Thread rolling machine

The rolling apparatus uses machine learning to accurately adjust the relative position and rolling pressure between flat dies, improving forging precision and reducing defects in screw production.

JP7868902B1Active Publication Date: 2026-06-02SANMEI SEISAKUSHO

Patent Information

Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
SANMEI SEISAKUSHO
Filing Date
2025-09-10
Publication Date
2026-06-02

AI Technical Summary

Technical Problem

Existing forging devices require manual adjustment of the relative position between fixed and movable flat dies, relying on skilled workers and multiple tests, which is inaccurate and labor-intensive.

Method used

A rolling apparatus with a fixed and movable flat die system that uses machine learning models to derive morphological information from captured images, adjusting the relative position and rolling pressure to form screws accurately without operator intuition.

Benefits of technology

Enables precise adjustment of the relative position and rolling pressure, reducing forming defects and ensuring high-quality screws conform to product specifications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007868902000001
    Figure 0007868902000001
  • Figure 0007868902000002
    Figure 0007868902000002
  • Figure 0007868902000003
    Figure 0007868902000003
Patent Text Reader

Abstract

The present invention provides a rolling apparatus that allows for accurate and easy adjustment of the relative opposing positions of a fixed flat die and a movable flat die, as well as the top dead center position of the movable flat die. A test piece is formed and a frontal image is captured. The amount of positional misalignment between the groove center of the fixed side groove and the groove center of the moving side groove is derived from the frontal image based on a machine learning model. The penetration depth of the fixed and moving side grooves is also derived from the frontal image based on the machine learning model. Then, a top dead center correction amount to correct the top dead center position of the moving flat die is calculated from the positional misalignment amount, and a penetration correction amount to correct the relative position of the fixed and moving flat dies in the opposing direction is calculated from the penetration depth. The positional misalignment correction amount and the penetration correction amount are then output. This allows for accurate and easy adjustment of the top dead center position of the moving flat die and the relative position of the fixed and moving flat dies without relying on experience or intuition.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention mainly relates to a forging device used for forging screws.

Background Art

[0002] In a forging device that performs forging, it includes a fixed flat die provided immovable in the forging direction and a movable flat die provided reciprocally movable in the forging direction. By the reciprocating movement of the movable flat die, a base material pinched between the movable die and the fixed die is forged. And, for example, in Patent Document 1, a forging monitoring device provided in such a forging device has been proposed. This forging monitoring device monitors the forging pressure applied to the base material to be forged and determines whether appropriate forging has been performed.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] By the way, in a forging device, it is required to adjust the relative position in the facing direction between the fixed flat die and the movable flat die with high accuracy. Here, by adjusting the relative position in the facing direction, the forging pressure between the fixed flat die and the movable flat die can be adjusted, and plastic deformation can be performed into a desired form. Conventionally, a large part of the adjustment of such a relative position in the facing direction depends on skilled workers. Furthermore, even for skilled workers, it was necessary to repeat multiple tests.

[0005] An object of the present invention is to provide a forging device that can accurately and easily adjust the relative position in the facing direction between a fixed flat die and a movable flat die.

Means for Solving the Problems

[0006] The present invention relates to a rolling apparatus comprising: a fixed flat die disposed on a base so as not to move in the rolling direction and having a fixed rolling surface; and a movable flat die disposed on the base so as to be reciprocally movable in the rolling direction and having a movable rolling surface facing the fixed rolling surface, wherein the movable flat die compresses and performs rolling by the reciprocating movement of the movable flat die, and at least one of the fixed flat die and the movable flat die is disposed so as to be able to change position along the opposing direction in which the fixed rolling surface and the movable rolling surface face each other, and the relative position of the fixed flat die and the movable flat die in the opposing direction is adjusted by opposing position adjustment means, and the movable flat die The rolling apparatus is characterized by comprising: a rolling process control means that includes control content to obtain a threaded material by rolling a base material by moving a die back and forth; an imaging means that captures a front image of the threaded material obtained by the rolling process control means; a thread information derivation means that derives morphological information of the threaded material from the front image of the threaded material captured by the imaging means based on a pre-learned machine learning model; and a rolling pressure adjustment amount output means that derives a rolling pressure adjustment amount to adjust the relative position of the fixed flat die and the movable flat die in the opposing direction from the morphological information of the threaded material derived by the thread information derivation means based on a pre-learned machine learning model, and outputs the rolling pressure adjustment amount.

[0007] Here, the morphological information of the male screw material refers to information regarding the dimensions and shape of the male screw, specifically information such as the outer diameter, root diameter, thread shape, and effective diameter, and one or more combinations of this information are preferred.

[0008] In this configuration, a front view of the threaded material processed by the thread rolling control means is captured, morphological information of the threaded material is derived from the front view using a machine learning model, and further, the amount of thread rolling pressure adjustment is derived from this morphological information using the machine learning model. Here, the machine learning model used to derive the morphological information can be, for example, a model that has learned data from the front views of a large number of threaded materials captured in advance and the morphological information shown in each front view. By using this machine learning model, the morphological information can be derived with high accuracy from the front view captured by the imaging means. In addition, the machine learning model used to derive the amount of thread rolling pressure adjustment can be, for example, a model that has learned the morphological information and the relative position data in the opposing direction that formed this morphological information. By using this machine learning model, the amount of thread rolling pressure adjustment can be derived with high accuracy.

[0009] Furthermore, by adjusting the relative positions of the fixed flat die and the movable flat die in the opposing direction using the opposing position adjustment means according to the amount of rolling pressure adjustment, the rolling pressure can be adjusted with even greater precision, and screw threads of the desired shape can be formed. This suppresses the occurrence of screw thread forming defects and the occurrence of slippage during the rolling process, thereby suppressing the occurrence of forming defects due to such slippage. As a result, products that conform to product specifications can be stably formed, and the probability of forming defects (hereinafter referred to as the forming defect rate) can be reduced. Thus, with this configuration, the rolling pressure can be adjusted without relying on the operator's experience or intuition, and the burden required for this adjustment work can be significantly reduced.

[0010] As described above, the configuration of the present invention allows for the highly accurate deriving of the rolling pressure adjustment amount from the front view of the male threaded material using a machine learning model. By adjusting the relative position in the opposing direction based on these values, the rolling accuracy can be dramatically improved. Furthermore, this adjustment can be performed without relying on the operator's experience or intuition, and the burden required for this adjustment work can be reduced. Therefore, screws conforming to desired product specifications can be formed with high precision and stability, and screws of superior quality can be manufactured stably.

[0011] The rolling apparatus of the present invention, as described above, includes a top dead center position adjustment means for adjusting the position of the top dead center at which the moving flat die begins to move in the rolling direction during rolling, and a die position adjustment means for adjusting the position of the top dead center and the relative position in the opposing direction before the rolling process is executed by the rolling process control means, wherein the die position adjustment means includes a forward movement of the moving flat die that moves the moving flat die by a predetermined amount from the top dead center position in the rolling direction to rotate the base material by half, and a return movement of the moving flat die that returns the moving flat die to the top dead center position. A test movement control means comprising control content to obtain a test piece in which groove-shaped fixed side striations formed by the fixed side rolling surface portion and groove-shaped moving side striations formed by the moving side rolling surface portion are formed on the base material by automatically controlling the return movement of the test movement control means, and a positional displacement amount derivation means that derives the amount of positional displacement along the axial direction of the test piece between the groove center of the fixed side striations and the groove center of the moving side striations of the test piece obtained by the test movement control means from a front image of the test piece captured by the imaging means based on a pre-learned machine learning model, and the test obtained by the test movement control means A means for deriving the radial penetration depth of the test piece in the fixed side groove and the moving side groove of the piece from a front image of the test piece captured by the imaging means, based on a pre-trained machine learning model; a means for calculating a top dead center correction amount in the rolling direction to correct the position of the top dead center of the moving flat die based on the positional displacement amount derived by the positional displacement amount derivet, and outputting the top dead center correction amount; and a means for determining the radial penetration depth of the test piece in the fixed side groove and the moving side groove based on the penetration depth derived by the penetration depth derivet, and the position of the fixed flat die and the moving flat die A configuration is proposed in which the rolling process control means calculates a biting correction amount in the opposing direction to correct the relative position in the opposing direction and outputs the biting correction amount, and based on the biting correction amount obtained from the top dead center correction amount output means, the top dead center position of the moving flat die is adjusted by the top dead center position adjustment means, and based on the biting correction amount obtained from the biting correction amount output means, the relative position of the fixed flat die and the moving flat die in the opposing direction is adjusted by the opposing position adjustment means, and the male screw material is rolled by the rolling process control means.

[0012] In this configuration, a front view of the test piece formed by the test movement control means is captured, and a machine learning model is used to derive the amount of positional displacement from the front view to obtain a top dead center correction amount. Simultaneously, a machine learning model is used to derive the penetration depth from the front view to obtain a penetration correction amount. Here, the machine learning model used to derive the amount of positional displacement can be, for example, a model that has learned the front views of a large number of test pieces captured in advance, as well as the groove center data of the fixed side streaks and the groove center data of the moving side streaks shown in each front view. By using such a machine learning model, the amount of positional displacement can be derived with high accuracy from the front view captured by the imaging means. Similarly, the machine learning model used to derive the penetration depth can be, for example, a model that has learned the front views of a large number of test pieces captured in advance, as well as the outer surface position and the valley position of the fixed side streaks (or moving side streaks) shown in each front view. By using such a machine learning model, the penetration depth can be derived with high accuracy from the front view captured by the imaging means.

[0013] Then, by adjusting the position of the top dead center of the movable flat die with the top dead center position adjustment means according to the top dead center correction amount obtained from the positional displacement, the position of the top dead center can be set to an appropriate position where the fixed side groove and the movable side groove coincide. In addition, by adjusting the relative position in the opposing direction between the fixed flat die and the movable flat die with the opposing position adjustment means according to the penetration correction amount obtained from the penetration depth, the relative position in the opposing direction can be set to a position that results in an appropriate penetration depth, thereby enabling adjustment of the screw pitch and rolling pressure.

[0014] As described above, the configuration of the present invention allows for the highly accurate deriving of the positional displacement and penetration depth from a front view of a test piece using a machine learning model. Based on these, the rolling accuracy can be further improved by adjusting the position of the top dead center of the moving flat die and its relative position in the opposing direction. Furthermore, the adjustment of the position of the top dead center and its relative position in the opposing direction can be performed without relying on the operator's experience or intuition, and the burden required for this adjustment work can be reduced. [Effects of the Invention]

[0015] According to the swaging device of the present invention, the relative position in the facing direction between the fixed flat die and the movable flat die can be appropriately adjusted without relying on the experience and intuition of an operator, and screws with high precision and excellent quality can be stably manufactured.

Brief Description of the Drawings

[0016] [Figure 1] It is a perspective view showing the swaging machine 1 of the embodiment from the side of the fixed flat die 3. [Figure 2] It is a perspective view showing the swaging machine 1 from the side of the movable flat die 4. [Figure 3] It is a plan view showing the fixed flat die 3 enlarged. [Figure 4] It is a schematic view showing the swaging device 35. [Figure 5] It is a block diagram showing the functions of the adjustment control device 36. [Figure 6] It is a flowchart showing the flow of adjusting the relative position between the fixed flat die 3 and the movable flat die 4. [Figure 7] It is an explanatory view showing a mode of forming the test piece 61 from the base material 60. [Figure 8] It is an enlarged view showing a part of the test piece 61 enlarged. [Figure 9] (A) It is an explanatory view showing the display mode of the top dead center correction amount on the touch panel 37 and (B) the display mode of the penetration correction amount. [Figure 10] It is an enlarged view showing a part of the male screw processed material 65 having an incomplete thread 81 enlarged. [Figure 11] It is an explanatory view showing the display mode of the swaging pressure adjustment amount on the touch panel 37.

Modes for Carrying Out the Invention

[0017] Embodiments of the present invention will be described according to examples. As shown in FIGS. 1 and 2, the forging machine 1 includes a base 2. On the base 2, a fixed flat die 3 having a fixed-side forging surface portion 11 and a movable flat die 4 having a movable-side forging surface portion 21 are disposed. The movable flat die 4 reciprocates in the forging direction. The movable flat die 4 stops at a standby position before forging while the forging machine 1 is not operating. Specifically, as shown in FIGS. 1 and 2, this standby position is a position spaced apart from the fixed flat die 3 in the forging direction and is different from the position of the start of forging (top dead center position) described later.

[0018] In the present embodiment, the direction orthogonal to the forging direction and in which the fixed-side forging surface portion 11 and the movable-side forging surface portion 21 face each other will be described as the facing direction (see FIGS. 1 and 2).

[0019] The movable flat die 4 is provided so as not to be movable in the vertical direction and the facing direction, and is provided so as to be reciprocally movable in the forging direction by a guide rail (not shown) disposed on the base 2 along the forging direction. A driving device (not shown) is disposed on the base 2, and the reciprocating movement of the movable flat die 4 is controlled by the driving device.

[0020] As described above, this movable flat die 4 stops at the standby position located on one side in the forging direction with respect to the fixed flat die 3 in a state where the forging machine 1 is not operating (FIGS. 1 and 2), and is moved to the position of the top dead center (hereinafter referred to as the top dead center position) described later when forging starts (FIG. 7(A)). During forging, the movable flat die 4 is reciprocally moved in the forging direction from the top dead center position, and after the forging process is completed, the movable flat die 4 returns to the standby position. Further, in the configuration of the present embodiment, a correction bolt 8 for adjusting the position of the top dead center position (position along the forging direction) of the movable flat die 4 is provided (FIG. 2), and by rotating the correction bolt 8 forward or backward, the position of the top dead center position of the movable flat die 4 can be adjusted along the forging direction.

[0021] As shown in Figures 1-3, the fixed flat die 3 is fixed to the die block 25 by a clamping means (not shown), and the die block 25 is mounted on the base 2 so as to be movable in opposing directions by a block position reference part 26 provided on the base 2. The die block 25 can be fastened to the base 2 by fastening bolts 10, 10, and when the fastening bolts 10, 10 are fastened, the die block 25 is fixed to the base 2 together with the fixed flat die 3 so as not to move. On the other hand, when the fastening bolts 10, 10 are loosened, the die block 25 becomes movable in the direction opposite to the block position reference part 26 (base 2), and can slide in the opposite direction in that state. As a result, with the fixed flat die 3 fixed to the die block 25, the fixed flat die 3 and the die block 25 can be moved integrally in opposing directions.

[0022] The block position reference section 26 is provided with position-changing extraction bolts 31, 31 at both ends in the rolling direction. Position-changing pressing bolts 32, 32 are provided at both ends of each position-changing extraction bolt 31, arranged vertically side-by-side. These position-changing extraction bolts 31 and position-changing pressing bolts 32 are arranged in opposing directions.

[0023] The position-changing extraction bolt 31 is rotatably attached to the block position reference section 26, with its tip protruding further toward the die block 25 than the block position reference section 26 and screwed into the die block 25. By rotating this position-changing extraction bolt 31 to one side (for example, to the left), the die block 25 can be moved away from the movable flat die 4 in the opposing direction. Each position-changing extraction bolt 31 can be rotated individually, and by rotating each one, the positions of the die block 25 (fixed flat die 3) toward both ends in the rolling direction can be changed, respectively, away from the movable flat die 4 in the opposing direction.

[0024] The position-changing pressure bolt 32 is rotatably attached to the block position reference section 26, with its tip protruding further toward the die block 25 than the block position reference section 26 and contacting the die block 25. By rotating this position-changing pressure bolt 32 in one direction (for example, to the right), the die block 25 can be pressed and moved toward the moving flat die 4 in the opposing direction. Each position-changing pressure bolt 32 can be rotated individually, and by rotating each one, the positions of the die block 25 (fixed flat die 3) toward both ends in the rolling direction can be changed toward the moving flat die 4 in the opposing direction.

[0025] Furthermore, in this embodiment, the fixed flat die 3 is provided so as to be slidable in the rolling direction relative to the block position reference section 26 when the fastening bolts 10, 10 are loosened. The block position reference section 26 is provided with an adjustment bolt (not shown) that changes the position of the fixed flat die 3 in the rolling direction, and by rotating this adjustment bolt, the fixed flat die 3 can be slid in the rolling direction.

[0026] Thus, in this embodiment, the fixed flat die 3 can be repositioned in the opposite direction by rotating the position-changing extraction bolt 31 and the position-changing pressing bolt 32 with the fastening bolts 10, 10 loosened, and can also be repositioned in the rolling direction by rotating the adjustment bolt. The fixed flat die 3 can then be positioned and fixed by fastening the fastening bolts 10, 10.

[0027] Furthermore, the mechanism for adjusting the opposing positions of the die block 25 (fixed flat die 3) by appropriately rotating the two position-shifting extraction bolts 31 and the four position-shifting pressing bolts 32 can be a commonly used mechanism, so its details will be omitted.

[0028] Furthermore, a guide channel 19 is provided on one side of the fixed flat die 3 for feeding the base material 60 (see Figure 7) between the fixed flat die 3 and the movable flat die 4. In this embodiment, the thread rolling machine 1 compresses the base material 60 fed in by this guide channel 19 between the fixed-side thread rolling surface 11 of the fixed flat die 3 and the movable-side thread rolling surface 21 of the movable flat die 4, and performs thread rolling by reciprocating the movable flat die 4 in the thread rolling direction. In this embodiment, the thread rolling machine 1 forms male threads by thread rolling a roughly round bar-shaped base material.

[0029] In this embodiment, the configuration for reciprocating the movable flat die 4 (guide rail, drive control device, etc.), the fixed rolling surface section 11, and the movable rolling surface section 21 can be configured using conventionally known configurations, so their details are omitted.

[0030] Next, the essential parts of the present invention will be described. As shown in Figure 4, the thread rolling apparatus 35 of this embodiment comprises the aforementioned thread rolling machine 1 and an adjustment control device 36 that controls the thread rolling machine 1. The adjustment control device 36 is composed of a computer and includes a central control unit (CPU), a storage device (database, etc.), a communication device, and a signal input / output device. A touch panel 37 is connected to the adjustment control device 36, and the device controls the display on the touch panel 36.

[0031] The rolling machine 35 includes opposing position detection sensors 33, 33, a rolling position detection sensor 34, a top dead center detection sensor 39, and a microscope 40 (see Figure 5), all of which are connected to the adjustment control device 36.

[0032] The opposing position detection sensors 33, 33 detect the opposing position of the die block 25 (fixed flat die 3) (hereinafter referred to as the opposing position), and are respectively positioned at the ends of the block position reference unit 26 in the rolling direction. These opposing position detection sensors 33, 33 detect the ends of the fixed flat die 3 in the rolling direction, and by this detection, the opposing positions of the ends of the fixed flat die 3 can be detected. The signals detected by these opposing position detection sensors 33, 33 are input to the adjustment control device 36, respectively.

[0033] The aforementioned rolling position detection sensor 34 detects the position of the fixed flat die 3 in the rolling direction (hereinafter referred to as the rolling direction position) and is installed in the block position reference unit 26. The signal detected by this rolling position detection sensor 34 is input to the adjustment control device 36.

[0034] The top dead center detection sensor 39 is used to detect the top dead center position (starting position of the rolling process) of the moving flat die 4 when the rolling process begins, and is installed on the base 2. The signal detected by this top dead center detection sensor 39 is input to the adjustment control device 36.

[0035] The microscope 40 is used to image the test piece 61 (see Figure 8) and the threaded material 65 (see Figure 10) processed by the thread rolling machine 1. In this embodiment, when the test piece 61 and the threaded material 65 are formed by the thread rolling machine 1, their frontal images are captured by the microscope 40. The image information (data) captured by the microscope 40 is input to the adjustment control device 36.

[0036] As shown in Figure 5, the adjustment control device 36 includes a moving flat die drive control means 71 for driving and controlling the moving flat die 4, a database 72 for storing data of a machine learning model described later, an adjustment amount derivation means 73 for deriving adjustment amounts for the fixed flat die 3 and the moving flat die 4, and a data output means 74 for outputting data to the touch panel 37. Here, the moving flat die drive control means 71 executes a process to drive and control the drive device of the rolling machine 1 according to a program stored in the storage device of the adjustment control device 36, and includes a test drive control means for executing the process to form the test piece 61, and a rolling drive control means for executing the process to form the male screw workpiece 65. The adjustment amount derivation means 73 executes a process to derive the adjustment amount according to the program and data stored in the database 72.

[0037] Various programs that execute each of these means 71, 73, and 74, as well as data used by the programs and data generated by the programs, are stored in the storage device of the adjustment control device 36, and the adjustment control device 36 executes the programs at an appropriate timing.

[0038] The database 72 stores product specification information of the screws to be rolled (information such as nominal diameter and pitch), front view data of numerous test pieces 61 (fixed side grooves 62 and movable side grooves 63) (hereinafter referred to as groove image data), groove center data of the fixed side grooves associated with each groove image data, groove center data of the movable side grooves associated with each groove image data, outer surface position data and root position data of the movable side grooves (or fixed side grooves) associated with each groove image data, front view data of numerous male screw workpieces 65 (hereinafter referred to as workpiece image data), outer diameter data, root diameter data and shape data of the screw threads 81 (see Figure 10) associated with each workpiece image data, and opposing direction position data of the fixed flat die 3 associated with the outer diameter data, root diameter data and shape data of the screw threads 81.

[0039] Each of these data is distributed and stored in a storage area provided in the database 72. As shown in Figure 5, the database 72 includes a streak image data area for storing the streak image data, a groove center data area for storing the groove center data, a streak depth data area for storing the outer surface position data and groove position data for the fixed side streak and the moving streak, a workpiece image data area for storing the workpiece image data, a workpiece diameter data area for storing the outer diameter data, groove diameter data, and shape data of the threads 81 of the male screw workpiece 65, and an opposing direction position data area for storing the opposing direction position data. In this embodiment, the outer diameter data of the base material 60 used for rolling is also stored in the workpiece diameter data area.

[0040] The adjustment control device 36 performs the following operations based on the various programs described above: drive control processing of the moving flat die 4, processing to derive positional deviation amount etc. from image data using a machine learning model, processing to calculate top dead center correction amount etc., and processing to display top dead center correction amount etc. on the touch panel 37.

[0041] The drive control process for the movable flat die 4 includes a drive control process (the test drive control means) for reciprocating the movable flat die 4 to form a test piece 61, and a drive control process (the rolling drive control means) for moving the movable flat die 4 to form a male screw workpiece 65. In each of these processes, the amount of reciprocating movement of the movable flat die 4 is calculated based on the product specification information described above, and the drive device of the rolling machine 1 is driven according to the calculated amount of movement to control the movement of the movable flat die 4.

[0042] The processes using the aforementioned machine learning model include a positional displacement amount derivation process that derives the positional displacement amount P (see Figure 8) from the front view of the test piece 61, a penetration depth derivation process that derives the penetration depth T (see Figure 8) from the front view of the test piece 61, a screw information derivation process that derives the outer diameter R1 and root diameter R2 (see Figure 10) of the screw thread 81 from the front view of the male screw workpiece 65, and a rolling pressure adjustment amount derivation process that derives the rolling pressure adjustment amount from the outer diameter R1, root diameter R2 and the shape of the screw thread 81 (see Figure 5).

[0043] In the displacement amount derivation process, a machine learning model trained on the aforementioned streak image data (see Figure 8) and groove center data of the fixed side streak 62 and moving side streak 63 associated with each streak image data is used to derive the groove center W of the fixed side streak 62 and the groove center V of the moving side streak 63 from the front image of the test piece 61 captured by the microscope 40 (Figure 8). Then, the displacement amount P of these groove centers W and V is calculated. Similarly, in the penetration depth derivation process, the outer surface position H and the valley position Q of the moving side streak 63 are derived from the front image of the test piece 61 (Figure 8) using the aforementioned streak image data and the outer surface position data and valley position data associated with each streak image data. Then, the penetration depth T is calculated from this outer surface position H and valley position Q. Furthermore, in the screw information derivation process, the outer diameter R1, root diameter R2, and thread shape of the screw thread 81 are derived from the front view of the male screw workpiece 65 captured by the microscope 40 (Figure 10) using a machine learning model trained on the workpiece image data (see Figure 10) and the outer diameter data, root diameter data, and thread shape data associated with each workpiece image data. In the rolling pressure adjustment amount derivation process, the opposing direction position requiring correction is derived from the outer diameter R1, root diameter R2, and thread shape of the male screw workpiece 65 and the opposing direction position data of the fixed flat die 3 using a machine learning model trained on the outer diameter data, root diameter data, and thread shape data of the male screw workpiece 65 and the opposing direction position data of the fixed flat die 3. Then, the rolling pressure adjustment amount (the amount to move the fixed flat die 3 in the opposing direction) to be adjusted from the current opposing direction position in which the male screw workpiece 65 was formed is calculated. Here, the outer diameter data of the base material 60 used for rolling is also used to calculate the rolling pressure adjustment amount. The current opposing position is detected by the opposing position detection sensors 33, 33 described above. As conventionally known methods can be applied to the machine learning model, details have been omitted.

[0044] Furthermore, in this embodiment, the groove center W of the fixed side groove 62, the groove center V of the movable side groove 63, and the outer surface position H and root position Q of the test piece 61 derived from each of these processes are associated with the front view of the test piece 61 and stored in the database 72. Then, each machine learning model is updated using these. Similarly, the outer diameter R1, root diameter R2, thread shape data of the male thread workpiece 65, and opposing direction position data are associated with the front view of the male thread workpiece 65 and stored in the database 72. Then, each machine learning model is updated using these.

[0045] The calculation process for the top dead center correction amount described above includes a top dead center correction amount calculation process for calculating the top dead center correction amount and a penetration correction amount calculation process for calculating the penetration correction amount (see Figure 5). Here, the top dead center correction amount calculation process uses the nominal diameter and pitch information from the product specification information described above, the top dead center position of the moving flat die 4 in the formation of the test piece 61, information regarding the height of the moving flat die 4 in the formation, the positional displacement amount P, and pi to calculate the top dead center correction amount to be adjusted from the current top dead center position. The current top dead center position is detected by the top dead center detection sensor 39 described above. In addition, the penetration correction amount calculation process uses the nominal diameter and pitch information from the product specification information described above, the opposing position of the fixed flat die 3 in the formation of the test piece 61, and the penetration depth T to calculate the penetration correction amount (the amount to move the fixed flat die 3 in the opposing direction) to be adjusted from the current opposing position. Here, the current opposing position is detected by the opposing position detection sensors 33, 33, as described above.

[0046] The processes displayed on the aforementioned touch panel 37 include switching between and displaying the top dead center correction amount and the penetration correction amount, and displaying the rolling pressure adjustment amount. Since conventionally known methods can be applied to these display processes, details have been omitted.

[0047] In the thread rolling apparatus 35 of this embodiment, the adjustment control device 36 is configured to drive and control only one thread rolling machine 1. That is, the thread rolling apparatus 35 consists of one thread rolling machine 1 and an adjustment control device 36 that drives and controls only that thread rolling machine 1. For this reason, for example, if multiple thread rolling machines 1 are installed in the factory, multiple adjustment control devices 36 are also installed to drive and control each thread rolling machine 1. When multiple thread rolling apparatuses 35 are installed in the factory, the data derived using the machine learning model described above (groove centers of fixed and moving side grooves, outer surface position and valley position of the test piece 61, outer diameter and valley diameter and thread shape of the male thread workpiece 65, and opposing direction position) is associated with the respective front views of the test piece 61 and the male thread workpiece 65 and output to the adjustment control device 36 of each thread rolling apparatus 35. Then, in the adjustment control device 36 of each thread rolling machine 35, when the aforementioned data (data derived using a machine learning model) is input from another adjustment control device 36, the respective machine learning model is updated using that data. In this way, when multiple thread rolling machines 35 are installed, information on the test piece 61 and the male screw processed material 65 is provided to all thread rolling machines 35, and the machine learning model in each thread rolling machine 35 is updated using that information.

[0048] Next, the process of adjusting the positions of the fixed flat die 3 and the movable flat die 4 in order to form the desired screw product that conforms to the product specifications will be explained with reference to Figure 6.

[0049] First, settings are made for the formation of the test piece 61 (S110), and the rolling machine 1 is driven according to these settings to form the test piece 61 (S120). In this embodiment, steps S110 and S120 are executed sequentially when the formation of the test piece 61 is instructed via the touch panel 37.

[0050] More specifically, when the operator selectively inputs the product specifications (nominal diameter, pitch, and thread direction) of the screw product to be manufactured via the touch panel 37, the reciprocating distance of the movable flat die 4 is set based on the product specifications information (S110). By controlling the drive device of the thread rolling machine 1 according to this reciprocating distance, the movable flat die 4 is automatically moved back and forth from the top dead center to form the test piece 61 (S120).

[0051] Here, as shown in Figure 7(a), the test piece 61 is formed by moving the movable flat die 4 in the forward direction of rolling until the base material 60 has rotated half a turn, starting from a state where the base material 60 is sandwiched between the fixed rolling surface 11 of the fixed flat die 3 and the movable rolling surface 21 of the movable flat die 4 at the top dead center position. This creates groove-shaped fixed-side streaks 62 based on the fixed-side rolling surface 11 and movable-side streaks 63 based on the movable-side rolling surface 21 on the base material 60 (see Figure 8). After the base material 60 has rotated half a turn, the movable flat die 4 is automatically returned to the top dead center position (reversal operation), as shown in Figure 7(b). The drive control device controls the automatic reciprocating movement of the movable flat die 4 in this way, thereby forming fixed-side streaks 62 and movable-side streaks 63 on the base material 60, and obtaining a test piece 61 having these streaks 62 and 63.

[0052] Once the test piece 61 is formed, a front view of the test piece 61 is captured by the microscope 40 (S130). The data of the front view of the test piece 61 captured by the microscope 40 is then transmitted to the adjustment control device 36. As shown in Figure 8, the test piece 61 has fixed side streaks 62 and movable side streaks 63 formed on it, and these fixed side streaks 62 and movable side streaks 63 appear in the front view.

[0053] When the adjustment control device 36 receives a front view of the test piece 61, it performs the positional displacement amount derivation process (S140) and the penetration depth derivation process (S160). In the positional displacement amount derivation process, as described above, a machine learning model is used to derive the groove center W of the fixed side streak 62 and the groove center V of the moving side streak 63 from the front view of the test piece 61. The positional difference between these groove centers W and V in the direction along the axis of the test piece 61 is then calculated as the positional displacement amount P (see Figure 8). On the other hand, in the penetration depth derivation process, as described above, a machine learning model is used to derive the outer surface position H and the valley position Q from the front view of the test piece 61, and the penetration depth T (see Figure 8) is calculated from these outer surface position H and valley position Q.

[0054] Once the positional displacement amount P is obtained through the positional displacement amount derivation process described above, the top dead center correction amount calculation process is executed (S150). In the top dead center correction amount calculation process, as described above, the top dead center correction amount to be adjusted from the current top dead center position input from the top dead center detection sensor 39 is calculated using the nominal diameter, screw pitch and screw direction, positional displacement amount P, etc. from the product specification information.

[0055] On the other hand, once the penetration depth T is obtained through the penetration depth derivation process, the penetration correction amount calculation process is executed (S170). In the penetration correction amount calculation process, as described above, the nominal diameter and pitch of the product specification information and the penetration depth T are used to calculate the penetration correction amount to be adjusted from the current opposing direction position input from the opposing position detection sensors 33, 33.

[0056] Once the top dead center correction amount and the biting correction amount are obtained, a process is performed to display them on the touch panel 37 (S180). In this process, the data for the top dead center correction amount and the biting correction amount are output to the touch panel 37 and displayed. Specifically, as shown in Figure 9(A), the screen of the touch panel 37 displays a standard display unit 41 that selectively displays the screw standard (nominal diameter, pitch, and screw direction, etc.), a height display unit 54 that displays information regarding the height of the input movable flat die 4, a position deviation amount display unit 55 that displays the position deviation amount P, a current value display unit 56 that displays the current top dead center position of the movable flat die 4, and a correction value display unit 57 that displays the corrected position of the top dead center. Here, the current value display unit 56 displays the top dead center position data input by the top dead center detection sensor 39. The correction value display unit 57 displays the top dead center position data corrected by the top dead center correction amount. Furthermore, the touch panel 37 displays a switching display unit 58, and by touching the switching display unit 58, the display mode switches from one that displays information on the top dead center position to one that displays information on the opposing position of the fixed flat die 3 (Figure 9(B)).

[0057] When the display mode is switched by operating the switching display unit 58, as shown in Figure 9(B), the screen of the touch panel 37 displays the aforementioned standard display unit 41, the biting amount display unit 44 that displays the biting depth T, the first to third current value display units 43a to 43c that display the current position of the fixed flat die 3 detected by the opposing position detection sensors 33, 33 and the rolling position detection sensor 34, and the first to third correction value display units 42a to 42c that display the correction position of the fixed flat die 3. Here, the first to third correction value display units 42a to 42c display the position data of the fixed flat die 3 corrected by the biting correction amount.

[0058] When the correction amounts for the top dead center position and the position of the fixed flat die 3 are displayed on the touch panel 37, the top dead center position and the position of the fixed flat die 3 are adjusted according to these correction amounts (S190). In this embodiment, these position adjustments are performed by an operator.

[0059] In other words, with the current value display unit 56 and the correction value display unit 57 of the top dead center position displayed on the touch panel 37 (Figure 9(A)), the operator can adjust the top dead center position of the movable flat die 4 to the correct position where the center of the movable side groove and the center of the fixed side groove coincide by operating the correction bolt 8 to match the current position information displayed on the current value display unit 56 with the correction position information displayed on the correction value display unit 57. In this way, the operator can visually recognize the amount of top dead center correction and operate the correction bolt 8. After adjusting the top dead center position, when the switching display unit 58 is operated, the first to third current value display units 43a to 43c and the first to third correction value display units 42a to 42c of the fixed flat die 3 are displayed on the touch panel 37 (Figure 9(B)). In this state, the operator loosens the fastening bolts 10, 10 that fix the fixed flat die 3, allowing the fixed flat die 3 to slide in the opposing direction and the rolling direction. Then, by operating the position change extraction bolt 31, the position change pressing bolt 32, and the adjustment bolt, the current position information displayed in the current value display units 43a to 43c, respectively, is matched with the correction position information shown in the correction value display units 42a to 42c, respectively, thereby adjusting the opposing direction position of the fixed flat die 3 to an appropriate position that allows for forming with the desired cutting depth (desired rolling pressure and pitch that matches the product specifications) during the rolling process. In this way, the operator can visually recognize the amount of biting correction of the fixed flat die 3 and operate the position change extraction bolt 31, the position change pressing bolt 32, and the adjustment bolt.

[0060] The adjustment is completed when the operator operates the completion indicator 49 on the touch panel 37 after adjusting the top dead center position and the position of the fixed flat die 3. This completes the adjustment. As a result, the male threaded material 65 is formed (S200). More specifically, when the completion indicator 49 is operated, the processing conditions for processing the screw product (such as the amount of movement of the movable flat die 4) are set based on the product specification information of the screw product to be manufactured, and the drive device of the thread rolling machine 1 is driven and controlled accordingly. As a result, the movable flat die 4 is advanced in the forward direction of the thread rolling from a state in which the base material 60 is sandwiched between the fixed side thread rolling surface 11 of the fixed flat die 3 and the movable side thread rolling surface 21 of the movable flat die 4 at the top dead center position, causing the base material 60 to rotate multiple times and form the male threaded material 65 shown in Figure 10. Furthermore, since the male threaded material 65 is formed under the processing conditions of screw products, it is formed in the same way as screw products, with the threads 81 arranged in a spiral shape.

[0061] Once the male threaded material 65 is formed, a front view of the male threaded material 65 is captured by the microscope 40 (S210). The data of the front view of the male threaded material 65 captured by the microscope 40 is then transmitted to the adjustment control device 36.

[0062] When the adjustment control device 36 receives a front view of the male threaded material 65, it executes the thread information derivation process (S220). In this thread information derivation process, as described above, a machine learning model is used to derive the outer diameter R1, root diameter R2, and thread shape 81 from the front view of the male threaded material 65. Once these outer diameter R1, root diameter R2, and thread shape 81 are obtained, the rolling pressure adjustment amount derivation process is executed (S230). In this rolling pressure adjustment amount derivation process, as described above, a machine learning model is used to derive the opposing position of the fixed flat die 3 that requires correction from the outer diameter R1 and root diameter R2 and thread shape 81 obtained in the thread information derivation process, and the rolling pressure adjustment amount to be adjusted from the current opposing position is calculated. Here, Rolling pressure adjustment amount derivation process When the outer diameter dimension of the base material 60 of the product to be rolled is entered, the appropriate rolling pressure adjustment amount for the base material 60 is calculated based on that outer diameter dimension.

[0063] Once the rolling pressure adjustment amount is obtained, the system performs a process to display the rolling pressure adjustment amount on the touch panel 37 (S240). As a result of this process, as shown in Figure 11, the touch panel 37 screen displays the adjustment amount display unit 59 which displays the rolling pressure adjustment amount, as well as the aforementioned standard display unit 41, first to third current value display units 43a to 43c, and first to third correction value display units 42a to 42c. Here, the first to third current value display units 43a to 43c display the current position of the fixed flat die 3 detected by the opposing position detection sensors 33, 33 and the rolling position detection sensor 34. The first correction value display unit 42a and the second correction value display unit 42b display the position data of the fixed flat die 3 corrected by the rolling pressure adjustment amount.

[0064] When the rolling pressure adjustment amount and other information are displayed on the touch panel 37, the operator adjusts the opposing position of the fixed flat die 3 accordingly. This completes the adjustment of the top dead center position of the movable flat die 4 and the opposing position of the fixed flat die 3 for manufacturing the desired screw product.

[0065] In this embodiment, the thread rolling apparatus 35 forms a test piece 61 and captures a front view image. From this front view image, a machine learning model is used to derive the misalignment amount P and the penetration depth T. The top dead center correction amount is calculated from the misalignment amount P, and the penetration correction amount is calculated from the penetration depth T. As a result, the misalignment amount P and the penetration depth T can be derived with high accuracy and stability. By adjusting the top dead center position of the moving flat die 4 and the opposing position of the fixed flat die 3 according to the top dead center correction amount and penetration correction amount calculated from these, the processing accuracy of the thread rolling process can be easily improved. Furthermore, by adjusting the top dead center position and the opposing position, a front view image of the male screw workpiece 65 is captured. From this front view image, a machine learning model is used to derive the outer diameter R1 and the root diameter R2. From the outer diameter R1, the root diameter R2, and the shape data of the screw threads 81, the machine learning model is used to derive the thread rolling pressure adjustment amount. This allows for the highly accurate and stable determination of the outer diameter R1, root diameter R2, and rolling pressure adjustment amount of the male thread material 65. By adjusting the opposing position of the fixed flat die 3 according to the rolling pressure adjustment amount, screw products conforming to product specifications can be formed with high precision and stability, while significantly reducing the occurrence of molding defects. In this embodiment, Rolling pressure adjustment amount Since a machine learning model that has learned the shape of the screw thread 81 is used to derive the result, screw products consisting of screw threads 81 with a perfect thread shape can be formed with even greater precision and stability. Thus, with the thread rolling apparatus 35 of this embodiment, the adjustment between the top dead center position of the moving flat die 4 and the opposing position (rolling pressure) of the fixed flat die 3 can be performed easily and with high precision without relying on the experience or intuition of the operator, thereby dramatically improving product quality and significantly reducing the burden on the operator required for the adjustment.

[0066] Furthermore, the configuration of this embodiment has the advantage of reducing the burden of quality control for screws and improving the lifespan of the dies, as it suppresses variations in quality due to differences in skill levels among workers. In addition, it has the advantage of further reducing the workload on workers by shortening the preparation time and improving the operating rate of the thread rolling machine. Moreover, it has the advantage of reducing the number of blank materials (base materials) required to manufacture screw products, thereby reducing the associated costs.

[0067] Furthermore, the configuration of this embodiment includes the positional displacement amount derivation process, the biting depth derivation process, the screw information derivation process, and The process for deriving the rolling pressure adjustment amount. Since the machine learning models are updated using the data derived in each step and the front view of the test piece 61 and the male screw-machined material 65 used in the derivation, the accuracy of data derivation in each of these processes can be further improved. As a result, each time the top dead center position and the opposing direction position are adjusted, the training data for the machine learning model increases, improving the accuracy of data derivation based on the machine learning model and further enhancing the effects described above. Furthermore, the configuration of this embodiment includes the positional displacement amount derivation process, the biting depth derivation process, the screw information derivation process, and The process for deriving the rolling pressure adjustment amount. The data derived from each of these processes, along with the front view data of the test pieces 61 and male screw-processed material 65 used in the derivation, are output to other rolling machines 35 in the factory. Each rolling machine 35 then updates its machine learning model based on the input data. This allows the machine learning model to be updated for each rolling machine 35 using the data derived from one rolling machine 35, enabling each rolling machine 35 to be updated with a suitable machine learning model and improving the accuracy of data derivation in each of the processes.

[0068] In the configuration of the embodiment described above, the position-changing extraction bolt 31 and the position-changing pressing bolt 32 correspond to the opposing position adjustment means according to the present invention, and the correction bolt 8 corresponds to the top dead center position adjustment means according to the present invention. The adjustment control device 36 (test drive control means for the moving flat die drive control means) and the drive device of the rolling machine 1 correspond to the test movement control means according to the present invention. The position deviation amount derivation process of the adjustment control device 36 (adjustment amount derivation means 73) corresponds to the position deviation amount derivation means according to the present invention, and the biting depth derivation process of the adjustment control device 36 (adjustment amount derivation means 73) corresponds to the biting depth derivation means according to the present invention. The top dead center correction amount calculation process and data output means 74 of the adjustment control device 36 (adjustment amount derivation means 73) correspond to the top dead center correction amount output means according to the present invention, and the biting correction amount calculation process and data output means 74 of the adjustment control device 36 (adjustment amount derivation means 73) correspond to the biting correction amount output means according to the present invention. The die position adjustment means according to the present invention is comprised of the test movement control means, position deviation amount derivation means, penetration depth derivation means, top dead center correction amount output means, and penetration correction amount output means. The microscope 40 corresponds to the imaging means according to the present invention. The adjustment control device 36 (rolling drive control means of the moving flat die drive control means) and the drive device of the rolling machine 1 correspond to the rolling process control means according to the present invention. The screw information derivation processing of the adjustment control device 36 (adjustment amount derivation means 73) corresponds to the screw information derivation means according to the present invention, and the rolling pressure adjustment amount derivation processing and data output means 74 of the adjustment control device 36 (adjustment amount derivation means 73) correspond to the rolling pressure adjustment amount output means according to the present invention. The outer diameter R1 and root diameter R2 of the male screw workpiece 65 and the shape of the screw threads 81 correspond to the morphological information of the male screw workpiece according to the present invention.

[0069] The present invention is not limited to the embodiments described above, and can be modified as appropriate without departing from the spirit of the invention. For example, the dimensions and shapes of each component constituting the rolling machine 1 can be changed as appropriate.

[0070] In the embodiment, the rolling apparatus displays the top dead center correction amount, the penetration correction amount, and the rolling pressure adjustment amount on a touch panel, allowing the operator to manually adjust the top dead center position of the moving flat die and the opposing position of the fixed flat die. However, the apparatus is not limited to this, and the top dead center position and the opposing position may be automatically adjusted and controlled. For example, the system is equipped with means for automatically rotating a correction bolt, and by automatically controlling this means according to the top dead center correction amount, the top dead center position of the movable flat die can be adjusted. Similarly, the system is equipped with means for automatically rotating a position-changing extraction bolt and a position-changing pressing bolt, and by automatically controlling these means according to the penetration correction amount and the rolling pressure adjustment amount, the opposing position of the fixed flat die can be adjusted.

[0071] The rolling apparatus in this embodiment is configured such that the fixed flat die can be moved in the opposite direction and its position in the opposite direction can be adjusted. However, it is not limited to this configuration, and the cutting depth may also be adjusted by moving the movable flat die in the opposite direction and adjusting its position in the opposite direction. Alternatively, both the fixed flat die and the movable flat die may be made movable in the opposite direction, and their respective positions in the opposite direction can be adjusted.

[0072] In the rolling apparatus of the embodiment, the location of the microscope can be set as appropriate. For example, it may be installed on the rolling machine, or it may be installed in a location close to the rolling machine, or it may be installed in a location separate from the rolling machine.

[0073] The rolling apparatus in this embodiment displays the top dead center correction amount, penetration correction amount, and rolling pressure adjustment amount via a touch panel, but the display format is not limited to this and can be changed as appropriate. For example, the top dead center correction amount (Figure 9(A)) and penetration correction amount (Figure 9(B)) may be displayed simultaneously.

[0074] The rolling apparatus in the embodiment is Screw information derivation process (screw information derivation means) This method is used to derive the outer diameter, root diameter, and thread shape, but is not limited to this. Screw information derivation processThe morphological information derived can be modified as needed. For example, Screw information derivation process The morphological information derived by this method may be the outer diameter and the valley diameter, or it may be the effective diameter. To elaborate on the configuration of another example in which the morphological information is used as the effective diameter, the configuration of this other example is as follows: Screw information derivation process Next, using a machine learning model trained on the processed material image data (see Figure 10) and the virtual vertex outer diameter data and virtual deepest point root diameter data associated with each processed material image data, the virtual vertex outer diameter and virtual deepest point root diameter are derived from the front view of the male screw processed material captured by a microscope (Figure 10), and the effective diameter is derived from these. Here, the virtual vertex indicates the intersection of the two bevels of the screw thread extending outward, and the outer diameter of this intersection is taken as the outer diameter of the virtual vertex. Similarly, the virtual deepest point indicates the intersection of the two bevels of the screw groove extending inward, and the diameter of this intersection is taken as the virtual deepest point root diameter. Then, half of the sum of the virtual vertex outer diameter and the virtual deepest point is calculated as the effective diameter. Furthermore, in the rolling pressure adjustment amount derivation process (rolling pressure adjustment amount output means), the machine learning model trained on the effective diameter data of the male screw processed material and the opposing direction position data of the fixed flat die is used, Screw information derivation process From the effective diameter derived, the opposing direction position requiring correction is derived, and the rolling pressure adjustment amount is calculated. Even in this alternative configuration, the rolling pressure adjustment amount can be derived with high accuracy and stability, similar to the embodiment described above. still, Screw information derivation process The method for deriving the effective diameter is not limited to using virtual vertices or virtual deepest points; other methods may also be applied. For example, positional data where the width of the thread and the width of the thread groove are equal. from Alternatively, the effective diameter may be derived using a machine learning model trained on image data of the processed material. [Explanation of Symbols]

[0075] 1. Rolling machine 2 Base 3 Fixed flat die 4 Moving Flat Dice 8. Correction bolts 11 Fixed side rolled surface part 21 Moving side rolled surface part 31 Position-changing extraction bolt 32 Position-changing press bolts 33 Opposing position detection sensor 35. Rolling machine 36 Adjustment control device 37 Touch panel 40 Microscopes 60 Base material 61 test pieces 62 Fixed side striations 63. Moving lateral striae 65 Male threaded material P Positional displacement R1 Outer diameter R2 Valley diameter T depth of penetration V,W groove center

Claims

1. A fixed flat die is positioned on a base table so as not to move in the rolling direction, and has a fixed rolling surface portion on the fixed side. A movable flat die is disposed on the base table so as to be able to reciprocate in the rolling direction, and has a movable rolling surface facing the fixed rolling surface. A rolling apparatus comprising the reciprocating movement of the moving flat die, which compresses the rolling surface of the moving flat die and the rolling surface of the fixed flat die, At least one of the fixed flat die and the movable flat die is arranged to be repositionable along the opposing directions in which the fixed rolling surface portion and the movable rolling surface portion face each other. The opposing position adjustment means adjusts the relative positions of the fixed flat die and the movable flat die in the opposing direction, A rolling process control means comprising control functions for rolling a base material to obtain a male screw-processed material by reciprocating the aforementioned movable flat die, An imaging means for capturing a front view of the threaded material obtained by the aforementioned thread rolling control means, A screw information derivation means derives morphological information of the male screw processed material from a front image of the male screw processed material captured by the imaging means, based on a pre-trained machine learning model. Based on the morphological information of the male threaded material derived by the screw information derivation means, a rolling pressure adjustment amount is derived from a pre-trained machine learning model to adjust the relative position of the fixed flat die and the movable flat die in the opposing direction, and the rolling pressure adjustment amount is output to a display device or opposing position adjustment means as a numerical value indicating the amount by which the fixed flat die and / or the movable flat die are moved in the opposing direction. A rolling apparatus characterized by being equipped with the following features.

2. A top dead center position adjustment means for adjusting the position of the top dead center at which the moving flat die starts moving in the rolling direction during the rolling process, Before the rolling process is executed by the rolling process control means, a die position adjustment means adjusts the position of the top dead center and the relative position in the opposing direction. Equipped with, The die position adjustment means is A test movement control means comprising control content for obtaining a test piece in which groove-shaped fixed-side streaks formed on the base material by automatically controlling the forward movement of the movable flat die, which moves the movable flat die by a predetermined amount in the rolling direction from the top dead center position to rotate the base material by half, and the reverse movement of the movable flat die, which returns the movable flat die to the top dead center position, A means for deriving the amount of positional displacement along the axial direction of the test piece, between the groove center of the fixed side groove and the groove center of the moving side groove of the test piece obtained by the test movement control means, from a front image of the test piece captured by the imaging means, based on a pre-trained machine learning model. A means for deriving the radial penetration depth of the test piece in the fixed side groove and the moving side groove obtained by the test movement control means from a front image of the test piece captured by the imaging means, based on a pre-trained machine learning model, Based on the positional displacement amount derived by the positional displacement amount derivation means, a top dead center correction amount in the rolling direction is calculated to correct the position of the top dead center of the moving flat die, and a top dead center correction amount output means outputs the top dead center correction amount. Based on the penetration depth derived by the penetration depth derivation means, a penetration correction amount in the opposing direction is calculated to correct the relative position of the fixed flat die and the movable flat die in the opposing direction, and a penetration correction amount output means outputs the penetration correction amount. It is equipped with, The top dead center position adjustment means is configured to adjust the top dead center position of the moving flat die based on the top dead center correction amount output from the top dead center correction amount output means. The rolling apparatus according to claim 1, characterized in that the opposing position adjustment means is configured to adjust the relative position of the fixed flat die and the movable flat die in the opposing direction based on the insertion correction amount output from the insertion correction amount output means.