A learning device and method for predicting adhesive force to electrodes, an electrode monitoring device and electrode manufacturing method using a predictive model learned therefrom, and a lithium secondary battery manufactured therefrom.

Abstract

The present invention relates to a learning device and method for predicting adhesive force to an electrode, an electrode monitoring device and electrode manufacturing method utilizing a prediction model learned using the same, and a lithium secondary battery manufactured therefrom. An exemplary embodiment of the present invention provides a learning device for predicting adhesive force to an electrode, which includes a memory storing the near-infrared spectrum of the electrode and measured values ​​for the adhesive force of the electrode, a prediction model that predicts the adhesive force of the electrode by receiving the differential mean of a plurality of wavenumber intervals containing characteristics for the adhesive force of the electrode in the near-infrared spectrum, and a processor that receives the near-infrared spectrum, performs a first derivative with respect to the near-infrared spectrum, extracts a plurality of wavenumber intervals from the first-differentiated near-infrared spectrum, calculates the differential mean of the plurality of wavenumber intervals and transmits it to the prediction model. The processor receives the predicted value predicted by the prediction model for the adhesive force of the electrode and can train the prediction model so that the predicted value approaches the measured value.

Description

Technical Field

[0001] [Cross - reference to Related Applications] This application claims the benefit of priority based on Korean Patent Application No. 10 - 2023 - 0179489 filed on December 12, 2023 and Korean Patent Application No. 10 - 2024 - 0154935 filed on November 5, 2024, and all the contents disclosed in the documents of the Korean patent applications are included as part of this specification.

[0002] The present invention relates to a learning device and method for predicting the adhesion force to an electrode, an electrode monitoring device and an electrode manufacturing method using a prediction model learned using the same, and a lithium secondary battery manufactured thereby.

Background Art

[0003] Conventionally, in order to measure the adhesion force of an electrode, a part of the electrode to be measured was cut to meet the standard, one side of the electrode was fixed to a base material using double - sided tape, and then the adhesion force of the electrode was measured while peeling the electrode with a physical property measuring instrument.

[0004] Such a destructive - type inspection has problems in that a part of the generated electrode is lost, the same operation needs to be repeated for each produced electrode sample, and a relatively long time is required for the electrode peeling process, so that the adhesion force of the electrode changing in the process cannot be measured in real - time.

[0005] On the other hand, there is a need for an electrode adhesion force prediction method that can monitor the adhesion force of an electrode in real - time in the process while reducing the loss of the produced electrode.

Summary of the Invention

Problems to be Solved by the Invention

[0006] An object of the present invention is to provide a learning method and device capable of predicting the adhesion force to an electrode in a non - destructive manner using near - infrared spectrum and machine learning. [Means for solving the problem]

[0007] A learning device for predicting adhesive force to an electrode according to an exemplary embodiment of the present invention includes: a memory storing a near-infrared spectrum for an electrode and a measured value for the adhesive force of the electrode; a prediction model that receives a differential mean of a plurality of wavenumber intervals containing characteristics for the adhesive force of the electrode from the near-infrared spectrum and predicts the adhesive force of the electrode; and a processor that receives the near-infrared spectrum, performs a first derivative with respect to the near-infrared spectrum, extracts a plurality of wavenumber intervals from the first-differentiated near-infrared spectrum, calculates a differential mean of the plurality of wavenumber intervals and transmits it to the prediction model. The processor receives the predicted value predicted by the prediction model for the adhesive force of the electrode and can train the prediction model so that the predicted value approaches the measured value.

[0008] The predictive model can predict the adhesive strength of the electrodes after the coating process or after rolling.

[0009] Multiple wavenumber intervals may include a moisture interval in which the characteristics of the near-infrared spectrum are altered by the moisture in the electrode, and a binder interval in which the characteristics of the near-infrared spectrum are altered by the binder in the electrode.

[0010] The processor may include a differentiator that uses a Savitsky-Golay filter to perform a first-order derivative of the near-infrared spectrum.

[0011] The processor receives input of near-infrared spectra and measured values, calculates multiple differential means for each of the multiple wavenumber intervals of the near-infrared spectrum, generates multiple differential means and measured values ​​as training data, and can build a dataset by generating training data for multiple electrodes.

[0012] The processor receives inputs of near-infrared spectrum, measured values, electrode thickness, and electrode active material loading value. It calculates multiple differential averages for each of the multiple wavenumber intervals of the near-infrared spectrum, and generates training data from these multiple differential averages, measured values, electrode thickness, and electrode active material loading value. This training data can then be generated for multiple electrodes to construct a dataset.

[0013] The predictive model can learn to predict electrode adhesion by applying multiple training data randomly selected from the dataset to a random forest learning model.

[0014] A learning method for predicting the adhesive force to an electrode according to an exemplary embodiment of the present invention may include the steps of: a processor performing a first derivative on a near-infrared spectrum input to the electrode, extracting a plurality of wavenumber intervals containing characteristics for the adhesive force of the electrode from the first derivative near-infrared spectrum, and calculating the differential mean of the plurality of wavenumber intervals; a prediction model predicting the adhesive force of the electrode by receiving the differential mean of the plurality of wavenumber intervals and a measured value input for the adhesive force of the electrode as input; and a processor training the prediction model so that the predicted value predicted by the prediction model for the adhesive force of the electrode approaches the measured value.

[0015] The electrode adhesion strength may include the adhesion strength after the electrode coating process or the adhesion strength after the electrode rolling process.

[0016] Multiple wavenumber intervals may include a moisture interval in which the characteristics of the near-infrared spectrum are altered by the moisture in the electrode, and a binder interval in which the characteristics of the near-infrared spectrum are altered by the binder in the electrode.

[0017] The step of extracting multiple wavenumber intervals may include a step in which the processor uses a Savitsky-Goray filter to perform a first derivative on the near-infrared spectrum.

[0018] The learning method may further include the steps of generating training data using a processor, which consists of differential mean values ​​and measured values ​​calculated for multiple wavenumber intervals of the near-infrared spectrum, and generating training data for multiple electrodes to construct a dataset.

[0019] The learning method may further include the steps of: generating training data using a processor, which includes multiple differential means calculated for each of multiple wavenumber intervals of the near-infrared spectrum, measured values, electrode thickness, and electrode active material loading value; and generating training data for multiple electrodes to construct a dataset.

[0020] The step of training the predictive model may include a step in which the predictive model learns to predict electrode adhesion by applying multiple training data randomly selected from the dataset to a random forest learning model.

[0021] An electrode manufacturing method utilizing a predictive model learned by the learning device described above may include the steps of: mixing an active material, a binder, and a conductive material in a mixer to produce a slurry; coating a support with the mixed slurry using a coating device and drying the slurry-coated support in a drying oven; irradiating the coated and dried electrode with near-infrared light using a near-infrared spectrometer to obtain a near-infrared spectrum, inputting the obtained near-infrared spectrum into the learned predictive model, and predicting the adhesion force to the coated and dried electrode; rolling the coated and dried electrode with a rolling roller; cutting the rolled electrode with a cutting device, and processing the cut electrode into a predetermined shape with a notching device.

[0022] The electrode manufacturing method using the prediction model learned by the above-described learning device includes a step of manufacturing a slurry by mixing an active material, a binder, and a conductive material with a mixer, a step of coating the mixed slurry on a support by a coating device, and a step of drying the support coated with the slurry by a drying oven, a step of rolling the coated and dried electrode by a rolling roller, a step of irradiating the rolled electrode with near-infrared light by a near-infrared spectrometer to obtain a near-infrared spectrum, inputting the obtained near-infrared spectrum into the learned prediction model to predict the adhesion force of the rolled electrode, a step of cutting the rolled electrode by a cutting device, and a step of processing the cut electrode into a predetermined form by a notch device.

[0023] An exemplary secondary battery according to an embodiment of the present invention may include a positive electrode manufactured by the above-described electrode manufacturing method, a negative electrode manufactured by the above-described electrode manufacturing method, and a separator interposed between the positive electrode and the negative electrode.

[0024] The electrode monitoring device using the prediction model learned by the above-described learning device includes a monitoring processor that calculates a differential average of a plurality of wavenumber intervals including characteristics of the adhesion force of the electrode from the near-infrared spectrum received from the prediction model and the near-infrared spectrometer and transmits it to the prediction model, and the prediction model can predict the adhesion force of the electrode by receiving an input of the differential average of the plurality of wavenumber intervals.

[0025] The near-infrared spectrum may be a spectrum obtained for the coated and dried electrode.

[0026] The near-infrared spectrum is a spectrum obtained for the rolled electrode.

Advantages of the Invention

[0027] According to an exemplary embodiment of the present invention, there is an advantage that the adhesion force to the electrode can be predicted in a non-destructive manner through machine learning.

[0028] In addition, by using the predicted adhesive force value, the change in the adhesive force of the electrode during production can be monitored in real time, and it is possible to discriminate between good and bad electrode quality.

Brief Description of the Drawings

[0029] [Figure 1] It is a block diagram of a learning device for predicting the adhesive force to an electrode according to an exemplary embodiment of the present invention. [Figure 2] It is a diagram for explaining a plurality of frequency intervals extracted by a processor. [Figure 3] It is a diagram for explaining a prediction model according to an exemplary embodiment of the present invention. [Figure 4] It is a flowchart of a learning method for predicting the adhesive force to an electrode according to an exemplary embodiment of the present invention. [Figure 5] It is a diagram for explaining the result of predicting the adhesive force of an electrode by using a prediction model according to an exemplary embodiment of the present invention. [Figure 6] It is an example of predicting the adhesive force of an electrode in real time by using a prediction model learned by a learning device or a learning method according to an exemplary embodiment of the present invention. [Figure 7] It is an example of predicting the adhesive force of an electrode in real time by using a prediction model learned by a learning device or a learning method according to an exemplary embodiment of the present invention. [Figure 8] It is a flowchart of an electrode manufacturing method using a prediction model learned by a learning device or a learning method according to an exemplary first embodiment of the present invention. [Figure 9] It shows an electrode manufacturing process according to an exemplary first embodiment of the present invention. [Figure 10] It is a flowchart of an electrode manufacturing method using a prediction model learned by a learning device or a learning method according to an exemplary second embodiment of the present invention. [Figure 11] It shows an electrode manufacturing process according to an exemplary second embodiment of the present invention. [Figure 12] This figure illustrates a lithium secondary battery manufactured by an electrode manufacturing method utilizing a predictive model learned by a learning device or learning method according to an exemplary embodiment of the present invention. [Modes for carrying out the invention]

[0030] In describing the exemplary embodiments disclosed herein, if a specific description of the relevant prior art is deemed to obscure the essence of the exemplary embodiments disclosed herein, such detailed description will be omitted. Furthermore, the accompanying drawings are provided for the purpose of easily understanding the exemplary embodiments disclosed herein, and it should be understood that the accompanying drawings do not limit the technical idea disclosed herein and include all modifications, equivalents, or substitutions that fall within the concept and technical scope of the present invention.

[0031] Terms including ordinal numbers such as "1st," "2nd," etc., can be used to describe various components, but the components are not limited by these terms. The terms are used solely for the purpose of distinguishing one component from another.

[0032] When one component is referred to as being "linked" or "connected" to another component, it is understood that it may be directly linked to or connected to the other component, but that other components may exist in between. On the other hand, when one component is referred to as being "directly linked" or "directly connected" to another component, it is understood that there are no other components in between.

[0033] In this application, terms such as “includes” or “have” are intended to identify the presence of features, figures, steps, actions, components, parts, or combinations thereof as described in the specification, and should be understood as not to preemptively exclude the possibility of the presence or addition of one or more other features, figures, steps, actions, components, parts, or combinations thereof.

[0034] The present invention will be described in detail below with reference to the attached drawings.

[0035] Figure 1 is a block diagram of a learning device for predicting the adhesive force to an electrode according to an exemplary embodiment of the present invention.

[0036] Referring to Figure 1, a learning device 1 for predicting the adhesive force to an electrode according to an exemplary embodiment of the present invention may include a memory 100, a prediction model 200, and a processor 300.

[0037] Memory 100 stores the near-infrared spectrum of the electrode and the measured value of the electrode's adhesive strength.

[0038] The near-infrared spectrum is measured on the surface of the electrode and can be acquired using the near-infrared spectrometer 440. The near-infrared spectrometer 440 acquires the near-infrared spectrum of the electrode by irradiating the electrode surface with near-infrared light in the wavelength range of 750 to 25000 nm and scanning it.

[0039] According to an exemplary embodiment, the near-infrared spectrum is 4000–13333 cm⁻¹. -1 It may also be a spectrum for the wavenumber region.

[0040] According to an exemplary embodiment, the near-infrared spectrum may include multiple characteristic segments related to the adhesive strength of the electrodes, and these multiple characteristic segments may be separated based on wavenumber.

[0041] The measured values ​​are obtained by measuring the adhesive strength of the electrodes, and can be obtained through conventionally known electrode adhesive strength measurement methods. For example, the measured values ​​may be values ​​obtained through a 90° peel test.

[0042] According to an exemplary embodiment, the measured values ​​can be separated by the processes used to produce the electrodes. For example, the measured values ​​may be the adhesive strength of the electrodes after the coating process, or the adhesive strength of the electrodes after the rolling process.

[0043] The near-infrared spectrum and measurements stored in memory 100 may be sent to processor 300 for training of prediction model 200.

[0044] The prediction model 200 can predict electrode adhesion by receiving a differential average input of multiple wavenumber intervals containing properties related to electrode adhesion from the near-infrared spectrum.

[0045] Since the properties of the near-infrared spectrum change depending on the adhesive strength of the electrodes, the prediction model 200 can predict the adhesive strength of the electrodes by analyzing the changes in the properties of the near-infrared spectrum. In other words, properties related to the adhesive strength of the electrodes are obtained from the near-infrared spectrum, and the prediction model 200 can learn the correlation between the changes in properties related to the adhesive strength of the electrodes and the adhesive strength of the electrodes from the near-infrared spectrum. Using the learned prediction model 200, the adhesive strength of the electrodes can be predicted from the properties related to the adhesive strength of the electrodes obtained from the near-infrared spectrum.

[0046] According to one exemplary embodiment, the predictive model 200 can predict the adhesive strength after the electrode coating process or after the electrode rolling process.

[0047] For example, the prediction model 200 can learn to predict the adhesion force to the electrode after the coating process by receiving the differential average of multiple wavenumber intervals in the near-infrared spectrum for the electrode after the coating process as input.

[0048] Furthermore, for example, if the prediction model 200 is trained using the near-infrared spectrum of the electrode after the rolling process, the trained prediction model 200 can predict the adhesion force to the electrode after the rolling process. In other words, depending on which process the near-infrared spectrum used to train the prediction model 200 was acquired from, it is possible to generate electrode adhesion force prediction models 200 for each process of the electrode.

[0049] According to one exemplary embodiment, the predictive model 200 can predict the adhesive strength of the electrode after a rolling process from the near-infrared spectrum acquired for the electrode after a coating process. For example, if the predictive model 200 is trained using the near-infrared spectrum acquired for the electrode after a coating process and a measured value of the adhesive strength of the electrode after a rolling process, a model can be generated that predicts the adhesive strength of the electrode after a rolling process from the near-infrared spectrum acquired for the electrode after a coating process. That is, the predictive model 200, trained using the near-infrared spectrum of the electrode after a coating process and a measured value of the adhesive strength of the electrode after a rolling process, can predict and output the adhesive strength of the electrode after a rolling process when a near-infrared spectrum is input.

[0050] The prediction model 200 can determine various prediction parameters necessary for prediction through learning that predicts the output based on the input. The prediction model 200 can be implemented as a program for multiple operations that use the determined prediction parameters to predict the output for a new input. The prediction model 200 can be implemented in hardware by a processor such as a CPU (Central Processing Unit) or GPU (Graphics Processing Unit) that executes the above program.

[0051] The processor 300 can extract multiple wavenumber intervals from the near-infrared spectrum, calculate the differential mean of the extracted wavenumber intervals, and transmit it to the prediction model 200.

[0052] According to an exemplary embodiment, the processor 300 can preprocess the near-infrared spectrum to extract multiple wavenumber intervals. This preprocessing of the near-infrared spectrum can be performed by the processor 300 receiving the near-infrared spectrum and performing a first derivative with respect to the near-infrared spectrum. The processor 300 can be implemented by a program containing multiple instructions that instruct multiple operations to perform the above-described operation, and hardware such as a CPU or GPU that executes this program. According to an exemplary embodiment, the processor 300 may include a differentiator 310, a Savitsky-Gorey filter 320, and a learning controller 330.

[0053] The differentiator 310 can perform a first derivative on the near-infrared intensity in each of several wavenumber intervals in the near-infrared spectrum, and generate each of the multiple derivative values ​​as one of several characteristic values.

[0054] According to an exemplary embodiment, the differentiator 310 can utilize the Savitsky-Gorey filter 320 to perform a first derivative of the near-infrared intensity in each of several wavenumber intervals.

[0055] The learning controller 330 can derive multiple near-infrared spectra for multiple electrodes from the database 2 and apply them to the differentiator 310. Each of the multiple near-infrared spectra may be converted into multiple characteristic values ​​for each of the multiple electrodes via the differentiator 310 and provided to the learning controller 330. The learning controller 330 can match the multiple characteristic values ​​for each of the multiple electrodes with the measured values ​​for each electrode and store them in the memory 100. The learning controller 330 can match all the derived multiple characteristic values ​​and measured values ​​for each of the multiple electrodes and store them in the memory 100 as a learning dataset.

[0056] The learning controller 330 can provide the prediction model 200 with characteristic values ​​and measured values ​​for any number of electrodes from the learning dataset as learning data. The prediction model 200 provides the learning controller 330 with the adhesive force values ​​(hereinafter referred to as "predicted values") predicted by learning using the learning data. The learning controller 330 can adjust the parameters of the prediction model 200 so that the predicted values ​​approach the measured values.

[0057] Figure 2 is a diagram illustrating the multiple wavenumber intervals extracted by the processor.

[0058] Referring to Figure 2, the multiple wavenumber intervals may be intervals that change depending on the adhesive strength of the electrodes in the near-infrared spectrum, or they may be specific intervals obtained experimentally.

[0059] For example, the multiple wavenumber intervals may refer to intervals in which peaks occur in the pre-processed near-infrared spectrum, such as by first derivative analysis, or they may be multiple pre-defined wavenumber intervals as shown in [Table 1] below. Here, the multiple pre-defined wavenumber intervals may be intervals in which the characteristic changes of the near-infrared spectrum due to adhesive force mainly occur.

[0060] [Table 1]

[0061] According to an exemplary embodiment, the wavenumber intervals may include water intervals and binder intervals. A water interval means an interval in which the properties of the near-infrared spectrum are altered by water in the electrode, and a binder interval means an interval in which the properties of the near-infrared spectrum are altered by a binder in the electrode.

[0062] For example, in [Table 1], sections 1, 2, 3, 6, 7, and 8 may be moisture sections, and sections 4, 5, 9, and 10 may be binder sections.

[0063] Figure 3 is a diagram illustrating a predictive model according to an exemplary embodiment of the present invention.

[0064] Referring to Figure 3, a processor 300 according to an exemplary embodiment of the present invention can construct a dataset 10 for training a prediction model 200 and randomly extract training data from the dataset 10. The processor 300 can also apply the extracted training data to the prediction model 200 to predict the adhesion strength of the electrodes. According to the exemplary embodiment, the prediction model 200 may be implemented as a random forest learning model.

[0065] According to an exemplary embodiment, the processor 300 receives a near-infrared spectrum and a measured value as input, calculates multiple differential means for each of the multiple wavenumber intervals extracted from the near-infrared spectrum, generates the multiple differential means and the measured value as training data, and constructs a dataset 10 by generating training data for multiple electrodes.

[0066] For example, as shown in Figure 3, the processor 300 can construct a dataset 10 using 25 training data points generated for each of the 25 electrodes. Here, each of the 25 training data points may include multiple differential means and measured values.

[0067] According to an exemplary embodiment, the processor 300 receives inputs of a near-infrared spectrum, measured values, electrode thickness, and electrode active material loading values, calculates multiple differential means for each of multiple wavenumber intervals extracted from the near-infrared spectrum, generates multiple differential means, measured values, electrode thickness, and electrode active material loading values ​​as training data, and constructs a dataset 10 by generating training data for multiple electrodes. In this case, each of the multiple training data included in the dataset 10 may include multiple differential means, measured values, electrode thickness, and electrode active material loading values.

[0068] According to an exemplary embodiment, the predictive model 200 can learn to predict the adhesive force of electrodes by applying multiple training data randomly extracted from the dataset 10 to the predictive model 200.

[0069] For example, the prediction model 200 can perform learning to predict the adhesive force of electrodes as follows. First, the learning controller 330 extracts learning data 21 from 1 to 10 from the dataset 10 and provides it to the prediction model 200. The prediction model 200 may be implemented as a random forest learning model. The prediction model 200 can perform learning using the learning data 21 and implement the first decision tree. The learning controller 330 extracts 10 arbitrary learning data 22 from the learning data 1 to 25 from the dataset 10 and provides it to the prediction model 200. The prediction model 200 can perform learning using the learning data 22 and implement the second decision tree. The learning controller 330 extracts 5 arbitrary learning data 23 from the learning data 1 to 25 from the dataset 10 and provides it to the prediction model 200. The prediction model 200 can perform learning using the learning data 22 and implement the third decision tree.

[0070] After training is complete, the prediction model 200 can determine the predicted value using three predicted values ​​derived by applying multiple characteristic values ​​for the input near-infrared spectrum to the first, second, and third decision trees. For example, the prediction model 200 can determine the predicted value as the average of the three predicted values.

[0071] On the other hand, the number of training data points extracted from dataset 10 and the number of decision trees that make up the random forest learning model are not limited to this example.

[0072] In other words, by changing the number of training data points used to generate the decision tree and the method of extracting the training data, diversity in the decision tree can be ensured, thereby improving the prediction accuracy of the random forest learning model.

[0073] Figure 4 is a flowchart of a learning method for predicting the adhesive force to an electrode according to an exemplary embodiment of the present invention.

[0074] Referring to Figure 4, a learning method for predicting the adhesive force to an electrode according to an exemplary embodiment of the present invention may include a wavenumber interval extraction step (S4100), an electrode adhesive force prediction step (S4200), and a prediction model learning step (S4300).

[0075] In the wavenumber interval extraction step (S4100), the processor 300 performs a first derivative with respect to the input near-infrared spectrum of the electrode, extracts multiple wavenumber intervals containing characteristics related to the electrode's adhesive strength from the first-derived near-infrared spectrum, and calculates the differential average of the multiple wavenumber intervals. At this time, the electrode's adhesive strength may include the adhesive strength after the electrode's coating process or the adhesive strength after the electrode's rolling process.

[0076] According to an exemplary embodiment, multiple wavenumber segments of the near-infrared spectrum can be divided into water segments, where the properties of the near-infrared spectrum are altered by the moisture in the electrode, and binder segments, where the properties of the near-infrared spectrum are altered by the binder in the electrode.

[0077] According to an exemplary embodiment, the wavenumber interval extraction step (S4100) may include a step (S4110) in which the processor 300 utilizes the Savitsky-Gorey filter 320 to perform a first derivative of the near-infrared spectrum.

[0078] In the electrode adhesion force prediction step (S4200), the prediction model 200 can predict the electrode adhesion force by receiving the differential mean over multiple wavenumber intervals and the input measured values ​​for the electrode adhesion force.

[0079] In the predictive model learning step (S4300), the processor 300 can train the predictive model 200 so that the predicted value it makes for the adhesive force of the electrodes approaches the measured value.

[0080] According to an exemplary embodiment, the predictive model learning step (S4300) may include a step in which the predictive model 200 learns to predict the adhesive force of electrodes by applying a random forest learning model to a plurality of training data randomly extracted from the dataset 10.

[0081] According to an exemplary embodiment, a learning method for predicting adhesion force to an electrode may further include the steps of generating differential mean and measured values ​​calculated by the processor 300 for multiple wavenumber intervals of the near-infrared spectrum as learning data, and generating learning data for multiple electrodes to construct a dataset 10.

[0082] According to an exemplary embodiment, a learning method for predicting adhesion to an electrode may further include the steps of generating a plurality of differential means, measured values, electrode thickness, and electrode active material loading values ​​calculated by the processor 300 for each of a plurality of wavenumber intervals of the near-infrared spectrum as learning data, and generating learning data for a plurality of electrodes to construct a dataset.

[0083] In other words, the training data may be generated using multiple differential means and measured values ​​calculated for each of multiple wavenumber intervals of the near-infrared spectrum, or it may be generated using multiple differential means, measured values, electrode thickness, and electrode active material loading value.

[0084] Figure 5 is a diagram illustrating the results of predicting the adhesive force of electrodes using a predictive model according to an exemplary embodiment of the present invention.

[0085] The graph in Figure 5 shows the results of comparing the predicted adhesive force predicted using the prediction model 200 with the actual adhesive force. In the graph, the horizontal axis represents the measured adhesive force of the electrodes, and the vertical axis represents the predicted adhesive force predicted using the prediction model 200 according to an exemplary embodiment of the present invention. The points displayed on the graph represent the predicted adhesive force predicted by the prediction model 200 in relation to the measured adhesive force. Therefore, the more linear the points displayed on the graph, the higher the prediction accuracy.

[0086] Referring to Figure 5, it can be seen that the electrode adhesion strength predicted using the prediction model 200 according to an exemplary embodiment of the present invention has an accuracy of 98%.

[0087] Figures 6 and 7 show an example in which the adhesive force of electrodes is predicted in real time using a predictive model learned by a learning device or learning method according to an exemplary embodiment of the present invention.

[0088] Referring to Figures 6 and 7, a predictive model learned by a learning device 1 or learning method according to an exemplary embodiment of the present invention can predict and output the electrode adhesion force that changes in the electrode process in real time.

[0089] According to an exemplary embodiment, the trained predictive model can receive input of a near-infrared spectrum measured in real time for an electrode, or input of the differential mean for each of a plurality of wavenumber intervals extracted from the near-infrared spectrum by the processor 300.

[0090] Here, the differential mean for each of the multiple wavenumber intervals extracted from the near-infrared spectrum can be calculated through the following processes: the near-infrared spectrometer 440 receives input of the near-infrared spectrum, which is repeatedly measured at regular time intervals on the electrode surface; the first derivative and filtering are performed on the near-infrared spectrum; a preset set of multiple wavenumber intervals are extracted; and the differential mean for each of the multiple wavenumber intervals is calculated.

[0091] According to an exemplary embodiment, the adhesive force predicted by the prediction model 200 is transmitted via the processor 300 to an external device such as a user terminal or display screen, and the results predicted by the prediction model 200 can be provided in the form of a graph or table.

[0092] For example, as shown in Figure 6, the data may be provided as a graph with time (sec) on the horizontal axis and adhesive strength on the vertical axis, or as shown in Figure 7, as a table showing the relationship between adhesive strength measurement time and measurement position. In Figure 7, "Top" and "Back" refer to the positions where the near-infrared spectrum was measured for a single electrode sample. In other words, the processor 300 can visualize the changes in adhesive strength occurring during the electrode process by providing them in the form of a table or graph.

[0093] Figure 8 is a flowchart of an electrode manufacturing method that utilizes a predictive model learned by a learning device or learning method according to an exemplary first embodiment of the present invention.

[0094] Referring to Figure 8, an exemplary electrode manufacturing method according to the first embodiment of the present invention may include a mixing step (S8100), a coating step (S8200), an electrode adhesion force prediction step (S8300), a rolling step (S8400), a slitting step (S8500), and a notching step (S8600).

[0095] Figure 9 shows an electrode manufacturing process according to an exemplary first embodiment of the present invention.

[0096] Referring to Figures 8 and 9, in the mixing step (S8100), the electrode material can be put into the mixer 410 to produce the slurry 41. That is, the mixer 410 can mix the electrode material that has been put in to produce the slurry 41. At this time, the electrode material used to produce the slurry 41 will vary depending on the capacity and output of the electrode to be produced, and one or more materials such as active material, conductive material, binder, and thickener can be mixed and used.

[0097] Here, the active material, which is responsible for the main electrochemical reactions of the positive electrode 610 and the negative electrode 620, may be lithium metal oxide, natural graphite, or artificial graphite, either alone or in combination. The conductive material is a substance that improves electrical conductivity, and carbon black or the like can be used. The binder is a substance that maintains the mechanical stability of the electrodes and acts as an adhesive to ensure that the active material adheres well to the support 42, and the adhesive strength of the electrodes can be adjusted by adjusting the amount of binder used in the manufacture of the slurry 41. The thickener is a chemical substance that increases the viscosity of the slurry 41, and the viscosity of the slurry 41 may be adjusted before coating by the amount of thickener contained in the slurry 41.

[0098] In the coating step (S8200), the mixed slurry 41 can be coated onto the support 42 and dried.

[0099] In the coating step (S8200), the support 42, which is wound in a roll, can be unwound in one direction. At this time, one of the ends of the support 42 being unwound in one direction may be transferred to the coating apparatus 420 by the transfer roller 480. Here, the support 42 is a component that supports the electrode material and provides an electrical connection, and may be, for example, aluminum foil or copper foil.

[0100] In the coating step (S8200), the coating apparatus 420 can coat at least one surface of the support 42 with the mixed slurry 41. At this time, the slurry 41 mixed in the mixer 410 may be transferred to the coating apparatus 420 via a transfer pipe 490 or the like. The support 42 coated with the slurry 41 may be transferred to the drying oven 430 by a transfer roller 480.

[0101] Here, the coating apparatus 420 can coat the support 42 using at least one slurry 41. For example, in the coating step (S8200), the support 42 can be coated using one type of slurry 41, or the support 42 can be multilayer coated using two or more slurries 41 having different compositions. When multilayer coating the support 42, the upper slurry 41 and the lower slurry 41 can be designed to have different types and composition ratios of active material, binder, thickener, and conductive material, depending on the capacity and output of the electrode to be manufactured.

[0102] The coating apparatus 420 can coat only one surface of the support 42 with slurry 41, or it can coat both surfaces of the support 42 with slurry 41. Figure 9 shows that slurry 41 is coated on both the upper and lower surfaces of the support 42, but the form and position of the slurry 41 coating on the support 42 are not limited to this. For example, slurry 41 may be coated only on the upper surface of the support 42, or only on the lower surface of the support 42. Also, in Figure 9, the cross-section of the coating apparatus 420 is shown as pentagonal and located on the upper part of the support 42, but the form and position of the coating apparatus 420 are not limited to this and can be freely changed according to the design requirements of the electrode manufacturing apparatus. For example, the coating apparatus 420 can employ a slot die coating method, which uniformly applies liquid slurry 41 onto the support 42 through a slit-shaped nozzle, or a comma coating method, which applies slurry 41 onto the support 42 between two rollers with a comma-shaped gap.

[0103] The coating apparatus 420 can adjust the amount of slurry 41 discharged and the coating speed, as well as the electrode thickness and loading amount.

[0104] In the coating step (S8200), the support 42 coated with slurry 41 can be dried in the drying oven 430. That is, the drying oven 430 dries the solvent in the slurry 41 coated on the support 42 and fixes the electrode material contained in the slurry 41 to the support 42.

[0105] The drying oven 430 can optimize the electrode drying process by adjusting the temperature and airflow of the drying oven according to the amount of slurry 41 discharged from the coating device 420, the coating speed of the slurry 41, the electrode thickness, and the loading amount, so that the electrode adhesion is uniform and stable.

[0106] According to an exemplary embodiment, in the coating step (S8200), a process is performed to coat and dry the slurry 41 on the upper surface of the support 42, and then to coat and dry the slurry 41 on the lower surface of the support 42, thereby producing an electrode in which the slurry 41 is coated and dried on both sides of the support 42.

[0107] For example, the coating step (S8200) may include a process of unwinding the rolled support 42 in one direction, a process of applying the slurry 41 to the upper surface of the support 42 to coat the support 42 with the slurry 41, a process of drying and fixing the slurry 41 coated on the upper surface of the support 42, a process of applying the slurry 41 to the lower surface of the support 42 to coat the support 42 with the slurry 41, and a process of drying and fixing the slurry 41 coated on the lower surface of the support 42.

[0108] In the electrode adhesion strength prediction step (S8300), near-infrared light is irradiated onto the coated and dried electrode to acquire a near-infrared spectrum, and the acquired near-infrared spectrum is input to the monitoring device 500 to predict the adhesion strength to the electrode.

[0109] In this case, the coated and dried electrodes can pass between two near-infrared spectrometers 440. In Figure 9, the coated and dried electrodes are shown passing between two near-infrared spectrometers 440 positioned above and below the transport path. However, the position and arrangement of the near-infrared spectrometers 440 are not limited to this, and can be freely changed depending on the location where the slurry 41 is coated onto the support 42 or the design requirements of the electrode manufacturing apparatus. For example, if the slurry 41 is coated only on the upper surface of the support 42, the near-infrared spectrometers 440 may be positioned above the transport path of the coated and dried electrodes.

[0110] The near-infrared spectrometer 440 can transmit the near-infrared spectrum acquired for the coated and dried electrodes to the monitoring device 500. According to an exemplary embodiment, the near-infrared spectrometer 440 can repeatedly measure the near-infrared spectrum of the electrode surface at regular time intervals.

[0111] The monitoring device 500 can repeatedly receive near-infrared spectra from the near-infrared spectrometer 440 at regular time intervals and can use the received near-infrared spectra to predict and provide adhesion force to electrodes. According to an exemplary embodiment, the monitoring device 500 may include a monitoring processor 510, a prediction model 520, and a display 530. In this case, the prediction model 520 may be a prediction model 520 learned by a learning device 1 according to an exemplary embodiment of the present invention.

[0112] The monitoring device 500 can input the near-infrared spectrum received from the near-infrared spectrometer 440 to the monitoring processor 510. The monitoring processor 510 can extract multiple wavenumber intervals from the near-infrared spectrum and can calculate the differential mean of the extracted wavenumber intervals and transmit it to the learned prediction model 520. For example, the monitoring processor 510 can extract a set number of wavenumber intervals by performing first differentiation and filtering on the received near-infrared spectrum and calculate the differential mean for each of the wavenumber intervals. Based on the differential mean for each of the wavenumber intervals, the learned prediction model 520 can predict the adhesion force to the coated and dried electrodes. The adhesion force predicted through the learned prediction model 520 may be transmitted through the monitoring processor 510 to a user terminal or an external device such as a display 530 and provided to the user.

[0113] Here, the electrode adhesion force predicted through the learned prediction model 520 may be the electrode adhesion force after the coating process or the electrode adhesion force after the rolling process. For example, if the learned prediction model 520 is a prediction model 520 that was learned using the near-infrared spectrum acquired for the electrode after the coating process and the measured value of the electrode adhesion force after the coating process, then the electrode adhesion force predicted in the electrode adhesion force prediction step (S8300) corresponds to the electrode adhesion force after the coating process. As another example, if the learned prediction model 520 is a prediction model 520 that was learned using the near-infrared spectrum acquired for the electrode after the coating process and the measured value of the electrode adhesion force after the rolling process, then the electrode adhesion force predicted in the electrode adhesion force prediction step (S8300) corresponds to the electrode adhesion force after the rolling process.

[0114] In the rolling step (S8400), the coated and dried electrode can be rolled to increase its density. At this time, the coated and dried electrode can pass between two rolling rollers 450. The rolling step (S8400) can optimize the thickness and density of the electrode after rolling by adjusting the distance between the two rolling rollers 450 and the temperature of the rolling rollers 450. By rolling the electrode in the rolling step (S8400), the gaps in the electrode can be minimized and the adhesive strength can be improved.

[0115] In the slitting step (S8500), the coated and rolled electrodes can be cut to the required size. At this time, the electrodes can be cut into various shapes by the cutting device 460, depending on the shape and size of the battery to be manufactured.

[0116] In Figure 9, the cutting device 460 is shown as a laser cutting machine that uses a laser beam to cut the electrode into a desired shape, but the type of cutting device 460 is not limited to this. For example, the cutting device 460 could be a rotary die cutter that uses a roller-shaped cutter to cut the electrode into a desired shape, or a scissors-type slitter that uses two blades that act like scissors to cut the electrode.

[0117] In the notching step (S8600), the cut electrode 43 can be processed into a specific shape that conforms to the design of the battery cell. For example, in the notching step (S8600), a specific portion of the electrode 43 cut by the notching device 470 may be precisely cut to form a tab 44 for connecting the positive electrode 610 or the negative electrode 620. In another example, in the notching step (S8600), the electrode 43 cut by the notching device 470 may be processed into a shape that conforms to the cell assembly (e.g., cylindrical, square plate type, etc.). According to an exemplary embodiment, the notching device 470 can be a punching machine or a laser machine, etc.

[0118] Figure 10 is a flowchart of an electrode manufacturing method utilizing a predictive model learned by a learning device or learning method according to an exemplary second embodiment of the present invention, and Figure 11 shows the electrode manufacturing process according to an exemplary second embodiment of the present invention.

[0119] Referring to Figures 10 and 11, an electrode manufacturing method utilizing a predictive model learned by a learning device or learning method according to an exemplary second embodiment of the present invention may include a mixing step (S9100), a coating step (S9200), a rolling step (S9300), an electrode adhesion force prediction step (S9400), a slitting step (S9500), and a notching step (S9600).

[0120] The electrode manufacturing method according to the exemplary second embodiment of the present invention is contrasted with the exemplary first embodiment of the present invention, in that the electrode adhesion force prediction step (S8300) is performed after the coating step (S8200), in that the electrode adhesion force prediction step (S9400) is performed after the rolling step (S9300). In other words, the electrode manufacturing methods according to the exemplary first and exemplary second embodiments of the present invention differ only in the order of some components (i.e., the coating step, the rolling step, and the electrode adhesion force prediction step), and are otherwise common. Hereinafter, only the parts of the electrode manufacturing method according to the exemplary second embodiment that differ from the exemplary first embodiment will be described.

[0121] In the coating step (S9200), the electrode that has been coated and dried may be transferred between the two rolling rollers 450 via a transfer device. In the rolling step (S9300), the electrode that has been rolled can be transferred to the near-infrared spectrometer 440 via a transfer device.

[0122] In the electrode adhesion force prediction step (S9400), near-infrared light is irradiated onto the rolled electrode to acquire a near-infrared spectrum, and the acquired near-infrared spectrum is input to the monitoring device 500 to predict the adhesion force to the electrode.

[0123] At this time, the rolled electrode can pass between two near-infrared spectrometers 440. In Figure 11, the coated and dried electrode is shown passing between two near-infrared spectrometers 440 positioned above and below the transport path. However, the position and arrangement of the near-infrared spectrometers 440 are not limited to this, and can be freely changed depending on the position where the slurry 41 is coated onto the support 42 or the design requirements of the electrode manufacturing apparatus. For example, if the slurry 41 is coated only on the upper surface of the support 42, the near-infrared spectrometers 440 may be positioned above the transport path of the coated and dried electrode.

[0124] The near-infrared spectrometer 440 can transmit the near-infrared spectrum acquired for the rolled electrode to the monitoring device 500. According to an exemplary embodiment, the near-infrared spectrometer 440 can repeatedly measure the near-infrared spectrum for the surface of the electrode at regular time intervals.

[0125] The electrode adhesion force predicted through the learned prediction model 520 in the electrode adhesion force prediction step (S9400) may be the electrode adhesion force after the coating process or the electrode adhesion force after the rolling process. For example, if the learned prediction model 520 is a prediction model 520 that was learned using the near-infrared spectrum acquired for the electrode after the rolling process and the measured value of the electrode adhesion force after the coating process, then the electrode adhesion force predicted in the electrode adhesion force prediction step (S9400) corresponds to the electrode adhesion force after the coating process. As another example, if the learned prediction model 520 is a prediction model 520 that was learned using the near-infrared spectrum acquired for the electrode after the rolling process and the measured value of the electrode adhesion force after the rolling process, then the electrode adhesion force predicted in the electrode adhesion force prediction step (S9400) corresponds to the electrode adhesion force after the rolling process.

[0126] Figure 12 illustrates a lithium secondary battery manufactured by an electrode manufacturing method utilizing a predictive model learned by a learning device or learning method according to an exemplary embodiment of the present invention.

[0127] Referring to Figure 12, a secondary battery according to an exemplary embodiment of the present invention may include a positive electrode 610, a negative electrode 620, a separator membrane 630, and an electrolyte (not shown in the drawings).

[0128] The positive electrode 610 may include a positive electrode current collector and a positive electrode active material disposed on at least one surface of the positive electrode current collector. The positive electrode active material is a substance that plays a role in accommodating and releasing lithium ions in a lithium secondary battery, and charging and discharging of the secondary battery occur in the process of lithium ions being released or inserted from the positive electrode active material. The positive electrode active material can be composed of lithium metal oxides such as lithium cobalt oxide, lithium iron phosphate, lithium nickel manganese cobalt oxide, and lithium manganese oxide. The positive electrode current collector plays a role in transmitting electrons generated in the positive electrode active material to the battery's external circuitry, smoothing the flow of electrons and improving the battery's efficiency. The positive electrode current collector can be composed of aluminum.

[0129] According to an exemplary embodiment, the positive electrode 610 can be manufactured by an electrode manufacturing method utilizing a predictive model 200 learned by a learning device and learning method according to an exemplary embodiment of the present invention.

[0130] The negative electrode 620 may include a negative electrode current collector and a negative electrode active material disposed on at least one surface of the negative electrode current collector. The negative electrode active material is a material that stores and releases lithium ions in a lithium secondary battery. During charging of the lithium secondary battery, lithium ions move from the positive electrode 610 to the negative electrode 620 and are stored in the negative electrode active material. During discharging of the lithium secondary battery, lithium ions move from the negative electrode 620 to the positive electrode 610 while an electric current is generated. The negative electrode active material can be made of graphite, silicon, lithium metal, metal oxides, and alloy materials. The negative electrode current collector plays a role in allowing electrons generated from the negative electrode active material to flow smoothly into the external circuit, providing a path for the electrons. The negative electrode current collector can be made of copper.

[0131] According to an exemplary embodiment, the negative electrode 620 can be manufactured by an electrode manufacturing method utilizing a predictive model 200 learned by a learning device and learning method according to an exemplary embodiment of the present invention.

[0132] The separation membrane 630 is a thin membrane located between the positive electrode 610 and the negative electrode 620, and has the property of allowing lithium ions to pass through but not electrons. Direct contact between the positive electrode 610 and the negative electrode 620 inside the lithium secondary battery is prevented via the separation membrane 630, thereby preventing short circuits. The separation membrane 630 can be made of a polymer such as polypropylene (PP) or polyethylene (PE).

[0133] The electrolyte is the medium that transfers lithium ions between the positive electrode 610 and the negative electrode 620 in a lithium secondary battery. The electrolyte allows lithium ions to move freely while the positive electrode 610 and the negative electrode 620 exchange electrons. The electrolyte can be composed of organic solvents such as ethylene carbonate, dimethyl carbonate, and ethyl methyl carbonate.

[0134] According to exemplary embodiments, the lithium secondary battery may be cylindrical, prismatic, or pouch-type secondary battery, but is not limited to any battery that corresponds to a charge / discharge device.

[0135] On the other hand, the above-described method can be created using a program that can be executed on a computer, and may be implemented on a general-purpose digital computer that runs the program using a computer-readable storage medium. The computer-readable storage medium may include magnetic storage media such as ROM (Read Only Memory), RAM (Random Access Memory), USB (Universal Serial Bus), floppy disks, and hard disks, or optical reading media such as CD (Compact Disc)-ROM and DVD (Digital Versatile Disc).

[0136] The scope of the present invention is indicated not by the detailed description but by the claims set forth below, and all modified or altered forms derived from the meaning and scope of the claims and the concept of equivalents thereto should be interpreted as being included in the scope of the present invention. [Explanation of symbols]

[0137] 1: Learning device 2: Database 10: Dataset 21, 22, 23: Training data 100: Memory 200: Predictive Model 300: Processor 310: Differentiator 320: Savitzky Gorey Filter 330: Learning Controller 41: Slurry 42:Support 43: Electrode 44: Tab 410: Mixer 420: Coating equipment 430: Drying oven 440:Near infrared spectrometer 450: Rolling roller 460: Cutting device 470: Notch device 480: Transfer roller 490:Transfer pipe 500: Monitoring device 510: Monitoring Processor 520: Predictive Model 530: Display 610: Positive electrode 620: Negative electrode 630: Separation membrane

Claims

1. A learning device for predicting the adhesive force to an electrode, A memory that stores the near-infrared spectrum of the electrode and the measured value of the adhesive strength of the electrode, A predictive model that predicts the adhesive strength of the electrode by receiving the differential average of multiple wavenumber intervals containing the characteristics of the adhesive strength of the electrode from the near-infrared spectrum, A processor that receives the near-infrared spectrum, performs a first derivative with respect to the near-infrared spectrum, extracts the plurality of wavenumber intervals from the first-derivative near-infrared spectrum, calculates the differential mean of the plurality of wavenumber intervals, and transmits it to the prediction model, The processor is a learning device that receives a predicted value predicted by the prediction model for the adhesive force of the electrode, and causes the prediction model to learn so that the predicted value approaches the measured value.

2. The learning device according to claim 1, wherein the predictive model predicts the adhesive strength of the electrode after a coating process or after a rolling process.

3. The aforementioned multiple wavenumber intervals are, The characteristics of the near-infrared spectrum are changed by the moisture content in the electrode in the moisture region, The learning device according to claim 1 or 2, comprising a binder section in which the characteristics of the near-infrared spectrum are changed by a binder in the electrode.

4. The learning apparatus according to claim 1, wherein the processor includes a differentiator that performs a first-order derivative of the near-infrared spectrum using a Savitsky-Gorey filter.

5. The learning device according to claim 1, wherein the processor receives the near-infrared spectrum and the measured value as input, calculates a plurality of differential means for each of a plurality of wavenumber intervals of the near-infrared spectrum, generates the plurality of differential means and the measured value as learning data, and generates the learning data for a plurality of electrodes to construct a dataset.

6. The learning device according to claim 1, wherein the processor receives the near-infrared spectrum, the measured value, the electrode thickness, and the active material loading value of the electrode as input, calculates a plurality of differential averages for each of a plurality of wavenumber intervals of the near-infrared spectrum, generates the plurality of differential averages, the measured value, the electrode thickness, and the active material loading value of the electrode as learning data, and generates the learning data for a plurality of electrodes to construct a dataset.

7. The learning device according to claim 5 or 6, wherein the prediction model applies a plurality of training data randomly extracted from the dataset to a random forest learning model to perform learning to predict the adhesive force of the electrodes.

8. A learning method for predicting the adhesive force to an electrode, The process involves a processor performing a first derivative on the near-infrared spectrum input to the electrode, extracting a plurality of wavenumber intervals from the first derivative near-infrared spectrum that contain characteristics related to the adhesive strength of the electrode, and calculating the differential mean of the plurality of wavenumber intervals. The prediction model predicts the adhesive force of the electrode by receiving inputs of the differential mean over the multiple wavenumber intervals and the measured value of the adhesive force of the electrode, A learning method comprising the step of causing the predictive model to learn using the processor such that the predicted value predicted by the predictive model for the adhesive force of the electrode approaches the measured value.

9. The learning method according to claim 8, wherein the adhesive force of the electrode includes the adhesive force of the electrode after a coating process or the adhesive force of the electrode after a rolling process.

10. The aforementioned multiple wavenumber intervals are, The characteristics of the near-infrared spectrum are changed by the moisture content in the electrode in the moisture region, The learning method according to claim 8, comprising a binder section in which the characteristics of the near-infrared spectrum are changed by the binder in the electrode.

11. The learning method according to claim 8, wherein the step of calculating the differential mean of the plurality of wavenumber intervals includes the step of the processor using a Savitsky-Gorey filter to perform a first derivative on the near-infrared spectrum.

12. The processor generates training data by combining the differential mean calculated for multiple wavenumber intervals of the near-infrared spectrum and the measured values. The learning method according to claim 8, further comprising the step of generating the aforementioned learning data for multiple electrodes and constructing a dataset.

13. The processor generates training data consisting of multiple differential averages calculated for each of the multiple wavenumber intervals of the near-infrared spectrum, the measured values, the thickness of the electrode, and the active material loading value of the electrode. The learning method according to claim 8, further comprising the step of generating the aforementioned learning data for multiple electrodes and constructing a dataset.

14. The learning method according to claim 12 or 13, wherein the step of training the predictive model includes the step of training the predictive model to predict the adhesive force of the electrodes by applying a plurality of training data randomly extracted from the dataset to a random forest learning model.

15. A method for manufacturing electrodes using a predictive model learned by the learning device described in claim 1, The process involves mixing the active material, binder, and conductive material in a mixer to produce a slurry, The process involves coating a support with a mixed slurry using a coating device, and drying the slurry-coated support in a drying oven. The steps include: using a near-infrared spectrometer to irradiate the coated and dried electrodes with near-infrared light to obtain a near-infrared spectrum, inputting the obtained near-infrared spectrum into the learned prediction model, and predicting the adhesion force to the coated and dried electrodes; The steps include rolling the coated and dried electrode with rolling rollers, An electrode manufacturing method comprising the steps of cutting a rolled electrode with a cutting device and processing the cut electrode into a predetermined shape with a notching device.

16. A method for manufacturing electrodes using a predictive model learned by the learning device described in claim 1, The process involves mixing the active material, binder, and conductive material in a mixer to produce a slurry, The process involves coating a support with a mixed slurry using a coating device, and drying the slurry-coated support in a drying oven. The steps include rolling the coated and dried electrode with rolling rollers, The steps include: irradiating the rolled electrode with near-infrared light using a near-infrared spectrometer to obtain a near-infrared spectrum, inputting the obtained near-infrared spectrum into the learned prediction model, and predicting the adhesion force to the rolled electrode; An electrode manufacturing method comprising the steps of cutting a rolled electrode with a cutting device and processing the cut electrode into a predetermined shape with a notching device.

17. A positive electrode manufactured by the electrode manufacturing method described in claim 15, A negative electrode manufactured by the electrode manufacturing method described in claim 15, A lithium secondary battery comprising a separator membrane interposed between the positive electrode and the negative electrode.

18. A positive electrode manufactured by the electrode manufacturing method described in claim 16, A negative electrode manufactured by the electrode manufacturing method described in claim 16, A lithium secondary battery comprising a separator membrane interposed between the positive electrode and the negative electrode.

19. A predictive model learned by the learning device described in claim 1, The system includes a monitoring processor that calculates the differential mean of multiple wavenumber intervals containing the properties of the electrode's adhesive strength from the near-infrared spectrum received from a near-infrared spectrometer and transmits it to the prediction model, The prediction model is a monitoring device that predicts the adhesive force of the electrode by receiving the differential average input of the multiple wavenumber intervals.

20. The monitoring apparatus according to claim 19, wherein the near-infrared spectrum is a spectrum obtained for coated and dried electrodes.

21. The monitoring apparatus according to claim 19, wherein the near-infrared spectrum is a spectrum acquired for a rolled electrode.

Images