Methods and apparatus for detecting circular runout

Abstract

This application provides a method and apparatus for detecting circular runout, which can effectively detect the circular runout of a coating roller. The method includes: acquiring data from multiple detection points on an electrode, wherein the data for each detection point includes the coating weight at each detection point and the angle through which the coating roller rotates when the coating weight is detected at each detection point; and determining the circular runout information of the coating roller based on the data from the multiple detection points.

Description

Technical Field

[0001] This application relates to the field of battery technology, and in particular, to a method and apparatus for detecting circular runout. Background Technology

[0002] Coating is a crucial process in lithium-ion battery manufacturing. The coating weight determines the battery's capacity, and weight consistency affects cycle life, safety, and other performance indicators. Currently, coating weight can be monitored using an areal density meter, but factors affecting coating weight consistency primarily originate from the coating equipment, such as the circular runout of the coating rollers. Therefore, effectively detecting the circular runout of the coating rollers is a problem that needs to be solved. Summary of the Invention

[0003] This application provides a method and apparatus for detecting circular runout, which can effectively detect the circular runout of a coating roller.

[0004] In a first aspect, a method for detecting circular runout is provided, the method comprising: acquiring data from multiple detection points on an electrode, wherein the data for each detection point includes the coating weight at each detection point and the angle through which the coating roller rotates when the coating weight at each detection point is detected; and determining the circular runout information of the coating roller based on the data from the multiple detection points.

[0005] In this embodiment, the electrode moves along the coating direction as the coating roller rotates. Multiple detection points on the electrode for detecting the coating weight are sequentially detected during the electrode's movement. The detection data for each detection point includes the coating weight at each detection point and the angle through which the coating roller rotates when the coating weight is detected at each detection point. By associating the coating weight of the electrode with the rotation angle of the coating roller, the circular runout information of the coating roller is determined based on the coating weight at each detection point on the electrode and the angle through which the coating roller rotates when the coating weight is detected at that detection point. This enables online detection of the circular runout of the coating roller without the need for additional circular runout detection equipment or sensing units, reducing costs and providing higher detection accuracy.

[0006] In one implementation, obtaining the angle through which the coating roller rotates when the coating weight is detected up to each detection point includes: determining the angle through which the coating roller rotates based on the coating speed of the electrode and the time elapsed when the coating weight is detected up to each detection point. This method allows for the correlation between the coating weight of the electrode and the rotation angle of the coating roller.

[0007] In one implementation, determining the angle rotated by the coating roller based on the coating speed of the electrode and the time elapsed to detect the coating weight at each detection point includes: multiplying the coating speed by the time to obtain the coating length; and dividing the coating length by the circumference of the coating roller to obtain the angle rotated by the coating roller. This method is easy to calculate and can accurately determine the rotation angle of the coating roller corresponding to the current detection point based on the time information at that detection point.

[0008] In one implementation, obtaining the coating weight of each detection point includes: determining the coating weight of each detection point based on the signal transmitted from and irradiated by rays at the detection point, wherein the rays reciprocate across the surface of the electrode sheet along its width direction during the coating process. The coating weight at that detection point on the electrode sheet can be accurately detected by utilizing the intensity of the signal transmitted from and irradiated by rays at that detection point.

[0009] In one implementation, the method further includes: statistically processing the coating weight in the data of each detection point based on the angle in the data of each detection point, and obtaining a detection matrix composed of multiple coating weights; determining the circular runout information of the coating roller based on the data of the multiple detection points includes: determining the circular runout information of the coating roller based on the detection matrix. Thus, by statistically processing the coating weight at each detection point based on the rotation angle of the coating roller corresponding to each detection point, the coating weight can be converted into the space of the rotation angle of the back roller, and a detection matrix representing the coating weight characteristics can be obtained.

[0010] In one implementation, the step of statistically processing the coating weight in the data of each detection point based on the angle in the data of each detection point to obtain a matrix composed of multiple coating weights includes: taking the remainder of the angle corresponding to each detection point modulo 2π to convert the angle corresponding to each detection point to the range of 0 to 2π; taking the coating weight of each detection point and the angle corresponding to each detection point as a set of data to obtain an N×M matrix composed of multiple sets of data from multiple detection points, where N is the number of single passes traversed when the ray reciprocates to scan the surface of the electrode, and M is the number of detection points located in the coating area on each single pass; dividing 0 to 2π into multiple intervals based on a preset angle value, and statistically analyzing the coating weight in the data group whose angle is in the same interval in each column of the N×M matrix to obtain a D×M detection matrix composed of multiple statistical data, where D is the ratio of 2π to the preset angle. This method is easy to calculate and can convert the coating weight at each detection point into the space of the rotation angle of the back roller, and obtain a detection matrix representing the coating weight characteristics.

[0011] In one implementation, the step of statistically analyzing the coating weight of data groups with angles within the same interval in each column of the N×M matrix to obtain a D×M matrix composed of coating weight statistics includes: performing K types of statistics on the coating weight of data groups with angles within the same interval in each column of the N×M matrix to obtain a K×D×M detection matrix composed of K types of coating weight statistics, where K is a positive integer, and the K types of statistics include at least one of mean statistics, variance statistics, and median statistics. Using statistical methods such as mean, variance, and median can simply and accurately represent the coating weight characteristics of the electrode.

[0012] In one implementation, determining the circular runout information of the coating roller based on the detection matrix includes: preprocessing the detection matrix to obtain a feature matrix; and using the feature matrix as input to a machine learning model to obtain the circular runout information of the coating roller output by the machine learning model.

[0013] The machine learning model can be trained using multiple feature matrices and their corresponding known circular oscillation information. Optionally, the machine learning model includes any of the following: logistic regression, Gaussian Naive Bayes, support vector machine, and decision tree.

[0014] In the above implementation, the detection matrix that can represent the coating weight characteristics of the electrode is preprocessed, and the preprocessed data is used as the input of the machine learning model. The output of the machine learning model is the circular runout information of the coating roller. The machine learning model can accurately and efficiently realize the online monitoring of the circular runout of the coating roller.

[0015] In one implementation, the data preprocessing of the detection matrix to obtain a feature matrix includes: normalizing the data in the detection matrix; filling null values ​​in each column of the detection matrix according to a preset value, such as the average value of the data in the column containing the null value; filling the columns of the detection matrix to give it a preset number of columns; and using principal component analysis (PCA) to reduce the dimensionality of the detection matrix to obtain the feature matrix. By processing the detection matrix through normalization, null value processing, data filling, and PCA dimensionality reduction, a corresponding feature matrix can be obtained, reducing the amount of data that needs to be input into the machine learning model.

[0016] Secondly, a device for detecting circular runout is provided. The device includes: a data acquisition module for acquiring data from multiple detection points on an electrode, wherein the data for each detection point includes the coating weight at each detection point and the angle through which the coating roller rotates when the coating weight at each detection point is detected; and a first data processing module for determining the circular runout information of the coating roller based on the data from the multiple detection points.

[0017] In one implementation, the data acquisition module is specifically used to: determine the angle through which the coating roller rotates based on the coating speed of the electrode sheet and the time elapsed when the coating weight is detected at each detection point.

[0018] In one implementation, the data acquisition module is specifically used to: multiply the coating speed by the time to obtain the coating length; and divide the coating length by the circumference of the coating roller to obtain the angle through which the coating roller has rotated.

[0019] In one implementation, the first data processing module is specifically used to: determine the coating weight of each detection point based on the signal transmitted from the detection point to the ray, wherein the ray reciprocates across the surface of the electrode along the width direction of the electrode during the coating process.

[0020] In one implementation, the device further includes a second data processing module, configured to: perform statistical processing on the coating weight in the data of each detection point based on the angle in the data of each detection point, and obtain a detection matrix composed of multiple coating weights; the step of determining the circular runout information of the coating roller based on the data of the multiple detection points includes: determining the circular runout information of the coating roller based on the detection matrix.

[0021] In one implementation, the second data processing module is specifically used to: take the remainder of the angle corresponding to each detection point modulo 2π to convert the angle corresponding to each detection point to the range of 0 to 2π; take the coating weight of each detection point and the angle corresponding to each detection point as a set of data to obtain an N×M matrix composed of multiple sets of data from multiple detection points, where N is the number of single passes traversed when the ray reciprocates to scan the surface of the electrode, and M is the number of detection points located in the coating area on each single pass; divide 0 to 2π into multiple intervals based on a preset angle value, and in each column of the N×M matrix, statistically analyze the coating weight of the data group whose angle is in the same interval to obtain a D×M detection matrix composed of multiple statistical data, where D is the ratio of 2π to the preset angle.

[0022] In one implementation, the second data processing module is specifically used to: perform K statistical methods on the coating weight of the data group whose angles are in the same interval in each column of the N×M matrix, to obtain the detection matrix of K×D×M composed of K statistical methods of coating weight, where K is a positive integer, and the K statistical methods include at least one of the average statistical method, variance statistical method and median statistical method.

[0023] In one implementation, the first data processing module is specifically used to: preprocess the detection matrix to obtain a feature matrix; and use the feature matrix as input to a machine learning model to obtain the circular runout information of the coating roller output by the machine learning model.

[0024] In one implementation, the first data processing module is specifically used to: normalize the data in the detection matrix; fill the empty values ​​in each column of the detection matrix according to a preset value; fill the columns of the detection matrix to give the detection matrix a preset number of columns; and use principal component analysis (PCA) to reduce the dimensionality of the detection matrix to obtain the feature matrix.

[0025] In one implementation, the preset value is the average value of the data in the column containing the null value.

[0026] In one implementation, the machine learning model includes any one of the following: logistic regression, Gaussian Naive Bayes, support vector machine, and decision tree.

[0027] In one implementation, the machine learning model is trained using multiple feature matrices and their corresponding known circular jump information.

[0028] Thirdly, an apparatus for detecting circular runout is provided, comprising a memory and a processor, the memory storing computer instructions, the processor invoking the computer instructions to cause the apparatus to implement the method according to the first aspect or any implementation thereof.

[0029] Fourthly, a computer-readable storage medium is provided for storing a computer program that, when executed by a computing device, causes the computing device to implement the method according to the first aspect or any implementation thereof.

[0030] Fifthly, a computer program product is provided that includes computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is executed in an electronic device, a processor in the electronic device performs the method according to the first aspect or any implementation thereof. Attached Figure Description

[0031] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0032] Figure 1 This is a schematic diagram showing the positional relationship between the coating roller and the slit.

[0033] Figure 2 This is a schematic flowchart of a method for detecting circular runout according to an embodiment of this application.

[0034] Figure 3 This is a schematic diagram showing the relationship between the rotation angle of the coating roller and the scanning trajectory of the electrode.

[0035] Figure 4 This is a schematic diagram showing the relationship between the rotation angle of the coating roller and the scanning trajectory of the electrode.

[0036] Figure 5 This is a schematic flowchart illustrating the data processing procedure in an embodiment of this application.

[0037] Figure 6 yes Figure 5 The diagram illustrates a specific implementation of the method shown.

[0038] Figure 7 yes Figure 5 The diagram illustrates a specific implementation of the method shown.

[0039] Figure 8 This is a schematic flowchart illustrating the data processing method of an embodiment of this application.

[0040] Figure 9 This is a diagram illustrating the selection of data samples for training a machine learning model.

[0041] Figure 10 This is a schematic block diagram of a device for detecting circular runout according to an embodiment of this application. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0043] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used in the description of this application is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms "comprising" and "having," and any variations thereof, in the description, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the description, claims, or accompanying drawings of this application are used to distinguish different objects, not to describe a specific order or hierarchy. "Vertical" is not strictly vertical, but within the allowable tolerance range. "Parallel" is not strictly parallel, but within the allowable tolerance range.

[0044] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.

[0045] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," "linking," and "attachment" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0046] In this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, in this application, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0047] In the embodiments of this application, the same reference numerals denote the same components, and for the sake of brevity, detailed descriptions of the same components are omitted in different embodiments. It should be understood that the thickness, length, width, and other dimensions of various components in the embodiments of this application shown in the accompanying drawings, as well as the overall thickness, length, width, and other dimensions of the integrated device, are merely illustrative and should not constitute any limitation on this application.

[0048] A battery typically refers to a single physical module comprising one or more individual battery cells to provide higher voltage and capacity. For example, a battery may include a battery module or a battery pack. Typically, a battery also includes a housing for encapsulating one or more battery cells. The housing prevents liquids or other foreign matter from affecting the charging or discharging of the battery cells.

[0049] A battery cell includes an electrode assembly and an electrolyte. The electrode assembly consists of a positive electrode, a negative electrode, and a separator. The battery cell primarily functions by the movement of metal ions between the positive and negative electrodes. The positive electrode includes a positive current collector and a positive active material layer. The positive active material layer is coated on the surface of the positive current collector, and the uncoated current collector protrudes beyond the coated current collector, serving as the positive electrode tab. Taking a lithium-ion battery as an example, the positive current collector can be made of aluminum, and the positive active material can be lithium cobalt oxide, lithium iron phosphate, ternary lithium, or lithium manganese oxide, etc. The negative electrode includes a negative current collector and a negative active material layer. The negative active material layer is coated on the surface of the negative current collector, and the uncoated current collector protrudes beyond the coated current collector, serving as the negative electrode tab. The negative current collector can be made of copper, and the negative active material can be carbon or silicon, etc. To reduce the probability of melting due to high current, the positive electrode tabs are multiple and stacked together, and the negative electrode tabs are multiple and stacked together.

[0050] The electrode coating process is a crucial step in the formation of battery cells. The coating weight determines the capacity of the lithium battery, and weight consistency affects its cycle life and safety indicators. Currently, the coating weight can be monitored using an areal density meter. However, factors affecting coating weight consistency primarily originate from the coating equipment, such as the runout of the coating roller. Runout issues caused by wear, deformation, and installation problems from long-term use result in periodic fluctuations in coating weight. Currently, measuring runout is necessary to assess the coating equipment's condition, significantly increasing monitoring costs and making it impossible to pinpoint the location of significant runout on the coating roller surface. This makes it difficult to accurately determine whether to repair or replace the coating roller during maintenance, leading to wasted maintenance costs.

[0051] For example, such as Figure 1The positional relationship between the coating roller 110 and the slit 120 is shown. The rotation of the coating roller 110 drives the electrode 130 to move along the Y direction, where X is the width direction of the electrode (i.e., the width direction), and Y is the direction of movement of the electrode (i.e., the coating direction). When the electrode 130 passes through the slit 120, the material fed out by the slit 120 is coated on the surface of the electrode 130. Generally, the circular runout of the coating roller 110 is mainly caused by three aspects: the coating roller 110 itself, the transmission, and the measurement. Specifically, regarding the coating roller 110, the circular runout is mainly affected by the roundness difference caused by the limited processing accuracy of the coating roller 110, as well as the deformation and wear of its surface caused by long-term use. Regarding the transmission, the circular runout is mainly caused by poor bearing installation accuracy, poor bearing processing accuracy, and bearing wear caused by use, resulting in a gap between the coating roller 110 and the bearing, causing the coating roller 110 to exhibit sinusoidal runout during rotation. The combined effect of these two factors causes circular runout of the coating roller 110, which in turn causes the slit 120 to also exhibit sinusoidal fluctuations, thus affecting the coating weight of the electrode 130. Therefore, the circular runout of the coating roller 110 affects the consistency of the coating weight. The coating roller 110 in this embodiment can also be referred to as a back roller or coating back roller.

[0052] Therefore, this application provides a method for detecting the circular runout of a coating roller. The coating weight of the electrode sheet is correlated with the rotation angle of the coating roller. The circular runout information of the coating roller is obtained based on the coating weight at each detection point on the electrode sheet and the angle through which the coating roller rotates when the coating weight is detected at that detection point. This enables online detection of the circular runout of the coating roller without the need for additional circular runout detection equipment or sensing units, thus reducing costs and achieving higher detection accuracy.

[0053] It should be understood that the electrode sheets described in the embodiments of this application include thin sheet materials in the lithium battery industry such as positive electrode sheets, negative electrode sheets, positive current collectors, negative current collectors, and separator materials. The method for detecting circular runout in the embodiments of this application can be applied to the coating process of any of the above-mentioned types of electrode sheets to detect the circular runout of the coating roller in real time.

[0054] Figure 2 A schematic flowchart illustrating a method for detecting circular runout according to an embodiment of this application is shown. Figure 2 The method 200 includes some or all of the following steps.

[0055] In step 210, data from multiple detection points on the electrode are acquired.

[0056] The data for each detection point includes the coating weight at each detection point, as well as the angle through which the coating roller rotates when the coating weight is detected at each detection point.

[0057] In step 220, the circular runout information of the coating roller is determined based on data from multiple detection points.

[0058] In this embodiment, the electrode moves along the coating direction as the coating roller rotates. Multiple detection points on the electrode for detecting the coating weight are sequentially detected during the electrode's movement. The detection data for each detection point includes the coating weight at each detection point and the angle through which the coating roller rotates when the coating weight is detected at each detection point. By associating the coating weight of the electrode with the rotation angle of the coating roller, the circular runout information of the coating roller is determined based on the coating weight at each detection point on the electrode and the angle through which the coating roller rotates when the coating weight is detected at that detection point. This enables online detection of the circular runout of the coating roller without the need for additional circular runout detection equipment or sensing units, reducing costs and providing higher detection accuracy.

[0059] The coating weight at each detection point can be accurately determined, for example, by using the intensity of the signal transmitted from the detection point to the point by radiation. Specifically, during the coating process, radiation scans the surface of the electrode back and forth along the width of the electrode. Based on the signal transmitted from the detection point to the point by radiation, the coating weight at each detection point can be determined.

[0060] The angle of rotation of the coating roller can be determined, for example, based on the coating speed of the electrode and the time elapsed to detect the coating weight at each detection point. In this way, the coating weight of the electrode can be correlated with the rotation angle of the coating roller.

[0061] For example, multiplying the coating speed by time yields the coating length; dividing the coating length by the circumference of the coating roller gives the angle through which the coating roller has rotated. This method is convenient for calculation and can accurately determine the rotation angle of the coating roller corresponding to the current detection point based on the time information at that detection point.

[0062] Combination Figure 3 and Figure 4 To elaborate, the emitting device and sensor are respectively positioned on both sides of the electrode and move at a speed of V2 along the width direction of the electrode, i.e., the scanning direction. As the coating roller rotates, the electrode moves at a speed of V1 along its length direction, i.e., the coating direction. The ray reciprocates along its scanning direction to scan the surface of the electrode. Figure 3 As can be seen, due to the superposition of velocities V1 and V2, the scanning trajectory of the ray on the surface of the electrode is Z-shaped, meaning that the distribution of all detection points on the electrode surface forms a Z-shape. Here, θ is the angle through which the coating roller rotates, that is, the angle between the coating direction of the electrode and the scanning direction of the ray.

[0063] For example, the angle through which the coating roller rotates can be determined by the following formula:

[0064]

[0065] like Figure 3 As shown, the ray scans back and forth on the surface of the electrode, with multiple detection points along each single-pass scan path. Here, T is the sum of the single-pass scan times preceding the current detection point, t is the single-pass scan time, and Y is the number of measurement points set on each single pass. As an example, Figure 3 Five single passes are shown, each with three detection points. i represents the position of the current detection point on the single pass, where 1 ≤ i ≤ Y. r is the radius of the coating roller.

[0066] In one implementation, method 200 further includes: statistically processing the coating weight in the data of each detection point based on the angle in the data of each detection point, and obtaining a detection matrix composed of multiple coating weights; then, in step 220, the circular runout information of the coating roller is determined based on the detection matrix.

[0067] In this way, by statistically processing the coating weight at each detection point according to the rotation angle of the coating roller corresponding to each detection point, the coating weight can be converted into the space of the rotation angle of the back roller, and a detection matrix that can represent the characteristics of the coating weight can be obtained.

[0068] For example, such as Figure 5 As shown, method 200 may further include steps 231, 232 and 233 for obtaining the detection matrix described above.

[0069] In step 231, the angle corresponding to each detection point is modulo 2π to convert the rotation angle of the coating roller corresponding to each detection point to the range of 0 to 2π.

[0070] In step 232, the coating weight of each detection point and the rotation angle of the coating roller corresponding to each detection point are taken as a set of data to obtain an N×M matrix composed of multiple sets of data from multiple detection points. Here, N is the number of single passes traversed when the X-ray reciprocates to scan the surface of the electrode, and M is the number of detection points located in the coating area on each single pass.

[0071] In step 233, the range from 0 to 2π is divided into multiple intervals based on a preset angle value. In each column of the N×M matrix, the coating weight of the data group whose angle is in the same interval is statistically analyzed to obtain a D×M detection matrix composed of multiple statistical data, where D is the ratio of 2π to the preset angle.

[0072] Through steps 231 to 233, the coating weight at each detection point can be converted into the space of the rotation angle of the back roller, and a detection matrix representing the coating weight characteristics can be obtained.

[0073] In fact, after calculating the angle of rotation of the back roller corresponding to each detection point, it can be regarded as the angle θ of each measurement point in the polar coordinate domain, and the coating weight data of that detection point can be regarded as the polar radius p of that measurement point in the polar coordinate domain. Below, we will use... Figure 6 For example, this section describes how to extract features of the polar radius p and angle θ of all detection points in the polar coordinate domain for a membrane roll according to steps 231 to 233. Assume the original data of the membrane region of the membrane roll is N rows × Y columns, where N is the number of single passes traversed by the ray when scanning the surface of the electrode, and Y is the number of detection points on each single pass.

[0074] like Figure 6 As shown, firstly, the non-membrane region data is removed from the original N×Y matrix, resulting in an N×M matrix. From... Figure 1 As can be seen, the coating area includes the membrane area and the non-membrane area located between the membrane areas. The coating weight of the non-membrane area is 0. Deleting the data of the detection points in the non-membrane area can reduce the amount of subsequent data processing.

[0075] Calculate the angle θ of the back roller rotation corresponding to each detection point, convert it into an angle in the range of 0 to 2π by taking its remainder with respect to 2π, and take this angle and the coating weight of the measurement point as a set of data, denoted as (Ai,j, θi,j), to form an N×M detection matrix composed of the data (Ai,j, θi,j) of each detection point, where 1≤i≤N and 1≤j≤M.

[0076] like Figure 7 As shown, a binning statistical method is used. Based on a preset angle value g, the range from 0 to 2π is divided into multiple intervals. All data in each column of the N×M detection matrix are statistically analyzed according to the various intervals after binning, changing the number of rows from N to D, where D is 2π / g. This interval can be, for example, left-open and right-closed or left-closed and right-open.

[0077] In one implementation, in step 233, K statistical methods can be performed on the coating weight of data groups with angles in the same interval in each column of the N×M matrix to obtain a K×D×M detection matrix composed of K statistical methods of coating weight, where K is a positive integer.

[0078] For example, the K types of statistical data include at least one of the mean statistical data, variance statistical data, and median statistical data. These statistical methods can be used to simply and accurately represent the coating weight characteristics of the electrode.

[0079] As an example, such as Figure 7 As shown, taking K=3 as an example, based on the preset angle value g, 0 to 2π is divided into D=2π / g intervals. The coating weight corresponding to the angle in the same interval is statistically analyzed by average value, variance value and median value. Finally, a 3×D×M detection matrix is ​​formed by the three statistical features of the coating weight of each statistical point.

[0080] In one implementation, in step 220, determining the circular runout information of the coating roller based on the detection matrix includes: performing data preprocessing on the detection matrix to obtain a feature matrix; using the feature matrix as input to a machine learning model to obtain the circular runout information of the coating roller output by the machine learning model.

[0081] The detection matrix representing the coating weight characteristics of the electrode is preprocessed, and the preprocessed data is used as the input of the machine learning model. The output of the machine learning model is the circular runout information of the coating roller. The machine learning model can accurately and efficiently achieve online monitoring of the circular runout of the coating roller.

[0082] Before inputting the detection matrix obtained after feature extraction into the machine learning model, data preprocessing is required to effectively extract data features and reduce the amount of data that needs to be input into the machine learning model.

[0083] For example, the data preprocessing process may include normalization, null value removal, data imputation, and Principal Component Analysis (PCA) dimensionality reduction. By processing the detection matrix obtained above through normalization, null value removal, data imputation, and PCA dimensionality reduction, the corresponding feature matrix can be obtained, reducing the amount of data that needs to be input into the machine learning model.

[0084] For example, such as Figure 8 As shown, normalizing the data in the detection matrix obtained above unifies data of different dimensions to the same dimension, resulting in a 3×D×M matrix. During the acquisition of this detection matrix, there may be cases where statistical values ​​are empty in different angle intervals. Therefore, it is necessary to fill the empty values ​​in each column of the detection matrix according to a preset value, such as the average value of the data in the column containing the empty value, resulting in a 3×D×M matrix. For different membrane rolls, the final detection matrix may have different numbers of columns. Therefore, column padding can be performed on the detection matrix to give it a preset number of columns H, resulting in a 3×D×H matrix. Finally, PCA is used to reduce the dimensionality of the padded detection matrix, resulting in a 3×E×E feature matrix.

[0085] Here, PCA is used to decouple a multidimensional dataset into a set of continuous orthogonal matrices to represent the variable with the largest variance in the original data. PCA processing can typically reduce the dimensionality of a dataset. In this embodiment, the matrix after data imputation can be reduced to a uniform dimension, and the influence of redundant information generated by the dimensionality expansion operation can be eliminated. Figure 8 As can be seen, after PCA processing, the 3×D×H matrix is ​​transformed into a 3×E×E matrix. The PCA process involves calculating eigenvalues ​​and eigenvectors using the covariance matrix. The arrangement of these eigenvalues ​​represents the ranking of the principal components of the eigenvectors. The matrix composed of E-dimensional principal component vectors is the dimensionality-reduced feature matrix. The feature matrix has a dimension of D×E, and the eigenvalues ​​have a dimension of E×H. The D×E feature matrix is ​​the portion used in training after dimensionality reduction. In most cases, D = E. In special cases, to further save server memory, the D×E feature matrix is ​​subjected to another PCA dimensionality reduction, resulting in an E×E feature matrix and D×E eigenvalues. In this case, E < 0. <D。

[0086] In this embodiment, the machine learning model can be trained using multiple feature matrices and their corresponding known circular runout information. These feature matrices serve as sample matrices, and the measured values ​​of the circular runout are manually labeled. Through supervised learning training of a neural network, a mathematical model for predicting the circular runout value of the coating roller from coating weight data is achieved, thus realizing the purpose of online monitoring of the circular runout of the coating roller.

[0087] For example, such as Figure 9As shown, to avoid the influence of other factors on coating quality, the dataset used to train the above machine learning model can be extracted from P film rolls of a coating device within a certain time period, for example, P = 3503. Based on a set threshold for circular runout, the detected circular runout values ​​can be divided into two categories. Circular runout values ​​above the threshold are marked as 1, indicating a high degree of circular runout; circular runout values ​​below the threshold are marked as 0, indicating a low degree of circular runout. In actual production, it is impossible to specifically detect the circular runout of every single film roll. Instead, the circular runout is detected once for the coating device at regular intervals, i.e., the circular runout is detected according to a certain cycle. Data on the coating weight of film rolls near the detection position of the circular runout can be used. For example, within each detection cycle of the circular runout, the first 10% of film rolls use the detection value of the previous circular runout, the last 10% of film rolls use the detection value of the next circular runout, and the middle 80% of film rolls are not sampled. Then, coating weight data from the detection points within the film area of ​​these 700 film rolls are extracted from the obtained database and processed using the method described above 200 to form a dataset. A portion of this dataset is randomly selected as the training set, and another portion as the test set. For example, 560 data samples are sampled as the training set, and the remaining 140 data samples as the test set. Using the coating weight data from the training and test sets as input to the machine learning model, and the corresponding circular runout detection values ​​as the output, the machine learning model is trained and tested, ultimately yielding a machine learning model capable of detecting the circular runout of the coating roll.

[0088] Because the feature data is small in scale and the dataset is small in size, the machine learning model can be any of the following: logistic regression, Gaussian Naive Bayes, Support Vector Machine (SVM), and Classification and Regression Tree (CART).

[0089] Table 1 shows the test results for these four machine learning models. It can be seen that SVM handles small sample datasets better and has higher accuracy. Furthermore, it has strong interpretability for features that affect classification results; for example, SVM can be used as a machine learning model to detect the circular runout of coating rollers.

[0090] Table 1

[0091] Model Name loss function Training set accuracy Test set accuracy Logistic Regression Cross-entropy 98.1％ 97.56％ Gaussian Naive Bayes CrossEntropy 97.23％ 96.53％ SVM CrossEntropy 98.21％ 98.1％ CART CrossEntropy 97.01％ 96.83％

[0092] As can be seen, by deploying a continuously trained and iterative machine learning model in an online production system, it is possible to predict in real time the type of circular runout of the coating roller as the film roll passes through, thereby determining whether the coating roller is currently in good health and the coating quality of the film roll. Here, the type of circular runout can be, for example, a large or small degree of circular runout, or a probability of a large degree of circular runout, or a probability of a small degree of circular runout, etc.

[0093] Furthermore, feature analysis of the machine learning model can be used to identify specific locations on the coating roller that significantly affect the circular runout, thereby determining whether to repair or replace the coating roller based on the location and the number of such locations.

[0094] This application also provides a device 300 for detecting circular runout, such as... Figure 10 As shown, the device 300 includes a data acquisition module 310 and a first data processing module 320.

[0095] The data acquisition module 310 is used to acquire data from multiple detection points on the electrode, wherein the data for each detection point includes the coating weight at each detection point and the angle through which the coating roller rotates when the coating weight at each detection point is detected; the first data processing module 320 is used to determine the circular runout information of the coating roller based on the data from the multiple detection points.

[0096] In one implementation, the data acquisition module 310 is specifically used to: determine the angle through which the coating roller rotates based on the coating speed of the electrode sheet and the time elapsed when the coating weight is detected at each detection point.

[0097] In one implementation, the data acquisition module 310 is specifically used to: multiply the coating speed by the time to obtain the coating length; and divide the coating length by the circumference of the coating roller to obtain the angle through which the coating roller has rotated.

[0098] In one implementation, the first data processing module 320 is specifically used to: determine the coating weight of each detection point based on the signal transmitted from the detection point to the ray, wherein the ray reciprocates across the surface of the electrode sheet along the width direction of the electrode sheet during the coating process.

[0099] In one implementation, the device further includes a second data processing module 330, which is configured to: perform statistical processing on the coating weight in the data of each detection point based on the angle in the data of each detection point, and obtain a detection matrix composed of multiple coating weights; the step of determining the circular runout information of the coating roller based on the data of the multiple detection points includes: determining the circular runout information of the coating roller based on the detection matrix.

[0100] In one implementation, the second data processing module 330 is specifically used to: take the remainder of the angle corresponding to each detection point modulo 2π to convert the angle corresponding to each detection point to the range of 0 to 2π; take the coating weight of each detection point and the angle corresponding to each detection point as a set of data to obtain an N×M matrix composed of multiple sets of data from multiple detection points, where N is the number of single passes traversed when the ray reciprocates to scan the surface of the electrode, and M is the number of detection points located in the coating area on each single pass; divide 0 to 2π into multiple intervals based on a preset angle value, and in each column of the N×M matrix, statistically analyze the coating weight of the data group whose angle is in the same interval to obtain a D×M detection matrix composed of multiple statistical data, where D is the ratio of 2π to the preset angle.

[0101] In one implementation, the second data processing module 330 is specifically used to: perform K statistical methods on the coating weight of the data group whose angles are in the same interval in each column of the N×M matrix, to obtain the detection matrix of K×D×M composed of K statistical methods of coating weight, where K is a positive integer, and the K statistical methods include at least one of the average statistical method, variance statistical method and median statistical method.

[0102] In one implementation, the first data processing module 320 is specifically used to: preprocess the detection matrix to obtain a feature matrix; and use the feature matrix as input to a machine learning model to obtain the circular runout information of the coating roller output by the machine learning model.

[0103] In one implementation, the first data processing module 320 is specifically used to: normalize the data in the detection matrix; fill the empty values ​​in each column of the detection matrix according to a preset value; fill the columns of the detection matrix so that the detection matrix has a preset number of columns; and use principal component analysis (PCA) to reduce the dimensionality of the detection matrix to obtain the feature matrix.

[0104] In one implementation, the preset value is the average value of the data in the column containing the null value.

[0105] In one implementation, the machine learning model includes any one of the following: logistic regression, Gaussian Naive Bayes, support vector machine, and decision tree.

[0106] In one implementation, the machine learning model is trained using multiple feature matrices and their corresponding known circular jump information.

[0107] This application also provides a device for detecting circular runout, including a memory and a processor. The memory stores computer instructions, and the processor invokes the computer instructions to cause the device to implement the method 200 in any of the above implementations.

[0108] This application also provides a computer-readable storage medium for storing a computer program, which, when executed by a computing device, causes the computing device to implement the method 200 described in any of the above implementations.

[0109] This application also provides a computer program product including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is running in an electronic device, the processor in the electronic device executes the method 200 described in any of the above implementations.

[0110] It should be noted that, without conflict, the various embodiments and / or technical features described in this application can be arbitrarily combined with each other, and the resulting technical solutions should also fall within the protection scope of this application.

[0111] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0112] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the above method embodiments, and will not be repeated here.

[0113] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0114] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

Claims

1. A method for detecting circular runout, characterized in that, The method includes: Data from multiple detection points on the electrode is acquired, wherein the data for each detection point includes the coating weight at each detection point and the angle through which the coating roller rotates when the coating weight at each detection point is detected; Based on the angle in the data of each detection point, the coating weight in the data of each detection point is statistically processed to obtain a detection matrix composed of multiple coating weights; The circular runout information of the coating roller is determined based on the detection matrix. The step of statistically processing the coating weight in the data of each detection point based on the angle in the data of each detection point to obtain a detection matrix composed of multiple coating weights includes: The angle corresponding to each detection point is modulo 2π to convert the angle corresponding to each detection point to the range of 0 to 2π. The coating weight of each detection point and the angle corresponding to each detection point are taken as a set of data to obtain an N×M matrix composed of multiple sets of data from multiple detection points, where N is the number of single passes when the ray reciprocates to scan the surface of the electrode, and M is the number of detection points located in the coating area on each single pass. The range from 0 to 2π is divided into multiple intervals based on a preset angle value. In each column of the N×M matrix, the coating weight of the data group whose angles are in the same interval is statistically analyzed to obtain the D×M detection matrix composed of multiple statistical data, where D is the ratio of 2π to the preset angle.

2. The method according to claim 1, characterized in that, The angle through which the coating roller rotates when obtaining the coating weight detected at each detection point includes: The angle through which the coating roller rotates is determined based on the coating speed of the electrode sheet and the time taken to detect the coating weight at each detection point.

3. The method according to claim 2, characterized in that, Determining the angle rotated by the coating roller based on the coating speed of the electrode sheet and the time elapsed to detect the coating weight at each detection point includes: Multiplying the coating speed by the time yields the coating length; Dividing the coating length by the circumference of the coating roller yields the angle through which the coating roller has rotated.

4. The method according to any one of claims 1 to 3, characterized in that, Obtaining the coating weight at each detection point includes: The coating weight of each detection point is determined based on the signal transmitted from the detection point to the ray, wherein the ray reciprocates across the surface of the electrode along the width direction of the electrode during the coating process.

5. The method according to any one of claims 1 to 3, characterized in that, In each column of the N×M matrix, the coating weight of data groups with angles in the same interval is statistically analyzed to obtain a D×M matrix composed of statistical data on coating weight, including: In each column of the N×M matrix, K types of statistics are performed on the coating weight of the data group whose angles are in the same interval, resulting in the detection matrix of K×D×M composed of K types of statistical data on coating weight, where K is a positive integer, and the K types of statistical data include at least one of the average statistical data, variance statistical data, and median statistical data.

6. The method according to any one of claims 1 to 3, characterized in that, Determining the circular runout information of the coating roller based on the detection matrix includes: The detection matrix is ​​preprocessed to obtain a feature matrix; Using the feature matrix as input to the machine learning model, the circular runout information of the coating roller is obtained from the output of the machine learning model.

7. The method according to claim 6, characterized in that, The step of preprocessing the detection matrix to obtain the feature matrix includes: The data in the detection matrix are normalized. According to preset values, empty values ​​in each column of the detection matrix are filled. The detection matrix is ​​padded with columns to give it a preset number of columns; The detection matrix is ​​reduced in dimensionality using principal component analysis (PCA) to obtain the feature matrix.

8. The method according to claim 7, characterized in that, The preset value is the average value of the data in the column containing the null value.

9. The method according to claim 6, characterized in that, The machine learning model includes any of the following: Logistic regression, Gaussian Naive Bayes, Support Vector Machine (SVM), and Decision Tree (CART).

10. A device for detecting circular runout, characterized in that, The device includes: The data acquisition module is used to acquire data from multiple detection points on the electrode, wherein the data for each detection point includes the coating weight at each detection point and the angle through which the coating roller rotates when the coating weight at each detection point is detected; The second data processing module performs statistical processing on the coating weight in the data of each detection point based on the angle in the data of each detection point, and obtains a detection matrix composed of multiple coating weights. The first data processing module is used to determine the circular runout information of the coating roller based on the detection matrix. Specifically, the second data processing module is used for: The angle corresponding to each detection point is modulo 2π to convert the angle corresponding to each detection point to the range of 0 to 2π. The coating weight of each detection point and the angle corresponding to each detection point are taken as a set of data to obtain an N×M matrix composed of multiple sets of data from multiple detection points, where N is the number of single passes when the ray reciprocates to scan the surface of the electrode, and M is the number of detection points located in the coating area on each single pass. The range from 0 to 2π is divided into multiple intervals based on a preset angle value. In each column of the N×M matrix, the coating weight data of the data group whose angles are in the same interval are statistically analyzed to obtain the D×M detection matrix composed of multiple statistical data, where D is the ratio of 2π to the preset angle.

11. The apparatus according to claim 10, characterized in that, The data acquisition module is specifically used for: The angle through which the coating roller rotates is determined based on the coating speed of the electrode sheet and the time taken to detect the coating weight at each detection point.

12. The apparatus according to claim 11, characterized in that, The data acquisition module is specifically used for: Multiplying the coating speed by the time yields the coating length; Dividing the coating length by the circumference of the coating roller yields the angle through which the coating roller has rotated.

13. The apparatus according to any one of claims 10 to 12, characterized in that, The first data processing module is specifically used for: The coating weight of each detection point is determined based on the signal transmitted from the detection point to the ray, wherein the ray reciprocates across the surface of the electrode along the width direction of the electrode during the coating process.

14. The apparatus according to any one of claims 10 to 12, characterized in that, The second data processing module is specifically used for: In each column of the N×M matrix, K types of statistics are performed on the coating weight data of the data group whose angles are in the same interval, to obtain the detection matrix of K×D×M composed of K types of statistical data of coating weight, where K is a positive integer, and the K types of statistical data include at least one of the average statistical data, variance statistical data and median statistical data.

15. The apparatus according to any one of claims 10 to 12, characterized in that, The first data processing module is specifically used for: The detection matrix is ​​preprocessed to obtain a feature matrix; Using the feature matrix as input to the machine learning model, the circular runout information of the coating roller is obtained from the output of the machine learning model.

16. The apparatus according to claim 15, characterized in that, The first data processing module is specifically used for: The data in the detection matrix are normalized. According to preset values, empty values ​​in each column of the detection matrix are filled. The detection matrix is ​​padded with columns to give it a preset number of columns; The detection matrix is ​​reduced in dimensionality using principal component analysis (PCA) to obtain the feature matrix.

17. The apparatus according to claim 16, characterized in that, The preset value is the average value of the data in the column containing the null value.

18. The apparatus according to claim 15, characterized in that, The machine learning model includes any of the following: Logistic regression, Gaussian Naive Bayes, Support Vector Machine (SVM), and Decision Tree (CART).

19. A device for detecting circular runout, characterized in that, The device includes a processor and a memory, the memory storing computer instructions, and the processor invoking the computer instructions to cause the device to perform the method according to any one of claims 1 to 9.

20. A computer-readable storage medium, characterized in that, Used to store a computer program, which, when executed by a computing device, causes the computing device to implement the method according to any one of claims 1 to 9.

21. A computer program product, characterized in that, Includes computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein, when the computer-readable code is executed in an electronic device, the processor in the electronic device performs the method of any one of claims 1 to 9.

Images