WO2019171498A1

WO2019171498A1 - Coating control system, coating control method, and coating control program

Info

Publication number: WO2019171498A1
Application number: PCT/JP2018/008786
Authority: WO
Inventors: 真典岡本; 林　新太郎; 英和塙; 浩樹渡邉; 憲一藤垣
Original assignee: 日立化成株式会社
Priority date: 2018-03-07
Filing date: 2018-03-07
Publication date: 2019-09-12

Abstract

A coating control system according to one embodiment of the present invention implements machine learning by acquiring from a database a dataset including a manufacturing parameter at at least two time points during a coating step for one lot, in which coating step a coating layer is formed by means of a coating material on a substrate flowing along a manufacturing line of a coating device, and using the dataset as an input, to generate a machine learning model for estimating a control value of the coating device, which is used to coat the coating material onto the substrate.

Description

Application control system, application control method, and application control program

One aspect of the present invention relates to a coating control system, a coating control method, and a coating control program.

Conventionally, a method for controlling sheet manufacturing is known. For example, Patent Document 1 describes a method for predicting and applying coating parameters. The method predetermines at least one paint parameter and inputs it to a paint parameter-response model parameter-response algorithm, and determines at least one target paint response based on the parameter-response algorithm and parameter. Process.

Special table 2007-508937

There are various factors that can affect the coating of the base material, and among these factors there are factors that change with changes in the environment and the passage of time in the coating process. For this reason, adjustment of the coating process largely depends on the operator's sense, skill, experience, or intuition. Therefore, it is desired to reduce the dependence on personal decisions in sheet manufacturing.

An application control system according to an aspect of the present invention includes at least one processor, and at least one processor forms an application layer for one lot in which an application material forms a coating layer on a substrate flowing through a production line of an application apparatus. The control value of the coating device for coating the coating material on the substrate is obtained by acquiring a data set including manufacturing parameters at at least two points from the database and executing machine learning using the data set as input. A machine learning model is generated for estimation.

A coating control method according to one aspect of the present invention is a coating control method executed by a coating control system including at least one processor, and a coating layer is formed from a coating material on a substrate flowing through a manufacturing line of a coating apparatus. In the application process for one lot, an acquisition step for acquiring a data set including manufacturing parameters at at least two points in time from the database, and machine learning using the data set as input, a coating material is applied on the substrate. A learning step of generating a machine learning model for estimating a control value of the coating apparatus for coating.

A coating control program according to an aspect of the present invention is a data set including manufacturing parameters at at least two points in a coating process for one lot in which a coating layer is formed from a coating material on a substrate flowing through a manufacturing line of a coating apparatus. A machine learning model for estimating a control value of a coating apparatus for applying a coating material on a base material by executing an acquisition step of acquiring data from a database and machine learning using a data set as an input. A learning step to be generated is executed by a computer.

In such an aspect, by executing machine learning using manufacturing parameters at a plurality of time points in one lot as input, considering changes in the application process of various elements that may affect the coating of the substrate. A machine learning model is generated. Therefore, by using this model, it is possible to reduce the dependence on personal decisions in sheet manufacturing.

According to one aspect of the present invention, it is possible to reduce the degree of dependence on personal decisions in sheet manufacturing.

It is a figure which shows typically an example of the whole structure of the coating system containing the coating control system which concerns on embodiment. It is a figure which shows an example of the hardware constitutions of the computer used with the application | coating control system which concerns on embodiment. It is a figure which shows an example of a function structure of the application | coating control system which concerns on embodiment. It is a figure which shows an example of a training data set. It is a flowchart which shows an example of operation | movement of the application | coating control system which concerns on embodiment. It is a figure which shows an example of a general neural network.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements are denoted by the same reference numerals, and redundant description is omitted.

[Configuration of coating system]
The application control system 10 is a computer system that controls the application apparatus 1 that manufactures a sheet material by forming a coating layer on a substrate. A sheet material is a member that spreads thinly, and includes both films and sheets. The kind of sheet material is not limited at all. For example, the sheet material may be a photosensitive film. The application control system 10 controls the application apparatus 1 to form an appropriate coating layer on the substrate. The appropriate coating layer refers to, for example, a coating layer whose thickness after drying is adjusted to an appropriate range.

FIG. 1 is a diagram schematically showing an example of the configuration of a coating system including a coating control system 10 and a coating apparatus 1. The coating system further includes a database 20 that stores data necessary for controlling the coating apparatus 1. As shown in FIG. 1, the application control system 10, the application apparatus 1, and the database 20 are connected by a wired or wireless communication line 30 and can transmit / receive data to / from each other.

[Configuration of coating device]
The configuration of the coating apparatus 1 will be described with reference to FIG. The coating apparatus 1 includes a production line 1a through which a substrate 90 to be processed flows. In addition, the kind of base material 90 is not limited at all, For example, a web base material may be sufficient. The production line 1a includes a supply roll 2 around which a base material 90 to be processed is wound, and a winding roll 3 that winds up the processed base material 90 (that is, a sheet material). The base material 90 is pulled out from the supply roll 2, and a sheet material is produced by forming a coating layer on the base material 90 while flowing on the production line 1 a, and the sheet material is taken up by the take-up roll 3. It is done. Below, the other component of the coating device 1 is demonstrated as it goes to the downstream from the production line 1a.

The base material 90 drawn from the supply roll 2 first reaches the coating unit 4. The coating unit 4 is connected by a supply path 6 and a tank 5 that stores a coating material composed of a solid component (nonvolatile component) and a solvent (volatile component), and the coating material supplied from the tank 5 is placed on the substrate 90. Apply to. “Application” refers to a process for forming a coating layer on a substrate. The application method is not limited at all, and may be spray or the like. The raw material of the coating material is not limited. For example, an organic solvent such as acetone, methyl ethyl ketone, toluene, or ethyl lactate may be used as a solvent, and a polymer resin may be used as a solid component.

In this embodiment, the coating unit 4 includes a lip coater 4a and a back roll 4b that face each other. The back roll 4b can adjust the gap with the lip coater 4a, and this adjustment is performed automatically by the actuator or manually. When the base material 90 passes through the gap between the opening of the lip coater 4a and the back roll 4b, the coating material is applied onto one main surface of the base material 90. In addition, the structure of the application part 4 is not limited to the lip coater 4a, For example, other structures, such as a die coater and a comma coater, may be employ | adopted.

In the middle of the supply path 6, a supply device 6a for controlling the supply amount of the coating material to the coating unit 4 is provided. The supply device 6a is constituted by a gear pump, for example.

The base material 90 that has passed through the coating unit 4 then enters the thickness measuring unit 8. The thickness measuring unit 8 is a device that measures the thickness before drying of the coating layer formed of the coating material on the base material 90 in real time as the wet thickness. Although the structure of the thickness measurement part 8 is not limited, For example, a spectroscopic thickness meter may be used. The spectroscopic thickness meter includes a spectroscopic unit and a sensor head. The spectroscopic thickness meter irradiates the coated surface of the substrate 90 with infrared light in a wide wavelength band from the spectral unit, and obtains the interference light intensity of reflected light on the coated surface and the reference surface inside the sensor head. Then, the wet thickness is measured. By providing the thickness measuring unit 8 in the vicinity of the coating unit 4, the wet thickness can be measured at an early stage.

The base material 90 that has passed through the thickness measuring unit 8 then enters the drying furnace 9. The drying furnace 9 is an apparatus that dries the coating material on the substrate 90. The coating material is lined on the substrate 90 by this drying. The sheet material is completed by drying the coating layer, and this sheet material is wound around the take-up roll 3.

[Configuration of application control system]
The application control system 10 is composed of one or more computers. In the case of using a plurality of computers, these computers are connected via a communication network such as the Internet or an intranet, so that one application control system 10 is logically constructed.

FIG. 2 is a diagram illustrating an example of a general hardware configuration of the computer 100 configuring the coating control system 10. The computer 100 includes a processor (e.g., CPU) 101 that executes an operating system, application programs, and the like, a main storage unit 102 that includes a ROM and a RAM, an auxiliary storage unit 103 that includes a hard disk, a flash memory, and the like. A communication control unit 104 configured by a network card or a wireless communication module, an input device 105 such as a keyboard and a mouse, and an output device 106 such as a monitor are provided.

Each functional element of the application control system 10 is realized by reading a predetermined program on the processor 101 or the main storage unit 102 and executing the program. The processor 101 operates the communication control unit 104, the input device 105, or the output device 106 in accordance with the program, and reads and writes data in the main storage unit 102 or the auxiliary storage unit 103. Data or a database necessary for processing is stored in the main storage unit 102 or the auxiliary storage unit 103.

FIG. 3 is a diagram illustrating an example of a functional configuration of the application control system 10. The application control system 10 controls the application apparatus 1 to form an appropriate coating layer on the substrate 90, and includes a learning unit 11, an acquisition unit 12, an estimation unit 13, and an instruction unit 14 for this control. .

The learning unit 11 is a functional element that generates a neural network used for estimating the control value of the coating apparatus 1 as a machine learning model (hereinafter also simply referred to as “model”). Machine learning is a technique for autonomously finding a rule or rule by repeatedly learning based on given information. A neural network is an information processing model that mimics the mechanism of the human cranial nervous system.

The acquisition unit 12 is a functional element that acquires data to be input to the generated model.

The estimation unit 13 is a functional element that estimates the control value of the coating apparatus 1 by inputting the data to the generated model and executing machine learning. The estimated control value is not limited in any way, but in the present embodiment, the estimation unit 13 estimates three types of control values: gap, bend, and coating amount. The gap is an interval between the lip coater 4a and the back roll 4b, and is used, for example, to control an actuator that adjusts the interval. The bend is a value indicating the deflection of the lip coater 4a in the width direction of the base material 90 (the direction perpendicular to the traveling direction of the base material 90 along the production line 1a), and prevents or reduces variation in the coating amount in the width direction. Used for. The coating amount is the amount of coating material supplied toward the substrate 90, and is used to control the supply device 6a (for example, to change the rotation speed of the gear pump). The gap, bend, and coating amount are all values related to the thickness of the coating layer after drying.

The instruction unit 14 is a functional element that transmits an instruction signal based on the estimated control value toward the coating apparatus 1. The instruction signal is a data signal for controlling application to the substrate 90 in order to form an appropriate coating layer on the substrate 90. In the present embodiment, the instruction unit 14 transmits an instruction signal based on the estimated gap and bend to the application unit 4, and transmits an instruction signal based on the estimated application amount to the supply device 6a.

When the application control system 10 is configured by a plurality of computers, it may be arbitrarily determined which processor executes which functional element. In any case, the application control system 10 including at least one processor functions as the learning unit 11, the acquisition unit 12, the estimation unit 13, and the instruction unit 14.

[Database]
The database 20 is a device that stores a training data set for generating a machine learning model. The training data set is a collection of one or more training samples (training data). Each training sample includes actual values of manufacturing parameters at at least two points in the coating process of one lot of sheet material. A “lot” is a unit that indicates a collection of articles that are produced or appear to have been produced under equal conditions.

The manufacturing parameter is an arbitrary variable that can change the control value of the coating apparatus 1. Actual values of a plurality of manufacturing parameters are collected at each time point. Each manufacturing parameter may be collected from an arbitrary device (for example, an arbitrary sensor) constituting the coating apparatus 1, or may be collected from an apparatus other than the coating apparatus 1 as a value related to an external factor. Examples of production parameters include the density, viscosity, and liquid temperature of the coating material before drying, the pressure of the decompression device, and the tension on the production line 1a (supply roll 2, inlet and outlet of the coating unit 4, the drying furnace 9 Tension at the outlet, the take-up roll 3 etc.), the speed of the production line 1a, the actual application amount and application width, the pump rotation speed of the supply device 6a, the internal pressure, the primary filtration pressure, and the secondary filtration pressure, The temperature in the drying furnace 9, the wind speed, the internal pressure, and the gas concentration, the temperature and humidity of the environment, the room pressure, the charge amount, the particles, and the state of the base material 90 are listed. Of course, the types of manufacturing parameters are not limited to these examples, and various factors may be set as manufacturing parameters. Furthermore, each training sample also includes actual values of gap, bend, and application amount estimated during operation.

FIG. 4 is a diagram schematically illustrating an example of a training data set stored in the database 20. The training data set includes training samples prepared for each of one or more types of application materials used in the application apparatus 1. FIG. 4 explicitly shows a set of training samples for each of the coating materials A and B. Multiple training samples are prepared for one application material, and FIG. 4 explicitly shows three training samples for each of application materials A and B. Specifically, for the coating material A, the training sample corresponding to the lot A1 whose manufacturing date is D _a1 , the training sample corresponding to the lot A2 whose manufacturing date is D _a2 , and the manufacturing date is D _a3 . There is a training sample corresponding to lot A3. The coating material B, a training sample manufacturing date corresponding to the lot B1 is D _b1, a training sample manufacturing date corresponding to the lot B2 is D _b2, manufacturing date corresponding to the lot B3 is D _b3 There is a training sample.

The training sample of lot A1 will be taken as a representative and its configuration will be described in detail. This training sample includes a set of manufacturing parameters at each of the three time points t ₁₁ , t ₁₂ , t ₁₃ . In this example, time t ₁₁ is the time around the beginning of the production lots A1, in other words, is a time corresponding to the initial value of the production parameters. Time t ₁₂ is the time in the vicinity of the middle of the production lots A1, in other words, is a time corresponding to an intermediate value of the production parameters. Time t ₁₃ is the time near the end of the production lots A1, in other words, is a time corresponding to the final value of the production parameters. The “time near the start of lot production” may be the start time of lot production, or any time in a time zone within the threshold Ta from the start time. The “time near the middle of manufacturing the lot” may be an intermediate time of manufacturing the lot (when half of the time required to manufacture the entire lot has elapsed), or in a time zone within the threshold Tb before and after the intermediate time. It may be at any time. The “time near the end of lot production” may be the end time of lot production, or may be any time in a time zone within the threshold Tc from the end time. The thresholds Ta, Tb, and Tc may be set arbitrarily.

Note that the timing at which manufacturing parameters are collected is not limited. For example, individual training samples may include manufacturing parameters at other times. The time point near the start of lot production, the time point near the middle of lot production, and the time point near the end of lot production are not essential, and at least one of these three time points may be omitted. . Thus, the training sample may include a set of production parameters at two time points, a time point near the start of lot production and a time point near the middle of lot production. Alternatively, the training sample may include a set of a plurality of production parameters at two time points, a time point near the start of lot production and a time point near the end of lot production. Alternatively, the training sample may include a set of a plurality of manufacturing parameters at two time points, a time point near the middle of manufacturing the lot and a time point near the end of manufacturing the lot.

FIG. 4 shows a plurality of manufacturing parameters collected at each time point as Param_a, Param_b, and the like. Each training sample may also include a label indicating the time point or a label indicating the type of time point as manufacturing parameters. Each training sample also includes gaps, bends, and application amounts used as correct answers in supervised learning.

Training samples in the series A1, the production parameters Param_a is a _{a 11} At time _{t 11,} a transition to _{a 12} At time _{t 12,} indicating that the transition to _{a 13} at time point _{t 13.} Param_b, for also other manufacturing parameters such as Param_c, time _{t 12} from the time point _{t 11,} the time _{t 13} value as time passes is to remain. Some of these manufacturing parameters may change in value over time and others may not change. Due to changes in at least some of the manufacturing parameters over time within a lot, at least one of the gap, bend, and application amount can also vary over time. Training samples in the series A1, the gap is a _{G 11} At time _{t 11,} a transition to _{G 12} at time _{t 12,} indicating that the transition to _{G 13} at time _{t 13.} Similarly, time _{t 12} from the time point _{t 11,} the bend as time passes and the time _{t 13} remained from _{B 11} and _B _{12, B 13,} the coating amount is to remain as _C _{12, C 13} from the _{C 11.}

The number of types of manufacturing parameters can be so large that it is very difficult for a person to formulate the relationship between the set of manufacturing parameters and control values (eg, gap, bend, and coating amount) or Impossible. Therefore, conventionally, the operator adjusts the control value based on his sense, skill, experience, or intuition. In the present embodiment, the coating control system 10 autonomously finds a model indicating the relationship between manufacturing parameters and control values in various situations by machine learning, and uses the model according to the environment surrounding the coating apparatus 1. Control values (eg, gap, bend, and coating amount) are estimated.

[Application control system operation]
The operation of the application control system 10 will be described with reference to FIGS. 5 and 6 and the application control method according to the present embodiment will be described. FIG. 5 is a flowchart showing an example of the operation of the application control system 10. FIG. 6 is a diagram illustrating an example of a general neural network.

In step S11, the learning unit 11 generates a machine learning model. This step is a learning phase in which the best neural network estimated to have the highest prediction accuracy (hereinafter simply referred to as “best neural network”) is generated as a model. It should be noted that the “best neural network” obtained by learning the neural network is not always “best in reality”. The learning unit 11 reads a training data set (one or more training samples) from the database 20 (acquisition step), and executes machine learning while sequentially inputting individual training samples to a neural network for learning (learning step). For example, the learning unit 11 may generate a model by executing deep learning using a multilayer neural network. The type of machine learning is not limited to deep learning, and the learning unit 11 may generate a model using another method.

The neural network illustrated in FIG. 6 includes a first layer that is an input layer, second and third layers that are intermediate layers, and a fourth layer that is an output layer. The first layer outputs an input vector x = (x ₀ , x ₁ , x ₂ ,... X _m ) having _m parameters as components to the second layer as it is. The second layer and the third layer convert the total input into an output by an activation function and pass the output to the next layer. The fourth layer also converts the total input into an output by an activation function, and this output is an output vector y = (y ₀ , y ₁ ,..., Y _n ) of a neural network having n parameters as components. This output vector y indicates an estimated value. If the correct answer of the output vector for one input vector x is d = (d ₀ , d ₁ ,..., D _n ), the weight w of the activation function is updated so that the output vector y is close to the correct d. . This is information processing called learning.

The learning unit 11 gives the manufacturing parameters of the training sample to the neural network as an input vector, and obtains an output vector indicating the gap, the bend, and the coating amount. Then, the learning unit 11 compares the output vector with the correct answer (gap, bend, and application amount indicated by the training sample), and adjusts the weight of the activation function. The learning unit 11 generates a model to be used in the operation phase by repeatedly updating the weights while giving many training samples to the neural network.

An example of processing by the learning unit 11 will be described on the assumption that the database 20 stores the training data set shown in FIG.

The learning unit 11 may perform learning using the training sample for the coating material A as follows. The learning unit 11 gives the entire training sample of the lot A1 as an input vector to the neural network (that is, by inputting the training sample of the lot A1 in a lump), and the time points t ₁₁ , t ₁₂ , and t ₁₃ To obtain an output vector indicating the gap, bend, and coating amount in each. Then, the learning unit 11 adjusts the weight of the activation function by comparing the output vector with the correct answer (gap, bend, and application amount at the time points t ₁₁ , t ₁₂ , and t ₁₃ of the training sample of the lot A1). To do.

Further, the learning unit 11 processes the training sample of lot A2 in the same manner as the training sample of lot A1. That is, the learning unit 11 obtains an output vector indicating the gap, the bend, and the application amount at each time point by giving the entire training sample of the lot A2 as an input vector to the neural network. Then, the learning unit 11 compares the output vector with the correct answer (gap, bend, and application amount at each time point of the training sample of the lot A2), and adjusts the weight of the activation function. Further, the learning unit 11 processes the training sample of the lot A3 in the same manner as the training samples of the lots A1 and A2. Further, the learning unit 11 processes the other training samples related to the coating material A in the same manner.

The learning unit 11 may perform further learning using the training sample for the coating material B as follows. The learning unit 11 gives the entire training sample of the lot B1 to the neural network as an input vector, thereby obtaining an output vector indicating the gap, bend, and application amount at each of the time points t ₂₁ , t ₂₂ , and t ₂₃ . Then, the learning unit 11 adjusts the weight of the activation function by comparing the output vector with a correct answer (gap, bend, and application amount at each of the time points t ₂₁ , t ₂₂ , and t ₂₃ of the training sample of the lot B1). To do.

Further, the learning unit 11 processes the training samples of lots B2 and B3 in the same manner as the training sample of lot B1. Furthermore, the learning unit 11 processes the other training samples related to the coating material B in the same manner.

Further, the learning unit 11 may process training samples for other coating materials in the same manner as the coating materials A and B.

The processing after step S12 is an operation phase in which control values (gap, bend, and coating amount) for applying the coating material to the base material 90 are estimated from the generated model.

In step S12, the acquisition unit 12 acquires manufacturing parameters. The acquisition unit 12 may collect manufacturing parameters from an arbitrary device (for example, an arbitrary sensor) constituting the coating apparatus 1, or collect manufacturing parameters from apparatuses other than the coating apparatus 1 as values relating to external factors. May be.

In step S13, the estimation unit 13 inputs manufacturing parameters to the generated model, and uses an output vector obtained from the model as an estimation result of control values (gap, bend, and coating amount).

In step S14, the instruction unit 14 transmits an instruction signal based on the estimated control value toward the coating apparatus 1 (instruction step). “Sending an instruction signal toward the coating apparatus” means transmitting an instruction signal to the coating apparatus 1 or to another apparatus corresponding to the coating apparatus 1. Transmission is an example of output. In the present embodiment, the instruction unit 14 transmits an instruction signal based on the estimated gap and bend to the application unit 4, and transmits an instruction signal based on the estimated application amount to the supply device 6a. As a result, the application to the base material 90 is adjusted so that the thickness of the coating layer becomes a desired value. Therefore, these instruction signals are signals for forming an appropriate coating layer on the substrate 90.

The processing in steps S12 to S14 (that is, the operation phase) can be executed again before starting to generate another lot of sheet material. Step S11 (that is, the learning phase) may be executed again at an arbitrary timing. For example, the application control system 10 generates a new model by executing machine learning using a new training data set in which the results accumulated in the operation phase are captured, and the new operation data is generated in a later operation phase. The control value may be estimated using a model.

The timing at which the operation phase is executed is not limited. For example, the application control system 10 may estimate an initial value in an application process for a new lot to be manufactured. In this case, before the coating process is started, the acquisition unit 12 acquires manufacturing parameters, and the estimation unit 13 inputs the manufacturing parameters into the model to control values (gap, bend, and coating amount). Is estimated. This control value is an initial value in the coating process for the new lot. Then, the instruction unit 14 transmits an instruction signal based on the estimated control value (initial value) toward the coating apparatus 1. After this series of processing, the coating apparatus 1 starts a coating process for a new lot according to the instruction signal.

The application control system 10 may estimate a control value in the application process for one lot that has already been started. In this case, in the course of the coating process, the acquisition unit 12 acquires manufacturing parameters, and the estimation unit 13 inputs the manufacturing parameters into the model to estimate control values (gap, bend, and coating amount). . This control value can be said to be a value for adjusting the operation of the coating apparatus 1 on the way. Then, the instruction unit 14 transmits an instruction signal based on the estimated control value toward the coating apparatus 1. In the coating device 1, the coating unit 4 or the supply device 6a is adjusted according to the instruction signal.

[program]
The application control program for causing the computer system to function as the application control system 10 includes program code for causing the computer system to function as the learning unit 11, the acquisition unit 12, the estimation unit 13, and the instruction unit 14. The application control program may be provided after being fixedly recorded on a tangible recording medium such as a CD-ROM, DVD-ROM, or semiconductor memory. Alternatively, the application control program may be provided via a communication network as a data signal superimposed on a carrier wave. The provided application control program is stored in the auxiliary storage unit 103, for example. The processor 101 reads out the application control program from the auxiliary storage unit 103 and executes the application control program, thereby realizing each functional element described above.

[effect]
As described above, the coating control system according to one aspect of the present invention includes at least one processor, and at least one processor forms a coating layer with a coating material on a base material flowing through a production line of the coating apparatus. In order to apply a coating material on a substrate, a data set including manufacturing parameters at at least two time points in a coating process for one lot is acquired from a database and machine learning is performed using the data set as an input. A machine learning model for estimating a control value of the coating apparatus is generated.

In such an aspect, by executing machine learning using manufacturing parameters at a plurality of time points in one lot as input, considering changes in the application process of various elements that may affect the coating of the substrate. A machine learning model is generated. Therefore, by using this model, it is possible to reduce the dependence on personal decisions in sheet manufacturing. Therefore, a sheet can be manufactured regardless of variations in individual workers' skills or skills, and the quality of the sheet can be stabilized. In addition, this machine learning can shorten the time required for adjusting the coating apparatus, reduce the loss of the coating material during the adjustment work, and improve the yield.

In the coating control system according to another aspect, at least two time points are a time point near the start of manufacturing of one lot, a time point near the middle of manufacturing of one lot, and a time point near the end of manufacturing of one lot. May be included. These three time points can be said to be representative timings of time passage in the coating process. Therefore, improvement in the accuracy of the generated model can be expected by considering the manufacturing parameters at at least one of these three time points.

In the coating control system according to another aspect, the control value may be an initial value in a coating process for a new lot. By estimating the control value at the start of manufacturing for one lot, the quality of the sheet material can be stabilized from the start of manufacturing.

In the application control system according to another aspect, the control value may include a value for controlling the supply device that supplies the application material toward the substrate. In this case, the supply of the coating material can be controlled without depending on the personal decision.

In the application control system according to another aspect, the value for controlling the supply device may include the application amount. In this case, it is possible to control the coating amount that directly affects the formation of the coating layer without depending on the determination of the individual.

In the coating control system according to another aspect, the control value may include a value for controlling a coating apparatus that coats the coating material on the substrate. In this case, the application to the substrate can be controlled without depending on the personal decision.

In the coating control system according to another aspect, the value for controlling the coating device may include at least one of a gap and a bend. In this case, gaps or bends that directly affect the formation of the coating layer can be controlled without relying on personal decisions.

In the coating control system according to another aspect, at least one processor further acquires a new manufacturing parameter related to the coating process for a new lot, and inputs a new manufacturing parameter to the generated model. A control value in the coating process for one lot may be estimated, and an instruction signal based on the estimated control value may be output. By estimating the control value of the coating process using a model obtained by machine learning in consideration of changes in various elements in the coating process, a sheet can be manufactured without depending on a personal decision.

[Modification]
The present invention has been described in detail based on the embodiments. However, the present invention is not limited to the above embodiment. The present invention can be variously modified without departing from the gist thereof.

The configuration of the coating apparatus is not limited to the above embodiment, and the coating control system can correspond to various coating apparatuses for manufacturing a sheet material. In this regard, the destination of the instruction signal is not limited to the application unit 4 and the supply device 6a. The indication signal may include any information for communicating the indication to any component that contributes to the formation of the coating layer.

Application control system including the learning unit 11 (that is, application control system that executes the learning phase), and application control system that includes the acquisition unit 12, the estimation unit 13, and the instruction unit 14 (that is, the application control system that executes the operation phase) ) May be separated. Each application control system separated in this way is also an application control system according to the present invention.

In the above embodiment, the database 20 is connected to the application control system 10 via the communication line 30, but the database 20 may be constructed in a computer constituting the application control system 10. In the above-described embodiment, the application control system 10 is connected to the application apparatus 1 via the communication line 30, but the application control system 10 may be constructed in a device constituting the application apparatus 1.

The processing procedure of the application control method executed by at least one processor is not limited to the example in the above embodiment. For example, some of the steps (processes) described above may be omitted, or the steps may be executed in a different order. Further, any two or more of the steps described above may be combined, or a part of the steps may be corrected or deleted. Alternatively, other steps may be executed in addition to the above steps.

When comparing the magnitude relationship between two values in the application control system, you can use either of the two criteria “greater than” or “greater than”, and the two criteria “less than” and “less than”. Either of these may be used. The selection of such a standard does not change the technical significance of the process of comparing the magnitude relationship between two numerical values.

DESCRIPTION OF SYMBOLS 1 ... Coating apparatus, 1a ... Production line, 2 ... Supply roll, 3 ... Winding roll, 4 ... Application | coating part, 4a ... Lip coater, 4b ... Back roll, 5 ... Tank, 6 ... Supply path, 6a ... Supply apparatus, 8 DESCRIPTION OF SYMBOLS ... Thickness measurement part, 9 ... Drying furnace, 10 ... Application | coating control system, 11 ... Learning part, 12 ... Acquisition part, 13 ... Estimation part, 14 ... Instruction part, 20 ... Database, 30 ... Communication line, 90 ... Base material.

Claims

At least one processor,
The at least one processor comprises:
A data set including manufacturing parameters at at least two points in a coating process for one lot in which a coating layer is formed by a coating material on a substrate flowing through a coating line of a coating apparatus is acquired from a database.
Generating a machine learning model for estimating a control value of the coating apparatus for applying the coating material on the substrate by performing machine learning with the data set as an input;
Application control system.
The at least two time points include at least one of a time point near the start of manufacturing of the one lot, a time point near the middle of manufacturing of the one lot, and a time point near the end of manufacturing of the one lot. ,
The coating control system according to claim 1.
The control value is an initial value in the coating process for a new lot.
The coating control system according to claim 1 or 2.
The control value includes a value for controlling a supply device that supplies the coating material toward the substrate.
The coating control system according to any one of claims 1 to 3.
The value for controlling the supply device includes a coating amount,
The coating control system according to claim 4.
The control value includes a value for controlling a coating apparatus that coats the coating material on the substrate.
The coating control system according to any one of claims 1 to 5.
The value for controlling the coating device includes at least one of a gap and a bend.
The coating control system according to claim 6.
The at least one processor further comprises:
Obtain new manufacturing parameters for the coating process for a new lot,
By inputting the new manufacturing parameters to the generated model, the control value in the coating process for the new one lot is estimated,
Outputting an instruction signal based on the estimated control value;
The coating control system according to any one of claims 1 to 7.
An application control method executed by an application control system comprising at least one processor,
An acquisition step of acquiring from a database a data set including manufacturing parameters at at least two points in an application process for one lot in which a coating layer is formed by a coating material on a substrate flowing through a production line of a coating apparatus;
A learning step of generating a machine learning model for estimating a control value of the coating apparatus for coating the coating material on the substrate by performing machine learning using the data set as an input An application control method comprising:
An acquisition step of acquiring from a database a data set including manufacturing parameters at at least two points in an application process for one lot in which a coating layer is formed by a coating material on a substrate flowing through a production line of a coating apparatus;
A learning step of generating a machine learning model for estimating a control value of the coating apparatus for coating the coating material on the substrate by performing machine learning using the data set as an input Application program for causing a computer to execute.