JP7006640B2 - Control devices, control methods, and control programs - Google Patents

Description

The present invention relates to a control device, a control method, and a control program.

A web carrier is used to transport a roll-shaped material (web) such as a film or foil. The web wound up in a roll shape is unwound by a web carrier, subjected to processing such as printing, coating, and packaging, and then wound up in a roll shape again. Through this process, products using the web are produced. In this web carrier, the friction characteristics between the web and the rollers change due to various factors such as changes in production conditions, equipment or environment, and the meandering of the web (change in position in the width direction) occurs. If this occurs, defects will occur in the product using the web, for example, wrinkles will occur on the web, processing accuracy will deteriorate, and so on.

In order to reduce the occurrence of this product defect, Patent Document 1 proposes a method of adjusting the position of the end portion of the web in consideration of the tension acting on the web. Specifically, the web transport device proposed in Patent Document 1 includes an edge sensor, a lateral position correction device, a pair of tension sensors, and a control device. The edge sensor detects the lateral deviation of the web. The horizontal position correction device corrects the horizontal position of the web by the guide roll. Each tension sensor detects the tension acting near each end of the web in the left-right direction. The control device performs feedback control of the lateral position correction device so that the edge of the web is placed at the target position according to the lateral position deviation detected by the edge sensor. During this feedback control, the control device changes the feedback gain according to the difference between the tension near the left end and the tension near the right end of the web detected by each tension sensor. According to this method, it is possible to prevent meandering of the web while suppressing an increase in the difference in tension acting on each end of the web, so that the occurrence of product defects can be reduced.

Japanese Unexamined Patent Publication No. 2009-220948

The present inventors have found that the conventional control method in web transport as described above has the following problems.

That is, there are various factors that cause the meandering of the web. However, in the conventional control method, the countermeasure is limited to a specific factor. For example, in the method of Patent Document 1, it is possible to suppress the occurrence of meandering due to the difference in tension acting on each end of the web, but it is difficult to suppress the occurrence of meandering due to other factors. .. Therefore, with the conventional control method, it is possible to suppress the occurrence of product defects due to a specific factor, but it is difficult to suppress the occurrence of product defects due to other factors. The present inventors have found that the conventional control method has such a problem.

The present invention has been made in view of such circumstances on one aspect, and an object thereof is a technique for improving the performance of web transport and reducing the probability that a product using the web will be defective. Is to provide.

The present invention adopts the following configuration in order to solve the above-mentioned problems.

That is, the control device according to one aspect of the present invention is a web transport device configured to transport the web, a sensor configured to detect the position in the width direction of the transported web, and a sensor. A control device that controls the operation of a web transport device including a correction mechanism configured to correct the position of the web to be transported in the width direction, and acquires a detected value of the position of the web in the width direction from the sensor. And a calculation unit that calculates at least one of the future detection value of the position in the width direction and the future command value for the correction mechanism from the acquired detection value using the prediction model. A determination unit that determines a command value for the correction mechanism according to the calculated predicted value, and an operation of the correction mechanism so as to correct the position of the web in the width direction based on the determined command value. It is provided with an operation control unit for controlling the above.

In the control device according to the configuration, in order to predict and control the operation of the correction mechanism, at least one of the future detection value of the width direction position and the future command value for the correction mechanism from the detection value of the position in the width direction of the web. A prediction model that predicts the predicted value of is used. The prediction model is composed of, for example, an autoregressive model. Such a prediction model is actual data obtained as a result of controlling the correction mechanism, that is, time series data showing the transition of the detected value of the position in the width direction of the web while controlling the operation of the correction mechanism (further). Can be generated from time-series data) showing the transition of the command value for the correction mechanism. At this time, by collecting actual data under various conditions, the meandering result of the web is reflected in the generation of the prediction model based on various factors. In other words, by using the actual data collected under various conditions to generate the predictive model, the generated predictive model is constructed so that the meandering of the web based on various factors can be predicted.

Therefore, according to the configuration, the meandering of the web can be predicted independently of the factors. Therefore, it is possible to improve the performance of web transport and reduce the probability that a product using the web will be defective. Furthermore, predictive control using a predictive model can reduce the frequency of maintenance (parameter adjustment) of the correction mechanism. As a result, it is possible to suppress human costs and reduce the probability of malfunction due to human factors (for example, mistakes in maintenance work).

The prediction model calculates (predicts) the predicted value of the widthwise position of the web at the second time in the future from the first time from the value (detected value or predicted value) of the widthwise position of the web at the first time. It is configured to do. That is, the prediction model is configured to output the predicted value of the position in the width direction of the web at the second time in response to the input of the value of the position of the position in the width direction of the web at the first time. The first time and the second time may be the time at one time point, or may be the time at a plurality of time points within a predetermined time interval, respectively. That is, the input and the output of the prediction model may be one value or a plurality of values, respectively. In addition, other data other than the value of the position in the width direction of the web may be input to the input of the prediction model. Other data may be, for example, a command value for the correction mechanism, individual information on the web, or the like.

The type of such a prediction model is not particularly limited and may be appropriately selected depending on the embodiment. The prediction model may be configured by a regression model such as an autoregressive model, or may be configured by a machine learning model such as a neural network, for example. The method of analyzing the actual data (time series data) in order to generate the prediction model may not be particularly limited and may be appropriately selected depending on the type of the prediction model. When the prediction model is composed of an autoregressive model, a known regression analysis method such as the least squares method may be used as the analysis method. When the prediction model is composed of a machine learning model such as a neural network, a known machine learning method such as an error backpropagation method may be used as the analysis method.

In the control device according to the above aspect, in the prediction model, with respect to the input of the value of the position in the width direction in each of the plurality of control cycles up to now, the detection value of the position in the width direction in the future control cycle and the modification thereof. It may be configured to output at least one predicted value of the command value for the mechanism. The acquisition unit may acquire the detection value in each of a plurality of control cycles. The calculation unit inputs each of the acquired detected values to the prediction model, executes arithmetic processing of the prediction model, and uses the output value output from the prediction model as the prediction value in a future control cycle. You may get it. The determination unit may determine a command value for the correction mechanism according to the acquired predicted value. The motion control unit may control the motion of the correction mechanism based on the determined command value. According to this configuration, it is possible to schedule the operation of the correction mechanism while synchronizing the prediction control by the prediction model with the actual environment in real time. It should be noted that the control by the command value determined according to the predicted value may be executed before the future control cycle in which the predicted value is acquired. This allows you to control the widthwise position of the web before it meanders.

In the control device according to the above aspect, the prediction model is a predicted value of a detected value of a position in the width direction in a future control cycle with respect to an input of a value of a position in the width direction in each of a plurality of control cycles up to now. May be configured to output. The calculation unit inputs each of the acquired detected values to the prediction model, executes arithmetic processing of the prediction model, and outputs a value output from the prediction model in the width direction in a future control cycle. It may be acquired as a predicted value of the detected value of the position. Even if the determination unit corrects the target value in the width direction according to the acquired predicted value and determines the command value for the correction mechanism according to the corrected target value and the detected value. good. The motion control unit may control the motion of the correction mechanism based on the determined command value. According to this configuration, conventional methods such as PID (Proportional-Integral-Differential) control and PI control can be adopted for the processing of the portion for determining the command value. Therefore, the introduction cost of the present invention can be suppressed. The command value is determined according to the target value and the detected value. Therefore, correcting the target value may include directly correcting the target value and indirectly correcting the target value by correcting the detected value. That is, correcting the target value includes correcting the target value relative to the detected value.

In the control device according to the above aspect, the prediction model sets the predicted value of the command value for the correction mechanism in the future control cycle for the input of the position value in the width direction in each of the plurality of control cycles up to now. It may be configured to output. The calculation unit inputs each of the acquired detected values to the prediction model, executes arithmetic processing of the prediction model, and outputs a value output from the prediction model to the correction mechanism in a future control cycle. It may be acquired as a predicted value of a command value. The determination unit calculates a reference value of a command value for the correction mechanism according to a target value at a position in the width direction and the detection value, and determines the calculated reference value according to the acquired predicted value. The command value may be determined by making corrections. The motion control unit may control the motion of the correction mechanism based on the determined command value.

In the control device according to the one aspect, inputting each of the detected values into the prediction model converts each of the acquired detected values into one or a plurality of feature quantities, and the obtained one or a plurality of features. It may be configured by inputting the feature amount of the above into the prediction model. According to this configuration, it is possible to appropriately execute the prediction by the prediction model, thereby improving the performance of web transport and reducing the probability that a product using the web will be defective.

The type of the feature amount may not be particularly limited as long as it can represent some feature of the detected value, and may be appropriately selected according to the embodiment. For example, the feature amount may be obtained by extracting the feature of the target to be used for prediction from the time series data of the detected value. Further, for example, the feature amount may be aggregated information (for example, a time series of average values for each unit interval), a series itself of one section from the time of prediction to the past, and the like. As a specific example, the feature amount may be a minimum value, a maximum value, an average value, a kurtosis, a differential value, or the like.

While controlling the operation of the web transport device, the control device according to the one aspect obtains at least one of the detected value of the position in the width direction acquired from the sensor and the determined command value as actual data. A data collection unit for updating the prediction model based on the collected actual data may be further provided. According to this configuration, based on the collected actual data, the prediction accuracy of the prediction model is improved, thereby improving the performance of web transport and reducing the probability of defects in web-based products. Can be done.

In at least one of the above aspects, a predicted value is calculated in the prediction model, and the target value at the position in the width direction or the command value for the correction mechanism is corrected according to the calculated predicted value. That is, the prediction process and the correction process are executed as separate processes. However, the scope of application of the present invention does not have to be limited to such examples. In each of the above aspects, the prediction process and the correction process may be executed by an integrated model.

For example, the control device according to one aspect of the present invention is a web transport device configured to transport the web, a sensor configured to detect the position in the width direction of the transported web, and a sensor. A control device that controls the operation of a web transport device including a correction mechanism configured to correct the position of the web to be transported in the width direction, and acquires a detected value of the position of the web in the width direction from the sensor. The reference value is calculated by the acquisition unit, the calculation unit that calculates the correction value for the reference value of the command value or the target value at the position in the width direction from the acquired detection value using the prediction model, and the calculated correction value. Based on the determination unit that determines the command value for the correction mechanism and the determined command value according to the target value of the position in the width direction corrected by the correction value or the calculated correction value. Further, the operation control unit for controlling the operation of the correction mechanism so as to correct the position of the web in the width direction is provided.

In the control device according to the above aspect, the prediction model is a reference value or a position in the width direction of a command value in a future control cycle with respect to an input of a position value in the width direction in each of a plurality of control cycles up to now. It may be configured to output a correction value for the target value of. The acquisition unit may acquire the detection value in each of a plurality of control cycles. The calculation unit inputs each of the acquired detected values to the prediction model, executes the arithmetic processing of the prediction model, and uses the output value output from the prediction model as the correction value in the future control cycle. You may get it. The determination unit calculates a reference value of the command value according to the target value and the detection value at a position in the width direction, and corrects the calculated reference value with the correction value, or is acquired. The command value for the correction mechanism may be determined according to the target value and the detection value corrected by the correction value. The motion control unit may control the motion of the correction mechanism based on the determined command value.

As another aspect of the control device according to each of the above embodiments, one aspect of the present invention may be an information processing method or a program that realizes each of the above configurations, or such a program may be used. It may be a stored storage medium that can be read by a computer or the like. Here, the storage medium that can be read by a computer or the like is a medium that stores information such as a program by electrical, magnetic, optical, mechanical, or chemical action.

For example, the control method according to one aspect of the present invention is a web transport device configured to transport the web, a sensor configured to detect the position of the transported web in the width direction, and a sensor. An information processing method for controlling the operation of a web transport device including a correction mechanism configured to correct the position of the web to be transported in the width direction, wherein a computer determines a detection value of the position of the web in the width direction. A step of calculating at least one of a future detection value of the position in the width direction and a future command value for the correction mechanism from the acquired detection value using the step of acquiring from the sensor and the step of using the prediction model. The step of determining the command value for the correction mechanism according to the calculated predicted value, and the correction mechanism so as to correct the position in the width direction of the web based on the determined command value. It is an information processing method that executes steps to control operations.

Further, for example, the control program according to one aspect of the present invention is a web transport device configured to transport the web, and is a sensor configured to detect the position in the width direction of the transported web. , And a program for controlling the operation of a web transport device including a correction mechanism configured to correct the widthwise position of the web to be transported, wherein the computer detects the widthwise position of the web. From the step of acquiring the value from the sensor and the detected value acquired using the prediction model, at least one of the predicted value of the future detection value of the position in the width direction and the future command value for the correction mechanism is calculated. The step to determine the command value for the correction mechanism according to the calculated predicted value, and the modification to correct the position in the width direction of the web based on the determined command value. It is a program for executing the steps for controlling the operation of the mechanism.

Further, for example, the control method according to one aspect of the present invention is a web transport device configured to transport the web, and a sensor configured to detect the position in the width direction of the transported web. An information processing method for controlling the operation of a web transport device including a correction mechanism configured to correct the widthwise position of the web to be transported, wherein a computer detects the widthwise position of the web. A step of acquiring a value from the sensor, a step of calculating a correction value for a reference value of a command value or a target value at a position in the width direction from the acquired detection value using a prediction model, and a calculated correction. A step of determining a command value for the correction mechanism by correcting the reference value with a value or according to a target value of a position in the width direction corrected by the calculated correction value, and the determined command. It is an information processing method that executes a step of controlling the operation of the correction mechanism so as to correct the position of the web in the width direction based on the value. The command value is determined according to the target value and the detected value. Therefore, the correction value for the target value may be applied directly to the target value, or may be indirectly applied to the target value by being applied to the detected value.

Further, for example, the control program according to one aspect of the present invention is a web transport device configured to transport the web, and is a sensor configured to detect the position in the width direction of the transported web. , And a program that controls the operation of a web transport device including a correction mechanism configured to correct the widthwise position of the web to be transported, the computer being informed of the detection value of the widthwise position of the web. With the step acquired from the sensor, the step of calculating the correction value for the reference value of the command value or the target value at the position in the width direction from the acquired detection value using the prediction model, and the calculated correction value. The step of determining the command value for the correction mechanism and the determined command value according to the target value of the position in the width direction corrected by the correction value or the calculated correction value. Based on this, it is a program for executing a step of controlling the operation of the correction mechanism so as to correct the position of the web in the width direction.

According to the present invention, it is possible to improve the performance of web transport and reduce the probability that a product using the web will be defective.

FIG. 1 schematically illustrates an example of a situation in which the present invention is applied. FIG. 2 schematically illustrates an example of the correction mechanism and the edge sensor according to the embodiment. FIG. 3 schematically illustrates an example of the hardware configuration of the control device according to the embodiment. FIG. 4 schematically illustrates an example of the software configuration of the control device according to the embodiment. FIG. 5 schematically illustrates an example of the relationship between the detected value and the predicted value of the end position. FIG. 6A illustrates an example of the processing procedure of the control device according to the embodiment. FIG. 6B illustrates an example of the processing procedure of the control device according to the embodiment. FIG. 7 schematically illustrates an example of a prediction control processing process using a prediction model according to a modified example. FIG. 8 schematically illustrates an example of a prediction control processing process using a prediction model according to a modified example. FIG. 9 schematically illustrates an example of a prediction control processing process using a prediction model according to a modified example. FIG. 10 schematically illustrates an example of a scene in which a model generator is applied. FIG. 11 schematically illustrates an example of the hardware configuration of the model generator. FIG. 12 schematically illustrates an example of the software configuration of the model generator. FIG. 13 illustrates an example of the processing procedure of the model generator.

Hereinafter, an embodiment according to one aspect of the present invention (hereinafter, also referred to as “the present embodiment”) will be described with reference to the drawings. However, the embodiments described below are merely examples of the present invention in all respects. Needless to say, various improvements and modifications can be made without departing from the scope of the present invention. That is, in carrying out the present invention, a specific configuration according to the embodiment may be appropriately adopted. The data appearing in the present embodiment are described in natural language, but more specifically, they are specified in a pseudo language, a command, a parameter, a machine language, etc. that can be recognized by a computer.

§1 Application example First, an example of a situation in which the present invention is applied will be described with reference to FIGS. 1 and 2. FIG. 1 schematically illustrates an example of an application scene of the control device 1 and the web transfer device 6 according to the present embodiment. FIG. 2 schematically illustrates an example of a scene in which the position of the end portion of the web 60 (hereinafter, also referred to as the end position) is adjusted in the web transport device 6. The end position is an example of the "position in the width direction" of the present invention.

First, the web transfer device 6 will be described. The web transport device 6 according to the present embodiment is configured to transport the web 60 while adjusting the position of the end portion of the web 60. Specifically, the web transport device 6 according to the present embodiment includes a supply roll 61 for supplying the web 60 and a collection roll 66 for collecting the supplied web 60. The type of the web 60 does not have to be particularly limited and may be appropriately selected depending on the embodiment. The web 60 may be, for example, a resin film such as a polyethylene film. Servo motors (611, 661) are attached to the rotating shafts of the supply roll 61 and the recovery roll 66, respectively. As a result, by driving each servomotor (611, 661), the web 60 is unwound from the supply roll 61, and the supplied web 60 is wound on the recovery roll 66.

Three driven rollers (62, 63, 641) are arranged between the supply roll 61 and the recovery roll 66. Of these, the driven roller 641 is provided with a correction mechanism 64 configured to correct (ie, displace) the position of the end of the web 60. As shown in FIG. 2, the correction mechanism 64 includes a driven roller 641 and an actuator 643. The driven roller 641 has a shaft 642 and is rotatably configured around the shaft 642 to convey the web 60. Further, an actuator 643 is attached to the shaft 642, and the driven roller 641 is configured to be slidable in the axial direction by driving the actuator 643. According to this correction mechanism 64, the position of the end portion of the web 60 can be corrected by sliding the driven roller 641 in the axial direction by driving the actuator 643. Specifically, the position of the end portion of the web 60 can be displaced in the sliding direction of the driven roller 641 according to the amount of sliding.

An edge sensor 65 is arranged near the correction mechanism 64. In the present embodiment, the edge sensor 65 is arranged on the recovery roll 66 side (downstream side) of the driven roller 641 provided with the correction mechanism 64 with respect to the transport direction of the web 60. The edge sensor 65 is configured to detect the position of the end of the conveyed web 60. The edge sensor 65 is an example of the "sensor" of the present invention. The type of the edge sensor 65 does not have to be particularly limited and may be appropriately selected depending on the embodiment. In the example of FIG. 2, the edge sensor 65 is an optical sensor and is configured to be able to measure the position of the end portion of the web 60 passing between the U-shapes.

The web transport device 6 according to the present embodiment includes the correction mechanism 64 and the edge sensor 65, so that the position of the end portion of the web 60 is detected and the position of the end portion of the web 60 is adjusted. It is configured to carry the web 60. However, the configuration of the web transfer device 6 does not have to be limited to such an example. Regarding the configuration of the web transfer device 6, components can be omitted, replaced, and added as appropriate according to the embodiment. For example, the number of driven rollers is not limited to three. Further, the arrangement of the correction mechanism 64 does not have to be limited to the example of FIG. 1, and may be appropriately selected according to the embodiment. The type of the correction mechanism 64 does not have to be limited to the example of FIG. 2, and may be appropriately selected depending on the embodiment. With respect to the configuration of the modification mechanism 64, components can be omitted, replaced, and added as appropriate according to the embodiment.

On the other hand, the control device 1 according to the present embodiment is a computer configured to control the operation of the web transfer device 6. Specifically, the control device 1 acquires the detection value 71 of the end position of the web 60 from the edge sensor 65. Subsequently, the control device 1 uses the prediction model M1 to calculate at least one predicted value 72 of the future detected value of the end position and the future command value for the correction mechanism 64 from the acquired detected value 71. do. In the present embodiment, the control device 1 calculates a predicted value 72 of a future detection value of the end position. Details will be described later. Next, the control device 1 determines the command value 77 for the correction mechanism 64 according to the calculated predicted value 72. The command value 77 defines, for example, the voltage of the actuator 643 for adjusting the position of the end of the web 60. Then, the control device 1 controls the operation of the correction mechanism 64 so as to correct the end position of the web 60 based on the determined command value 77.

As described above, in the present embodiment, in order to predict and control the operation of the correction mechanism 64, a prediction model that predicts the predicted value 72 of the future detected value of the end position from the detected value 71 of the end position of the web 60. M1 is used. The prediction model M1 is a detection value of the end position of the web 60 while controlling the operation of the correction mechanism 64, which is obtained as a result of controlling the actual machine of the correction mechanism 64 or simulating the operation of the correction mechanism 64. It can be generated from time-series data (also called actual data) showing the transition of. By collecting time series data under various conditions, the meandering result of the web 60 is reflected in the generation of the prediction model M1 based on various factors. That is, by using the time series data collected under various conditions, it is possible to construct a prediction model M1 that can predict the meandering of the Web 60 based on various factors.

Therefore, according to the configuration, the meandering of the web 60 can be predicted independently of the factors. Therefore, it is possible to improve the transport performance of the web 60 and reduce the probability that a defect will occur in the product using the web 60. Further, by predictive control using the predictive model M1, the frequency of maintenance (adjustment of parameters) of the correction mechanism 64 can be reduced. As a result, it is possible to suppress human costs and reduce the probability of malfunction due to human factors (for example, mistakes in maintenance work).

It should be noted that the control of the operation of the web transport device 6 as in the control device 1 does not have to be limited to the control of the end position of the web 60. The operation amount of the control device 1 does not have to be limited to the voltage of the actuator 643 of the correction mechanism 64. Further, the control amount does not have to be limited to the end position of the web 60. For example, in addition to these, each servomotor (611, 661) of each roll (61, 66) may be treated as an operation amount, and the transport speed of the web 60 may be treated as a control amount. That is, the control device 1 controls the transfer speed of the web 60 by operating each servomotor (611, 661) of each roll (61, 66) in addition to controlling the correction mechanism 64 of the web transfer device 6. You may control it.

§2 Configuration example [Hardware configuration]
Next, an example of the hardware configuration of the control device 1 according to the present embodiment will be described with reference to FIG. FIG. 3 schematically illustrates an example of the hardware configuration of the control device 1 according to the present embodiment.

As shown in FIG. 3, the control device 1 according to the present embodiment is a computer to which the control unit 11, the storage unit 12, the external interface 13, the input device 14, the output device 15, and the drive 16 are electrically connected. .. In FIG. 3, the external interface is described as "external I / F".

The control unit 11 includes a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), and the like, which are hardware processors, and is configured to execute information processing based on a program and various data. To. The storage unit 12 is an example of a memory, and is composed of, for example, a RAM, a ROM, or the like. In the present embodiment, the storage unit 12 stores various information such as the control program 81 and the learning result data 125.

The control program 81 is a program for causing the control device 1 to execute information processing (FIGS. 6A and 6B) described later that controls the operation of the web transfer device 6. The control program 81 includes a series of instructions for the information processing. The training result data 125 shows information about the trained prediction model M1. Details will be described later.

The external interface 13 is, for example, a USB (Universal Serial Bus) port, a dedicated port, or the like, and is an interface for connecting to an external device. The type and number of the external interfaces 13 may be appropriately selected according to the type and number of connected external devices. In the present embodiment, the control device 1 is connected to the web transfer device 6 via the external interface 13. As a result, the control device 1 can control the operation of the web transfer device 6 by transmitting a command value to the web transfer device 6. Further, the control device 1 can acquire the detection value (measured value) of the edge sensor 65 from the web transfer device 6. The method of connecting to the web transport device 6 does not have to be limited to such an example. The control device 1 may be connected to the web transfer device 6 via, for example, a communication interface or the like. In this case, the external interface 13 may be omitted.

The input device 14 is, for example, a device for inputting a mouse, a keyboard, or the like. Further, the output device 15 is, for example, a device for outputting a display, a speaker, or the like. The operator can operate the control device 1 by using the input device 14 and the output device 15. The input device 14 and the output device 15 may be integrally configured by a touch panel display or the like.

The drive 16 is, for example, a CD drive, a DVD drive, or the like, and is a drive device for reading a program stored in the storage medium 91. The type of the drive 16 may be appropriately selected according to the type of the storage medium 91. At least one of the control program 81 and the learning result data 125 may be stored in the storage medium 91.

The storage medium 91 transfers the information of the program or the like by electrical, magnetic, optical, mechanical or chemical action so that the computer or other device, the machine or the like can read the information of the recorded program or the like. It is a medium to accumulate. The control device 1 may acquire at least one of the control program 81 and the learning result data 125 from the storage medium 91.

Here, in FIG. 3, as an example of the storage medium 91, a disc-type storage medium such as a CD or a DVD is illustrated. However, the type of the storage medium 91 is not limited to the disc type, and may be other than the disc type. Examples of storage media other than the disk type include semiconductor memories such as flash memories.

Regarding the specific hardware configuration of the control device 1, components can be omitted, replaced, or added as appropriate according to the embodiment. For example, the control unit 11 may include a plurality of hardware processors. The hardware processor may be composed of a microprocessor, an FPGA (field-programmable gate array), a DSP (digital signal processor), or the like. The storage unit 12 may be composed of a RAM and a ROM included in the control unit 11. The storage unit 12 may be configured by an auxiliary storage device such as a hard disk drive or a solid state drive. At least one of the external interface 13, the input device 14, the output device 15, and the drive 16 may be omitted. The control device 1 may further include a communication interface for performing data communication with another computer via a network. Further, the control device 1 is an information processing device designed exclusively for the provided service, as well as an information processing device such as a PLC (programmable logic controller), a desktop PC (Personal Computer), or a tablet PC, depending on the object to be controlled. May be replaced with.

[Software configuration]
Next, an example of the software configuration of the control device 1 according to the present embodiment will be described with reference to FIG. FIG. 4 schematically illustrates an example of the software configuration of the control device 1 according to the present embodiment.

The control unit 11 of the control device 1 expands the control program 81 stored in the storage unit 12 into the RAM. Then, the control unit 11 controls each component by interpreting the control program 81 expanded in the RAM by the CPU and executing a series of instruction groups included in the control program 81. As a result, as shown in FIG. 4, the control device 1 according to the present embodiment includes the acquisition unit 111, the calculation unit 112, the determination unit 113, the operation control unit 114, the data collection unit 115, and the model update unit 116 as software modules. Operates as a computer. That is, in this embodiment, each software module is realized by the control unit 11 (CPU).

The acquisition unit 111 acquires the detection value 71 of the end position of the web 60 from the edge sensor 65. The calculation unit 112 includes the prediction model M1 by holding the learning result data 125. The calculation unit 112 sets the trained prediction model M1 with reference to the learning result data 125. The calculation unit 112 uses the prediction model M1 to calculate the predicted value 72 of the future detection value of the end position from the acquired detection value 71. The determination unit 113 determines the command value 77 for the correction mechanism 64 according to the calculated predicted value 72. The motion control unit 114 controls the motion of the correction mechanism 64 so as to correct the end position of the web 60 based on the determined command value 77.

Here, the process of deriving the command value 77 from the predicted value 72 calculated by the predicted model M1 will be described in detail with reference to FIG. FIG. 5 schematically illustrates an example of the relationship between the detected value 71 and the predicted value 72 at the end position. As shown in FIG. 5, in the present embodiment, the prediction model M1 predicts the detected value of the end position in the future control cycle with respect to the input of the value of the end position in each of the plurality of control cycles up to now. It may be configured to output a value. The control cycle is a cycle in which a command value 77 is given to the web transport device 6 (correction mechanism 64).

The acquisition unit 111 acquires the detection value 71 in each of the plurality of control cycles. The calculation unit 112 inputs each of the acquired detection values 71 into the prediction model M1 and executes arithmetic processing of the prediction model M1. As a result, the calculation unit 112 acquires the output value output from the prediction model M1 as the predicted value 72 of the detected value of the end position in the future control cycle. The number of control cycles involved in the input (that is, the number of detected values 71 to be input) and the number of controlled cycles to be predicted (that is, the number of predicted values 72 to be output) may not be particularly limited, respectively. It may be appropriately determined according to the form of.

The determination unit 113 determines the command value 77 for the correction mechanism 64 according to the acquired predicted value 72. In the present embodiment, the determination unit 113 corrects the target value 70 of the end position according to the acquired predicted value 72. For example, the determination unit 113 may calculate the correction value 75 according to the difference between the target value 70 and the predicted value 72, and correct the target value 70 by the calculated correction value 75. That is, the determination unit 113 may acquire the corrected target value 76 by applying the correction value 75 to the target value 70 (for example, adding or subtracting). Then, the determination unit 113 determines the command value 77 for the correction mechanism 64 according to the corrected target value 76 and the detection value 71. The method for determining the command value 77 does not have to be particularly limited and may be appropriately selected depending on the embodiment. The command value 77 may be determined by a known control method such as, for example, PID control or PI control. In this case, the determination unit 113 determines the command value 77 according to the difference between the corrected target value 76 and the detection value 71. The operation control unit 114 controls the operation of the correction mechanism 64 based on the determined command value 77.

The prediction model M1 includes arithmetic parameters for calculating the predicted value from the input value. The type of such a prediction model M1 does not have to be particularly limited, and may be appropriately selected depending on the embodiment. For the prediction model M1, for example, a regression model such as an autoregressive model may be used. The autoregressive model is represented by a regression equation that sums one or more terms including the autoregressive coefficient. When the prediction model M1 is configured by this autoregressive model, the autoregressive coefficient and the like in each term of the autoregressive model are examples of calculation parameters.

Alternatively, for the prediction model M1, for example, a machine learning model such as a neural network may be used. A neural network comprises one or more layers containing one or more neurons (nodes). When a plurality of layers are provided, typically, each layer is arranged from the input side to the output side, and neurons included in the adjacent layers are appropriately connected to each other. A weight (bonding load) is set for each bond. A threshold is set for each neuron, and basically, the output of each neuron is determined by whether or not the sum of the products of each input and each weight exceeds the threshold. When the prediction model M1 is configured by this neural network, for example, the weight of the connection between each neuron, the threshold value of each neuron, and the like are examples of arithmetic parameters.

This prediction model M1 may be generated as appropriate. As an example, time-series data showing the transition of the detected value (measured value or calculated value) of the end position of the web 60 while the web transport device 6 is appropriately transporting the web 60 is acquired. The time series data may be acquired by operating the actual machine or by simulation. Then, by analyzing the acquired time-series data, it is possible to generate a prediction model M1 that predicts the detected value of the end position of the web 60. The method of analyzing the time series data (that is, the method of generating the prediction model M1) for generating the prediction model M1 does not have to be particularly limited, and may be appropriately selected depending on the type of the prediction model M1. ..

For example, when the prediction model M1 is composed of an autoregressive model, the prediction model M1 can be obtained by deriving the values of arithmetic parameters such as the autoregressive coefficient from time series data by a known regression analysis method such as the least squares method. Can be generated. Further, for example, when the prediction model M1 is composed of a machine learning model such as a neural network, the time series data is used as training data, and the values of the calculation parameters of the machine learning model are derived by a known machine learning method. Then, the prediction model M1 can be generated.

As an example, when the prediction model M1 is configured by a neural network, a sample of time series data at an arbitrary time is used as training data, and a sample at a time later than an arbitrary time is used as teacher data for prediction. Perform machine learning for model M1. Specifically, by back-propagating the gradient of the error between the output value for the training data and the teacher data by the error back-propagation method, the values of the calculation parameters such as the weight of the connection between each neuron and the threshold value of each neuron are adjusted. do. As a result, the prediction model M1 is trained to output an output value that matches the teacher data with respect to the input of the training data. As a result of this machine learning, it is possible to generate a prediction model M1 that has acquired the ability to predict the predicted value of the detected value of the end position at a time later than the arbitrary time from the detected value of the end position at an arbitrary time. can.

The prediction model M1 may be configured to accept input of data other than the command value for the correction mechanism 64, the detection value of the end position such as the individual information of the web 60, and the like. In this case, data other than the detected value may be further acquired. Then, in the above analysis process, other acquired data may be further used. For example, when the prediction model M1 is composed of an autoregressive model, other acquired data may be further used for deriving the autoregressive coefficient. Further, for example, when the prediction model M1 is configured by a neural network, other acquired data may be further used as training data. As a result, it is possible to generate a prediction model M1 constructed so as to output a predicted value of the detected value of the end position with respect to the input of the detected value of the end position and other data.

Further, inputting each detection value 71 into the prediction model M1 converts each acquired detection value 71 into one or a plurality of feature quantities 710, and the obtained one or a plurality of feature quantities 710 is a prediction model. It may be configured by inputting to M1. The type of the feature amount 710 is not particularly limited as long as it can represent some feature of the detected value 71, and may be appropriately selected depending on the embodiment. For example, the feature amount 710 may be obtained by extracting the feature of the target to be used for prediction from the time series data of the detected value 71. Further, for example, the feature amount 710 may be aggregated information (for example, a time series of average values for each unit interval), a series itself of one section from the prediction time to the past, and the like. As a specific example, the feature amount 710 may be a minimum value, a maximum value, an average value, a kurtosis, a differential value, or the like. In this case, the detected value of the end position indicated by the time series data may be converted into a feature amount, and the obtained feature amount may be used for the above analysis process. As a result, it is possible to generate a prediction model M1 constructed so as to output a predicted value of the detected value of the end position with respect to the input of the feature amount derived from the detected value of the end position.

The process of generating the prediction model M1 may be executed by the control device 1 or may be executed by a computer other than the control device 1. The training result data 125 is generated as information indicating the trained prediction model M1. The generated learning result data 125 may be provided to the control device 1 at an arbitrary timing. For example, the control device 1 may acquire the learning result data 125 from another computer. Further, for example, the generated learning result data 125 may be stored in a predetermined storage area. The predetermined storage area may be, for example, a storage unit 12, a storage medium 91, an external storage device, or the like. The external storage device may be a network server such as NAS (Network Attached Storage), or may be an external storage device connected to the control device 1. In this case, the control device 1 may acquire the learning result data 125 from a predetermined storage area. Further, for example, the learning result data 125 may be incorporated in the control device 1 in advance.

When the prediction model M1 is configured by an autoregressive model, the training result data 125 shows, for example, the values of arithmetic parameters such as the autoregressive coefficient derived in each term. When the prediction model M1 is configured by a neural network, the training result data 125 may be, for example, a value of a calculation parameter (for example, a weight of a connection between each neuron, a threshold value of each neuron, etc.) and a structure of the neural network (for example). For example, the number of layers, the number of neurons contained in each layer, the connection relationship between neurons in adjacent layers, etc.) are shown. The calculation unit 112 reproduces the trained prediction model M1 with reference to the learning result data 125, and calculates the prediction value 72 from each detection value 71.

While controlling the operation of the web transfer device 6, the data collection unit 115 collects the detection value 71 of the end position of the web 60 acquired from the edge sensor 65 as the actual data 79.
The model update unit 116 updates the prediction model M1 based on the collected actual data 79. The method for updating the prediction model M1 may be the same as the method for generating the prediction model M1.

Each software module of the control device 1 will be described in detail in an operation example described later. In this embodiment, an example in which each software module of the control device 1 is realized by a general-purpose CPU is described. However, some or all of the above software modules may be implemented by one or more dedicated processors. Further, regarding the software configuration of the control device 1, the software module may be omitted, replaced, or added as appropriate according to the embodiment.

§3 Operation example Next, an operation example of the control device 1 according to the present embodiment will be described with reference to FIGS. 6A and 6B. 6A and 6B are flowcharts showing an example of the processing procedure of the control device 1 according to the present embodiment. The processing procedure of the control device 1 described below is an example of the "control method" of the present invention. However, the processing procedure described below is only an example, and each processing may be changed as much as possible. Further, with respect to the processing procedure described below, steps can be omitted, replaced, and added as appropriate according to the embodiment. The order of processing without dependencies may be changed as appropriate.

(Step S101)
In step S101, the control unit 11 operates as the acquisition unit 111 to acquire the target value 70, which is the target of the end position of the web 60.

The acquisition destination of the target value 70 may be appropriately selected according to the embodiment. The information indicating the target value 70 may be stored in a storage area such as a storage unit 12, a storage medium 91, or an external storage device. The external storage device may be, for example, a data server such as NAS, or an external storage device connected to the control device 1. The control unit 11 can acquire the target value 70 by accessing the storage area.

The target value 70 of the end position may be appropriately determined so that the web 60 can be appropriately conveyed. For example, the target value 70 of the end position may be determined based on the result of transporting the web 60 to the web transport device 6 by an actual machine or a simulation. Further, for example, the target value 70 of the end position may be given by the operator's designation. In this case, the control device 1 can acquire the target value 70 of the end position by accepting the input of the operator via the input device 14 or acquiring the information input by the operator in another computer. ..

The target value 70 may be given for each control cycle. In this case, the target value 70 may be a different value for each control cycle, or may be a constant value in at least a part of the control cycles. Further, the control unit 11 may acquire the target value 70 in a plurality of control cycles at once, or may acquire the target value 70 of the target control cycle for each control cycle. When the target value 70 is acquired, the control unit 11 proceeds to the next step S102.

(Step S102)
In step S102, the control unit 11 operates as the acquisition unit 111 and acquires the detection value 71 of the end position of the web 60 from the edge sensor 65. The detection of the end position may be performed on a simulation, or may be performed by using an actual machine of the web transfer device 6. Further, the control unit 11 may directly acquire the detection value 71 of the end position from the edge sensor 65, or may use, for example, the edge sensor 65 via an external storage device, a storage medium 91, another computer, or the like. It may be obtained indirectly from. When the detection value 71 is acquired, the control unit 11 proceeds to the next step S103.

(Step S103)
In step S103, the control unit 11 operates as the operation control unit 114, and determines whether or not to perform the predictive control of the web transfer device 6.

Whether or not to carry out predictive control of the web transfer device 6 may be appropriately determined. For example, whether or not to perform predictive control of the web transfer device 6 may be determined by instructions of a computer other than the control device 1 or an operator. In this case, the control unit 11 may determine whether or not to perform predictive control in response to instructions from another computer or operator.

Further, in the present embodiment, the prediction model M1 outputs the predicted value of the detected value of the end position in the future control cycle for the input of the value of the end position in each of the plurality of control cycles up to the present. May be configured in. In this case, the control unit 11 may determine whether or not to execute the prediction control depending on whether or not the detection values 71 for a plurality of control cycles input to the prediction model M1 have been acquired. That is, when the detection values 71 for a plurality of control cycles are acquired, the control unit 11 may determine that the prediction control is executed. On the other hand, if this is not the case, the control unit 11 may determine that the predictive control is not performed.

If it is determined that the predictive control is to be performed, the control unit 11 proceeds to the next step S104. On the other hand, if it is determined that the prediction control is not performed, the control unit 11 omits the processing of step S104 and step S105 and proceeds to the next step S106.

(Step S104)
In step S104, the control unit 11 operates as the calculation unit 112, refers to the learning result data 125, and sets the trained prediction model M1. Then, the control unit 11 inputs the detection value 71 into the set prediction model M1 and executes the arithmetic processing of the prediction model M1. As a result, the control unit 11 can calculate the predicted value 72 of the future detection value of the end position from the acquired detection value 71 by using the prediction model M1.

In the present embodiment, by repeating the processes of steps S101 to S108 in step S101 or in step S108 described later, the control unit 11 can acquire the detection value 71 in each of the plurality of control cycles. The control unit 11 inputs each of the acquired detection values 71 into the prediction model M1 and executes arithmetic processing of the prediction model M1. As a result, the control unit 11 can acquire the output value output from the prediction model M1 as the predicted value 72 of the detected value of the end position in the future control cycle.

As an example of arithmetic processing, when the prediction model M1 is configured by an autoregressive model, the control unit 11 sets the regression equation of the prediction model M1 by referring to the learning result data 125. Next, the control unit 11 inputs each detection value 71 into each term of the regression equation and executes the arithmetic processing of the regression equation. Based on the result of this calculation, the control unit 11 can acquire the predicted value 72 of the detected value of the end position.

Further, as another example of the arithmetic processing, when the prediction model M1 is configured by the neural network, the control unit 11 sets the trained neural network by referring to the learning result data 125. Next, the control unit 11 inputs each detection value 71 to the input layer of the neural network, and determines the firing of each neuron included in each layer in order from the input side. Based on the result of this calculation, the control unit 11 can acquire an output value (that is, a predicted value 72 of the detected value of the end position) corresponding to the result of predicting the detected value of the end position from the output layer.

When the prediction model M1 is configured so that other data such as a command value for the correction mechanism 64 and individual information of the web 60 can be further input, the control unit 11 may perform the step S104 before executing the step S104. The other data may be further acquired. Then, the control unit 11 inputs each of the acquired detection values 71 and other data to the prediction model M1 and executes the arithmetic processing of the prediction model M1. As a result, the control unit 11 can acquire the output value output from the prediction model M1 as the prediction value 72.

Further, when the prediction model M1 is configured to accept the input of one or a plurality of feature quantities 710, the control unit 11 converts each acquired detection value 71 into one or a plurality of feature quantities 710. Then, the control unit 11 inputs the obtained one or a plurality of feature quantities 710 to the prediction model M1 and executes the arithmetic processing of the prediction model M1. As a result, the control unit 11 can acquire the output value output from the prediction model M1 as the prediction value 72.

When the predicted value 72 of the detected value of the end position is acquired, the control unit 11 proceeds to the next step S105.

(Step S105)
In step S105, the control unit 11 operates as the determination unit 113 and corrects the target value 70 of the end position according to the calculated predicted value 72. As a result, the control unit 11 acquires the corrected target value 76.

The relationship between the predicted value 72 and the correction amount with respect to the target value 70 may not be particularly limited and may be appropriately determined according to the embodiment. The fact that the predicted value 72 obtained by the prediction model M1 deviates from the target value 70 means that even if the correction mechanism 64 is controlled as it is based on the target value 70, the behavior deviates from the target value 70 due to the influence of disturbance or the like. It may correspond to the prediction that the correction mechanism 64 will operate. Therefore, the relationship between the predicted value 72 and the correction amount may be given, for example, so as to reduce this deviated behavior. As an example, the relationship between the predicted value 72 and the correction amount is such that the target value 70 is corrected to a larger value when the predicted value 72 is smaller than the target value 70, and the target value 72 is larger than the target value 70. It may be determined to modify 70 to a smaller value. Thereby, the target value 70 can be corrected so that the correction mechanism 64 can be operated with the behavior desired by the original target value 70. The calculation formula for calculating such a correction amount may be given by, for example, the following formula 1.

なお、Ｒ（ｔ）は、時刻ｔにおける補正量の値（補正値７５）を示す。ｒは、比例定数を示す。Ｅ（ｔ）は、時刻ｔにおける目標値７０を示す。Ｐ（ｔ）は、時刻ｔにおける端部位置の予測値７２を示す。ｓは、定数項を示す。

In addition, R (t) shows the value (correction value 75) of the correction amount at time t. r indicates a constant of proportionality. E (t) indicates a target value 70 at time t. P (t) indicates a predicted value 72 of the end position at time t. s indicates a constant term.

In this calculation formula, the degree of correction is determined by the values of the proportionality constant r and the constant term s. The proportionality constant r and the constant term s are examples of correction weights, respectively. However, the calculation formula for determining the correction amount is not limited to the example of the formula 1, and may be appropriately determined according to the embodiment.

The values of the proportionality constant r and the constant term s may be given as appropriate. For example, the values of the proportionality constant r and the constant term s may be given by the operator's specification. Further, for example, the values of the proportional constant r and the constant term s may be determined based on the result of transporting the web 60 to the web transport device 6 by an actual machine or a simulation. At least one of the proportionality constant r and the constant term s may differ between one control cycle and the other control cycle. That is, different weights may be set according to the control cycle.

The control unit 11 substitutes the acquired predicted value 72 (P (t)) and target value 70 (E (t)) into each term of Equation 1 and executes the arithmetic processing of Equation 1 to correct the value. 75 (R (t)) is calculated. That is, the control unit 11 calculates the difference between the target value 70 and the predicted value 72, and determines the correction value 75 based on the calculated difference and the weight. Specifically, the control unit 11 calculates the product of the calculated difference and the proportionality constant r, and adds the constant term s to the calculated product to calculate the correction value 75. Then, the control unit 11 acquires the corrected target value 76 by adding the calculated correction value 75 to the target value 70. This addition process may be replaced with a subtraction process. When the corrected target value 76 is acquired, the control unit 11 proceeds to the next step S106.

(Step S106)
In step S106, the control unit 11 operates as the determination unit 113, and determines the command value 77 for the correction mechanism 64 according to the command target value and the detection value 71 of the end position. When predictive control is implemented, the corrected target value 76 is used as the command target value. As a result, the control unit 11 determines the command value 77 for the correction mechanism 64 according to the calculated predicted value 72. On the other hand, when the predictive control is not performed, the target value 70 is used as the command target value. In the present embodiment, the control unit 11 can determine the command value 77 for the correction mechanism 64 according to the command target value and the detection value 71 of the end position.

The method for determining the command value 77 does not have to be particularly limited and may be appropriately selected depending on the embodiment. As a method for determining the command value 77, for example, a known control method such as PID control or PI control may be adopted. In this case, the control unit 11 calculates the difference between the command target value and the detection value 71, and determines the command value 77 according to the calculated difference. Specifically, when predictive control is being performed, the control unit 11 determines the command value 77 according to the difference between the corrected target value 76 and the detected value 71. On the other hand, when the predictive control is not performed, the control unit 11 determines the command value 77 according to the difference between the target value 70 and the detected value 71. The number of command values 77 determined in one process may be appropriately determined according to the embodiment. When the command value 77 is determined, the control unit 11 proceeds to the next step S107.

(Step S107)
In step S107, the control unit 11 controls the operation of the correction mechanism 64 so as to correct the end position of the web 60 based on the determined command value 77. The end position of the web 60 is adjusted based on the operation of the correction mechanism 64. In the present embodiment, the control unit 11 can control the operation of the correction mechanism 64 based on the determined command value 77. The control of the operation of the correction mechanism 64 based on the determined command value 77 may be executed in the control cycle before the future control cycle in which the predicted value 72 is calculated. Thereby, when the meandering of the web 60 is predicted, the position of the end portion of the web 60 can be adjusted so as to prevent the meandering of the web 60.

The command value 77 defines, for example, the voltage of the actuator 643 for adjusting the position of the end of the web 60. The control unit 11 controls the position of the end portion of the web 60 in the web transfer device 6 by operating the actuator 643 with the voltage specified by the command value 77. The control of the operation may be performed on a simulation, or may be performed by using an actual machine of the web transfer device 6. When the operation of the correction mechanism 64 is controlled according to the command value 77, the control unit 11 proceeds to the next step S108.

(Step S108)
In step S108, the control unit 11 operates as the operation control unit 114, and determines whether or not to end the control of the operation of the web transfer device 6.

Whether or not to end the operation control of the web transfer device 6 may be appropriately determined. For example, whether or not to end the operation control of the web transfer device 6 may be determined by an instruction of a computer other than the control device 1 or an operator. In this case, the control unit 11 may determine whether or not to end the operation control of the web transfer device 6 in response to an instruction from another computer or an operator.

Further, for example, the control unit 11 may determine whether or not to end the operation control of the web transfer device 6 according to the amount of the web 60 remaining in the supply roll 61. As an example, when the amount of the web 60 remaining on the supply roll 61 is equal to or greater than the threshold value, the control unit 11 may determine that the operation control of the web transfer device 6 is not terminated (that is, continued). On the other hand, when the amount of the web 60 remaining on the supply roll 61 is less than the threshold value, the control unit 11 may determine that the operation control of the web transport device 6 is terminated. The threshold may be determined as appropriate. The amount of the web 60 remaining on the supply roll 61 may be detected by a sensor (not shown) or may be calculated from the amount of the web 60 conveyed.

Further, for example, when the web 60 is transported for manufacturing a product, it may be determined whether or not the operation control of the web transport device 6 is terminated according to the amount of the manufactured product. As an example, when the quantity of the manufactured product is equal to or less than the threshold value, the control unit 11 may determine that the operation control of the web transfer device 6 is continued. On the other hand, when the amount of the manufactured product exceeds the threshold value, the control unit 11 may determine that the operation control of the web transfer device 6 is terminated. The threshold may be determined as appropriate. The amount of the manufactured product may be specified as appropriate.

When it is determined that the operation control of the web transfer device 6 is terminated, the control unit 11 proceeds to the next step S109. On the other hand, when it is determined that the operation control of the web transfer device 6 is not terminated, that is, the operation control of the web transfer device 6 is to be continued, the control unit 11 returns the process to step S101 and repeats the processes of steps S101 to S108. .. The timing at which the treatment is repeated may be appropriately determined according to the embodiment. In the present embodiment, the control unit 11 can repeat the processes of steps S101 to S108 for each control cycle. As a result, the control unit 11 can schedule the operation of the web transfer device 6 while synchronizing the prediction control by the prediction model M1 with the actual environment in real time.

(Step S109)
In step S109, the control unit 11 determines the detection value 71 of the end position of the web 60 acquired from the edge sensor 65 by step S102 while the data collection unit 115 controls the operation of the web transfer device 6. collect. As a result, the control unit 11 acquires the actual data 79 indicating the detected value 71 of the end position of the web 60 while controlling the operation of the web transport device 6 in chronological order. That is, the actual data 79 is time-series data showing the transition of the detected value 71 while the web transport device 6 is transporting the web 60.

The control unit 11 may store the acquired actual data 79 in a predetermined storage area. The predetermined storage area may be, for example, a storage unit 12, a storage medium 91, an external storage device, another computer, or the like. The actual data 79 may include not only the data obtained by the control processing of steps S101 to S108, but also data obtained by other than the above control processing such as data collected in advance. When the actual data 79 is acquired, the control unit 11 proceeds to the next step S110.

(Step S110)
In step S110, the control unit 11 determines whether or not to update the prediction model M1. Whether or not to update the prediction model M1 may be appropriately determined. For example, whether or not to update the prediction model M1 may be determined by the instruction of a computer or an operator other than the control device 1. In this case, the control unit 11 may determine whether or not to update the prediction model M1 in response to an instruction from another computer or an operator.

Further, for example, when the processing procedure according to this operation example is repeated, the control unit 11 may update the prediction model M1 periodically or irregularly. As an example, the control unit 11 may determine that the update process is to be executed when a predetermined time has elapsed since the update process of the prediction model M1 was executed before, and if not, the control unit 11 may determine that the update process is to be executed. It may be determined that the update process is omitted. The predetermined time (threshold value) may be appropriately determined.

Further, for example, the control unit 11 may determine whether or not to update the prediction model M1 according to the amount of the collected actual data 79. That is, the control unit 11 may determine that the update process is to be executed when a sufficient amount of actual data 79 is obtained for the update, and may determine that the update process is omitted if not. good. As an example, the control unit 11 compares the data amount of the actual data 79 with the threshold value. The control unit 11 may determine that the update process is executed when the data amount of the actual data 79 is equal to or greater than the threshold value, and may determine that the update process is omitted if not.

When it is determined that the prediction model M1 is to be updated (that is, the update process is executed), the control unit 11 proceeds to the next step S111. On the other hand, when it is determined that the prediction model M1 is not updated, the control unit 11 ends the process related to this operation example.

(Step S111)
In step S111, the control unit 11 operates as the model update unit 116 and updates the prediction model M1 based on the acquired actual data 79.

The method for updating the prediction model M1 does not have to be particularly limited, and may be appropriately determined according to the embodiment. For example, the control unit 11 can derive a new prediction model by deriving the value of the calculation parameter by analyzing at least a part of the time series data and the actual data 79 of the detected value used for generating the prediction model M1. Generate. The method for generating a new prediction model may be the same as the above method. In the actual data 79, the portion used to generate the new prediction model may be limited to the portion where the web transport device 6 appropriately transports the web 60.

The control unit 11 may replace the existing prediction model used in step S104 with a new prediction model. Alternatively, the control unit 11 may save a new prediction model separately from the existing prediction model, and change the prediction model M1 used in step S104 from the existing prediction model to the new prediction model. As a result, the control unit 11 can update the prediction model M1. When the processing procedure according to this operation example is repeated thereafter, the control unit 11 can use the updated prediction model M1 in the prediction control. When the prediction model M1 is updated, the control unit 11 ends the process related to this operation example.

[feature]
As described above, in the present embodiment, in the prediction control of steps S104 to S106, the prediction model M1 that predicts the predicted value 72 of the future detected value of the end position from the detected value 71 of the end position of the web 60 is used. We are using. The prediction model M1 is generated based on time-series data showing the transition of the detected value of the end position of the web 60 while controlling the operation of the correction mechanism 64. By collecting time-series data under various conditions, the meandering result of the web 60 can be reflected in the prediction model M1 based on various factors. That is, by using the time series data collected under various conditions, it is possible to construct a prediction model M1 that can predict the meandering of the Web 60 based on various factors.

For example, when the process of transporting the web 60 includes an intermittent operation, acceleration / deceleration of the transport speed of the web 60 due to the intermittent operation may cause meandering of the web 60. Further, for example, when a series of manufacturing processes includes a step of laminating a plurality of sheet materials to form a web, the attributes of the web change before and after laminating the sheet materials. Even in these cases, the meandering of the web 60 can be appropriately predicted by using the prediction model M1 generated based on the time-series data reflecting the conditions. Further, in the present embodiment, the detection value 71, which is the input (explanatory variable) of the prediction model M1, is acquired from the edge sensor 65 arranged near the correction mechanism 64. That is, since the range of the prediction process is closed in the process of transporting the web 60, even if the process of transporting the web 60 is independent of other processes of the manufacturing process, the prediction model M1 allows the web 60 to be conveyed. The meandering can be predicted appropriately.

Therefore, according to the configuration, in step S104, the meandering of the web 60 can be predicted without depending on the factor. Therefore, it is possible to improve the transport performance of the web 60 and reduce the probability that a defect will occur in the product using the web 60. Further, by predictive control using the predictive model M1, the frequency of maintenance (adjustment of parameters) of the correction mechanism 64 can be reduced. As a result, it is possible to suppress human costs and reduce the probability of malfunction due to human factors (for example, mistakes in maintenance work).

§4 Modifications Although the embodiments of the present invention have been described in detail above, the above description is merely an example of the present invention in all respects. Needless to say, various improvements and modifications can be made without departing from the scope of the present invention. For example, the following changes can be made. In the following, the same reference numerals will be used for the same components as those in the above embodiment, and the same points as in the above embodiment will be omitted as appropriate. The following modifications can be combined as appropriate.

<4.1>
In the above embodiment, the control device 1 is composed of one computer. However, the number of computers constituting the control device 1 does not have to be limited to such an example. The control device 1 may be composed of a plurality of computers. For example, the control device 1 is configured to control the operation of the first information processing device configured to determine the command target value and the web transfer device 6 according to the determined command target value. It may be composed of two information processing devices. Each information processing device may be composed of one or a plurality of computers.

In this case, the first information processing apparatus determines the command target value by executing the processes from step S101 to step S105. Then, the first information processing device transmits the determined command value 77 to the second information processing device. The second information processing apparatus determines the command value 77 according to the received command target value by executing the steps S106 and S107, and the web transport device 6 determines the command value 77 based on the determined command value 77. It controls the operation of the correction mechanism 64. According to this configuration, a conventional controller can be used for the second information processing device. Therefore, the introduction cost of the present invention can be suppressed. The first information processing device may be referred to as an "upper controller", and the second information processing device may be referred to as a "lower controller".

<4.2>
In the above embodiment, the prediction model M1 is configured to calculate the prediction value 72 of the detection value of the end position of the web 60 from the acquired detection value 71. However, the prediction model M1 does not have to be limited to such an example. The prediction model M1 may be configured to calculate at least one of the future detection value of the end position of the web 60 and the future command value for the correction mechanism 64 from the acquired detection value 71. In response to this, in step S109, the control unit 11 may collect at least one of the detection value 71 of the end position acquired from the edge sensor 65 and the determined command value 77 as actual data 79.

FIG. 7 schematically illustrates an example of a predictive control processing process using the predictive model M2 according to the present modification in the software configuration of the control device 1 according to the above embodiment. In this modification, the prediction model M2 is configured to calculate the predicted value 72A of the future command value for the correction mechanism 64 from the acquired detection value 71. Similar to the above embodiment, the prediction model M2 outputs the predicted value of the command value for the correction mechanism 64 in the future control cycle for the input of the value of the end position in each of the plurality of control cycles up to now. May be configured in.

In the generation of the prediction model M2, in addition to the time-series data of the detected value of the end position, the time-series data of the command value for the correction mechanism 64 is also used. The time-series data of the command value may be obtained by controlling or simulating the actual machine in the same manner as the time-series data of the detected value of the end position. In analysis methods such as regression analysis and machine learning using the error back propagation method, the time series data of the detected value is used as input data (training data), and the time series data of the command value is used as teacher data (correct answer data). do. This makes it possible to generate a prediction model M2 trained to output a predicted value of a command value in response to an input of a value at the end position. The training result data 125A is generated as information indicating the trained prediction model M2. In addition, except for these points, the prediction model M2 may be the same as the prediction model M1.

The control device 1 predicts the correction mechanism 64 by using the prediction model M2 by the same processing procedure as that of the above embodiment except that the processing of steps S104 to S106, step S109, and step S111 is changed. Control can be carried out. Specifically, in this modification, the control unit 11 calculates the predicted value 72A of the future command value for the correction mechanism 64 by executing the next process instead of the process of step S104 in the above process procedure. do. That is, the control unit 11 operates as the calculation unit 112, and sets the trained prediction model M2 with reference to the learning result data 125A. Then, the control unit 11 inputs the detection value 71 into the set prediction model M2, and executes the arithmetic processing of the prediction model M2. As a result, the control unit 11 can calculate the predicted value 72A of the future command value for the correction mechanism 64 from the acquired detection value 71 by using the prediction model M2.

In this modification, the control unit 11 can acquire the detection value 71 in each of the plurality of control cycles, as in the above embodiment. The control unit 11 inputs each of the acquired detection values 71 into the prediction model M2, and executes the arithmetic processing of the prediction model M2. As a result, the control unit 11 can acquire the output value output from the prediction model M2 as the predicted value 72A of the command value in the future control cycle.

Further, in this modification, the control unit 11 executes the following process instead of the process of step S105 and step S106, and the command value 77 for the correction mechanism 64 is executed according to the calculated predicted value 72A. To decide. That is, the control unit 11 operates as the determination unit 113, and calculates the reference value 73 of the command value for the correction mechanism 64 according to the target value 70 and the detection value 71 of the end position. The method for calculating the reference value 73 may be the same as in step S106. In this modification, the control unit 11 can determine the command value 77 by correcting the calculated reference value 73 according to the acquired predicted value 72A.

The relationship between the predicted value 72A and the correction amount with respect to the reference value 73 may not be particularly limited as in the above embodiment, and may be appropriately determined according to the embodiment. For example, the value of the correction amount (correction value 75A) with respect to the reference value 73 may be determined so as to reduce the discrepancy between the predicted value 72A and the reference value 73. As an example, the control unit 11 may calculate the difference between the predicted value 72A and the reference value 73, and multiply the calculated difference by a proportionality constant. Next, the control unit 11 may calculate the correction value 75A by adding a constant term to the obtained seats. Then, the control unit 11 may acquire the corrected command value (that is, the command value 77) by adding the correction value 75A to the reference value 73. When the command value 77 is determined, the control unit 11 proceeds to step S107 in the above processing procedure.

Further, in this modification, the control unit 11 detects the end position of the web 60 acquired from the edge sensor 65 by step S102 while the control unit 11 controls the operation of the web transport device 6 in step S109. In addition to 71, the command value 77 determined by the above processing is collected as actual data 79A. In step S111, the control unit 11 updates the prediction model M2 based on the actual data 79A. The method of updating the prediction model M2 may be the same as that of the above embodiment. The control unit 11 may execute the processing other than these in the same manner as in the above embodiment. Thereby, in this modification, the predictive control of the correction mechanism 64 can be realized as in the above embodiment.

<4.3>
In the above-described embodiment and modification, the predicted value (72, 72A) is calculated by the prediction model (M1, M2), and the correction value (75, 75A) is calculated according to the calculated predicted value (72, 72A). is doing. That is, the prediction model (M1, M2) is configured to indirectly derive the correction value (75, 75A). However, the process of deriving the correction value (75, 75A) does not have to be limited to such an example. For example, the prediction model (M1, M2) may be configured to directly derive the correction value (75, 75A).

FIG. 8 schematically illustrates an example of a predictive control processing process using the predictive model M3 according to the present modification in the software configuration of the control device 1 according to the above embodiment. In this modification, the prediction model M3 is configured to calculate a correction value 75 with respect to the target value 70 of the end position from the acquired detection value 71. Similar to the above embodiment, the prediction model M3 outputs a correction value 75 for the target value 70 of the end position in the future control cycle for the input of the value of the end position in each of the plurality of control cycles up to now. It may be configured to do so.

In the generation of the prediction model M3, in addition to the time-series data of the detected value of the end position, the data indicating the corresponding correction value is used. In the learning stage, the correction value for the target value of the end position may be appropriately determined so that the web transport device 6 appropriately transports the web 60 on the actual machine or in the simulation. In analysis methods such as regression analysis and machine learning using the error back propagation method, the time series data of the detected values are used as input data (training data), and the corresponding correction values are used as teacher data (correct answer data). This makes it possible to generate a prediction model M3 trained to output a correction value of the target value for the input of the value of the end position. The training result data 125B is generated as information indicating the trained prediction model M3. In addition, except for these points, the prediction model M3 may be the same as the prediction model M1.

By using this prediction model M3, the control device 1 can directly derive the correction value 75 in the steps S104 and S105. That is, the control unit 11 operates as the calculation unit 112, and sets the trained prediction model M3 with reference to the learning result data 125B. Then, the control unit 11 inputs the detection value 71 into the set prediction model M3, and executes the arithmetic processing of the prediction model M3. As a result, the control unit 11 can calculate the correction value 75 with respect to the target value 70 of the end position from the acquired detection value 71 by using the prediction model M3.

In this modification, the control unit 11 can acquire the detection value 71 in each of the plurality of control cycles, as in the above embodiment. The control unit 11 inputs each of the acquired detection values 71 into the prediction model M3, and executes the arithmetic processing of the prediction model M3. As a result, the control unit 11 can acquire the output value output from the prediction model M3 as the correction value 75 in the future control cycle. The control unit 11 obtains the corrected target value 76 by correcting the target value 70 at the end position to the calculated correction value 75.

Other processing procedures in predictive control are the same as those in the above embodiment. In step S106, the control unit 11 operates as the determination unit 113, and determines the command value 77 for the correction mechanism 64 according to the target value 76 of the corrected end position. In step S107, the control unit 11 operates as the operation control unit 114, and controls the operation of the correction mechanism 64 so as to correct the end position of the web 60 based on the determined command value 77. In this modification, in step S106, the control unit 11 can determine the command value 77 for the correction mechanism 64 according to the corrected target value 76 and the detection value 71. In step S107, the control unit 11 can control the operation of the correction mechanism 64 based on the determined command value 77.

Further, in this modification, the control unit 11 detects the end position of the web 60 acquired from the edge sensor 65 by step S102 while the control unit 11 controls the operation of the web transport device 6 in step S109. In addition to 71, the correction value 75 calculated by the above processing is collected as the actual data 79B. In step S111, the control unit 11 updates the prediction model M3 based on the actual data 79B. The method of updating the prediction model M3 may be the same as that of the above embodiment. The control unit 11 may execute the processing other than these in the same manner as in the above embodiment. Thereby, in this modification, the predictive control of the correction mechanism 64 can be realized as in the above embodiment.

FIG. 9 schematically illustrates an example of a predictive control processing process using the predictive model M4 according to the present modification in the software configuration of the control device 1 according to the above embodiment. In this modification, the prediction model M4 is configured to calculate a correction value 75A with respect to the reference value 73 of the command value from the acquired detection value 71. Similar to the above-described embodiment and modification, the prediction model M4 has a correction value 75A with respect to the reference value 73 of the command value in the future control cycle for the input of the end position value in each of the plurality of control cycles up to now. May be configured to output.

Similar to the above modification, data indicating the corresponding correction value is used in addition to the time series data of the detected value of the end position for the generation of the prediction model M4. In the learning stage, the correction value of the command value with respect to the reference value may be appropriately determined so that the web transport device 6 appropriately transports the web 60 on the actual machine or in the simulation. In analysis methods such as regression analysis and machine learning using the error back propagation method, the time series data of the detected values are used as input data (training data), and the corresponding correction values are used as teacher data (correct answer data). As a result, it is possible to generate a prediction model M4 trained to output a correction value of a reference value for an input of a value at the end position. The training result data 125C is generated as information indicating the trained prediction model M4. In addition, except for these points, the prediction model M4 may be the same as the prediction model M2.

By using this prediction model M4, the control device 1 can directly derive the correction value 75A in the processing procedure of the modified example using the prediction model M2. That is, the control unit 11 operates as the calculation unit 112, and sets the trained prediction model M4 with reference to the learning result data 125C. Then, the control unit 11 inputs the detection value 71 into the set prediction model M4, and executes the arithmetic processing of the prediction model M4. As a result, the control unit 11 can calculate the correction value 75A with respect to the reference value 73 of the command value from the acquired detection value 71 by using the prediction model M4.

In this modification, the control unit 11 can acquire the detection value 71 in each of the plurality of control cycles, as in the above modification. The control unit 11 inputs each of the acquired detection values 71 into the prediction model M4, and executes the arithmetic processing of the prediction model M4. As a result, the control unit 11 can acquire the output value output from the prediction model M4 as the correction value 75A in the future control cycle.

Other processing procedures in the prediction control are the same as the modification using the prediction model M2. The control unit 11 operates as a determination unit 113, and corrects the reference value 73 with the calculated correction value 75A to determine the command value 77 for the correction mechanism 64. The control unit 11 operates as an operation control unit 114, and controls the operation of the correction mechanism 64 so as to correct the end position of the web 60 based on the determined command value 77. In this modification, the control unit 11 can calculate the reference value 73 of the command value according to the target value 70 and the detection value 71 of the end position. The control unit 11 can determine the command value 77 for the correction mechanism 64 by correcting the calculated reference value 73 with the correction value 75A. The control unit 11 can control the operation of the correction mechanism 64 based on the determined command value 77.

Further, in this modification, the control unit 11 detects the end position of the web 60 acquired from the edge sensor 65 by step S102 while the control unit 11 controls the operation of the web transport device 6 in step S109. In addition to 71, the correction value 75A calculated by the above processing is collected as the actual data 79C. In step S111, the control unit 11 updates the prediction model M4 based on the actual data 79C. The method for updating the prediction model M4 may be the same as the above modification. Thereby, in this modification, the prediction control of the correction mechanism 64 can be realized as in the modification using the prediction model M2.

<4.4>
In the above-described embodiment and modification, the generation and update of the prediction models M1 to M4 may be executed by a computer different from the control device 1. This other computer may be referred to as a model generator. In the above embodiments and modifications, when updating the prediction models M1 to M4 on another computer, or when updating the prediction models M1 to M4 is omitted, the processing of steps S109 to S111 is performed in the above processing procedure. May be omitted. Further, the data acquisition unit 115 and the model update unit 116 may be omitted from the software configuration of the control device 1.

FIG. 10 schematically illustrates an example of a scene in which the model generator 4 is applied. The model generator 4 is a computer configured to generate a prediction model M1 from time series data. The control system may be configured by the control device 1 and the model generation device 4. In the example of FIG. 10, the control device 1 and the model generation device 4 are connected to each other via a network. In this case, the control device 1 further includes a communication interface (not shown) in the hardware configuration. The type of network may be appropriately selected from, for example, the Internet, a wireless communication network, a mobile communication network, a telephone network, a dedicated network, and the like. However, the method of exchanging data between the control device 1 and the model generation device 4 does not have to be limited to such an example, and may be appropriately selected depending on the embodiment. For example, data may be exchanged between the control device 1 and the model generation device 4 by using a storage medium.

[Hardware configuration]
FIG. 11 schematically illustrates an example of the hardware configuration of the model generator 4 according to this modification. As shown in FIG. 11, in the model generation device 4 according to this modification, the control unit 41, the storage unit 42, the external interface 43, the input device 44, the output device 45, the drive 46, and the communication interface 47 are electrically connected. It is a computer that has been used. The control units 41 to drive 46 of the model generation device 4 may be configured in the same manner as the control units 11 to drive 16 of the control device 1. In FIG. 11, the communication interface is described as "communication I / F".

The storage unit 42 of the model generation device 4 according to this modification stores various information such as the generation program 84, the time series data 421, and the learning result data 125. The generation program 84 is a program for causing the model generation device 4 to execute information processing (FIG. 13) described later regarding the generation of the prediction model M1. The generation program 84 includes a series of instructions for the information processing. The time-series data 421 shows the transition of the detected value (measured value or calculated value) of the end position of the web 60 while the web transport device 6 appropriately transports the web 60. The training result data 125 shows information about the trained prediction model M1. The learning result data 125 is obtained as a result of executing the generation program 84.

At least one of the generation program 84 and the time series data 421 may be stored in the storage medium 94. Further, the model generation device 4 may acquire at least one of the generation program 84 and the time series data 421 from the storage medium 94. The type of the storage medium 94 may be a disc type or may be other than the disc type.

The communication interface 47 is, for example, a wired LAN (Local Area Network) module, a wireless LAN module, or the like, and is an interface for performing wired or wireless communication via a network. By using this communication interface 47, the model generation device 4 can perform data communication via the network with another computer (for example, the control device 1).

Regarding the specific hardware configuration of the model generation device 4, components can be omitted, replaced, or added as appropriate according to the embodiment. For example, the control unit 41 may include a plurality of hardware processors. The hardware processor may be composed of a microprocessor, FPGA, DSP and the like. The storage unit 42 may be composed of a RAM and a ROM included in the control unit 41. The storage unit 42 may be configured by an auxiliary storage device such as a hard disk drive or a solid state drive. At least one of the external interface 43, the input device 44, the output device 45, the drive 46, and the communication interface 47 may be omitted. The model generator 4 may be composed of a plurality of computers. In this case, the hardware configurations of the computers may or may not match. Further, the model generation device 4 may be a general-purpose server device, a general-purpose PC, or the like, in addition to an information processing device designed exclusively for the provided service.

[Software configuration]
FIG. 12 schematically illustrates an example of the software configuration of the model generator 4 according to this modification. Similar to the control device 1, the software configuration of the model generation device 4 is realized by executing the generation program 84 by the control unit 41. That is, the control unit 41 expands the generation program 84 in the RAM, and executes the expanded generation program 84 by the CPU. As a result, as shown in FIG. 12, the model generation device 4 according to this modification operates as a computer including a data acquisition unit 411, a model generation unit 412, and a storage processing unit 413 as software modules. That is, in this modification, each software module is realized by the control unit 41 (CPU).

The data acquisition unit 411 acquires time series data 421. The model generation unit 412 generates (updates) the prediction model M1 based on the time series data 421. The storage processing unit 413 generates information about the generated (updated) prediction model M1 as learning result data 125, and stores the generated learning result data 125 in a predetermined storage area.

In this modification, an example in which each software module of the model generator 4 is realized by a general-purpose CPU is described. However, some or all of the above software modules may be implemented by one or more dedicated processors. Further, with respect to the software configuration of the model generator 4, software modules may be omitted, replaced, or added as appropriate according to the embodiment.

[Operation example]
Next, an operation example of the model generation device 4 according to the present modification will be described with reference to FIG. FIG. 13 is a flowchart showing an example of the processing procedure of the model generation device 4 according to this modification. However, the processing procedure described below is only an example, and each processing may be changed as much as possible. Further, with respect to the processing procedure described below, steps can be omitted, replaced, and added as appropriate according to the embodiment. The order of processing without dependencies may be changed as appropriate.

(Step S401)
In step S401, the control unit 41 operates as a data acquisition unit 411, and the detected value (measured value or calculated value) of the end position of the web 60 while the web transport device 6 appropriately transports the web 60. Show the transition. The time-series data 421 may be obtained by operating the actual machine or may be obtained by simulation.

The time series data 421 may be generated by the model generation device 4 or may be generated by a computer other than the model generation device 4. When generating the time series data 421, the model generation device 4 may be connected to the web transfer device 6 or a web transfer device of the same type thereof via the external interface 43. The control unit 41 can generate time-series data 421 by operating the actual machine of the web transfer device 6 or simulating the operation of the web transfer device 6. As a result, the time series data 421 may be acquired. Alternatively, when the time-series data 421 is generated by another computer, the control unit 41 directly or indirectly from the other computer via the network, the storage medium 94, the external storage device, or the like, the time-series data 421. May be obtained.

When the prediction model M1 is configured to further accept the input of data other than the value of the end position, the control unit 41 may further acquire a sample of the other data. Further, when the prediction model M1 is configured to accept the input of one or a plurality of feature quantities calculated from the detected value of the end position, the detected value of the end position indicated by the time series data 421 is converted into the feature quantity. May be. Alternatively, the control unit 41 may convert the detected value of the end position indicated by the time series data 421 into a feature amount. When the time series data 421 is acquired, the control unit 41 proceeds to the next step S402.

(Step S402)
In step S402, the control unit 41 operates as the model generation unit 412 and generates the prediction model M1 by analyzing the time series data 421. As described above, as a method for analyzing the time series data 421, a known method such as regression analysis or machine learning using an error backpropagation method may be used. When the prediction model M1 is generated, the control unit 41 proceeds to the next step S403.

(Step S403)
In step S403, the control unit 41 operates as the storage processing unit 413 and generates information about the generated prediction model M1 as learning result data 125. Then, the control unit 41 stores the generated learning result data 125 in a predetermined storage area.

The predetermined storage area may be, for example, a RAM in the control unit 41, a storage unit 42, a storage medium 94, an external storage device, or a combination thereof. The external storage device may be, for example, a data server such as NAS. In this case, the control unit 41 may store the learning result data 125 in the data server via the network by using the communication interface 47. Further, the external storage device may be, for example, an external storage device connected to the model generation device 4. When the learning result data 125 is saved, the control unit 41 ends the process related to this operation example.

The generated learning result data 125 may be provided to the control device 1 at any timing. For example, the control unit 41 may transfer the learning result data 125 to the control device 1 as the process of step S403 or separately from the process of step S403. The control device 1 may acquire the learning result data 125 by accepting this transfer. Further, for example, the control device 1 may acquire the learning result data 125 by accessing the model generation device 4 or the data server via the network. Further, for example, the control device 1 may acquire the learning result data 125 via the storage medium 91. The learning result data 125 may be incorporated in the control device 1 in advance.

Further, the model generation device 4 may update the prediction model M1 by using the actual data 79 collected by the control device 1 together with the time series data 421 or instead of the time series data 421. The updated prediction model M1 (learning result data 125) may be provided to the control device 1 at an arbitrary timing.

Further, the model generation device 4 may be similarly applied to each of the above-mentioned modifications. That is, the model generator 4 may be configured to perform at least one of the generation and update of the prediction models M2 to M4. According to the modification, the processing load of the control device 1 can be reduced by the model generation device 4 being in charge of information processing related to the generation and update of the prediction models M1 to M4.

<4.5>
The end position of the web 60 in the above embodiment is an example of the "position in the width direction of the web" of the present invention. However, the reference of the position in the width direction does not have to be limited to such an end portion, and may be set arbitrarily. For example, the position in the width direction of the web may be detected using the entire width of the web as a reference. Further, for example, the position in the width direction of the web may be detected by using the center in the width direction of the web as a reference. Further, the edge sensor 65 is an example of the "sensor" of the present invention. However, the sensor configured to detect the position of the web in the width direction is not limited to this edge sensor. As another example of the sensor, for example, a laser sensor, an image sensor, a capacitance sensor, a pressure sensor and the like may be used.

Further, in the above embodiment, the corrected target value 76 is acquired by adding the calculated correction value 75 to the target value 70 in step S105. Directly correcting the target value 70 is an example of "correcting the target value" of the present invention. However, correcting the target value does not have to be limited to such an example. In the above embodiment, the command value 77 is determined according to the corrected target value 76 and the detected value 71. Therefore, correcting the target value may include indirectly correcting the target value 70 by correcting the detected value 71. That is, the correction of the target value 70 may be achieved by correcting the target value 70 relative to the detected value 71. The correction amount of the detection value 71 may be determined so that the correction mechanism 64 can be operated with the behavior desired for the original target value 70, similarly to the correction amount of the target value 70. Similarly, in the above modification, the correction of the target value 70 indirectly corrects the target value 70 with respect to the detected value 71 by applying the correction value 75 calculated by the prediction model M3 to the detected value 71. It may be achieved by that.

1 ... Control device,
11 ... Control unit, 12 ... Storage unit, 13 ... External interface,
14 ... input device, 15 ... output device, 16 ... drive,
111 ... acquisition unit, 112 ... calculation unit, 113 ... determination unit,
114 ... Operation control unit,
115 ... Data collection department, 116 ... Model update department,
125 ... Learning result data,
81 ... control program, 91 ... storage medium,
6 ... Web transfer device,
60 ... Web,
61 ... Supply roll, 611 ... Servo motor,
62.63 ... Driven roller,
64 ... Correction mechanism,
641 ... Roller, 642 ... Shaft, 643 ... Actuator,
65 ... Edge sensor,
66 ... Recovery roll, 661 ... Servo motor,
70 ... target value, 71 ... detection value,
72 ... predicted value, 75 ... corrected value,
76 ... target value, 77 ... command value,
79 ... Actual data,
M1 ... Predictive model,
4 ... Model generator,
41 ... Control unit, 42 ... Storage unit, 43 ... Communication interface,
44 ... Input device, 45 ... Output device, 46 ... Drive,
411 ... Data acquisition unit, 412 ... Model generation unit,
413 ... Preservation processing unit,
84 ... Generation program, 94 ... Storage medium

Claims (8)

A web carrier configured to transport a web, a sensor configured to detect the widthwise position of the web being transported, and a widthwise position of the web being transported. A control device that controls the operation of a web transport device equipped with a correction mechanism configured as such.
An acquisition unit that acquires the detection value of the position in the width direction of the web from the sensor, and
A calculation unit that calculates at least one of the future detection value of the position in the width direction and the future command value for the correction mechanism from the acquired detection value using the prediction model.
A determination unit that determines a command value for the correction mechanism according to the calculated predicted value, and
An operation control unit that controls the operation of the correction mechanism so as to correct the position of the web in the width direction based on the determined command value.
Equipped with
The prediction model predicts at least one of the detection value of the width direction position in the future control cycle and the command value for the correction mechanism for the input of the width direction position value in each of the plurality of control cycles up to now. Configured to output a value,
The acquisition unit acquires the detection value in each of the plurality of control cycles, and obtains the detection value.
The calculation unit inputs each of the acquired detected values to the prediction model, executes arithmetic processing of the prediction model, and uses the output value output from the prediction model as the prediction value in a future control cycle. Acquired,
The determination unit determines a command value for the correction mechanism according to the acquired predicted value.
The motion control unit controls the motion of the correction mechanism based on the determined command value.
Control device. The prediction model is configured to output a predicted value of a detected value of a position in the width direction in a future control cycle in response to an input of a value of a position in the width direction in each of a plurality of control cycles up to now.
The calculation unit inputs each of the acquired detected values to the prediction model, executes arithmetic processing of the prediction model, and outputs a value output from the prediction model in the width direction in a future control cycle. Obtained as a predicted value of the detected value of the position,
The determination unit corrects the target value at the position in the width direction according to the acquired predicted value, and determines the command value for the correction mechanism according to the corrected target value and the detected value.
The motion control unit controls the motion of the correction mechanism based on the determined command value.
The control device according to claim 1 . The prediction model is configured to output a predicted value of a command value for the correction mechanism in a future control cycle in response to an input of a position value in the width direction in each of a plurality of control cycles up to now.
The calculation unit inputs each of the acquired detected values to the prediction model, executes arithmetic processing of the prediction model, and outputs a value output from the prediction model to the correction mechanism in a future control cycle. Obtained as a predicted value of the command value,
The determination unit calculates a reference value of a command value for the correction mechanism according to a target value at a position in the width direction and the detection value, and determines the calculated reference value according to the acquired predicted value. By correcting, the command value is determined.
The motion control unit controls the motion of the correction mechanism based on the determined command value.
The control device according to claim 1 . Entering each of the detection values into the prediction model
Converting each of the acquired detected values into one or more feature quantities, and inputting the obtained one or more feature quantities into the prediction model.
Consists of,
The control device according to any one of claims 1 to 3 . A data collection unit that collects at least one of the detected value at the position in the width direction acquired from the sensor and the determined command value as actual data while controlling the operation of the web transfer device.
A model update unit that updates the prediction model based on the collected actual data,
Further prepare,
The control device according to any one of claims 1 to 4 . A web carrier configured to transport a web, a sensor configured to detect the widthwise position of the web being transported, and a widthwise position of the web being transported. A control device that controls the operation of a web transport device equipped with a correction mechanism configured as such.
An acquisition unit that acquires the detection value of the position in the width direction of the web from the sensor, and
A calculation unit that calculates a correction value for the reference value of the command value or the target value of the position in the width direction from the acquired detection value using the prediction model.
A determination unit that determines a command value for the correction mechanism by correcting the reference value with the calculated correction value or according to the target value of the position in the width direction corrected by the calculated correction value.
An operation control unit that controls the operation of the correction mechanism so as to correct the position of the web in the width direction based on the determined command value.
Equipped with
The prediction model outputs a correction value for a reference value of a command value in a future control cycle or a target value for a position in the width direction for input of a position value in the width direction in each of a plurality of control cycles up to now. Is configured to
The acquisition unit acquires the detection value in each of the plurality of control cycles, and obtains the detection value.
The calculation unit inputs each of the acquired detected values to the prediction model, executes the arithmetic processing of the prediction model, and uses the output value output from the prediction model as the correction value in the future control cycle. Acquired,
The determination unit calculates a reference value of the command value according to the target value and the detection value at a position in the width direction, and corrects the calculated reference value with the correction value, or is acquired. The command value for the correction mechanism is determined according to the target value and the detection value corrected by the correction value.
The motion control unit controls the motion of the correction mechanism based on the determined command value.
Control device. A web transport device configured to transport a web, a sensor configured to detect the widthwise position of the web being transported, and a widthwise position of the web being transported. It is a control method for controlling the operation of a web transport device having a correction mechanism configured as described above.
The computer
The step of acquiring the detection value of the position in the width direction of the web from the sensor, and
Using the prediction model, the step of calculating at least one of the future detection value of the position in the width direction and the future command value for the correction mechanism from the acquired detection value, and
A step of determining a command value for the correction mechanism according to the calculated predicted value, and
A step of controlling the operation of the correction mechanism so as to correct the position in the width direction of the web based on the determined command value.
And run
The prediction model predicts at least one of the detection value of the width direction position in the future control cycle and the command value for the correction mechanism for the input of the width direction position value in each of the plurality of control cycles up to now. Configured to output a value,
The acquisition step is configured by acquiring the detection value in each of the plurality of control cycles.
In the calculation step, each of the acquired detected values is input to the prediction model, and the output value output from the prediction model by executing the arithmetic processing of the prediction model is the predicted value in the future control cycle. Consists of getting as
The determination step is configured by determining a command value for the correction mechanism according to the acquired predicted value.
The control step is configured by controlling the operation of the correction mechanism based on the determined command value.
Control method. A web transport device configured to transport a web, a sensor configured to detect the widthwise position of the web being transported, and a widthwise position of the web being transported. A control program for controlling the operation of a web carrier equipped with a correction mechanism configured as such.
On the computer
The step of acquiring the detection value of the position in the width direction of the web from the sensor, and
Using the prediction model, the step of calculating at least one of the future detection value of the position in the width direction and the future command value for the correction mechanism from the acquired detection value, and
A step of determining a command value for the correction mechanism according to the calculated predicted value, and
A step of controlling the operation of the correction mechanism so as to correct the position in the width direction of the web based on the determined command value.
Is a control program for executing
The prediction model predicts at least one of the detection value of the width direction position in the future control cycle and the command value for the correction mechanism for the input of the width direction position value in each of the plurality of control cycles up to now. Configured to output a value,
The acquisition step is configured by acquiring the detection value in each of the plurality of control cycles.
In the calculation step, each of the acquired detected values is input to the prediction model, and the output value output from the prediction model by executing the arithmetic processing of the prediction model is the predicted value in the future control cycle. Consists of getting as
The determination step is configured by determining a command value for the correction mechanism according to the acquired predicted value.
The control step is configured by controlling the operation of the correction mechanism based on the determined command value.
Control program.

Images