Control device, winding device, control system constructing method, control system, control method, program, and recording medium

Abstract

To provide a control device, a winding device, a control system constructing method, a control system, a control method, a program, and a storage medium that are able to bring the amount of control from a control object closer to a target value.SOLUTION: A control device according to an embodiment controls a winding mechanism capable of winding an object. The control device includes a first control unit, a first calculation unit, and a second calculation unit. The first control unit acquires, from the winding mechanism, the amount of control, which is to be controlled. The first control unit acquires a first amount of operation corresponding to the amount of operation in a case where the amount of control is output from the winding mechanism by inputting the amount of control to a model that has been machine-learned in advance. The first calculation unit calculates a difference between the first amount of operation and a second amount of operation transmitted to the winding mechanism. The second calculation unit inputs, to the winding mechanism, a third amount of operation obtained by adding the difference to the second amount of operation.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] An embodiment of the present invention relates to a control device, a winding device, a control system construction method, a control system, a control method, a program, and a storage medium. [Background technology]

[0002] There is a device that controls a controlled object. For this control device, there is a demand for a technology that can bring the controlled amount from the controlled object closer to a target value. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6813231 Summary of the Invention [Problem to be solved by the invention]

[0004] The problem that the present invention aims to solve is to provide a control device, a winding device, a control system construction method, a control system, a control method, a program, and a storage medium that are capable of bringing the control amount from the controlled object closer to a target value. [Means for solving the problem]

[0005] A control device according to an embodiment controls a winding mechanism capable of winding up an object. The control device includes a first control unit, a first calculation unit, and a second calculation unit. The first control unit acquires a control amount, which is an object to be controlled, from the winding mechanism. The first control unit inputs the control amount into a model that has been machine-learned in advance, to acquire a first operation amount that corresponds to an operation amount when the control amount is output from the winding mechanism. The first calculation unit calculates a difference between the first operation amount and a second operation amount transmitted to the winding mechanism. The second calculation unit inputs a third operation amount, which is the second operation amount plus the difference, to the winding mechanism. [Brief description of the drawings]

[0006] [Figure 1] FIG. 1 is a block diagram showing the configuration of a winding device according to an embodiment. [Diagram 2] FIG. 2 is a perspective view showing the winding device according to the embodiment. [Diagram 3] FIG. 3 is a block diagram showing the configuration of a winding device 1r according to a reference example. [Figure 4] 4(a) and 4(b) show data indicating characteristics of the control system according to the embodiment. [Diagram 5] FIG. 5 is a block diagram showing a configuration of a winding device according to a modified example of the embodiment. [Figure 6] FIG. 6 is a schematic diagram showing a hardware configuration. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0007] Hereinafter, each embodiment of the present invention will be described with reference to the drawings. The drawings are schematic or conceptual, and the relationship between the thickness and width of each part, the size ratio between parts, etc. are not necessarily the same as the actual ones. Even when the same part is shown, the dimensions and ratios of each part may be shown differently depending on the drawing. In this specification and each drawing, elements similar to those already explained are given the same reference numerals, and detailed explanations are omitted as appropriate.

[0008] FIG. 1 is a block diagram showing the configuration of a winding device according to an embodiment. As shown in Fig. 1, a winding device 1 according to an embodiment includes a control device 10 and a winding mechanism 20. The winding mechanism 20 is capable of winding up an object. The object is a foil, a wire, or the like. The control device 10 controls the operation of the winding mechanism 20.

[0009] The control device 10 includes a first control unit 11, a second control unit 12, a first calculation unit 13, a second calculation unit 14, and a calculation unit 15. The first control unit 11 acquires a control amount y (output value) from the winding mechanism 20. The "control amount" refers to an amount that is an object to be controlled among amounts that belong to a control target. When the first control unit 11 acquires the control amount y, it refers to a model 16. The model 16 is machine-trained in advance so that, when a control amount is input, it estimates an operation amount corresponding to the control amount. The "operation amount" refers to an amount that is applied to a control target in a control system to control the control amount. The definitions of the control amount and the operation amount can be those defined in JIS Z8116-1994.

[0010] For example, the model 16 includes a decision tree or a neural network. The model 16 may include a random forest as a decision tree. If a neural network is used, the model 16 preferably includes a recurrent neural network (RNN). The RNN may include a Long Short-Term Memory structure.

[0011] The first control unit 11 inputs the control amount y to the model 16 and obtains a first operation amount u1 estimated by the model 16. The first operation amount u1 is an operation amount corresponding to the control amount y. The control amount y output from the winding mechanism 20 corresponds to the control amount when the first operation amount u1 is input to the winding mechanism 20. In other words, the first operation amount u1 corresponds to the operation amount when the control amount y is output from the winding mechanism 20.

[0012] The calculation unit 15 receives a target value r for the operation of the winding mechanism 20 from a higher-level system. A "target value" refers to a quantity given as a target so that a controlled variable takes that value in a control system. The definition of the target value can also be that defined in JIS Z8116-1994. The calculation unit 15 calculates an error e between the target value r and the controlled variable y. The second control unit 12 calculates a second manipulated variable u2 to compensate for the error e, and transmits it to the winding mechanism 20.

[0013] When the first control unit 11 estimates the first manipulated variable u1, a target value r may be used in addition to the controlled variable y. In this case, the model 16 is machine-trained in advance to output the manipulated variable in response to the input of the controlled variable and the target value. The first control unit 11 inputs the controlled variable y and the target value r to the model 16 to obtain the first manipulated variable u1.

[0014] The first calculation unit 13 acquires the first operation amount u1 and the second operation amount u2, and calculates a difference d1 between the first operation amount u1 and the second operation amount u2. The difference d1 corresponds to a disturbance estimated based on a control amount from the winding mechanism 20.

[0015] The second calculation unit 14 calculates a third operation amount u3 by adding the difference d1 to the second operation amount u2. The second calculation unit 14 inputs the third operation amount u3 to the winding mechanism 20. The winding mechanism 20 operates in accordance with the third operation amount u3.

[0016] When a disturbance d2 is applied to the winding mechanism 20, the controlled variable y changes in response to the disturbance d2. In response to the change in the controlled variable y, the first manipulated variable u1 also changes, and as a result, the third manipulated variable u3 also changes.

[0017] As a specific example, the control amount y decreases due to the disturbance d2. In response to the decrease in the control amount y, the first operation amount u1 also decreases. The difference d1 between the first operation amount u1 and the second operation amount u2 increases. The difference d1 is added to the second operation amount u2. As a result, the operation amount for the winding mechanism 20 is corrected in response to the disturbance d2 so that the target value is output.

[0018] FIG. 2 is a perspective view showing the winding device according to the embodiment. For example, as shown in Fig. 2, the winding mechanism 20 includes a first roller 21, a second roller 22, an intermediate roller 23, a sensor 24, a movable part 25, and a movable part 26. In the example shown in Fig. 2, the object O is a foil (thin film).

[0019] The first roller 21 winds up the object O by rotation. Specifically, the first roller 21 includes a frame body 21a, a rotating shaft 21b, and a motor 21c. The frame body 21a is a member around which the object O is wound. In the illustrated example, the frame body 21a is circular, but the specific shape of the frame body 21a is arbitrary. The rotating shaft 21b is attached to the center of the frame body 21a. The motor 21c rotates the rotating shaft 21b. The operation of the motor 21c causes the frame body 21a to rotate around the rotating shaft 21b.

[0020] The object O is wound around the second roller 22 in advance. The second roller 22 sends the object O to the first roller 21 by rotating. The second roller 22 can have the same configuration as the first roller 21. For example, the second roller 22 includes a frame body 22a, a rotating shaft 22b, and a motor 22c. The frame body 22a is a member around which the object O is wound in advance. The frame body 22a may be circular or may have another shape. The rotating shaft 22b is attached to the center of the frame body 22a. In the illustrated example, the rotating shaft 21b and the rotating shaft 22b are parallel to each other. The motor 22c rotates the rotating shaft 22b. The frame body 22a rotates around the rotating shaft 22b by the operation of the motor 22c.

[0021] For the sake of explanation, the direction parallel to rotation axis 21b or rotation axis 22b is referred to as the "X direction." A direction intersecting with the X direction is referred to as the "Y direction." A direction intersecting with the XY plane is referred to as the "Z direction."

[0022] While the object O travels from the second roller 22 to the first roller 21, the object O comes into contact with a number of intermediate rollers 23. Each intermediate roller 23 is rotatable around the X direction. Each intermediate roller 23 changes the transport direction of the object O and applies an appropriate tension to the object O. In the illustrated example, ten intermediate rollers 23, intermediate rollers 23a to 23j, are provided between the object O and the first roller 21 and the second roller 22.

[0023] The sensor 24 detects parameters related to the object O. For example, the sensor 24 includes an encoder 24a, a force sensor 24b, a force sensor 24c, an optical sensor 24d, an optical sensor 24e, and a camera (image sensor) 24f.

[0024] The encoder 24a detects the rotation speed of the intermediate roller 23a. Specifically, the encoder 24a is an optical or magnetic encoder, and counts optical or magnetic signals detected in response to the rotation of the intermediate roller 23a. The encoder 24a calculates the rotation speed of the intermediate roller 23a from the number of rotations of the intermediate roller 23a per unit time. The rotation speed of the intermediate roller 23a corresponds to the conveying speed of the object O.

[0025] The force sensor 24b is attached to the intermediate roller 23b. A force corresponding to the tension of the object O is applied to the intermediate roller 23b in a direction intersecting the X direction. The force sensor 24b detects the force applied to the intermediate roller 23b as the tension of the object O.

[0026] The force sensor 24c is attached to the intermediate roller 23i. A force corresponding to the tension of the object O is applied to the intermediate roller 23i in a direction intersecting the X direction. The force sensor 24c detects the force applied to the intermediate roller 23i as the tension of the object O.

[0027] Optical sensors 24d and 24e detect the position of object O in a direction perpendicular to the conveying direction. In the illustrated example, optical sensor 24d detects the position of object O in the X direction between intermediate rollers 23c and 23d. Optical sensor 24e detects the position of object O in the X direction between intermediate rollers 23g and 23h.

[0028] The camera 24f detects the amount of positional deviation of the object O in a direction perpendicular to the conveying direction and the amount of angular deviation of the object O in the conveying direction. For example, the end of the foil in the perpendicular direction is detected from an image acquired by the camera 24f. The distance between the position of the end in the image and a preset position is the amount of positional deviation. Also, the angle between a preset line and a line indicating the end (edge) of the foil is the amount of angular deviation. The preset line is set parallel to the direction connecting the adjacent rollers. In the illustrated example, the camera 24f detects the amount of positional deviation and the inclination of the object O between the intermediate rollers 23h and 23i.

[0029] The movable part 25 includes an actuator and can drive the first roller 21 in the X direction. The operation of the movable part 25 changes the position in the X direction of the object O wound around the first roller 21. In the illustrated example, the movable part 25 includes a base frame 25a and a stage 25b. The base frame 25a is a plate-like member parallel to the YZ plane. The stage 25b is a plate-like member parallel to the XY plane. The stage 25b is movable in the X direction and supports the base frame 25a from below. The first roller 21, the intermediate rollers 23f to 23j, the force sensor 24c, the optical sensor 24e, and the camera 24f are attached to the base frame 25a. The movement of the stage 25b in the X direction changes the position in the X direction of each element attached to the base frame 25a.

[0030] The movable unit 26 includes an actuator and can drive the second roller 22 in the X direction. The operation of the movable unit 26 changes the position in the X direction of the object O sent out from the second roller 22. The movable unit 26 can have the same configuration as the movable unit 25. For example, the movable unit 26 includes a base frame 26a and a stage 26b. The base frame 26a is a plate-like member parallel to the YZ plane. The stage 26b is a plate-like member parallel to the XY plane. The stage 26b is movable in the X direction and supports the base frame 26a from below. The second roller 22, the intermediate rollers 23a to 23e, the encoder 24a, the force sensor 24b, and the optical sensor 24d are attached to the base frame 26a. The movement of the stage 26b in the X direction changes the position in the X direction of each element attached to the base frame 26a.

[0031] The winding mechanism 20 shown in Fig. 2 outputs values ​​detected by the sensors 24a to 24f. That is, the control amount y shown in Fig. 1 includes the conveying speed detected by the encoder 24a, the tension detected by the force sensor 24b, the tension detected by the force sensor 24c, the position detected by the optical sensor 24d, the position detected by the optical sensor 24e, and the positional deviation amount and inclination detected by the camera 24f. The control device 10 receives these control amounts.

[0032] The control device 10 also calculates operation amounts for the motor 21c, the motor 22c, the movable part 25, and the movable part 26. That is, the first operation amount u1 to the third operation amount u3 include the rotation speed of the motor 21c, the rotation speed of the motor 22c, the position of the first roller 21 (the movable part 25), and the position of the second roller 22 (the movable part 26).

[0033] The model 16 used to estimate the first manipulated variable u1 is machine-learned in advance. In the machine learning, a plurality of learning data sets are used. Each learning data set includes a controlled variable and a manipulated variable. The model 16 is repeatedly trained to output a manipulated variable in response to an input of a controlled variable. For example, the learning data set is collected by operating the controlled object in a state where there is little or substantially no disturbance to the controlled object. The model 16 is trained using the collected learning data set.

[0034] One advantage of the embodiment will now be described. During operation of the winding mechanism 20, disturbance d2 may occur due to various factors. For example, disturbance d2 may occur due to a change in the material of the object O, partial damage to the object O, twisting of the object O, or changes in the winding mechanism 20 over time. When disturbance d2 occurs, the conveying speed of the object O, the tension of the object O, or the position of the object O changes. When these values ​​change, the behavior of the object O deviates from the desired state, and problems such as wrinkles or twists in the foil or breakage of the foil may occur.

[0035] FIG. 3 is a block diagram showing the configuration of a winding device 1r according to a reference example. The winding device 1r shown in FIG. 3 includes a control device 10r instead of the control device 10. The control device 10r does not include the first control unit 11, the first calculation unit 13, and the second calculation unit 14, as compared with the control device 10. In the winding device 1r, a second operation amount u2 corresponding to the target value r and the control amount y is transmitted from the second control unit 12 to the winding mechanism 20. This controls the winding mechanism 20 so that the difference between the control amount y and the target value r becomes smaller. However, when a disturbance d2 occurs, it may be difficult to bring the control amount y closer to the target value r even if the second operation amount u2 is input to the winding mechanism 20.

[0036] In order to bring the controlled variable y closer to the target value r, it is effective to estimate the first controlled variable u1, which corresponds to the actual input to the winding mechanism 20, from the controlled variable y. By estimating the first controlled variable u1, the second controlled variable u2 can be corrected and the third controlled variable u3 can be calculated. On the other hand, two or more types of controlled variables, such as the position of the first roller 21 (movable part 25), the position of the second roller 22 (movable part 26), the rotation speed of the first roller 21, and the rotation speed of the second roller 22, are input to the winding mechanism 20. In addition, two or more types of controlled variables, such as the conveying speed of the object O, the tension of the object O, and the position of the object O, are output from the winding mechanism 20. The multiple controlled variables input to the winding mechanism 20 have a nonlinear relationship with the multiple controlled variables output from the winding mechanism 20. For this reason, it is practically impossible to analyze the relationship between the controlled variable y and the first controlled variable u1 by regression analysis or the like. Therefore, with regard to the winding mechanism 20, the controlled variable y has conventionally been adjusted by inputting the second manipulated variable u2 corresponding to the error e to the winding mechanism 20.

[0037] Regarding this problem, the control device 10 according to the embodiment includes a first control unit 11, a first calculation unit 13, and a second calculation unit 14. The first control unit 11 estimates a first operation amount u1 corresponding to a control amount y using a model 16. The model 16 includes a decision tree or a neural network, and is machine-learned in advance. By using the model 16, it becomes possible to estimate the first operation amount u1, which has a nonlinear relationship with the control amount y, from the control amount y. By obtaining the first operation amount u1, it becomes possible to calculate a difference d1 according to the disturbance d2. The third operation amount u3 can be calculated using the difference d1. As a result, it becomes possible to bring the control amount y closer to the target value r.

[0038] Another advantage of the embodiment will now be described. It is preferable that the control device 10 according to the embodiment can be introduced to a control target after the start of operation. For example, it is preferable that the first control unit 11, the first calculation unit 13, and the second calculation unit 14 can be retrofitted to the winding device 1r after the operation of the winding device 1r shown in FIG. 3 is started. Regarding the learning of the model 16, the operation amount to the winding mechanism 20 and the control amount from the winding mechanism 20 are collected in a state where the operation of the winding mechanism 20 has already started. The collected data is used for learning the model 16 as a learning data set. After the learning of the model 16, the control device 10 according to the embodiment can be realized by adding the first control unit 11, the first calculation unit 13, and the second calculation unit 14 to the control device 10r.

[0039] 1, it is also possible to add a control device 10b (second control device) including a first control unit 11, a first calculation unit 13, a second calculation unit 14, and a model 16 to a control device 10a (first control device) including a second control unit 12 and a model 16. By adding a new control device 10b to a control system including an existing control device 10a, a new control system can be constructed.

[0040] The embodiment of the present invention is also applicable to a control object other than a winding device. That is, the control object by the control device 10 is arbitrary. The first control unit 11 acquires a control amount from the control object and inputs the control amount to the model 16 to acquire a first operation amount u1. The first calculation unit 13 calculates a difference d1 between the first operation amount u1 and a second operation amount u2 transmitted to the control object. The second calculation unit 14 inputs a third operation amount u3 obtained by adding the difference d1 to the second operation amount u2 to the control object. This makes it possible to bring the control amount y of the control object closer to the target value r. In addition, the control device 10 according to the embodiment can be realized by adding the control device 10b to an existing device.

[0041] The embodiment is particularly effective when the operation amount and the control amount have a nonlinear relationship. When the operation amount and the control amount have a nonlinear relationship, manual regression analysis is difficult. By using the model 16, it becomes possible to accurately estimate the operation amount based on the control amount.

[0042] More preferably, the model 16 includes an RNN, and the first control unit 11 inputs the time series data of the controlled variable to the model 16. As a result of verification by the present inventors, it has been confirmed that, when the controlled variable and the controlled variable have a nonlinear relationship, the error of the estimated controlled variable can be reduced by using the RNN and the time series data.

[0043] 4(a) and 4(b) show data indicating characteristics of the control system according to the embodiment. FIG. 4(a) shows the characteristics when the model includes a neural network other than an RNN. FIG. 4(b) shows the characteristics when the model includes an RNN. In the graphs of FIG. 4(a) and FIG. 4(b), the horizontal axis represents the number of times estimation is performed. The vertical axis represents one of the operation amounts estimated by the model. The solid line represents the operation amount estimated by the model, and the dashed line represents the correct operation amount. The control amount input to the model includes the conveying speed detected by the encoder 24a, the tension detected by the force sensor 24b, the tension detected by the force sensor 24c, the position detected by the optical sensor 24d, the position detected by the optical sensor 24e, and the positional deviation amount and inclination detected by the camera 24f. The operation amount estimated by the model includes the position of the first roller 21, the position of the second roller 22, the rotational speed of the first roller 21, and the rotational speed of the second roller 22.

[0044] The results shown in Figure 4(a) show that the model including the neural network is generally able to estimate the operation amount close to the correct answer. The root mean square deviation (RMSE) between the estimated value and the correct answer was 70.010. The mean absolute error (MAE) was 46.833. These results confirm that the model including the neural network is able to estimate the operation amount with high accuracy.

[0045] The results shown in Figure 4(b) show that the model including the RNN is able to estimate the operation amount with higher accuracy than the model including a general neural network. Specifically, when the model including the RNN is used, the RMSE was 56.164, and the MAE was 46.512. A large improvement in accuracy of about 20% was confirmed for the RMSE.

[0046] From this result, it is preferable that the model includes an RNN and that time series data is input to the model. This can further improve the accuracy of estimating the manipulated variable. By improving the accuracy of estimating the manipulated variable, the accuracy of the difference d1 corresponding to the disturbance is also improved. As a result, it is possible to make the controlled variable output from the controlled object closer to the target value.

[0047] (Modification) FIG. 5 is a block diagram showing a configuration of a winding device according to a modified example of the embodiment. 5 includes a control device 10x and a winding mechanism 20. Compared to the control device 10, the control device 10x further includes a determination unit 17. The functions of the elements in the control device 10x other than the determination unit 17 are similar to those of the elements in the control device 10.

[0048] The determination unit 17 receives the difference d1 from the first calculation unit 13. In addition, the determination unit 17 transmits the difference d1 to the second calculation unit 14. Upon receiving the difference d1, the determination unit 17 determines whether or not the winding mechanism 20 is normal based on the difference d1. For example, the determination unit 17 compares the difference d1 with a preset threshold value. If the difference d1 is greater than the threshold value, the determination unit 17 determines that the winding mechanism 20 is abnormal.

[0049] If it is determined that the winding mechanism 20 is abnormal, the determination unit 17 outputs the determination result. For example, the determination unit 17 transmits the determination result to a preset terminal device. If the winding mechanism 20 is equipped with an output device (speaker, light, or monitor), the determination unit 17 may cause the output device to output an output indicating the determination result. If it is determined that the winding mechanism 20 is abnormal, the determination unit 17 may stop the operation of the winding mechanism 20.

[0050] If the difference d1 is excessively large, there is a possibility that an abnormality has occurred in the winding mechanism 20 or the object O. Providing the determination unit 17 makes it possible to detect an abnormality in the winding device 2 at an earlier stage. For example, it is possible to avoid continuing to process the object O in an undesirable state.

[0051] FIG. 6 is a schematic diagram showing a hardware configuration. 6 is used as the first control unit 11, the second control unit 12, and the determination unit 17. The computer 90 includes a CPU 91, a ROM 92, a RAM 93, a storage device 94, an input interface 95, an output interface 96, and a communication interface 97.

[0052] The ROM 92 stores a program for controlling the operation of the computer 90. The ROM 92 stores a program necessary for causing the computer 90 to realize each of the above-mentioned processes. The RAM 93 functions as a storage area in which the programs stored in the ROM 92 are expanded.

[0053] The CPU 91 includes a processing circuit. The CPU 91 uses a RAM 93 as a working memory and executes a program stored in at least one of a ROM 92 and a storage device 94. During execution of the program, the CPU 91 controls each component via a system bus 98 and executes various processes.

[0054] The storage device 94 stores data necessary for executing the programs and data obtained by executing the programs.

[0055] The input interface (I / F) 95 can connect the computer 90 and the input device 95a. The input I / F 95 is, for example, a serial bus interface such as USB. The CPU 91 can read various data from the input device 95a via the input I / F 95.

[0056] The output interface (I / F) 96 can connect the computer 90 and the output device 96a. The output I / F 96 is, for example, a video output interface such as a Digital Visual Interface (DVI) or a High-Definition Multimedia Interface (HPMI (registered trademark)). The CPU 91 can transmit data to the output device 96a via the output I / F 96 and cause the output device 96a to display an image.

[0057] The communication interface (I / F) 97 can connect the computer 90 to a server 97a outside the computer 90. The communication I / F 97 is, for example, a network card such as a LAN card. The CPU 91 can read various data from the server 97a via the communication I / F 97.

[0058] The storage device 94 includes one or more selected from a hard disk drive (HDD) and a solid state drive (SSD). The input device 95a includes one or more selected from a mouse, a keyboard, a microphone (audio input), and a touch pad. The output device 96a includes one or more selected from a monitor, a projector, a printer, and a speaker. A device having the functions of both the input device 95a and the output device 96a, such as a touch panel, may be used.

[0059] Each process executed by the first control unit 11, the second control unit 12, and the determination unit 17 may be realized by one computer 90, or may be realized by cooperation of a plurality of computers 90. For example, a computer 90 functioning as the second control unit 12 and the determination unit 17 may be added to a control system including a computer 90 functioning as the first control unit 11.

[0060] For example, a general-purpose arithmetic circuit inside a programmable logic controller (PLC) is used as the first arithmetic unit 13, the second arithmetic unit 14, and the arithmetic unit 15. An electric arithmetic circuit may be used as the first arithmetic unit 13, the second arithmetic unit 14, and the arithmetic unit 15. A computer 90 functioning as the first control unit 11 or the second control unit 12 may have the function of the first arithmetic unit 13, the second arithmetic unit 14, or the arithmetic unit 15.

[0061] The various data processing operations described above may be recorded as a program that can be executed by a computer on a magnetic disk (such as a flexible disk or a hard disk), an optical disk (such as a CD-ROM, CD-R, CD-RW, DVD-ROM, DVD±R, DVD±RW), a semiconductor memory, or other non-transitory computer-readable storage medium.

[0062] For example, information recorded on a recording medium can be read by a computer (or an embedded system). The recording medium may have any recording format (storage format). For example, the computer reads a program from the recording medium and causes a CPU to execute instructions described in the program based on the program. The computer may acquire (or read) the program via a network.

[0063] Embodiments of the invention include the following features. (Feature 1) A control device for controlling a winding mechanism capable of winding up an object, a first control unit that acquires a control amount, which is an object to be controlled, from the winding mechanism, and inputs the control amount into a model that has been machine-learned in advance, thereby acquiring a first operation amount that corresponds to an operation amount when the control amount is output from the winding mechanism; a first calculation unit that calculates a difference between the first operation amount and a second operation amount transmitted to the winding mechanism; a second calculation unit that inputs a third operation amount, which is the sum of the second operation amount and the difference, to the winding mechanism. (Feature 2) the winding mechanism includes a first roller that winds up the object, 2. The control device according to feature 1, wherein the first operation amount, the second operation amount, and the third operation amount include a rotation speed of the first roller. (Feature 3) the winding mechanism includes a first roller that winds up the object, The first roller rotates about an axial direction, 3. The control device according to feature 1 or 2, wherein the first operation amount, the second operation amount, and the third operation amount include a position of the first roller in the axial direction. (Feature 4) the winding mechanism includes a second roller that feeds out the object, 4. The control device according to any one of features 1 to 3, wherein the first operation amount, the second operation amount, and the third operation amount include a rotation speed of the second roller. (Feature 5) The control device according to any one of features 1 to 4, wherein the control amount includes two or more selected from a tension of the object, a conveying speed of the object, and a position of the object in a direction perpendicular to a conveying direction of the object. (Feature 6) 6. The control device according to any one of features 1 to 5, wherein the model includes a decision tree or a neural network. (Feature 7) the model includes a recurrent neural network; 6. The control device according to any one of features 1 to 5, wherein the first control unit inputs time-series data of the controlled variable to the model. (Feature 8) The control device according to any one of features 1 to 7, further comprising a second control unit that calculates the second operation amount based on a target value for the winding mechanism. (Feature 9) The control device according to any one of features 1 to 8, further comprising a determination unit that determines whether the winding mechanism is normal based on the first operation amount. (Feature 10) A control device according to any one of features 1 to 9; A winding device comprising the winding mechanism.

[0064] According to the embodiments described above, a control device, a winding device, a method for constructing a control system, a control system, a control method, a program, and a storage medium are provided that are capable of bringing the control amount from the controlled object closer to the target value.

[0065] As used herein, "or" indicates that "at least one or more" of the items listed in the sentence may be employed.

[0066] Although some embodiments of the present invention have been illustrated above, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, modifications, etc. can be made without departing from the gist of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are included in the scope of the invention and its equivalents described in the claims. In addition, the above-mentioned embodiments can be implemented in combination with each other. [Explanation of symbols]

[0067] 1, 1r, 2: winding device, 10, 10a, 10b, 10r, 10x: control device, 11: first control unit, 12: second control unit, 13: first calculation unit, 14: second calculation unit, 15: calculation unit, 16: model, 17: determination unit, 20: winding mechanism, 21: first roller, 21a: frame, 21b: rotating shaft, 21c: motor, 22: second roller, 22a: frame, 22b: rotating shaft, 22c: motor, 23, 23a to 23j: intermediate roller, 24: sensor, 24a: encoder, 24b, 24c: force sensor, 24d, 24e: optical sensor, 24f: camera, 25: moving part, 25a: base frame, 25b: stage, 26: moving part, 26a: base frame, 26b: stage, 90: computer, 91: CPU, 92: ROM, 93: RAM, 94: storage device, 95: input interface, 95a: input device, 96: output interface, 96a: output device, 97: communication interface, 97a: server, 98: system bus, O: object, d1: difference, d2: disturbance, e: error, r: target value, u1: first manipulated variable, u2: second manipulated variable, u3: third manipulated variable, y: controlled variable

Claims

1. A control device for controlling a winding mechanism capable of winding up an object, a first control unit that acquires a control amount, which is an object to be controlled, from the winding mechanism, and inputs the control amount into a model that has been machine-learned in advance, thereby acquiring a first operation amount that corresponds to an operation amount when the control amount is output from the winding mechanism; a first calculation unit that calculates a difference between the first operation amount and a second operation amount transmitted to the winding mechanism; a second calculation unit that inputs a third operation amount, which is the sum of the second operation amount and the difference, to the winding mechanism.

2. the winding mechanism includes a first roller that winds up the object, The control device according to claim 1 , wherein the first operation amount, the second operation amount, and the third operation amount include a rotation speed of the first roller.

3. the winding mechanism includes a first roller that winds up the object, The first roller rotates about an axial direction, The control device according to claim 1 , wherein the first operation amount, the second operation amount, and the third operation amount include a position of the first roller in the axial direction.

4. The winding mechanism includes a second roller that delivers the object, The control device according to claim 1 , wherein the first operation amount, the second operation amount, and the third operation amount include a rotation speed of the second roller.

5. The control device according to claim 1 or 2, wherein the controlled variables include two or more selected from a tension of the object, a conveying speed of the object, and a position of the object in a direction perpendicular to a conveying direction of the object.

6. The control device according to claim 1 or 2, wherein the model comprises a decision tree or a neural network.

7. the model includes a recurrent neural network; The control device according to claim 1 , wherein the first control unit inputs time-series data of the controlled variable to the model.

8. The control device according to claim 1 , further comprising a second control unit that calculates the second operation amount based on a target value for the winding mechanism.

9. The control device according to claim 1 , further comprising a determination unit that determines whether the winding mechanism is normal based on the first operation amount.

10. The control device according to claim 1 or 2; A winding device comprising the winding mechanism.

11. For a control system including a first control device and a control target controlled by the first control device, a first control unit that acquires a control amount, which is an object to be controlled, from the control object, and inputs time-series data of the control amount into a model that has been machine-learned in advance, thereby acquiring a first manipulated variable corresponding to a manipulated variable when the control amount is output from the control object; a first calculation unit that calculates a difference between the first manipulated variable and a second manipulated variable transmitted from the control device to the controlled object; a second calculation unit that inputs a third manipulated variable obtained by adding the difference to the second manipulated variable to the controlled object; A method for constructing a control system, comprising: adding a second control device including the

12. The method of claim 11 , wherein the model comprises a recurrent neural network.

13. A control device for controlling a control target, a first control unit that acquires a control amount, which is an object to be controlled, from the control object, and inputs time-series data of the control amount into a model that has been machine-learned in advance, thereby acquiring a first manipulated variable corresponding to a manipulated variable when the control amount is output from the control object; a first calculation unit that calculates a difference between the first manipulated variable and a second manipulated variable transmitted to the control target; a second calculation unit that inputs a third manipulated variable obtained by adding the difference to the second manipulated variable to the controlled object; A control device comprising:

14. A control device according to claim 13; The control system includes the control object.

15. A control method for controlling a winding mechanism capable of winding up an object, comprising: A control amount to be controlled is obtained from the winding mechanism, and the control amount is input to a machine-learned model to obtain a first operation amount corresponding to an operation amount when the control amount is output from the winding mechanism; Calculating a difference between the first operation amount and a second operation amount transmitted to the winding mechanism; a third operation amount obtained by adding the difference to the second operation amount is input to the winding mechanism.

16. A program causing a computer to execute the control method according to claim 15.

17. A storage medium storing the program according to claim 16.

Images