JP7583635B2 - Conveying device, processing device, conveying method, and processing method - Google Patents

Description

The present invention relates to a technology for transporting a long strip-shaped substrate in the longitudinal direction.

Conventionally, there is known an inkjet printing device that prints an image on a long strip-shaped substrate by ejecting ink from multiple heads while transporting the substrate in the longitudinal direction. Inkjet printing devices eject ink of different colors from multiple heads. Then, a multi-color image is printed on the surface of the substrate by superimposing single-color images formed by each color of ink. A conventional printing device is described, for example, in Patent Document 1.

JP 2019-116341 A

In this type of printing device, the substrate is transported while controlling the tension applied to the substrate so that it remains approximately constant. For example, the tension applied to the substrate is measured and the transport mechanism is feedback-controlled based on the measured value. However, feedback control is a reactive control in which the measured tension value fluctuates and then the control value is corrected based on the amount of fluctuation. For this reason, feedback control is vulnerable to external disturbances.

In order to suppress the effects of disturbances, it is possible to combine the above feedback control with feedforward control. In feedforward control, the characteristics of the transport mechanism are formulated and the effects of disturbances are predicted based on this formulation. However, when the transport mechanism is large and complex, it becomes difficult to formulate the characteristics of the transport mechanism. Even if it is possible to formulate it, there is a problem that it can only handle specific disturbances that are assumed in advance. Furthermore, if the installation environment of the device changes or if the device deteriorates over time, the accuracy of the prediction results based on the formula decreases.

The present invention has been developed in consideration of these circumstances, and aims to provide a technology that can accurately suppress tension fluctuations in a substrate in a device that transports a long strip of substrate.

In order to solve the above problem, the first invention of the present application is a conveying device that includes a conveying mechanism that conveys a long strip-shaped substrate in the longitudinal direction along a predetermined conveying path, a plurality of sensors provided on the conveying path, and a tension prediction unit that inputs the measurement values of the plurality of sensors into a prediction model obtained in advance by performing machine learning using the measurement values of the plurality of sensors as input variables and the tension applied to the substrate at a time a predetermined time after the measurement time of the measurement values as an objective variable, to obtain a predicted value of tension output from the prediction model, and a control unit that controls the conveying mechanism based on the predicted value obtained by the tension prediction unit.

The second invention of the present application is the conveying device of the first invention, further comprising a tension measuring unit that measures the tension applied to the substrate, and a learning unit that generates the predictive model by performing machine learning using the measurement values of the multiple sensors as input variables and the actual tension measured by the tension measuring unit at a predetermined time after the measurement time of the measurement value as an objective variable.

The third invention of the present application is a conveying device according to the first or second invention, in which the conveying mechanism has a motor and a motor control unit that inputs a command value to the motor, and the input variable includes the command value output from the motor control unit.

The fourth invention of the present application is a conveying device according to any one of the first to third inventions, in which the control unit controls the conveying mechanism by PID control based on the deviation between the predicted value and the target tension value, the derivative of the deviation, and the integral of the deviation.

The fifth invention of the present application is a processing device that includes a conveying device according to any one of the first to fourth inventions, and a processing section that performs a predetermined process on the surface of a substrate conveyed by the conveying mechanism.

The sixth invention of the present application is the processing device of the fifth invention, in which the processing section is a printing section having multiple heads that eject ink onto the surface of the substrate.

The seventh invention of the present application is a method for transporting a long strip-shaped substrate in the longitudinal direction along a predetermined transport path by a transport mechanism, comprising the steps of: a) acquiring measurement values from a plurality of sensors provided on the transport path; b) inputting the measurement values acquired in step a) into a prediction model previously obtained by performing machine learning, with the measurement values of the plurality of sensors as input variables and the tension applied to the substrate at a time a predetermined time after the measurement time of the measurement values as an objective variable, to obtain a predicted value of tension output from the prediction model; and c) controlling the transport mechanism based on the predicted value acquired in step b).

The eighth invention of the present application is the conveying method of the seventh invention, further comprising, prior to step a), a step of performing machine learning to generate the prediction model using the measurement values of the multiple sensors as input variables and the actual tension value measured by the tension measuring unit at a predetermined time after the measurement time of the measurement value as an objective variable.

The ninth aspect of the present invention is the conveying method of the seventh or eighth aspect, in which the conveying mechanism has a motor and a motor control unit that inputs a command value to the motor, and the input variable includes the command value of the motor control unit.

The tenth invention of the present application is a conveying method according to any one of the seventh to ninth inventions, in which in step c), the conveying mechanism is controlled by PID control based on the deviation between the predicted value and the target tension value, the derivative of the deviation, and the integral of the deviation.

The eleventh invention of the present application is a processing method, in which a predetermined processing is performed on the surface of the substrate while the substrate is being transported by the transport method of any one of the seventh to tenth inventions.

The twelfth aspect of the present invention is the processing method of the eleventh aspect, in which the predetermined processing is a processing of ejecting ink from multiple heads onto the surface of the substrate.

According to the first to twelfth aspects of the present application, a prediction model obtained by machine learning is used to obtain a predicted value of the tension applied to the substrate. Then, the conveying mechanism is controlled based on the predicted value. This makes it possible to suppress the tension fluctuation of the substrate with high accuracy.

FIG. 1 illustrates a configuration of a printing device. FIG. 2 is a partial top view of the printing device in the vicinity of the printing unit. 2 is a block diagram showing connections between a control unit and each unit of the printing device. FIG. FIG. 2 is a block diagram conceptually showing the functions of a control unit. 13 is a flowchart showing the flow of a learning process. 13 is a flowchart showing the flow of a printing process.

The following describes an embodiment of the present invention with reference to the drawings.

1. Configuration of the Printing Device
1 is a diagram showing the configuration of a printing device 1 including a transport device according to one embodiment of the present invention. This printing device 1 is an apparatus that prints an image on the surface of a long strip-shaped substrate 9 by ejecting ink droplets from a plurality of heads 21 to 24 toward the substrate 9 while transporting the substrate 9. The substrate 9 may be printing paper or a resin film. The substrate 9 may also be a substrate made of cardboard, metal foil, or glass.

As shown in FIG. 1, the printing device 1 includes a transport mechanism 10, a printing unit 20, a number of sensors 30, a tension measuring unit 40, a camera 50, and a computer 60. The parts of the printing device 1 other than the printing unit 20 are an example of a transport device according to the present invention.

The transport mechanism 10 transports the substrate 9 in a transport direction along its longitudinal direction. The transport mechanism 10 of this embodiment has an unwinding section 11, multiple transport rollers 12, and a winding section 13. The substrate 9 is unwound from the unwinding section 11 and transported along a transport path formed by multiple transport rollers 12. Each transport roller 12 rotates around an axis extending perpendicular to the transport direction, thereby guiding the substrate 9 downstream of the transport path. After transport, the substrate 9 is collected in the winding section 13. In addition, tension is applied to the substrate 9 in the transport direction. This suppresses sagging and wrinkling of the substrate 9 during transport.

The transport mechanism 10 has a motor 14 that rotates some of the rollers (hereinafter referred to as "drive rollers"), and a motor control unit 15 that controls the motor 14. The transport mechanism 10 may have multiple drive rollers. The motor control unit 15 is electrically connected to a computer 60, which will be described later. When the motor 14 is driven, the motor control unit 15 calculates a command value for driving the motor 14 according to a control signal from the computer 60. Then, a drive signal corresponding to the calculated command value is input to the motor 14. Then, the motor 14 drives according to the command value, and the drive roller rotates. As a result, the substrate 9 is transported from the unwinding unit 11 to the winding unit 13.

The printing unit 20 is a processing unit that ejects ink droplets (hereinafter referred to as "ink droplets") onto the substrate 9 transported by the transport mechanism 10. The printing unit 20 of this embodiment has a first head 21, a second head 22, a third head 23, and a fourth head 24. The first head 21, the second head 22, the third head 23, and the fourth head 24 are arranged at intervals along the transport direction of the substrate 9. The substrate 9 is transported below the four heads 21 to 24 with the printing surface facing upward.

Figure 2 is a partial top view of the printing device 1 near the printing unit 20. As indicated by the dashed lines in Figure 2, the underside of each head 21-24 is provided with a plurality of nozzles 201 arranged parallel to the width direction of the substrate 9. Each head 21-24 ejects ink droplets of each color, K (black), C (cyan), M (magenta), and Y (yellow), which are the color components of a multicolor image, from the plurality of nozzles 201 toward the upper surface of the substrate 9.

That is, the first head 21 ejects K ink droplets onto the top surface of the substrate 9 at the first printing position P1 on the transport path. The second head 22 ejects C ink droplets onto the top surface of the substrate 9 at the second printing position P2 downstream of the first printing position P1. The third head 23 ejects M ink droplets onto the top surface of the substrate 9 at the third printing position P3 downstream of the second printing position P2. The fourth head 24 ejects Y ink droplets onto the top surface of the substrate 9 at the fourth printing position P4 downstream of the third printing position P3.

A fixing section for fixing the ink to the printing surface of the substrate 9 may be further provided downstream of the heads 21 to 24 in the transport direction. The fixing section, for example, blows heated gas toward the substrate 9 to dry the ink adhering to the substrate 9. However, the fixing section may also cure the ink by irradiating UV-curable ink with ultraviolet light.

The multiple sensors 30 are measuring instruments that measure the transport state of the substrate 9. The multiple sensors 30 are provided at multiple measurement points on the transport path of the substrate 9. The sensors 30 obtain measurement values at each measurement point. Measurement items of the sensors 30 can include, for example, the rotation speed of the motor 14, the torque of the motor 14, the rotation speed of some of the transport rollers 12, the vertical displacement of the substrate 9 (the amount of displacement in a direction perpendicular to the substrate 9), the width direction position of the edge of the substrate 9, and the like. Note that sensors 30 that measure the same item may be disposed at multiple positions on the transport path. While the substrate 9 is being transported by the transport mechanism 10, the multiple sensors 30 constantly measure the state of each measurement point. The multiple sensors 30 then output signals indicating the obtained measurement values to the computer 60.

The tension measuring unit 40 is a measuring device that measures the tension applied to the substrate 9 in the transport direction. The tension measuring unit 40 has a load cell connected to the rotation shaft of some of the transport rollers 12. The load cell measures the load applied to the rotation shaft of the transport roller 12. The tension measuring unit 40 calculates the actual measured value of the tension applied to the substrate 9 based on the load measured by the load cell. While the substrate 9 is being transported by the transport mechanism 10, the tension measuring unit 40 constantly measures the current tension of the substrate 9. The tension measuring unit 40 then outputs a signal indicating the obtained actual measured value of tension to the computer 60.

The camera 50 is an imaging device that captures an image of the top surface of the substrate 9 that has passed through the printing unit 20. The camera 50 is positioned facing the printed surface of the substrate 9 at an imaging position P5 downstream of the four heads 21-24 on the transport path. The camera 50 is, for example, a line sensor in which multiple imaging elements such as CCDs or CMOSs are arranged in the width direction. While the substrate 9 is being transported by the transport mechanism 10, the camera 50 constantly captures an image of the printed surface of the substrate 9, thereby acquiring image data of the printed substrate 9. The camera 50 then transmits the acquired image data to the computer 60.

The computer 60 is an information processing device for controlling the printing device 1. FIG. 3 is a block diagram showing the connection between the computer 60 and each part of the printing device 1. As conceptually shown in FIG. 3, the computer 60 has a processor 601 such as a CPU, a memory 602 such as a RAM, and a storage unit 603 such as a hard disk drive. The storage unit 603 stores a computer program 80 for executing the learning process and printing process described below.

As shown in FIG. 3, the computer 60 is communicatively connected to the motor control unit 15, the four heads 21-24, the multiple sensors 30, the tension measuring unit 40, and the camera 50. The computer 60 controls the operation of each of these units according to a computer program 80 and various data. This allows the transport of the substrate 9 and the printing process to proceed.

<2. Computer Functions>
The computer 60 of this printing device 1 has a function of predicting future tension applied to the substrate 9 and controlling the transport of the substrate 9 based on the predicted value. Fig. 4 is a block diagram conceptually showing the function of the computer 60. As shown in Fig. 4, the computer 60 has a data accumulation unit 61, a learning unit 62, a tension prediction unit 63, and a control unit 64. The functions of the data accumulation unit 61, the learning unit 62, the tension prediction unit 63, and the control unit 64 are realized by the computer 60 operating in accordance with a computer program 80.

The data accumulation unit 61 is a processing unit for accumulating data used in the learning process described below. Measurement values D1 input from the multiple sensors 30 to the computer 60 are stored in the data accumulation unit 61. In addition, actual tension values D2 input from the tension measurement unit 40 to the computer 60 are stored in the data accumulation unit 61. The data accumulation unit 61 stores the measurement values D1 of these sensors 30 and the actual tension values D2 together with the measurement times.

The learning unit 62 is a processing unit for learning the relationship between the measurement values D1 of the multiple sensors 30 and the tension applied to the substrate 9. The learning unit 62 reads out learning data D3 from the data accumulation unit 61. The learning data D3 includes the measurement values D1 of the multiple sensors 30 at a certain time t1 and the actual measured value D2 of the tension of the substrate 9 at a time t2 a predetermined time (e.g., several milliseconds) after the time t1. The learning unit 62 reads out a large number of such learning data D3 from the data accumulation unit 61 at different times.

The learning unit 62 performs learning processing using a supervised machine learning algorithm with the measurement values D1 of the multiple sensors 30 as input variables and the actual measured value D2 of the tension of the substrate 9 at a predetermined time after the measurement time as the objective variable. The learning unit 62 repeats this learning processing until a predetermined termination condition is satisfied. As a result, the learning unit 62 generates a prediction model M that can output a predicted value D4 indicating the tension of the substrate 9 after a predetermined time based on the measurement values D1 of the multiple sensors 30. Details of the learning processing will be described later.

The machine learning algorithm used in the learning unit 62 may be a simple regression algorithm or a tree model such as random forest or gradient boosting. The machine learning algorithm used in the learning unit 62 may also be a deep learning algorithm such as a convolutional neural network.

The tension prediction unit 63 is a processing unit for predicting the tension of the substrate 9 after a predetermined time based on the measurement values D1 of the multiple sensors 30. The tension prediction unit 63 predicts the tension of the substrate 9 using the prediction model M generated by the learning unit 62. The tension prediction unit 63 inputs the current measurement values D1 acquired from the multiple sensors 30 to the prediction model M. The prediction model M then outputs a predicted value D4 of the tension of the substrate 9 corresponding to the measurement value D1. The tension prediction unit 63 inputs the predicted value D4 output from the prediction model M to the control unit 64.

The control unit 64 is a processing unit for controlling the operation of the transport mechanism 10 and the four heads 21 to 24. The control unit 64 calculates a control value by performing feedback control based on the predicted value D4 obtained from the tension prediction unit 63 and a preset target value D5. The control unit 64 then outputs a control signal D6 including the calculated control value to the motor control unit 15 of the transport mechanism 10. This controls the operation of the motor 14 of the transport mechanism 10. The control unit 64 also controls the operation of the four heads 21 to 24 based on the image data to be printed. This causes ink to be ejected from the four heads 21 to 24.

3. Learning process
Next, a description will be given of the learning process executed in the above-mentioned printing device 1. Fig. 5 is a flow chart showing the flow of the learning process.

As shown in FIG. 5, when performing the learning process, first, the transport mechanism 10 is operated to transport the substrate 9 while accumulating data necessary for learning (step S1). Here, the data accumulation unit 61 of the computer 60 stores the measurement values D1 sent from the multiple sensors 30 together with the measurement times. The data accumulation unit 61 also stores the actual tension values D2 sent from the tension measurement unit 40 together with the measurement times.

When a sufficient amount of data required for learning has been accumulated, the learning unit 62 then acquires learning data D3 from the data accumulation unit 61 (step S2). The learning data D3 includes the measurement values D1 of the multiple sensors 30 at a certain time t1, and the actual measurement value D2 of the tension of the substrate 9 at a time t2 that is a predetermined time (e.g., several milliseconds) after the time t1.

Then, the learning unit 62 performs a learning process using a supervised machine learning algorithm with the measurement value D1 at a certain measurement time included in the learning data D3 as an input variable and the actual measured value D2 of the tension of the substrate 9 a predetermined time after the measurement time included in the learning data D3 as an objective variable. At this time, the learning unit 62 prepares a prediction model M for predicting the tension of the substrate 9 after the predetermined time based on the measurement value D1 of the sensor 30. The prediction model M outputs a prediction result of the tension applied to the substrate 9 after the predetermined time based on the input measurement value D1. The learning unit 62 adjusts the parameters of the prediction model M so that the prediction result output from the prediction model M approaches the actual measured value D2, which is the objective variable (step S3).

Then, the learning unit 62 determines whether a predetermined termination condition is satisfied (step S4). The termination condition may be, for example, that the difference between the prediction result and the actual measurement value D2 is smaller than a preset threshold value. The termination condition may also be that the number of repetitions of steps S2 to S3 reaches a preset threshold value. If the termination condition is not satisfied (step S4: No), the learning unit 62 repeats the above-mentioned processing of steps S2 to S3. At that time, the learning data D3 used may be from a time different from the previous time.

By repeating the learning process of steps S2 and S3, the prediction accuracy of the prediction model M improves. When the termination condition is eventually satisfied (step S4: Yes), the learning unit 62 ends the learning process. This generates a trained prediction model M that can accurately output a predicted value D4 of the tension of the substrate 9 after a predetermined time based on the measurement values D1 of the multiple sensors 30. The learning unit 62 provides the generated prediction model M to the tension prediction unit 63.

Note that a predetermined number of pieces of training data D3 may be treated as one training dataset. Then, the training process may be performed for multiple training datasets. In this case, while the processes of steps S2 to S3 are being performed for a predetermined number of pieces of training data D3 included in one training dataset, the processes of steps S2 to S3 may be repeated without determining whether the termination condition is met in step S4. Then, when the processes of steps S2 to S3 are completed for all training data D3 included in one training dataset, the termination condition in step S4 may be determined. If the termination condition is not met in step S4, the training process of steps S2 to S3 may be performed for another training dataset.

<4. Printing process>
Next, a description will be given of the printing process executed in the printing device 1 after the above-mentioned learning process. Fig. 6 is a flow chart showing the flow of the printing process.

As shown in FIG. 6, the printing device 1 first starts the operation of the transport mechanism 10 (step S5). This starts the transport of the substrate 9. When the transport of the substrate 9 starts, the multiple sensors 30 measure the state of each measurement point. Then, the multiple sensors 30 transmit the current measurement value D1 to the computer 60 (step S6).

The tension prediction unit 63 of the computer 60 inputs the current measurement values D1 transmitted from the multiple sensors 30 to the prediction model M. The prediction model M then outputs a predicted value D4 of the tension applied to the substrate 9 after a predetermined time (step S7). The tension prediction unit 63 inputs the obtained predicted value D4 to the control unit 64.

The control unit 64 controls the operation of the transport mechanism 10 based on the predicted tension value D4 input from the tension prediction unit 63 (step S8). A target value D5 of the tension of the substrate 9 is preset in the control unit 64. The control unit 64 feedback controls the transport mechanism 10 so that the predicted tension value D4 input from the tension prediction unit 63 approaches the target value D5.

For example, PID control is used for the feedback control. That is, the control unit 64 calculates a control value based on the deviation between the predicted value D4 and the target value D5, the derivative of the deviation, and the integral of the deviation. Then, a control signal D6 including the control value is supplied to the motor control unit 15. The motor control unit 15 drives the motor 14 in accordance with the control signal D6. This controls the conveying operation of the substrate 9.

The control unit 64 also controls the operation of the four heads 21 to 24 while transporting the substrate 9 according to the above-mentioned control. Each of the four heads 21 to 24 ejects ink droplets toward the upper surface of the substrate 9. This causes an image to be printed on the upper surface of the substrate 9 (step S9).

The control unit 64 then determines whether or not to end the printing process (step S10). If image data to be printed remains, the control unit 64 continues the printing process (step S10: No). In this case, the computer 60 executes the processes of steps S6 to S9 again. In this way, the computer 60 repeats the processes of steps S6 to S9 to progress the printing process while transporting the substrate 9 with high precision.

When there is no more image data to be printed, the control unit 64 ends the printing process (step S10: Yes). In this case, the printing device 1 stops the operation of the transport mechanism 10 and ends the transport of the substrate 9 (step S11).

As described above, in this printing device 1, the tension applied to the substrate 9 after a predetermined time is predicted based on the current measurement values D1 of the multiple sensors 30. Then, the operation of the transport mechanism 10 is controlled based on the predicted value D4. This makes it possible to suppress the tension fluctuation of the substrate 9 with high accuracy. As a result, a high-quality printout can be obtained with little misalignment of the printed image on the top surface of the substrate 9.

When using feedforward control as in the past, the characteristics of the transport mechanism 10 must be re-formulated every time the state of the printing device 1 or the type of substrate 9 changes. This formulation task requires consideration from various perspectives, which places a heavy burden on the user. In contrast, the printing device 1 of this embodiment can re-create the prediction model M by performing a certain learning process, even if the state of the printing device 1 or the type of substrate 9 changes. This makes it easy to respond to changes in conditions.

In addition, in the printing device 1 of this embodiment, it is also possible to additionally execute the learning process of FIG. 5 while executing the printing process of FIG. 6. In this way, the prediction model M can be updated at any time while executing the printing process. Therefore, even if the installation environment of the printing device 1 changes or the state of the printing device 1 changes, the prediction accuracy of the prediction model M can always be maintained at a high level.

5. Modifications
Although one embodiment of the present invention has been described above, the present invention is not limited to the above embodiment.

In the above embodiment, only the measurement values D1 of the multiple sensors 30 were used as input variables for the prediction model M. However, the input variables for the prediction model M may include variables other than the measurement values D1 of the multiple sensors 30. For example, the command value output from the motor control unit 15 may be included as an input variable. The command value is a variable that more directly affects the conveying speed of the substrate 9. Therefore, by including the command value as an input variable, a more accurate prediction model M can be generated.

The printing device 1 may further include a sensor that measures environmental conditions such as temperature and humidity. Information on the environmental conditions obtained by the sensor may then be included in the input variables of the prediction model M.

In addition, the amount of misalignment of the printed image on the upper surface of the substrate 9 may be detected based on the captured image obtained by the camera 50, and the amount of misalignment may be included in the input variables. The amount of misalignment of the printed image is a variable that directly affects the quality of the printed result. Therefore, by including the amount of misalignment in the input variables, a prediction model M that is more useful for improving the quality of the printed result can be generated.

In addition, in the above embodiment, the measurement values D1 of the multiple sensors 30 are used as input variables as they are. However, processing such as filtering may be performed on these measurement values D1, and the measurement values D1 after processing may be used as input variables. Also, from the multiple measurement values D1, the measurement values D1 that have a large influence on tension fluctuations may be selected in advance, and only the selected measurement values D1 may be used as input variables. This allows a more accurate prediction model M to be generated.

In the above embodiment, PID control is given as an example of feedback control by the control unit 64. However, the feedback control performed by the control unit 64 may be another method such as PI control.

In the above embodiment, as shown in FIG. 2, the nozzles 201 are arranged in a row in the width direction in each of the heads 21 to 24. However, the nozzles 201 may be arranged in two or more rows in each of the heads 21 to 24.

The printing device 1 in the above embodiment was equipped with four heads 21 to 24. However, the number of heads equipped in the printing device 1 may be one to three, or five or more. For example, the printing device 1 may be equipped with a head that ejects ink of a special color in addition to the colors K, C, M, and Y.

In the above embodiment, a printing device 1 that ejects ink onto the surface of a substrate 9 has been described. That is, in the above embodiment, the printing unit 20, which is an example of a processing unit, ejects ink, which is an example of processing. However, the processing device of the present invention may perform processing other than printing on the surface of a substrate transported by a transport mechanism. For example, the processing device may perform other processing, such as applying a resist liquid, exposing a pattern, or drawing with a laser, on the surface of a substrate transported by a transport mechanism.

Furthermore, the elements appearing in the above embodiments and variations may be combined as appropriate to the extent that no contradictions arise.

REFERENCE SIGNS LIST 1 Printing device 9 Substrate 10 Transport mechanism 14 Motor 15 Motor control unit 20 Printing unit 21 First head 22 Second head 23 Third head 24 Fourth head 30 Sensor 40 Tension measuring unit 50 Camera 60 Computer 61 Data storage unit 62 Learning unit 63 Tension prediction unit 64 Control unit 80 Computer program D1 Measurement value D2 Actual measured tension value D3 Learning data D4 Prediction value D5 Target value D6 Control signal M Prediction model

Claims (12)

A conveying mechanism that conveys a long strip-shaped base material in a longitudinal direction along a predetermined conveying path;
A plurality of sensors provided on the conveying path;
a tension prediction unit that inputs the measurement values of the plurality of sensors into a prediction model previously obtained by performing machine learning, the measurement values of the plurality of sensors being used as input variables and the tension applied to the base material at a predetermined time after the measurement time of the measurement value being used as an objective variable, and obtains a predicted value of the tension output from the prediction model;
a control unit that controls the transport mechanism based on the predicted value obtained by the tension prediction unit;
A conveying device comprising: 2. The conveying device according to claim 1,
A tension measuring unit that measures the tension applied to the base material;
A conveying device further comprising a learning unit that generates the predictive model by performing machine learning using measurement values of the plurality of sensors as input variables and an actual measurement value of tension measured by the tension measuring unit at a predetermined time after the measurement time of the measurement value as an objective variable. The conveying device according to claim 1 or 2,
The transport mechanism includes:
A motor;
a motor control unit that inputs a command value to the motor;
having
The input variables include the command values output from the motor control unit. A conveying device according to any one of claims 1 to 3,
The control unit controls the transport mechanism by PID control based on the deviation between the predicted value and the target value of the tension, a derivative of the deviation, and an integral of the deviation. A conveying device according to any one of claims 1 to 4,
a processing section for performing a predetermined process on a surface of the base material transported by the transport mechanism;
A processing device comprising: 6. The processing device according to claim 5,
The processing unit is a printing unit having a plurality of heads that eject ink onto a surface of the substrate. A conveying method for conveying a long strip-shaped substrate in a longitudinal direction along a predetermined conveying path by a conveying mechanism, comprising:
a) acquiring measurement values from a plurality of sensors provided on the conveying path;
b) inputting the measurement values obtained in step a) into a prediction model previously obtained by machine learning using the measurement values of the plurality of sensors as input variables and the tension applied to the substrate at a predetermined time after the measurement of the measurement values as an objective variable, and obtaining a predicted value of the tension output from the prediction model;
c) controlling the transport mechanism based on the predicted value obtained by the step b);
The transport method comprises: The conveying method according to claim 7,
Prior to step a),
The conveying method further includes a step of generating the prediction model by performing machine learning using the measurement values of the plurality of sensors as an input variable and an actual measurement value of tension measured by a tension measuring unit at a predetermined time after the measurement time of the measurement value as an objective variable. The conveying method according to claim 7 or 8, further comprising the steps of:
The transport mechanism includes:
A motor;
a motor control unit that inputs a command value to the motor;
having
The conveying method, wherein the input variables include the command values of the motor control unit. The conveying method according to any one of claims 7 to 9,
In the step c), the conveying mechanism is controlled by PID control based on a deviation between the predicted value and a target value of the tension, a derivative of the deviation, and an integral of the deviation. A processing method in which a predetermined processing is performed on the surface of the substrate while the substrate is being transported by the transport method according to any one of claims 7 to 10. 12. The method of claim 11, further comprising the steps of:
The predetermined process is a process of ejecting ink from a plurality of heads onto a surface of a substrate.

Images