognizes that the existing technology pulling mechanism always has the phenomenon of roller-material interface slippage, and the pole piece has warping and detection errors when measuring the position of the pole ear, resulting in a certain difference between the actual length of the pole piece sent by the pulling mechanism and the set theoretical length. As the speed increases, the difference also increases, and the production yield is affected by the increase in the difference and reduced. By introducing a closed-loop size detection CCD to collect the actual length of the pole piece in real time, a "target-measured" deviation data closed loop is formed; an autoregressive model is used to dynamically learn the historical deviation sequence, predict future deviation trends, and actively compensate for the feeding amount of the main driving force of the pulling material; after compensation, the actual deviation of the new pole piece is collected again to realize the iterative optimization of "prediction-compensation-verification", and the effective data is continuously injected back into the model to suppress random errors caused by slippage, warping, etc.; a continuous error exceeding the standard alarm mechanism is set, and when the model becomes inaccurate due to sudden changes in working conditions, it is stopped in time to prompt retraining to prevent defective pole pieces from continuing to circulate. With this compensation, the error value of the pole piece caused by unavoidable objective factors such as positioning detection error and roller-material interface slip is proactively compensated, and the production accuracy is improved from ±0.25mm to ±0.2mm, with an accuracy improvement of 25%, achieving the effect of improving the process level. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A flow chart of a die-cutting drawing motion control method provided in one embodiment of the present application. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0030] In the description of the present application, unless otherwise specified, expressions such as “include”, “comprising”, “having”, etc. also mean “not limited to” (certain units, components, materials, steps, etc.).

[0031] The purpose of this application is to provide a method for predicting and compensating for the difference in the size of the drawn material. Based on the autoregressive model (AR), a linear combination of historical data is used to predict the future difference in the size of the pole piece. The main drive of the drawn material is compensated according to the predicted value to improve the dimensional accuracy of the cut pole piece.

[0032] In one embodiment, a method for controlling die-cutting drawing motion is provided, comprising:

[0033] The electrode size data is detected by the size detection CCD, and the difference between the target length and the actual measured length of the completed electrode is counted;

[0034] The autoregressive model is used to calculate and predict the future difference between the statistical target length of the pole piece and the actual measured length; the main drive of the pulling material is compensated according to the predicted value;

[0035] A new pole piece is made by using the compensated main drive for drawing material; the actual deviation value of the new pole piece detected by the size detection CCD is read, and a prediction error judgment is performed. If the prediction error is too large, the number of times the error exceeds the standard is increased by one. If the number of times the error exceeds the standard exceeds the set alarm number, the alarm is sounded and the machine is shut down, and a reminder is given as to whether the autoregressive model needs to be retrained; if the error is normal, the number of times the error exceeds the standard is cleared, and the current actual deviation value is input into the autoregressive model as new historical data for subsequent prediction calculation of future differences.

[0036] In this embodiment, the pulling main drive is compensated according to the predicted value, mainly by calculating the circumference of the main drive through the diameter of the main drive roller, and then converting the compensation value into the number of turns of the main drive. The real-time cam real-time proportional modification function is used to modify the proportion of the cam to realize the predicted value compensation of the main drive.

[0037] In this embodiment, a four-step closed-loop control process of "measure first, calculate second, compensate third, and calibrate" is implemented. First, the actual length of each electrode is measured using a CCD, and the difference from the target length is calculated. Next, an autoregressive model is used to "guess" the deviation from the target length for the next electrode based on historical differences. This predicted deviation is then communicated to the main drive of the material feeder in advance, allowing it to automatically adjust the length when feeding the material. After the new electrode is cut, the CCD is used to verify the accuracy of the cut. Based on the results, a decision is made as to whether the data should be fed to the model for self-evolution or whether an alarm should be issued to shut down the machine and prompt manual retraining of the model.

[0038] The solution provided by this embodiment has the following benefits: the scrap rate is significantly reduced. The compensation action is completed before the "next cut", and the machine adjustment waste is reduced from hundreds of meters to ten meters. The more the equipment is used, the more accurate it is. The model continuously absorbs new data, and the length Cpk can naturally climb over time, offsetting long-term drifts such as tool wear and material thickness fluctuations. Quality is traceable. The measured deviation, compensation amount, and prediction error of each electrode are automatically recorded by the system, and in the event of a customer complaint, the problem batch can be located in minutes. Early warning of risks. The continuous error exceeding the standard alarm is equivalent to giving the process personnel a red light that "the model has failed", preventing large-scale defective products from flowing into the subsequent process.

[0039] In one embodiment, the calculating and predicting of the future difference based on the statistical difference between the target pole piece length and the actual measured length using the autoregressive model specifically includes:

[0040] The autoregressive model formula is

[0041]

[0042] Among them, yt represents the predicted value; φ i represents the autoregressive coefficient; c represents a constant, ∈ t represents noise; p represents the class of the autoregressive model, which is a preset value; i represents the lag index, which ranges from 1 to p, indicating that the i-th of the past p historical data points is used; t represents the position of the current time point in the time series. For example, when predicting the sixth data, t = 6. In the formula, y t-i Represents the historical value of the i-th step before time t;

[0043] Training the autoregressive model in an external tool based on historical data of the device includes: using the least squares method in the external tool to solve and obtain the coefficient φ i and the constant term c;

[0044] The coefficient φ obtained by training the model i and the constant term c are solidified into the PLC program;

[0045] Input the deviation value of the previous pole piece, call the PLC program to perform autoregressive model calculation, and read the predicted value output by the PLC program as the predicted result of the deviation value of the next pole piece.

[0046] In this embodiment, the order p should not be too large, otherwise it will overfit the noise and reduce the prediction accuracy. Before least squares training, outliers (such as CCD false detection) must be eliminated, otherwise the mutation points will be learned as normal fluctuations. If the equipment speed or material batch changes significantly, the curing coefficient should be retrained and updated.

[0047] In one embodiment, after inputting the deviation value of the previous electrode, a step of judging the deviation value is also included: judging whether the deviation value is within the set process requirement range. If it is outside the range, the unreasonable data is eliminated and replaced with 0, wherein the process requirement range is determined according to different process standards and electrode types.

[0048] In this embodiment, the process personnel predetermine the maximum allowable positive and negative deviation window (e.g. ±0.15mm) based on the electrode type, material thickness, and customer specifications. This window is automatically synchronized to the PLC as the work order is switched. The PLC immediately compares the measured deviation of each electrode with the window after receiving it:

[0049] If it falls within the interval, the original value enters the shift register for normal operation of the AR model;

[0050] If it falls outside the interval, it is determined to be an abnormal point, directly set to 0, and an "abnormal rejection" event code is sent to the upper computer for easy tracing.

[0051] The model considers a value of 0 to be "unbiased," effectively telling the AR model, "This observation is unreliable; please rely more on historical trends." This prevents outliers from biasing subsequent forecasts. Due to the autoregressive nature of AR, the impact of a single point of 0 is rapidly diluted with the order p and does not interfere with the system over the long term.

[0052] Through this embodiment, "false compensation" caused by CCD misdetection or electrode damage can be reduced, avoiding continuous waste of films; maintaining the statistical consistency of model input data and extending the interval between two retrainings; retaining rejection records to facilitate process engineers to backtrack and optimize threshold settings.

[0053] In one embodiment, the step of inputting the deviation value of the previous electrode piece further includes: if the electrode pieces have just been produced, the deviation values ​​of the first n electrodes are defaulted to 0. In this embodiment, n is 5.

[0054] In this embodiment, when the system detects that the current electrode batch is in the initial stage and has not yet obtained sufficient valid historical deviation values, the deviation values ​​of the first n electrode sheets are uniformly set to 0 as the initial filling. The value of n here is the order p of the autoregressive model, or slightly greater than p (such as p+1), to ensure that the shift register is filled before the first formal prediction, thereby avoiding the AR model from generating invalid or violently oscillating prediction values ​​due to "empty windows". After the n+1th electrode sheet, the system begins to collect and update the true deviation value, gradually replacing the initial 0, and entering the normal closed-loop prediction-compensation process.

[0055] In one embodiment, the order of the autoregressive model is determined based on the production capacity requirements of the die-cutting machine and the high-speed response requirements of the PLC to avoid overfitting or excessive computation.

[0056] In one embodiment, the predictive calculation of future differences by the autoregressive model is performed in the PLC.

[0057] In this embodiment, the predictive calculation of future differences through the autoregressive model is completed in the PLC, which is independent of external software. Network interruption or PC failure will not affect production; all intermediate variables and coefficient registers are retained during power failure and do not need to be re-downloaded after restart; on-site debugging can be completed by simply monitoring the PLC registers, greatly reducing the maintenance threshold.

[0058] The following describes the solution of the present application from another perspective.

[0059] To improve the cutting accuracy of the electrode, we collect, count, and calculate the difference between the target length and the actual measured length of the electrode by collecting and analyzing the dimensional data detected by the size detection CCD. We then use the autoregressive model (ARModel) to calculate and predict the future difference and compensate for it to reduce the difference and improve the cutting accuracy of the electrode.

[0060] Reference Figure 1 As shown in the figure, the processing flow of the compensation method of the present application includes the following steps.

[0061] Step one: Autoregressive model formula

[0062]

[0063] Where y t represents the predicted value, φ i represents the autoregressive coefficient, c represents the constant, ∈ t represents the noise, i represents the lag index, which ranges from 1 to p, and represents the i-th historical data point used in the past p; t represents the position of the current time point in the time series, such as predicting the 6th data t = 6, in the formula, y t-i represents the historical value at the i-th step before time t. p represents the order of the autoregressive model, which is calculated in the PLC (Programmable Logic Controller) according to the formula, the number of operations is determined by p, high-order AR is prone to overfitting and has large calculation amount, in order to meet the production capacity requirements of the die cutting machine and the high-speed response requirements of the PLC, we choose the order p = 5.

[0064] Step two: Offline training of AR model: In external tools (such as Python, MATLAB), train the AR(5) model according to the historical data of the device, and use the least squares method to solve for the coefficients φ1 to φ5 and the constant term c.

[0065] The example code for solving the coefficients φ1 to φ5 and the constant term c using the least squares method is as follows:

[0066]

[0067]

[0068] Step three: Parameter solidification: Write φ1 to φ5 and the constant term c obtained from the trained model into the initial values of phi[1..5] and c in the PLC program, respectively.

[0069] Step four: Input the deviation value of the last sheet (if you just started production, the deviation values of the first five sheets are set to 0 by default), and determine whether the deviation value is within the set process requirement range (different process standards and sheet types have different ranges). If it is outside the range, eliminate unreasonable data and replace it with 0.

[0070] Step five: Call the prediction method: As shown in the figure, call the PLC program to perform autoregressive model calculation. Where sum and i are temporary variables, InputArray[] is a historical data array with a length of 5, and Prediction is the output predicted value.

[0071] The example code is as follows:

[0072] / / Main calculation logic

[0073] sum:=0.0;

[0074] FORi:=1TO 5DO

[0075] sum:=sum+phi[i]*InputArray[i]; / / Calculate ∑(φ_i*x_i)

[0076] END_FOR;

[0077] Prediction:=sum+c; / / Output predicted value y=c+∑(φ_i*x_i)

[0078] Step 6: Get the predicted value: Read the Prediction as the predicted result of the next pole piece deviation value, and use the predicted value to compensate for the pulling material.

[0079] Step 7: After the film is produced, the actual deviation value of the size detection is read to determine the prediction error. If the prediction error is too large, the number of times the error exceeds the standard is recorded plus one. If the number of times the error exceeds the set alarm number, the alarm will be stopped and a reminder will be given whether the AR model needs to be retrained. If the error is normal, the number of times the standard is exceeded is reset to zero, and the process returns to step 4 to input the deviation value and predict the deviation value of the next film to enter the loop.

[0080] Key points of this application:

[0081] This application uses an autoregressive model (AR) with its high efficiency, lightweight and ease of use to predict the pole piece deviation value based on historical deviation values, and performs material pulling compensation based on the predicted value, thereby improving the production accuracy of the equipment.

[0082] The autoregressive fifth-order model used is trained externally using the least squares method with external tools (such as Python and MATLAB) based on the historical data of the equipment. The remaining predictive calculations in the PLC are computationally low, and a single prediction takes less than 1ms, which is suitable for the millisecond-level response of the PLC, making it very suitable for die-cutting machines that require high-speed production.

[0083] Data range check: The input data needs to be range checked to ensure that the data is within the set process requirements to prevent overflow and affect the accuracy of the predicted value.

[0084] During the blanking phase, the actual electrode deviation is measured using a dimension detection CCD. This data is then used in the next round of deviation prediction, achieving a closed-loop prediction and compensation system. The actual deviation is then compared with the predicted value to determine whether the predicted error exceeds the specified limit. The final compensation effect is then tested and evaluated to ensure its effectiveness.

[0085] Compared to existing technologies, the proposed method, using an autoregressive model, predicts the deviation of the pole piece in advance, thereby performing predictive compensation and realizing a closed-loop prediction and compensation system. This compensation method proactively compensates for unavoidable objective factors such as positioning errors and roller-material interface slip, improving production accuracy from ±0.25mm to ±0.2mm, a 25% improvement in accuracy and ultimately enhancing the process.

[0086] In one embodiment, a die-cutting pull compensation system is also provided, comprising:

[0087] Data acquisition and processing module: used to detect the electrode size data through the size detection CCD, and to collect statistics on the difference between the target length and the actual measured length of the completed electrode;

[0088] Prediction and compensation module: used to calculate and predict the future difference based on the difference between the statistical pole piece target length and the actual measured length using the autoregressive model; compensate the main drive of the pulling material according to the predicted value;

[0089] Pole piece production and error monitoring feedback module: used to produce new pole pieces using the compensated pulling main drive; read the actual deviation value of the size detection CCD for the new pole piece detection, and make a prediction error judgment. If the prediction error is too large, the number of continuous error exceeding the standard will be increased by one. If the number of continuous error exceeding the standard exceeds the set alarm number, the alarm will be shut down and a reminder will be given whether the autoregressive model needs to be retrained; if the error is normal, the number of continuous error exceeding the standard will be cleared, and the current actual deviation value will be input into the autoregressive model as the new historical data for subsequent prediction calculation of future differences.

[0090] In one embodiment, a computer program product is also provided, including a computer program / instruction, which implements the steps of the above-mentioned embodiment method when executed by a processor.

[0091] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A die-cutting material drawing motion control method, characterized in that: include: The electrode size data is detected by the size detection CCD, and the difference between the target length and the actual measured length of the completed electrode is counted; The autoregressive model is used to calculate and predict the future difference between the statistical target length of the pole piece and the actual measured length; Compensate the main drive of the pulling material according to the predicted value; Use the compensated pulling main drive to make new pole pieces; Read the actual deviation value of the size detection CCD for the new pole piece detection, and make a prediction error judgment. If the prediction error is too large, the number of times the error exceeds the standard continuously is increased by one. If the number of times the error exceeds the standard continuously exceeds the set alarm number, the alarm is shut down and a reminder is given as to whether the autoregressive model needs to be retrained. If the error is normal, the number of times the error exceeds the standard continuously is reset to zero, and the current actual deviation value is input into the autoregressive model as the new historical data for subsequent prediction calculation of future differences.

2. The die-cutting material drawing motion control method according to claim 1, characterized in that: The method of calculating and predicting the future difference based on the statistical difference between the target length of the pole piece and the actual measured length using the autoregressive model specifically includes: The autoregressive model formula is Among them, y t represents the predicted value; φ i represents the autoregressive coefficient; c represents a constant, ∈ t represents noise; p represents the class of the autoregressive model, which is the preset value; i represents the lag index, which ranges from 1 to p, indicating that the i-th of the past p historical data points is used; t represents the position of the current time point in the time series, and y t-i Represents the historical value of the i-th step before time t; Training the autoregressive model in an external tool based on historical data of the device includes: using the least squares method in the external tool to solve and obtain the coefficient φ i and the constant term c; The coefficient φ obtained by training the model i and the constant term c are solidified into the PLC program; Input the deviation value of the previous pole piece, call the PLC program to perform autoregressive model calculation, and read the predicted value output by the PLC program as the predicted result of the deviation value of the next pole piece.

3. The die-cutting drawing motion control method according to claim 2, characterized in that: After inputting the deviation value of the previous electrode, the process also includes a step of judging the deviation value: judging whether the deviation value is within the set process requirement range. If it is outside the range, the unreasonable data is eliminated and replaced with 0, wherein the process requirement range is determined according to different process standards and electrode types.

4. The die-cutting drawing motion control method according to claim 2, characterized in that: The step of inputting the deviation value of the previous electrode piece also includes: if the electrode piece has just started to be produced, the deviation values ​​of the first n pieces are defaulted to 0.

5. The die-cutting drawing motion control method according to claim 1 or 2, characterized in that: The order of the autoregressive model is determined according to the production capacity requirements of the die-cutting machine and the high-speed response requirements of the PLC to avoid overfitting or excessive calculation.

6. The die-cutting drawing motion control method according to claim 1, characterized in that: The predictive calculation of future differences by the autoregressive model is performed in the PLC.

7. A die-cutting material pulling compensation system, characterized in that: include: Data acquisition and processing module: used to detect the electrode size data through the size detection CCD, and to collect statistics on the difference between the target length and the actual measured length of the completed electrode; Prediction and compensation module: used to calculate and predict the future difference based on the difference between the statistical pole piece target length and the actual measured length using the autoregressive model; Compensate the main drive of the pulling material according to the predicted value; Pole piece production and error monitoring feedback module: used to produce new pole pieces using the compensated pulling main drive; Read the actual deviation value of the size detection CCD for the new pole piece detection, and make a prediction error judgment. If the prediction error is too large, the number of times the error exceeds the standard continuously is increased by one. If the number of times the error exceeds the standard continuously exceeds the set alarm number, the alarm is shut down and a reminder is given as to whether the autoregressive model needs to be retrained. If the error is normal, the number of times the error exceeds the standard continuously is reset to zero, and the current actual deviation value is input into the autoregressive model as the new historical data for subsequent prediction calculation of future differences.

8. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to claim 1 are implemented.

Images