JP7667481B2 - Proposed device, proposed method, and method for producing electrode mixture and program - Google Patents

Description

This disclosure relates to a proposed device, a proposed method, a method for producing an electrode mixture, and a program.

There is a technology that uses a machine learning model to suggest a compounding that improves the properties of a substance. For example, Patent Document 1 discloses a substance search device that uses a machine learning model to identify a predicted range of the spectral properties of a specific substance that will improve the evaluation value of the properties, and searches for a specific substance that has that predicted range of spectral properties. Patent Document 2 discloses a prediction method that uses a machine learning technique to predict the properties of a paint after adjustment when raw materials for adjusting the paint properties are added to the paint before adjustment.

JP 2021-149449 A International Publication No. 2022/131180

However, the prior art does not propose a formulation for the electrode mixture to be used in the production of electrodes.

This disclosure provides a technology that proposes a formulation for an electrode mixture whose physical properties meet target values.

The proposal device according to the first aspect of the present disclosure is a proposal device having a control unit that proposes a formulation of an electrode mixture, and the control unit generates a candidate formulation of the electrode mixture, predicts physical property values of the electrode mixture based on the candidate formulation, and outputs the candidate formulation in which the physical property values satisfy target values.

According to the first aspect of the present disclosure, it is possible to propose a formulation for an electrode mixture whose physical properties meet target values.

A second aspect of the present disclosure is a proposed device according to the first aspect, in which the control unit predicts the physical property value based on a trained model.

A third aspect of the present disclosure is a proposed device according to the first or second aspect, in which the electrode mixture is a mixture of an active material, a conductive material, and a binder.

A fourth aspect of the present disclosure is the proposed device according to the third aspect, in which the binder includes a polymer.

A fifth aspect of the present disclosure is the proposed device according to the fourth aspect, in which the polymer includes a fluororesin.

A sixth aspect of the present disclosure is a proposal device according to any one of the first to fifth aspects, in which the control unit predicts two or more of the physical property values and outputs the candidate formulation in which all of the two or more of the physical property values satisfy target values.

In the proposal method according to the seventh aspect of the present disclosure, a control unit of a proposal device that proposes a formulation of an electrode mixture generates a candidate formulation of the electrode mixture, predicts physical property values of the electrode mixture based on the candidate formulation, and outputs the candidate formulation in which the physical property values satisfy target values.

The method for producing an electrode mixture according to the eighth aspect of the present disclosure includes a step of mixing an active material, a conductive material, and a binder according to the candidate composition output by the proposed method according to the seventh aspect.

The program according to the ninth aspect of the present disclosure generates candidate formulations of the electrode mixture in a control unit of a proposal device that proposes a formulation of the electrode mixture, predicts physical property values of the electrode mixture based on the candidate formulations, and outputs the candidate formulations in which the physical property values satisfy target values.

1A to 1C are diagrams illustrating an example of a manufacturing process for an electrode sheet. FIG. 2 is a block diagram showing an example of a hardware configuration of the proposed device. 13 is a flowchart showing an example of a processing procedure of a learning method. FIG. 1 is a diagram showing an example of a viscosity change rate model after one day. FIG. 1 is a diagram showing an example of a model of viscosity change rate after 7 days. FIG. 2 is a diagram showing an example of a coating resistance model. FIG. 1 is a diagram illustrating an example of a three-point bending strength model. FIG. 13 is a diagram showing an example of a peel strength model. 11 is a flowchart illustrating an example of a processing procedure of the proposed method. FIG. 11 is a diagram illustrating an example of condition information. FIG. 11 is a diagram illustrating an example of proposal information.

Each embodiment will be described below with reference to the accompanying drawings. Note that in this specification and the drawings, components having substantially the same functional configuration are designated by the same reference numerals, and redundant description will be omitted.

[Embodiment]
An embodiment of the present disclosure is a proposal device that proposes a formulation of a mixture for an electrode. The mixture for an electrode in this embodiment is, for example, an electrode slurry.

The electrode slurry is a material used to prepare an electrode sheet for use in a secondary battery such as a lithium ion battery. The electrode slurry is a slurry containing an active material, a conductive additive, and a binder. The binder may be a mixture of multiple types of binders.

For example, a lithium ion battery is manufactured by filling a container with a laminate of a separator sandwiched between an electrode sheet that will become a positive electrode and an electrode sheet that will become a negative electrode, injecting an electrolyte solution, and sealing the container. Hereinafter, the electrode sheet may be referred to simply as "electrode." Note that the electrode slurry in this embodiment may be used to create either an electrode sheet that will become a positive electrode or an electrode sheet that will become a negative electrode.

Figure 1 is a diagram showing an example of the manufacturing process of an electrode sheet. As shown in Figure 1, the manufacturing process of an electrode sheet can be broadly divided into a slurry preparation process and an electrode preparation process. In the slurry preparation process, an active material, a conductive assistant, and a binder are mixed to prepare an electrode slurry. In the electrode preparation process, an electrode sheet is prepared using the electrode slurry prepared in the slurry preparation process.

In the slurry preparation process, first, the binder solution and conductive additive are put into a stirrer and kneaded. Next, the active material and solvent are added to the stirrer and the kneading is repeated again. For example, N-methylpyrrolidone (NMP) is used as the solvent. Then, more solvent is put into the stirrer and kneaded.

In the electrode preparation process, first, the electrode slurry is applied to a current collector such as aluminum foil. Next, the current collector to which the electrode slurry has been applied is heat-treated and dried. The heat treatment temperature may be increased stepwise, for example, to 90°C, 110°C, and 120°C. The completed electrode sheet is cut into a shape appropriate for the application and used to manufacture secondary batteries.

The composition of the electrode mixture significantly affects the physical properties of the electrode using that electrode mixture and the performance of the secondary battery using that electrode. Here, the composition of the electrode mixture includes the mixing ratio of the materials (i.e., the active material, the conductive additive, and the binder) and the mixing ratio of the binder.

Conventionally, the composition of electrode mixtures has been determined based on the designer's knowledge. For example, the designer first makes a hypothesis about a composition that meets the required specifications based on his or her own experience or knowledge. The designer then prototypes the electrode mixture with that composition and measures its physical properties through experiments. The designer may also use the electrode mixture to prototype an electrode or secondary battery and measure its performance. The designer then repeats the prototypes and experiments until an electrode mixture that meets the required specifications is obtained.

The purpose of this embodiment is to propose a formulation for an electrode mixture that satisfies the required specifications. In particular, this embodiment predicts the physical property values of the electrode mixture, and proposes a formulation for an electrode mixture whose physical property values meet the target values without performing prototypes or experiments.

In this embodiment, one or more trained models are used to predict physical property values related to the electrode mixture. The physical property values related to the electrode mixture include physical property values measured from the electrode mixture and physical property values measured from an electrode using the electrode mixture.

The physical properties measured from the electrode mixture include physical properties that indicate slurry stability. Examples of physical properties that indicate slurry stability include viscosity change rate, rheology, and sedimentation evaluation. In this embodiment, as an example, a machine learning model that predicts the viscosity change rate after 1 day and the viscosity change rate after 7 days will be described.

The physical properties measured from the electrodes include a physical property value indicating dispersibility. Examples of the physical property value indicating dispersibility include coating resistance and blackness. In this embodiment, as an example, a machine learning model that predicts coating resistance will be described.

The physical properties measured from the electrodes include a physical property value indicating flexibility. Examples of the physical property value indicating flexibility include three-point bending test results and mandrel test results. In this embodiment, as an example, a machine learning model for predicting three-point bending strength is described.

The physical properties measured from the electrodes include physical properties that indicate adhesion. Examples of physical properties that indicate adhesion include peel test results. In this embodiment, a machine learning model that predicts peel strength will be described as an example.

In one aspect, this embodiment makes it possible to efficiently determine the composition of an electrode mixture that satisfies required specifications. This makes it possible to shorten the lead time in the development of electrode mixtures, electrodes, and secondary batteries.

<Hardware Configuration>
Fig. 2 is a block diagram showing an example of a hardware configuration of the proposed device 100 in this embodiment. As shown in Fig. 2, the proposed device 100 includes a processor 101, a memory 102, an auxiliary storage device 103, an operation device 104, a display device 105, a communication device 106, and a drive device 107. Each piece of hardware of the proposed device 100 is connected to each other via a bus 108.

The processor 101 has various computing devices such as a CPU (Central Processing Unit). The processor 101 reads various programs installed in the auxiliary storage device 103 onto the memory 102 and executes them.

The memory 102 has a primary storage device such as a ROM (Read Only Memory) or a RAM (Random Access Memory). The processor 101 and the memory 102 form a so-called computer (hereinafter also referred to as a "control unit"), and the processor 101 executes various programs read onto the memory 102, causing the computer to realize various functions.

The auxiliary storage device 103 (hereinafter also referred to as the "storage unit") stores various programs and various data used when the various programs are executed by the processor 101.

The operation device 104 is an operation device that allows a user of the proposed device 100 to perform various operations. The display device 105 is a display device that displays the results of various processes executed by the proposed device 100.

The communication device 106 is a communication device for communicating with external devices via a communication network.

The drive device 107 is a device for setting the storage medium 109. The storage medium 109 here includes media that store information optically, electrically, or magnetically, such as CD-ROMs, flexible disks, and magneto-optical disks. The storage medium 109 may also include semiconductor memory that stores information electrically, such as ROMs and flash memories.

The various programs to be installed in the auxiliary storage device 103 are installed, for example, by setting the distributed storage medium 109 in the drive device 107 and reading the various programs stored in the storage medium 109 by the drive device 107. Alternatively, the various programs to be installed in the auxiliary storage device 103 may be installed by downloading them from a network via the communication device 106.

<Learning method flow>
3 is a flowchart showing an example of the flow of a learning method executed by the proposal device 100 in this embodiment. The learning method is a method for learning a prediction model used to propose an electrode mixture.

In step S1, the control unit of the proposed device 100 stores the learning data in the memory unit. The learning data is data used to learn the prediction model. The prediction model is a machine learning model for predicting physical property values related to the electrode mixture. Examples of the prediction models in this embodiment include a viscosity change rate model after 1 day, a viscosity change rate model after 7 days, a coating resistance model, a three-point bending strength model, and a peel strength model.

The training data includes parameters that are explanatory variables of the prediction model and parameters that are objective variables of the prediction model. The training data can be collected by conducting experiments using electrode mixtures prepared with various compositions. The training data may also be collected from information published in papers, databases, etc.

The learning data in this embodiment has the following data items.
(1) Material composition ratio (2) Binder compounding ratio (3) Solid content concentration (4) Initial viscosity (5) Basic binder properties (6) Electrode information (7) Viscosity change rate after 1 day (8) Viscosity change rate after 7 days (9) Coating resistance (10) Three-point bending strength (11) Peel strength

(1) The material composition ratio further includes the following data items:
(1-1) Weight ratio of active material (1-2) Weight ratio of conductive additive (1-3) Weight ratio of binder

(2) The binder mixing ratio is the weight ratio of each type of binder used in preparing the electrode mixture. (3) The solids concentration is the solids concentration when the electrode mixture is an electrode slurry. (4) The initial viscosity is the viscosity measured immediately after preparing the electrode mixture.

(5) The basic physical properties of the binder further include the following data items:
(5-1) Intrinsic viscosity (5-2) Elastic modulus (5-3) Polymerization method (5-4) Presence or absence of functional group modification (5-5) Mw/Mn
(5-6) Solution viscosity (5-7) Amount of modification (5-8) Degree of branching (5-9) Abnormal bond degree (5-10) Random ratio

(6) The electrode information further includes the following data items:
(6-1) Amount of slurry applied (6-2) Density before pressing (6-3) Density after pressing

(7) The viscosity change rate after one day is the ratio of the initial viscosity to the viscosity measured after one day. (8) The viscosity change rate after seven days is the ratio of the initial viscosity to the viscosity measured after one day. (9) The coating resistance may be measured using a measuring device on an electrode sheet made using the electrode mixture. (10) The three-point bending strength may be measured by performing a three-point bending test on an electrode sheet made using the electrode mixture. (11) The peel strength may be measured by performing a peel test on an electrode sheet made using the electrode mixture.

(1) Material composition ratio, (2) binder compounding ratio, (3) solids concentration, (4) initial viscosity, (5) binder basic properties, and (6) electrode information are used as explanatory variables of the prediction model. (7) Viscosity change rate after 1 day, (8) Viscosity change rate after 7 days, (9) coating resistance, (10) 3-point bending strength, and (11) peel strength are used as objective variables of the prediction model.

When collecting learning data, constraints on the composition of the electrode mixture or binder may be set based on the expected required specifications. While there are a huge number of combinations of electrode mixtures and binders, the combinations that can actually be used in the manufacture of secondary batteries are limited. By limiting the learning data to areas that are unlikely to be actually used, the amount of calculations required for learning can be reduced and prediction accuracy can be improved.

In this embodiment, the active material and conductive additive are fixed to one type. Two types of binders are mixed. One of the two types of binders is selected from a predetermined number of binder candidates. The number of binder candidates is not limited, but in this embodiment, ten types are used as an example.

The other of the two types of binders is a binder containing a polymer. An example of a polymer is a fluororesin. Hereinafter, a binder containing a fluororesin will be referred to as a "fluororesin binder."

The weight ratio of the active material is in the range of 96.0 to 98.0 wt%. The weight ratio of the conductive additive is in the range of 1.0 to 3.0 wt%. The weight ratio of the binder is in the range of 1.0 to 3.0 wt%. The weight ratio of the fluororesin binder to the total binder is in the range of 0 to 20 wt%.

In step S2, the control unit of the proposed device 100 learns a one-day viscosity change rate model based on the learning data stored in the memory unit. The one-day viscosity change rate model is a machine learning model that predicts the one-day viscosity change rate of the electrode mixture. The control unit of the proposed device 100 stores the learned one-day viscosity change rate model in the memory unit.

Figure 4 shows an example of a one-day viscosity change rate model. As shown in Figure 4, the one-day viscosity change rate model 300 inputs the material composition ratio, binder mixing ratio, initial viscosity, solid content concentration, and binder physical properties, and outputs the one-day viscosity change rate. The one-day viscosity change rate is classified into four classes: "A: 0-120%", "B: 121-150%, "C: 151-300%, "D: 301% or more".

The binder properties input to the one-day viscosity change rate model 300 are the intrinsic viscosity, elastic modulus, polymerization method, and the presence or absence of functional group modification among the binder property items included in the learning data.

The viscosity change rate model after one day 300 includes a classification model 301 and a regression model 302. The classification model 301 inputs the material composition ratio, binder mixing ratio, initial viscosity, solid content concentration, and binder physical properties, and classifies the viscosity change rate after one day into "300% or less" or "301% or more." The structure of the classification model 301 can be, for example, a random forest.

Formulations classified as having a viscosity change rate after one day of "300% or less" in classification model 301 are input to regression model 302. On the other hand, formulations classified as having a viscosity change rate after one day of "301% or more" in classification model 301 are classified into the "D: 301% or more" class.

The absolute value of the viscosity change rate after one day tends to be small because the time that has elapsed since the electrode mixture was prepared is short. Therefore, there is little training data for which the viscosity change rate after one day is 300% or more, and there is a risk of reduced prediction accuracy. The viscosity change rate after one day model 300 is configured to verify whether the formulation has a viscosity change rate after one day that is within a realistic range before making a prediction using the regression model 302.

The regression model 302 takes the material composition ratio, binder mixing ratio, initial viscosity, solids concentration, and binder physical properties as input, and outputs the viscosity change rate after one day. The regression model 302 classifies the viscosity change rate after one day into "A: 0-120%", "B: 121-150%", and "C: 151-300%". The structure of the regression model 302 can be, for example, Gaussian process regression.

Returning to FIG. 3, in step S3, the control unit of the proposed device 100 learns a 7-day viscosity change rate model based on the learning data stored in the memory unit. The 7-day viscosity change rate model is a machine learning model that predicts the 7-day viscosity change rate of the electrode mixture. The control unit of the proposed device 100 stores the learned 7-day viscosity change rate model in the memory unit.

Figure 5 is a diagram showing an example of a viscosity change rate model after 7 days. As shown in Figure 5, the viscosity change rate model after 7 days 310 inputs the material composition ratio, binder mixing ratio, initial viscosity, solid content concentration, and binder physical properties, and outputs the viscosity change rate after 7 days. The viscosity change rate after 7 days is classified into four classes: "A: 0-120%", "B: 121-150%, "C: 151-300%, "D: 301% or more". The structure of the viscosity change rate model after 7 days 310 can be, for example, Gaussian process regression.

The binder properties input to the 7-day viscosity change rate model 310 are the intrinsic viscosity, elastic modulus, polymerization method, and the presence or absence of functional group modification, among the binder property items included in the learning data, similar to the 1-day viscosity change rate model 300.

The absolute value of the viscosity change rate after 7 days tends to be large because a long time has passed since the electrode mixture was prepared. Therefore, unlike the viscosity change rate after 1 day model 300, the viscosity change rate after 7 days model 310 can be configured to directly predict the viscosity change rate after 7 days for all formulations.

Returning to FIG. 3, in step S4, the control unit of the proposed device 100 learns a coating resistance model using the learning data stored in the memory unit. The coating resistance model is a machine learning model that predicts the coating resistance of an electrode produced using the electrode mixture. The control unit of the proposed device 100 stores the learned coating resistance model in the memory unit.

Figure 6 is a diagram showing an example of a coating resistance model. As shown in Figure 6, the coating resistance model 320 inputs the material composition ratio, binder mixing ratio, binder properties, and electrode information, and outputs the coating resistance. The structure of the coating resistance model 320 can be, for example, Gaussian process regression.

The binder properties input to the coating resistance model 320 are the intrinsic viscosity, elastic modulus, polymerization method, and the presence or absence of functional group modification, among the various binder property items included in the learning data. The electrode information input to the coating resistance model 320 is the density before the pressing process, among the various electrode information items included in the learning data.

Returning to FIG. 3, in step S5, the control unit of the proposed device 100 uses the learning data stored in the memory unit to learn a three-point bending strength model. The three-point bending strength model is a machine learning model that predicts the three-point bending strength of an electrode produced using the electrode mixture. The control unit of the proposed device 100 stores the learned three-point bending strength model in the memory unit.

Figure 7 is a diagram showing an example of a three-point bending strength model. As shown in Figure 7, the three-point bending strength model 330 inputs the material composition ratio, binder mixing ratio, binder properties, and electrode information, and outputs the three-point bending strength. The structure of the three-point bending strength model 330 can be, for example, Gaussian process regression.

The binder properties input to the three-point bending strength model 330 are the intrinsic viscosity, modulus of elasticity, Mw/Mn, solution viscosity, amount of modification, degree of branching, degree of abnormal bonding, and random rate, which are all items of the binder properties included in the training data. The electrode information input to the three-point bending strength model 330 is the amount of slurry applied, the density before the pressing process, and the density after the pressing process, which are all items of the electrode information included in the training data.

Returning to FIG. 3, in step S6, the control unit of the proposed device 100 learns a peel strength model using the learning data stored in the memory unit. The peel strength model is a machine learning model that predicts the peel strength of an electrode produced using the electrode mixture. The control unit of the proposed device 100 stores the learned peel strength model in the memory unit.

Figure 8 is a diagram showing an example of a peel strength model. As shown in Figure 8, the peel strength model 340 inputs the material composition ratio, binder mixing ratio, binder properties, and electrode information, and outputs the peel strength. The structure of the peel strength model 340 can be, for example, Gaussian process regression.

The binder properties input to the peel strength model 340 are the intrinsic viscosity, modulus of elasticity, Mw/Mn, solution viscosity, amount of modification, degree of branching, degree of abnormal bonding, and random rate, which are all items of the binder properties included in the learning data. The electrode information input to the peel strength model 340 is the amount of slurry applied, the density before the pressing process, and the density after the pressing process, which are all items of the electrode information included in the learning data.

<Proposal method flow>
7 is a flowchart showing an example of the flow of a proposal method executed by the proposal device 100 in this embodiment. The proposal method is a method of proposing a formulation of a mixture for an electrode by using a trained prediction model.

In step S11, the control unit of the proposal device 100 accepts input of condition information. The condition information is information indicating the prerequisites when proposing the formulation of the electrode mixture. The condition information can also be said to be information indicating the required specifications for the electrode mixture or the electrode. The condition information may be input, for example, in CSV (Comma-Separated Values) format or the data format of a desired spreadsheet software.

Figure 10 is a diagram showing an example of condition information. As shown in Figure 10, the condition information includes blending conditions, fixed conditions, and target values.

The blending conditions include active material ratio, conductive additive ratio, binder ratio, fluororesin binder ratio, and solids concentration. Each blending condition includes a minimum value, a maximum value, and an interval. The minimum value, maximum value, and interval are used to generate candidate blends for the electrode mixture.

The fixed conditions include the amount of slurry applied, density after the pressing process, density before the pressing process, and initial viscosity. The fixed conditions are used to specify the values to be set for the explanatory variables of each prediction model.

Target values include peel strength, three-point bending strength, viscosity change rate after 1 day, viscosity change rate after 7 days, and coating resistance. The target values are used to specify the values that the physical properties predicted by each prediction model should satisfy.

Returning to FIG. 9, in step S12, the control unit of the proposed device 100 generates candidate formulations for the electrode mixture according to the condition information received in step S11. Specifically, the control unit of the proposed device 100 first generates options for the formulation ratio of each material according to the formulation conditions included in the condition information. For example, if the formulation conditions for the active material ratio are set to a minimum of 97, a maximum of 98, and an interval of 1, the options for the active material are 97 wt% or 98 wt%. Also, for example, if the formulation conditions for the conductive assistant ratio are set to a minimum of 1, a maximum of 3, and an interval of 1, the options for the conductive assistant are 1 wt%, 2 wt%, and 3 wt%.

Next, the control unit of the proposed device 100 extracts, from among all combinations of the compounding ratios of each material, combinations in which the sum of the compounding ratios of each material is 100 wt%. For example, a combination in which the compounding ratio of active material is 97 wt%, the compounding ratio of conductive additive is 1 wt%, and the compounding ratio of binder is 2 wt% is extracted because the sum of the compounding ratios is 100 wt%. Also, for example, a combination in which the compounding ratio of active material is 98 wt%, the compounding ratio of conductive additive is 2 wt%, and the compounding ratio of binder is 2 wt% is not extracted because the sum of the compounding ratios is 102 wt%.

Next, the control unit of the proposal device 100 generates all combinations of candidate binder options, fluororesin binder ratio options, and solids concentration options for each extracted formulation. This generates candidate formulations for the electrode mixture.

In step S13, the control unit of the proposed device 100 calculates the binder properties to be set as explanatory variables of each prediction model for each candidate blend generated in step S12. In this embodiment, since two types of binders are mixed, the binder properties are calculated by taking a weighted average of the basic properties of each binder according to the binder blend ratio. Note that the basic physical properties are the physical properties of the binder alone. The basic physical properties can be measured in advance and stored in the memory unit.

In a formulation in which only one type of binder is used without mixing, the basic physical properties of the binder can be used as is. In a formulation in which three or more types of binders are mixed, the weighted average can be calculated according to the binder mixing ratio, just as in the case of mixing two types of binders.

In step S14, the control unit of the proposal device 100 reads out the trained one-day viscosity change rate model 300 from the memory unit. Next, the control unit of the proposal device 100 predicts the one-day viscosity change rate for each candidate formulation based on the trained one-day viscosity change rate model 300.

Specifically, the control unit of the proposal device 100 inputs the material composition ratio, binder mixing ratio, initial viscosity, solids concentration, and binder properties (intrinsic viscosity, elastic modulus, polymerization method, and presence or absence of functional group modification) for each candidate formulation into the classification model 301 of the one-day viscosity change rate model 300. The classification model 301 classifies each candidate formulation into a class in which the one-day viscosity change rate is "300% or less" or "301% or more."

The control unit of the proposed device 100 outputs the class "D: 301% or more" as the predicted result of the viscosity change rate after one day for the candidate formulation classified by the classification model 301 as a class of "301% or more" in terms of the viscosity change rate after one day.

The control unit of the proposal device 100 inputs the material composition ratio, binder mixing ratio, initial viscosity, solids concentration, and binder properties (intrinsic viscosity, elastic modulus, polymerization method, and presence or absence of functional group modification) for candidate formulations classified by the classification model 301 as a class in which the viscosity change rate after one day is "300% or less" into the regression model 302 of the viscosity change rate after one day model 300. The regression model 302 outputs a predicted value of the viscosity change rate after one day for each candidate formulation. The control unit of the proposal device 100 then classifies the predicted value of the viscosity change rate after one day output from the regression model 302 into one of the classes "A: 0-120%, " "B: 121-150%, " "C: 151-300%, " and outputs it as the predicted result of the viscosity change rate after one day.

In step S15, the control unit of the proposed device 100 reads out the learned 7-day viscosity change rate model 310 from the memory unit. Next, the control unit of the proposed device 100 predicts the 7-day viscosity change rate for each candidate formulation based on the learned 7-day viscosity change rate model 310.

Specifically, the control unit of the proposed device 100 inputs the material composition ratio, binder mixing ratio, initial viscosity, solids concentration, and binder properties (intrinsic viscosity, elastic modulus, polymerization method, and presence or absence of functional group modification) for each candidate formulation into the 7-day viscosity change rate model 310. The 7-day viscosity change rate model 310 outputs a predicted value of the viscosity change rate after 7 days for each candidate formulation. The control unit of the proposed device 100 then classifies the predicted value of the viscosity change rate after 7 days output from the 7-day viscosity change rate model 310 into one of "A: 0-120%", "B: 121-150%, "C: 151-300%, "D: 301% or more", and outputs it as the predicted result of the viscosity change rate after 7 days.

In step S16, the control unit of the proposed device 100 reads the trained coating resistance model 320 from the memory unit. Next, the control unit of the proposed device 100 predicts the coating resistance for each candidate formulation based on the trained coating resistance model 320.

Specifically, the control unit of the proposed device 100 inputs the material composition ratio, binder mixing ratio, binder physical properties (intrinsic viscosity, elastic modulus, polymerization method, and presence or absence of functional group modification) and electrode information (density before the pressing process) for each candidate formulation into the coating resistance model 320. The coating resistance model 320 outputs a predicted value of the coating resistance for each candidate formulation. The control unit of the proposed device 100 then outputs the predicted value of the coating resistance output from the coating resistance model 320 as the predicted result of the coating resistance.

In step S17, the control unit of the proposed device 100 reads the trained three-point bending strength model 330 from the memory unit. Next, the control unit of the proposed device 100 predicts the three-point bending strength for each candidate composition based on the trained three-point bending strength model 330.

Specifically, the control unit of the proposed device 100 inputs the material composition ratio, binder mixing ratio, binder physical properties (intrinsic viscosity, elastic modulus, Mw/Mn, solution viscosity, modification amount, branching degree, abnormal bonding degree, and random rate) and electrode information (slurry application amount, density before the pressing process, and density after the pressing process) for each candidate formulation into the three-point bending strength model 330. The three-point bending strength model 330 outputs a predicted value of the three-point bending strength for each candidate formulation. Then, the control unit of the proposed device 100 outputs the predicted value of the three-point bending strength output from the three-point bending strength model 330 as the predicted result of the three-point bending strength.

In step S18, the control unit of the proposed device 100 reads the trained peel strength model 340 from the memory unit. Next, the control unit of the proposed device 100 predicts the peel strength for each candidate formulation based on the trained peel strength model 340.

Specifically, the control unit of the proposed device 100 inputs the material composition ratio, binder mixing ratio, binder physical properties (intrinsic viscosity, elastic modulus, Mw/Mn, solution viscosity, modification amount, branching degree, abnormal bonding degree, and random rate) and electrode information (slurry application amount, density before the pressing process, and density after the pressing process) for each candidate formulation into the peel strength model 340. The peel strength model 340 outputs a predicted value of the peel strength for each candidate formulation. The control unit of the proposed device 100 then outputs the predicted value of the peel strength output from the peel strength model 340 as the predicted result of the peel strength.

In step S19, the control unit of the proposal device 100 extracts, from among the candidate formulations, a candidate formulation in which the physical property values related to the electrode mixture predicted in steps S14 to S18 satisfy the target values specified in the condition information. Then, the control unit of the proposal device 100 outputs proposal information including the extracted candidate formulations. The proposal information may be output, for example, in CSV format or a data format of a desired spreadsheet software.

FIG. 11 is a diagram showing an example of proposed information. As shown in FIG. 11, the proposed information includes a material composition ratio, a binder mixing ratio, a viscosity change rate after 1 day, a viscosity change rate after 7 days, a coating resistance, a three-point bending strength, and a peel strength. The material composition ratio includes an active material ratio, a conductive additive ratio, and a binder ratio. The binder mixing ratio includes a binder type and ratio, and a fluororesin binder type and ratio. The viscosity change rate after 1 day includes a predicted value, a standard deviation, and a label indicating a class. The viscosity change rate after 7 days includes a predicted value, a standard deviation, and a label indicating a class. The coating resistance, the three-point bending strength, and the peel strength each include a predicted value and a standard deviation.

The user of the proposed device 100 can manufacture an electrode mixture that meets the required specifications based on the proposed information output by the proposed device 100. Specifically, the user of the proposed device 100 can manufacture an electrode mixture by mixing an active material, a conductive assistant, and a binder according to a desired formulation of the electrode mixture contained in the proposed information. The electrode mixture manufactured in this way is predicted to have physical property values that meet the target values, so there is a high possibility that it will be an electrode mixture that meets the required specifications.

<Modification>
In this embodiment, a configuration has been described in which the viscosity change rate after 1 day, the viscosity change rate after 7 days, the coating resistance, the three-point bending strength, and the peel strength are predicted using a trained model, but these physical property values are merely examples, and the physical property values predicted by the proposed device 100 are not limited to these. This embodiment may be configured to predict any physical property value as long as it is a physical property value of an electrode mixture or an electrode using an electrode mixture. In addition, the present embodiment is not limited to a configuration in which multiple physical property values are predicted, and may be configured to predict one physical property value.

<Summary>
As described above, according to one embodiment of the present disclosure, it is possible to propose a formulation of an electrode mixture whose physical property values satisfy the target values. For example, the proposal device 100 predicts physical property values related to the electrode mixture based on the candidate formulations of the electrode mixture, and outputs the candidate formulations whose physical property values satisfy the target values. Therefore, according to one embodiment of the present disclosure, it is possible to propose a formulation of an electrode mixture whose physical property values satisfy the target values.

The electrode mixture is a mixture containing an active material, a conductive member, and a binder, and the binder may contain a binder containing a polymer such as a fluororesin. Therefore, according to one embodiment of the present disclosure, it is possible to propose a formulation of an electrode mixture containing a binder containing a fluororesin, an active material, and a conductive member.

The proposal device 100 may predict two or more physical property values of the electrode mixture and output a candidate formulation in which all of the two or more physical property values meet the target values. Therefore, according to one embodiment of the present disclosure, it is possible to propose a formulation of an electrode mixture in which multiple physical property values simultaneously meet the target values.

A user of the proposed device 100 may manufacture an electrode mixture by mixing an active material, a conductive material, and a binder according to a candidate formulation included in the proposed information output from the proposed device 100. Since the candidate formulation included in the proposed information is predicted to have physical property values that satisfy target values, an electrode mixture manufactured according to the candidate formulation is considered to satisfy the required specifications. Therefore, according to one embodiment of the present disclosure, an electrode mixture, an electrode, or a secondary battery that satisfies the required specifications can be manufactured with a short lead time.

[supplement]
Each function of the above-described embodiments can be realized by one or more processing circuits. Here, the term "processing circuit" in this specification includes a processor programmed to execute each function by software, such as a CPU (Central Processing Unit) or GPU (Graphics Processing Unit) implemented by an electronic circuit, and an ASIC (Application Specific Integrated Circuit), DSP (Digital Signal Processor), FPGA (Field Programmable Gate Array), conventional circuit module, and other devices designed to execute each function described above.

Although the embodiments have been described above, it will be understood that various changes in form and details are possible without departing from the spirit and scope of the claims.

100 Proposed device

Claims (8)

A proposal device having a control unit that proposes a blend of an electrode mixture,
The control unit is
generating candidate formulations for the electrode mixture;
Predicting physical properties of an electrode manufactured using the electrode mixture based on the trained model and the candidate formulation;
Outputting the candidate formulation in which the physical property value satisfies the target value;
The electrode mixture is a mixture containing an active material, a conductive material, and a plurality of binders,
The trained model includes, as explanatory variables, physical property values obtained by weighting the physical property values of the plurality of binders according to the blending ratio of the plurality of binders included in the candidate blend.
Proposed device. The trained model includes the candidate formulation and the manufacturing conditions of the electrode as explanatory variables.
The proposal device according to claim 1 . The binder comprises a polymer.
The proposal device according to claim 1 or 2 . The polymer includes a fluororesin.
The proposal device according to claim 3 . The control unit is
Predicting two or more of the physical properties;
outputting the candidate formulation in which all of the two or more physical property values satisfy the target values;
The proposal device according to claim 1 or 2. A control unit of a proposal device for proposing a blend of an electrode mixture,
generating candidate formulations for the electrode mixture;
Predicting physical properties of an electrode manufactured using the electrode mixture based on the trained model and the candidate formulation;
outputting the candidate formulation in which the physical property value satisfies the target value;
Execute the process,
The electrode mixture is a mixture containing an active material, a conductive material, and a plurality of binders,
The trained model includes, as explanatory variables, physical property values obtained by weighting the physical property values of the plurality of binders according to the blending ratio of the plurality of binders included in the candidate blend.
Proposal method. A method for producing an electrode mixture, comprising the step of mixing an active material, a conductive material, and a plurality of binders according to a candidate mixture output by the proposal method according to claim 6 . A control unit of a proposal device for proposing a blend of an electrode mixture,
generating candidate formulations for the electrode mixture;
Predicting physical properties of an electrode manufactured using the electrode mixture based on the trained model and the candidate formulation;
outputting the candidate formulation in which the physical property value satisfies the target value;
Execute the process ,
The electrode mixture is a mixture containing an active material, a conductive material, and a plurality of binders,
The trained model includes, as explanatory variables, physical property values obtained by weighting the physical property values of the plurality of binders according to the blending ratio of the plurality of binders included in the candidate blend.
program.

Images