Design support method of scrap chute and design support device of scrap chute

Abstract

To provide a scrap chute design support method and a scrap chute design support device capable of supporting a high-quality scrap chute design which prevents a scrap clogging.SOLUTION: A design support method of a scrap chute 40 includes: a learned model construction step that a computer 11 acquires the morphological data of a scrap 72 and the morphological data of the scrap chute 40 as learning data, runs the simulation to drop the scrap into the scrap chute, and gives the reward to a learning part 36 on the condition that the scrap is discharged to the outside of a press mold, and the learning part performs the reinforcement learning to optimize the scrap chute morphological data so that the reward is maximized; and an output step that the computer outputs the morphological data optimal for the scrap chute using the learned model constructed in the learned model construction step.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a method for supporting the design of a scrap chute and a device for supporting the design of a scrap chute, and particularly to a method for supporting the design of a scrap chute and a device for supporting the design of a scrap chute using a simulation of dropping scrap into the scrap chute.

Background Art

[0002] In the manufacturing process of press products, after a workpiece, which is a plate material, is drawn and formed by a press die provided in a press device, a trimming process is performed to cut and discard the scrap portion of the workpiece. The scrap cut in the trimming process is discharged outside the press die through the scrap chute of the press device. At this time, an unexpected dropping behavior may occur in the scrap, and the scrap may remain in the scrap chute. If the scrap chute becomes clogged due to the remaining scrap, it may cause defective products in subsequent press product processing or cause damage to the press die.

[0003] In order to solve such problems, Patent Document 1 describes a structure in which, in a press die, posture correction protrusions are provided in a region from when the scrap is separated from the workpiece until it drops into the scrap chute to align the naturally dropping scrap in a posture along the discharge direction of the scrap chute. In this structure, by controlling the posture of the scrap until it drops into the scrap chute by the posture correction protrusions, clogging of the scrap in the scrap chute can be suppressed.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, there is still a current situation where sufficient consideration has not been given to changing the form of the scrap chute itself to prevent clogging of the scrap chute. Since air resistance and the like act on the scrap when it falls into the scrap chute, the magnitude and direction of the force received by the scrap during falling change each time, and simply providing the posture correction protrusions before reaching the scrap chute as in the past may not be sufficient to eliminate clogging.

[0006] Designers are required to design the form of the scrap chute by predicting the force received by the scrap when it falls. However, since the forms of the scraps are also various, it is difficult for inexperienced designers to make the above prediction, and thus the design may be prone to problems.

[0007] The present invention has been made in view of the above problems, and an object thereof is to provide a design support method for a scrap chute and a design support device for a scrap chute that can assist in designing a high-quality scrap chute capable of suppressing clogging of the scrap.

Means for Solving the Problems

[0008] In order to achieve the above object, an embodiment of the present invention is A design support method for a scrap chute provided in the discharge process of scrap generated from a workpiece processed by a press die, A computer acquires, as learning data, scrap form data and scrap chute form data, and based on the acquired learning data, performs a simulation in which gravity and a force of a random magnitude within a predetermined range are applied to the scrap to cause it to fall into the scrap chute, and rewards are given to a learning unit on the condition that the scrap is discharged to the outside of the press die through the scrap chute, and the learning unit performs reinforcement learning to optimize the scrap chute form data so as to maximize the reward. A learned model construction step, An output step in which the computer outputs optimal form data of a scrap chute where scrap generated from the workpiece falls, using the learned model constructed in the learned model construction step; characterized by including the above.

[0009] In addition, in order to achieve the above object, an embodiment of the present invention is A design support device for a scrap chute provided in the process of discharging scrap generated from a workpiece processed by a press die, As learning data, form data of scrap and form data of a scrap chute are acquired, and based on the acquired learning data, a simulation is performed in which gravity and a force of a random magnitude within a predetermined range are applied to the scrap to cause it to fall onto the scrap chute. A learned model construction unit that performs reinforcement learning to optimize the form data of the scrap chute so that a reward is given to the learning unit on the condition that the scrap is discharged outside the press die through the scrap chute, and the learning unit maximizes the reward; A design support processing unit that outputs optimal form data of a scrap chute where scrap generated from the workpiece falls, using the learned model constructed by the learned model construction unit; characterized by comprising the above.

Effect of the Invention

[0010] According to the scrap chute design support method and the scrap chute design support device according to the present invention, it is possible to support the design of a high-quality scrap chute that can prevent clogging of scrap.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5A

Figure 5B

Figure 6

Figure 7

Figure 8

Figure 9

Mode for Carrying Out the Invention

[0012] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. The scrap chute design support method and the scrap chute design support device according to an embodiment of the present invention are suitably used for the design support of a press device that performs press working. FIG. 1 is a schematic diagram of a press device 50. The press device 50 presses a workpiece 70 which is a plate material, and performs trimming for cutting the scrap portion of the workpiece 70. The press device 50 includes a press die 52 and a scrap chute 40 that guides the scrap 72 cut from the workpiece 70 to the outside of the press die 52.

[0013] The press die 52 is a part for performing pressing and trimming processes on the workpiece 70, and includes a lower die 54 fixed to the installation site and an upper die 56 that can move up and down with respect to the lower die 54. The lower die 54 includes a lower cutting blade 55 supported by a lower die body 54a. The upper die 56 includes a pad 56a for pressing the workpiece 70 placed on the lower die 54, a cam slider 56b, and an upper cutting blade 57. The cam slider 56b abuts against a cam driver 54b provided on the lower die body 54a by the downward movement of the upper die 56 and moves forward toward the workpiece 70 side, and moves backward away from the workpiece 70 by the upward movement of the upper die 56. The upper cutting blade 57 is provided on the cam slider 56b so as to face the lower cutting blade 55. The workpiece 70 is cut by the upper cutting blade 57 and the lower cutting blade 55 of the press die 54, and a part of the cut workpiece 70 falls as scrap 72 into the scrap chute 40 of the press device 50.

[0014] The scrap chute 40 is disposed below the lower cutting blade 55 of the press die 52 and extends obliquely with respect to the horizontal direction. In the illustrated example, the scrap chute 40 and the lower die body 54a are integrally formed. As shown in FIG. 2, the scrap chute 40 has a first side wall 41 and a second side wall 42 that extend in the vertical direction with a predetermined interval therebetween, and a bottom wall 44 connecting the lower end portions of the side walls 41 and 42, and is open at the top. The scrap 72 slides downward along the upper surface of the bottom wall 44 in the inclined direction and is discharged to the outside of the press device 50. FIG. 1 shows a state where the scrap 72 is discharged into the scrap discharge pit 47 through the scrap chute 40.

[0015] Next, a design support device 10 according to an embodiment of the present invention will be described. The design support device 10 is used for the design support of the scrap chute 40 provided in the discharge process of the scrap 72 generated from the workpiece 70 processed by the press die 52. As shown in FIG. 3, the design support device 10 includes a computer 11 having a known hardware configuration. The computer 11 includes a processor 12, a RAM 13, a ROM 14, a storage 15, a communication device 16, an input device 17, and an output device 18.

[0016] The processor 12 executes an operating system and application programs. The storage 15 can be a known storage device that stores information, such as a hard disk drive (HDD) or a solid state drive (SSD). The communication device 16 is a transmission / reception device for performing communication between computers 11 via at least one of a wired network and a wireless network, and is composed of, for example, a network device, a network controller, a network card, a wireless communication module, etc. The input device 17 is composed of a keyboard, a mouse, a touch panel, a microphone, etc. The output device 18 is composed of a display, a speaker, etc.

[0017] As shown in FIG. 4, the design support device 10 includes, as functional units, a storage unit 20, a communication unit 21, an input unit 22, an output unit 23, and an information processing unit 24.

[0018] The storage unit 20 includes the storage 15 and stores information for constructing a learned model and information for executing design support. The communication unit 21 includes the communication device 16 and enables data transmission and reception between the device outside the design support device 10. The information stored in the storage unit 20 includes a program for constructing a learned model and a design support program. These programs may be provided after being recorded on a tangible recording medium such as a CD-ROM, a DVD-ROM, or a semiconductor memory, or may be provided as a data signal via the communication unit 21. The input unit 22 includes the input device 17 and acquires information input by the user. The output unit 23 includes the output device 18 and outputs information to the user. The output unit 23 outputs information such as an image, text, and voice so that the user can recognize the information using vision and hearing.

[0019] The information processing unit 24 has a learned model construction unit 26 and a design support processing unit 28. The information processing unit 24 includes the processor 12, and the learned model construction unit 26 and the design support processing unit 28 are configured by software executed by the processor 12.

[0020] The learned model construction unit 26 constructs a learned model by performing learned model construction processing. The learned model construction unit 26 can be configured to include a neural network, which is a model of information processing that imitates the mechanism of the neural network in the human brain. The learned model construction unit 26 includes a three-dimensional model construction unit 31, a random force setting unit 32, a force application unit 33, a deceleration processing unit 34, a reward application unit 35, and a learning unit 36.

[0021] The three-dimensional model construction unit 31 reads the shape data of the press device 50 including the press die 52 and the scrap chute 40, the workpiece 70, and the scrap 72, and constructs these three-dimensional models. The shape data can be acquired via the input unit 22. In the present embodiment, a three-dimensional model is created using CAD data used when designing the press device 50 and the workpiece 70. The three-dimensional model can be, for example, a three-dimensional mesh model representing the shapes of the press device 50 and the workpiece 70 with a polygon mesh.

[0022] In the three-dimensional model, as shown in FIG. 2, the scrap chute 40 can be composed of a first side wall 41, a second side wall 42, and a bottom wall 44 forming the bottom surface. Note that in the three-dimensional model, the scrap chute 40 may have a structure without the first side wall 41 and the second side wall 42. In the present embodiment, in the three-dimensional model, as shown in FIGS. 5A and 5B, the bottom surface of the scrap chute 40 is composed of a first flat portion 44A and a second flat portion 44B adjacent to each other in the width direction of the scrap chute 40. The first and second flat portions 44A and 44B are surfaces forming the inner bottom surface of the scrap chute 40. The first flat portion 44A and the second flat portion 44B are connected via a connecting portion 45 so that the angle formed by each other is adjustable. In FIGS. 5A and 5B, the X direction, the Y direction, and the Z direction are directions orthogonal to each other, the X direction coincides with the width direction of the scrap chute 40, and the Y direction coincides with the vertical direction, which is the height direction of the scrap chute 40. In the simulation using the three-dimensional model, the Y direction coincides with the vertical direction, and the ZX plane coincides with the horizontal plane.

[0023] In this embodiment, the shape data of the scrap chute 40 includes the position C of the connection portion 45 of the first and second flat portions 44A and 44B in the width direction of the discharge port of the scrap chute 40. The position C of the connection portion of the first and second flat portions 44A and 44B is, inside the scrap chute 40, the first distance P1 which is the distance from the first end in the width direction of the scrap chute 40 (the end on the side of the first flat portion 44A) to the connection portion 45, and the second distance P2 which is the distance from the second end in the width direction (the end on the side of the second flat portion 44B) to the connection portion 45, and can be defined by these. The shape data of the scrap chute 40 further includes a first angle P3 which is the angle formed by the horizontal plane and the first flat plate portion 44A with the connection portion 45 as the top, a second angle P4 which is the angle formed by the horizontal plane and the second flat portion 44B with the connection portion 45 as the top, and a third angle P5 which is the inclination angle of the scrap chute 40 with respect to the horizontal plane. When the first and second angles P3 and P4 are such that P3 = P4 = 0 degrees, the scrap chute 40 has a flat bottom like the bottom wall 44 shown in FIG. 2. In the three-dimensional model, the width dimension of the scrap chute 40 is set to be constant over the length direction of the scrap chute 40. When constructing the learned model, when acquiring the shape data, the ranges of the respective distances P1, P2 and the respective angles P3, P4, P5 which are parameters are set. The distances P1 and P2 are appropriately set according to the width dimension of the scrap chute 40. The angles P3 and P4 are set such that 0° ≤ P3, P4 < 90°, and the angle P5 is set such that 0 < P5 < 90°.

[0024] The three-dimensional model further includes a collection box 46 for collecting the scrap 72. The collection box 46 is arranged below the discharge port of the scrap chute 40. In the three-dimensional model of this embodiment, a gate 48 shown by a virtual line in FIG. 5A is provided between the scrap chute 40 and the collection box 46, and the opening 48a of the gate 48 forms the discharge port of the scrap chute 40. In FIGS. 5A and 5B, dots are added to the bottom wall 46a forming the inner bottom surface of the collection box 46.

[0025] The random force setting unit 32 sets a force F of a random magnitude to act at the initial stage of the fall of the scrap 72 within a range of a predetermined magnitude. This force F of a random magnitude (hereinafter, also referred to as "random force F") can be represented by a random function shown in the following formula (1). force=(upper limit value, lower limit value) ····· formula (1)

[0026] In the present embodiment, as shown in the following formulas (2), (3), and (4), random forces F are set for the respective directions of the X direction, Y direction, and Z direction that are orthogonal to each other. X - direction force=(upper limit value x up , lower limit value x low ) ····· formula (2) Y - direction force=(upper limit value y up , lower limit value y low ) ····· formula (3) Z - direction force=(upper limit value z up , lower limit value z low ) ····· formula (4)

[0027] In the present embodiment, as an example, the positive direction of the Y direction is set as the gravity direction, and the lower limit value y low = 0 is set. Also, the upper limit value x up , the upper limit value y up , and the upper limit value z up are set to positive values, and the lower limit value x low and the lower limit value z low are set to negative values. The random force F acting on the scrap 72 is the resultant force of the forces in the X direction, Y direction, and Z direction set by formulas (2) to (4). The range of the magnitude of the random force F is set so as to be smaller than the magnitude of the gravity acting on the scrap 72.

[0028] The force applying unit 33 applies gravity and the random force F set by the random force setting unit 32 to the scrap 72. The random force F is applied at the initial stage of the fall of the scrap 72 when the scrap 72 is generated from the work 70.

[0029] The deceleration processing unit 34 performs deceleration processing on the scrap 72 when the falling speed v of the scrap 72 exceeds a predetermined threshold value v th (hereinafter also referred to as "speed threshold value v th "). The deceleration processing is performed according to a preset rule. In this embodiment, when the speed threshold value v th is exceeded, as shown in the following formula (5), the process of halving the magnitude of the falling speed v of the scrap 72 is performed. The falling speed v' after deceleration processing = v / 2 ····· Formula (5)

[0030] The reward giving unit 35 gives a reward to the learning unit 36 based on the result of the simulation of the scrap fall. Specifically, a reward is given when the scrap 72 is dropped into the scrap chute 40 and discharged outside the scrap chute 40. In this embodiment, the inner bottom surface of the collection box 46 is set as the target, and a positive reward is given to the learning unit 36 on the condition that the scrap 72 collides with the inner bottom surface of the collection box 46 through the scrap chute 40. Also, in the simulation, when the scrap 72 does not collide with the inner bottom surface of the collection box 46, the reward giving unit 35 does not give a reward to the learning unit 36 or gives a negative reward.

[0031] The learning unit 36 generates and acquires a learned model by performing reinforcement learning. The learning unit 36 has the function of an agent in reinforcement learning, and generates a learned model by learning so as to maximize the reward given by the reward giving unit 35. In this embodiment, the learning unit 36 acquires the states of the first and second distances P1, P2 that define the position C of the connection part 45, the first angle P3, the second angle P4, and the third angle P5, performs a falling simulation of the scrap 72, and updates the states of the distances P1, P2 and the angles P3, P4, P5 so as to maximize the reward given by the reward giving unit 35.

[0032] The design support processing unit 28 is configured by software executed by the processor 12, and performs design support for the scrap shoot 40 using the learned model constructed by the learned model construction unit 26.

[0033] The design support apparatus 10 may be configured by one computer 11, or may be configured by a plurality of computers 11. When using a plurality of computers 11, each computer 11 is connected via a communication network such as the Internet or an intranet.

[0034] Next, the construction process of the learned model for design support using the design support apparatus 10 will be described with reference to the flowchart of FIG. 6. The learned model of the present embodiment is constructed by performing reinforcement learning using a simulation in which the three-dimensional model construction unit 31 and the learned model construction unit 26 three-dimensionally model the scrap 72 and the scrap shoot 40 on the computer 11 and drop the scrap 72 onto the scrap shoot 40.

[0035] FIG. 6 is a flowchart showing the procedure of the learned model construction process. First, in step S10, various form data and calculation condition data are input via the input unit 22, whereby the computer 11 acquires the form data and the calculation condition data. The form data includes form data of the press device 50 including the scrap shoot 40, the work 70, the scrap 72, and the collection box 46 for collecting the discharged scrap 72. The calculation condition data includes the numerical ranges of the first and second distances P1 and P2 shown in FIGS. 5A and 5B, which are the form data of the scrap shoot 40, and the numerical ranges of the first to third angles P3, P4, and P5. Further, the calculation condition data includes the pressing force of the press die 54, the time Te (for example, 4 seconds) for dropping the scrap 72, the number of frames M of the scene obtained by equally dividing the dropping time Te (M is an integer of 2 or more), the upper and lower limit values of the above-described random function, the magnitude of gravity, and the condition of the reaction force received from the scrap shoot 56 when the scrap 72 abuts on the scrap shoot 40.

[0036] In the next step S11, from the read form data, a three-dimensional model of the press device 50 including the scrap chute 40, the workpiece 70, the scrap 72, and the collection box 46 is created by the three-dimensional model construction unit 31.

[0037] In the next step S12, the number of dropping times N (N is an integer of 2 or more) for performing the dropping operation of the scrap 72 is set. The number of dropping times N can be the number instructed by the input unit 22. In the next step S13, the number of times n (n is an integer from 1 to N) of the dropping operation of the scrap 72 on the computer 11 is reset and set to n = 1. In the next step S14, the learned model construction unit 26 performs a simulation of dropping the scrap 72 onto the scrap chute 72.

[0038] FIG. 7 is a flowchart showing the process of the dropping simulation of the scrap 72 performed in step S14. First, in step S31, the scene m = 1 is set for the dropping operation of the scrap 72. Here, m is the number of each scene of the number of frames M read in the calculation condition data and is an integer from 1 to M. In this embodiment, scene 1 is set to the scene where the scrap 72 is generated by cutting the workpiece 70, that is, the scene at the initial time of dropping the scrap 72.

[0039] In the next step S32, the random force F is set by the random force setting unit 32 based on the random functions of the above-described formulas (2) to (4). In the next step S33, the gravitational force and the random force F are applied to the scrap 72 in scene m = 1. Due to these forces, the dropping operation of the scrap 72 is started.

[0040] In the next step S34, it is determined whether or not the dropping speed v of the scrap 72 exceeds a predetermined speed threshold v th . If it exceeds the speed threshold v th , in the subsequent step S35, the deceleration processing unit 34 performs deceleration processing of the dropping speed v.

[0041] After the deceleration process is performed, in the subsequent step S36, it is determined whether the falling time of the scrap 72 exceeds the set falling time Te of the scrap 72, that is, whether the scene m has reached the number of frames M. The falling time Te is set to the time until the scrap 72 is discharged to the outside through the scrap chute 40 without remaining in the press device 50. In step S34, when the falling speed v is the speed threshold v th In the following cases, proceed to step S36 without performing the deceleration process.

[0042] In step S36, when the time Te is not exceeded, in the subsequent step S37, the scene m is counter-processed, and the process continues from step S34 again. In step S36, when the time Te is exceeded, it is determined that the falling operation of the scrap 72 is completed once, and the process proceeds to step S15 of the flowchart shown in FIG. 6.

[0043] In step S15, it is determined by the reward giving unit 35 whether the scrap 72 was discharged in the simulation performed in step S14. In the present embodiment, a collection box 46 is arranged as a target for determination below the discharge port of the scrap chute 40. When the scrap 72 collides with the inner bottom surface 46a of the collection box 46, it is determined that the scrap 72 has been discharged, and when it does not collide, it is determined that it has not been discharged.

[0044] In step S15, when the scrap 72 is discharged, the process proceeds to step S16, and the reward giving unit 35 gives a positive reward to the learning unit 36 and proceeds to step S18. In step S15, when the scrap 72 has not been discharged, that is, when the scrap 72 remains in the scrap chute 40 or is discharged outside the collection box 46, the process proceeds to step S17, and the reward giving unit 35 gives a negative reward to the learning unit 36 and proceeds to step S18. Give. Note that the reward giving unit 35 may be configured not to give a reward when the scrap 72 has not been discharged.

[0045] In step S18, it is determined whether the number of times n of the dropping operation of the scrap 72 is equal to or greater than a set number of drops N. If the number of times n is less than the number of drops N, in the subsequent step S19, the number of times n is counter-processed, and the process proceeds to step S20. The learning unit 36 updates the distances P1, P2 and angles P3, P4, P5, which are parameters of the shape data of the scrap shoot 72, so that the reward increases, and the process continues again from step S14 (that is, the process of newly dropping the scrap 72 from the dropping initial position in scene 1 continues).

[0046] FIG. 8 is a diagram for explaining the random force F applied to the scrap 72. In the dropping operation with the number of times n = 1, a random force F1 is applied at the initial stage of the drop of the scrap 72 using a random function. Also, in the dropping operation with the number of times n = 2, a random force F2 is applied at the initial stage of the drop of the scrap 72. Thus, by using a random function, the random force F acting on the scrap 72 changes every time the number of times n changes. Note that the gravity acting on the scrap 72 at the initial stage of the drop is constant.

[0047] In step S18, when the number of times n is equal to or greater than the set number of drops N, the process proceeds to step S21.

[0048] In the subsequent step S21, it is determined whether the end condition of the reinforcement learning determined in advance by the learned model construction unit 26 is satisfied. The end condition of the reinforcement learning can be, for example, that the respective values of the distances P1, P2 and angles P3, P4, P5, which are the shape data of the scrap shoot 72, converge to one value by reinforcement learning. In the initial stage of the reinforcement learning, large variations occur in the respective values of the distances P1, P2 and angles P3, P4, P5 that the learning unit 36 updates to maximize the reward, but it can be determined that the reinforcement learning has ended when this variation becomes small and converges to one value. The determination of whether the end condition is satisfied can be made by the user using the design support device 10 based on a graph or the like showing the degree of variation of the values. Alternatively, a configuration may be adopted in which the computer 11 determines whether the end condition is satisfied.

[0049] If the termination condition is not satisfied in step S21, the process proceeds to step S22, where the user modifies mainly the calculation condition data, such as the setting values of the reinforcement learning algorithm, etc. Then, again, the process from step S10 is continued, and an attempt is made for the respective values of distances P1, P2 and angles P3, P4, P5 to converge to one value by reinforcement learning. If it is determined in step S21 that the termination condition is satisfied, the learning unit 36 acquires the learned model and terminates the construction process of the learned model.

[0050] To construct a learned model, it is sufficient to have form data for at least one scrap 72. In the present embodiment, a learned model is constructed by performing reinforcement learning using the above-described drop simulation of the scrap 72 for six scraps 72 having different forms.

[0051] By performing the above-described learned model construction process, for example, when 10 values are set in advance for each of the distances P1, P2 and angles P3, P4, P5 as the form data of the scrap shoot 40, it is possible to derive one value that maximizes the reward from among the 10 values.

[0052] Next, with reference to FIG. 9, a method for supporting the design of the scrap shoot 40 using the design support apparatus 10 will be described. The design support method includes a learned model construction step S41, an initial data acquisition step S42, and an output step S43.

[0053] In the trained model construction process S41, according to the flowchart shown in FIG. 6, the computer 11 that constitutes the design support device 10 performs a trained model construction process based on calculation condition data, form data, etc. input by the user. That is, the computer 11 acquires, as learning data, the form data of the scrap 72 and the form data of the scrap chute 40, and based on the acquired learning data, performs a simulation in which gravity and a force of a random magnitude within a predetermined range are applied to the scrap 72 to cause it to fall onto the scrap chute 40. In the computer 11, the trained model construction unit 26 gives a positive reward to the learning unit 36 on the condition that the scrap 72 has been discharged outside the press die 52 through the scrap chute 40, and the learning unit 36 performs reinforcement learning to optimize the form data of the scrap chute 40 so that the reward is maximized. When the trained model construction unit 26 satisfies a predetermined end condition of the reinforcement learning, it acquires the generated trained model.

[0054] In the initial data acquisition process S42, the computer 11 acquires initial data related to the design of the scrap chute 40. The acquisition of the initial data is performed by the user inputting the initial data into the computer 11 via the input unit 22. The initial data can be, for example, the form data of the scrap 72 generated from the workpiece.

[0055] In the output process S43, the computer 11 uses the trained model constructed in the trained model construction process S41 and outputs the optimized form data of the scrap chute 40 onto which the scrap 72 falls based on the initial data acquired in the initial data acquisition process S42. In the present embodiment, the form data of the scrap chute 72 estimated to have the highest discharge efficiency of the scrap 72 derived by reinforcement learning is output to the display of the computer 11, which is the output unit 22. Specifically, the values of the first distance P1, the second distance P2, the first angle P3, the second angle P4, and the third angle P5 at which the reward is maximized in the reinforcement learning are output to the display.

[0056] As described above, in the design support apparatus 10 and the design support method of the present embodiment, when the form data of the scrap 72 and the scrap chute 40 are input to the computer 11 as learning data by the user, the learning unit 36 performs reinforcement learning so as to maximize the reward. That is, the form data of the scrap chute 40 is optimized so that the scrap 72 is discharged to the outside of the press die 52 through the scrap chute 40 in the simulation. In this simulation, since the forces such as the air resistance received by the scrap 72 when it falls are applied as forces F with random magnitudes, a movement close to the actual falling movement of the scrap is reproduced, and reinforcement learning is performed so as to derive a form of the scrap chute 40 suitable for this, whereby a learned model can be constructed.

[0057] In the present embodiment, in the simulation when constructing the learned model, as a condition for giving a reward, it is set that the scrap 72 collides with the inner bottom surface of the collection box 46. Thereby, a situation in which the scrap 72 remains in the scrap chute 40 or jumps out of the outside of the collection box 46 can be excluded, and reinforcement learning can be performed so that the scrap 72 is collected in the collection box 46. Further, thereby, in the actual press die 52, the scrap chute 40 can be designed and supported so that the scrap 72 is collected in the scrap discharge pit 47 through the scrap chute 40.

[0058] In addition, in the design support device 10 of the present embodiment, after the learned model is constructed, when the user inputs the form data of the scrap 72 that falls onto the scrap shoot 40 for which design support is sought into the computer 11, the computer 11 uses the learned model to derive and output the form data of the optimal scrap shoot 40 that can suppress clogging with respect to the scrap 72. In the actual press die 52, the forms of the plurality of scraps 72 generated from the work 70 are various, and it is difficult to derive the form of the high-quality scrap shoot 40 while predicting the action of air resistance or the like on each scrap 72. In the design support device 10 of the present embodiment, by using the learned model learned to derive the optimal form data of the scrap shoot 40 by reinforcement learning, it is possible to obtain the form data of the scrap shoot 40 that is optimized in a short time. As a result, even a designer with little experience can design a high-quality scrap shoot 40.

[0059] In order to design the scrap shoot 40 with high discharge efficiency of the scrap 72, first, what angle should the inclination angle of the scrap shoot 40 itself with respect to the horizontal plane be, and second, whether to flatten the bottom surface of the scrap shoot 40 or make it V-shaped are the key points. In the above-described design support device 10, the optimal inclination angle of the scrap shoot 40 itself with respect to the horizontal plane can be obtained by the third angle P5 output in the output step S43. Also, based on the values of the first angle P3 and the second angle P4 output in the output step S43, it can be inferred that when P3 = P4 = 0 degrees, it is optimal to flatten it, and in other cases, it is optimal to make it V-shaped.

[0060] Furthermore, when designing the bottom surface of the scrap chute 40 in a V shape, the key points are where to bend it in the width direction of the scrap chute 40 and what value to set for the inclination angle of the V shape. In this embodiment, the optimal bending position when forming a V shape can be obtained based on the values of the first distance P1 and the second distance P2 that define the position of the connecting portion 45 output in the output step S43. Also, the optimal inclination angle of the V shape can be obtained based on the values of the output first angle P3 and the second angle P4.

[0061] Note that the present invention is not limited to the above-described embodiments, and various modifications are possible without departing from the spirit of the invention.

Explanation of Reference Numerals

[0062] 10 Design support device 11 Computer 20 Storage unit 21 Communication unit 22 Input unit 23 Output unit 24 Information processing unit 26 Learned model construction unit 28 Design support processing unit 40 Scrap chute 44 Bottom wall of the scrap chute 44A First flat portion 44B Second flat portion 45 Connecting portion 46 Collection box 46a Bottom wall of the collection box 50 Press device 52 Press die 70 Workpiece 72 Scrap

Claims

1. A method for assisting in the design of a scrap chute provided in the process of discharging scrap generated from a workpiece processed by a press die, comprising: A learning step of obtaining, by a computer, as learning data, scrap shape data and scrap chute shape data, performing a simulation in which gravity and a force of a random magnitude within a predetermined range are applied to the scrap based on the obtained learning data to cause the scrap to fall onto the scrap chute, and giving a reward to a learning unit on the condition that the scrap is discharged outside the press die through the scrap chute, and the learning unit performing reinforcement learning to optimize the scrap chute shape data so as to maximize the reward; An output step of outputting, by the computer, optimized shape data of the scrap chute onto which the scrap generated from the workpiece falls, using the learned model constructed in the learned model construction step; A method for assisting in the design of a scrap chute, characterized by including the above.

2. In the simulation, the bottom surface of the scrap chute is composed of a first flat portion and a second flat portion adjacent to each other in the width direction of the scrap chute; The first flat portion and the second flat portion are connected via a connecting portion so that the angle formed by them can be adjusted; The scrap chute shape data in the learning data includes the position of the connecting portion in the width direction, a first angle which is the angle formed by the first flat portion with respect to the horizontal plane with the connecting portion as the top, a second angle which is the angle formed by the second flat portion with respect to the horizontal plane with the connecting portion as the top, and a third angle which is the inclination angle of the scrap chute with respect to the horizontal plane; In the reinforcement learning, the learning unit obtains the states of the position of the connecting portion, the first angle, the second angle, and the third angle, and updates the states so as to maximize the reward. The method for assisting in the design of a scrap chute according to claim 1.

3. In the simulation, a collection box for collecting the scrap is arranged below the discharge port of the scrap chute; The method for supporting the design of a scrap chute according to claim 1 or 2, wherein, in the step of constructing the learned model, the computer gives a positive reward to the learning unit on the condition that the scrap collides with the inner bottom surface of the collection box.

4. A design support device for a scrap chute provided in the process of discharging scrap generated from a workpiece processed by a press die, A learned model construction unit that acquires, as learning data, scrap form data and scrap chute form data, and based on the acquired learning data, performs a simulation in which gravity and a force of a random magnitude within a predetermined range are applied to the scrap to cause it to fall onto the scrap chute, and gives a reward to the learning unit on the condition that the scrap is discharged to the outside of the press die through the scrap chute, and performs reinforcement learning in which the learning unit optimizes the form data of the scrap chute so that the reward is maximized; A design support processing unit that outputs optimized form data of a scrap chute onto which scrap generated from the workpiece falls, using the learned model constructed by the learned model construction unit; A design support device for a scrap chute, characterized by comprising the above.

Images