Modeling planning support system, learning device, computing device, modeling planning support method, and program

Abstract

The molding planning support system (30) includes a learning device (31) comprising: a first data acquisition unit that acquires molding conditions, which are conditions for additive manufacturing, and evaluation values ​​that indicate an evaluation of additive manufacturing; a learning unit that learns a prediction function and a predicted variance of evaluation values ​​for predicting evaluation values ​​corresponding to molding conditions based on learning data including the corresponding molding conditions and evaluation values, and outputs a prediction function and a predicted variance inferred from the input molding conditions and the input evaluation values; a second data acquisition unit that acquires the prediction function and predicted variance output by the learning device (31); and a molding condition determination unit that determines the molding conditions for acquiring a target evaluation value, which is the target of the evaluation value, based on the prediction function and the predicted variance.

Description

Technical Field

[0005] , , ,

[0001] The present disclosure relates to a modeling plan support system, a learning device, an arithmetic device, a modeling plan support method, and a program for assisting in formulating a plan for additive manufacturing by an additive manufacturing apparatus.

Background Art

[0002] As one of the technologies for manufacturing three-dimensional objects, the technology of additive manufacturing (AM) is known. In additive manufacturing, since the modeling conditions have a great influence on the internal structure of the object to be modeled and the characteristics of the object to be modeled, it is necessary to adjust the modeling conditions in order to manufacture an object to be modeled having a desired internal structure and desired characteristics. In the conventional technology, the optimization of the modeling conditions may be performed by the trial and error of an operator. However, since many parameters included in the modeling conditions affect each other, and the internal structure and characteristics of the manufactured object to be modeled and each parameter of the modeling conditions are complicatedly related to each other, it is difficult to optimize the modeling conditions by trial and error.

[0003] Patent Document 1 discloses a welding system that generates a model for estimating the size of a defect from the modeling conditions by learning the relationship between the size of a defect generated inside an object to be modeled and the modeling conditions, and repeatedly corrects the modeling conditions until the size of the defect estimated by the model becomes less than or equal to a reference value.

Prior Art Documents

Patent Documents

[0006] This disclosure is made in view of the above, and aims to provide a modeling planning support system that enables easy optimization of modeling conditions. [Means for solving the problem]

[0007] To solve the above-mentioned problems and achieve the objective, the molding planning support system according to this disclosure is a molding planning support system that assists in formulating a plan for additive manufacturing using an additive manufacturing apparatus. The molding planning support system according to this disclosure includes a learning device comprising: a first data acquisition unit that acquires molding conditions, which are conditions for additive manufacturing, and evaluation values ​​that indicate the evaluation of additive manufacturing; a learning unit that learns a prediction function and a predicted variance of evaluation values ​​for predicting evaluation values ​​corresponding to molding conditions based on learning data including the molding conditions and evaluation values ​​that are associated with each other, and outputs a prediction function and a predicted variance inferred from the input molding conditions and the input evaluation values; a second data acquisition unit that acquires the prediction function and predicted variance output by the learning unit; and a molding condition determination unit that determines the molding conditions for acquiring a target evaluation value, which is the target of the evaluation value, based on the prediction function and the predicted variance. [Effects of the Invention]

[0008] The molding planning support system described herein has the effect of easily optimizing molding conditions. [Brief explanation of the drawing]

[0009] [Figure 1] A diagram showing an example configuration of an additive manufacturing apparatus according to Embodiment 1. [Figure 2] This figure shows an example of the configuration of the molding planning support system according to Embodiment 1. [Figure 3] A diagram showing an example configuration of the learning device according to Embodiment 1. [Figure 4] A diagram illustrating the overview of learning using the learning device according to Embodiment 1. [Figure 5] A diagram showing an example configuration of the arithmetic unit according to Embodiment 1. [Figure 6] A diagram illustrating the outline of the process for determining the molding conditions using the calculation device according to Embodiment 1. [Figure 7] A flowchart showing an example of the processing procedure by the molding planning support system and additive manufacturing apparatus according to Embodiment 1. [Figure 8] This figure shows an example of the configuration of the control circuit according to Embodiment 1. [Modes for carrying out the invention]

[0010] The following describes in detail, with reference to the drawings, the modeling planning support system, learning device, computing device, modeling planning support method, and program according to the embodiment.

[0011] Embodiment 1. Figure 1 shows an example of the configuration of an additive manufacturing apparatus 10 according to Embodiment 1. The additive manufacturing apparatus 10 manufactures metal additive manufactured objects, which are three-dimensional additive manufactured objects made of metal material.

[0012] In Embodiment 1, the additive manufacturing method used by the additive manufacturing apparatus 10 is assumed to be Directed Energy Deposition (DED). However, the additive manufacturing method used by the additive manufacturing apparatus 10 is not limited to DED, and may also be Powder Bed Fusion (PBF), or other methods.

[0013] The additive manufacturing apparatus 10 supplies material to a commanded position and forms an object 22 using the material melted by the beam. The object 22 is a solidified product obtained when the molten material solidifies. The object 22 is also called a bead.

[0014] The additive manufacturing apparatus 10 forms a molten pool 23 by melting material together with a portion of the workpiece. A molded object 22 is formed when the molten pool 23 solidifies. The additive manufacturing apparatus 10 manufactures a three-dimensional additively manufactured object by sequentially stacking layers of the molded object 22. The workpiece is an object to which the molten material is added, and is either a base material 21 or a molded object 22.

[0015] In Embodiment 1, the beam that serves as the heat source for melting the material is a laser beam L. The heat source beam is not limited to a laser beam L, but may also be an electron beam or an arc. In Embodiment 1, the material is a metal wire W. The material is not limited to a wire W, but may also be a metal powder.

[0016] The X, Y, and Z axes are three axes perpendicular to each other. The X and Y axes are two horizontal axes. The Z axis is a vertical axis. In each of the X, Y, and Z axis directions, the direction indicated by the arrow is considered positive, and the direction opposite to the arrow is considered negative. The positive Z direction is considered to be the vertically upward direction. The layers of the fabricated object 22 are stacked in the positive Z direction.

[0017] The additive manufacturing apparatus 10 includes a stage 11, a beam irradiation device 12, a gas injection device 13, a wire supply device 14, a head drive device 15, a laser oscillator 16, an evaluation value acquisition device 17, a temperature adjustment device 18, and a control device 19. In FIG. 1, the dashed arrows represent signal lines.

[0018] On the stage 11, a base material 21, which is the base of the three-dimensional additive manufactured object, is placed. In the example shown in FIG. 1, the stage 11 is a plate material. The stage 11 may be other than a plate material.

[0019] The beam irradiation device 12 irradiates a workpiece with a laser beam L. When the laser beam L is irradiated, the workpiece and the wire W are melted, and a molten pool 23 is formed. The additive manufactured object 22 is formed by cooling and solidifying the molten pool 23.

[0020] The gas injection device 13 injects a shielding gas G toward the processing point, which is the irradiation position of the laser beam L. An example of the shielding gas G is an inert gas such as nitrogen or argon. By injecting the shielding gas G, oxidation of the workpiece at the processing point is prevented, and the workpiece is cooled.

[0021] In the example shown in FIG. 1, the gas injection device 13 is installed on the outer peripheral surface of the beam irradiation device 12. The gas injection device 13 injects the shielding gas G in the direction of the central axis of the laser beam L emitted from the beam irradiation device 12. That is, the beam irradiation device 12 and the gas injection device 13 are arranged coaxially with each other. In the example shown in FIG. 1, the beam irradiation device 12 and the gas injection device 13 are integrated with each other to form a processing head.

[0022] The gas injection device 13 is provided to inject shielding gas G into the area including the processing point in order to prevent oxidation of the workpiece at the processing point irradiated by the laser beam L. The gas injection device 13 may inject shielding gas G toward the processing point from a direction oblique to the Z-axis. The gas injection device 13 does not have to be integrated with the beam irradiation device 12. The gas injection device 13 may inject shielding gas G oblique to the central axis of the laser beam L emitted from the beam irradiation device 12.

[0023] The wire supply device 14 is a device that supplies wire W toward the machining point. In one example, the wire supply device 14 includes a wire spool around which the wire W is wound, a rotary motor that rotates the wire spool around a rotation axis, and a wire nozzle that advances the wire W from the wire spool toward the machining point. In Figure 1, the wire spool, rotary motor, and wire nozzle are not shown.

[0024] The head drive unit 15 moves the beam irradiation device 12 in the X-axis direction, the Y-axis direction, and the Z-axis direction, respectively. The head drive unit 15 includes a servo motor that constitutes an operating mechanism for moving the beam irradiation device 12 in the X-axis direction, a servo motor that constitutes an operating mechanism for moving the beam irradiation device 12 in the Y-axis direction, and a servo motor that constitutes an operating mechanism for moving the beam irradiation device 12 in the Z-axis direction. The head drive unit 15 is an operating mechanism that enables translational motion in the X-axis direction, the Y-axis direction, and the Z-axis direction, respectively. Figure 1 omits the illustration of each servo motor.

[0025] The additive manufacturing apparatus 10 moves the irradiation position of the laser beam L on the workpiece by moving the beam irradiation device 12 using a head drive device 15. The head drive device 15 is not limited to moving the beam irradiation device 12 in three mutually perpendicular axes. A multi-axis drive device such as a robot arm may be used for the head drive device 15. In the example shown in Figure 1, the beam irradiation device 12 and the gas injection device 13 are integrated together to form a processing head. In the example shown in Figure 1, the head drive device 15 drives the processing head.

[0026] In the example shown in Figure 1, the beam irradiation device 12 emits a laser beam L in the Z-axis direction. The wire supply device 14 is located at a distance from the beam irradiation device 12 in the XY plane. The wire supply device 14 advances the wire W toward the processing point from a direction oblique to the Z-axis. The wire nozzle of the wire supply device 14 may be fixed to the beam irradiation device 12. The wire nozzle of the wire supply device 14 may be fixed to the beam irradiation device 12 such that the direction in which the wire W advances is the Z-axis direction.

[0027] The laser oscillator 16 is a beam source such as a solid-state laser, gas laser, fiber laser, or semiconductor laser. The laser oscillator 16 emits a laser beam L. The laser oscillator 16 and the beam irradiation device 12 are connected by a fiber optic cable. The fiber optic cable is an optical transmission path that propagates the laser beam L from the laser oscillator 16 to the beam irradiation device 12.

[0028] The evaluation value acquisition device 17 is a device that performs measurements to acquire evaluation values. Details of the evaluation values ​​will be described later. The measurement area of ​​the evaluation value acquisition device 17 is the area that includes the entire area in which the molded object 22 is formed. When the molded object 22 is formed on the substrate 21, the measurement area includes the entire area in which the molded object 22 is formed and the substrate 21. The evaluation value acquisition device 17 measures, for example, at least one of the crack length, defect size, and temperature distribution of the molded object 22. Examples of the evaluation value acquisition device 17 include an inspection device that inspects cracks or defects using laser ultrasonic methods or acoustic emission (AE) sensors, or a temperature measuring instrument equipped with an infrared camera.

[0029] The temperature control device 18 heats or cools the substrate 21 so that its temperature is maintained at a temperature set by the control device 19. Examples of the temperature control device 18 include heating devices that perform heating by high-frequency induction, heating devices such as hot plates with heating wires, cooling devices such as heat sinks that perform water cooling, and cooling devices that use thermoelectric cooling elements such as Peltier modules. Alternatively, the temperature control device 18 may be a combination of the above heating device and the above cooling device. In the example shown in Figure 1, the temperature control device 18 is provided on the stage 11 to heat or cool the lower surface of the substrate 21.

[0030] The control device 19 controls the additive manufacturing apparatus 10 according to the processing program. Specifically, the control device 19 controls the beam irradiation device 12, the gas injection device 13, the wire supply device 14, the head drive device 15, the laser oscillator 16, and the temperature control device 18. An example of the control device 19 is a numerical control device. The control device 19 controls the head drive device 15 by outputting a movement command to the head drive device 15. When a movement command is input to the head drive device 15, it moves the beam irradiation device 12 according to the movement command.

[0031] The control device 19 controls the laser oscillation of the laser oscillator 16 by outputting commands to the laser oscillator 16 according to the conditions of the beam output, which is the output of the laser beam L. The control device 19 controls the wire supply device 14 by outputting commands to the wire supply device 14 according to the conditions of the amount of material supplied. The control device 19 adjusts the feeding speed of the wire W toward the processing point by controlling the drive of the wire supply device 14, specifically the drive of the rotary motor. The feeding speed represents the amount of material supplied per unit time.

[0032] The control device 19 controls the amount of shielding gas G injected from the gas injection device 13 by outputting commands to the gas injection device 13 according to the conditions for the amount of shielding gas G supplied.

[0033] Next, the molding planning support system according to Embodiment 1 will be described. Figure 2 is a diagram showing an example of the configuration of the molding planning support system 30 according to Embodiment 1. The molding planning support system 30 assists in formulating a plan for additive manufacturing by the additive manufacturing apparatus 10.

[0034] The molding planning support system 30 adjusts the molding conditions to obtain a target evaluation value. The molding planning support system 30 uses machine learning techniques to adjust the molding conditions. The molding planning support system 30 comprises a learning device 31 and a computing device 32. The learning device 31 is responsible for the learning phase, which generates a trained model by learning the molding conditions and evaluation values. The computing device 32 is responsible for the search phase, which estimates the molding conditions to obtain the target evaluation value.

[0035] Figure 3 shows an example of the configuration of a learning device 31 according to Embodiment 1. The learning device 31 comprises a first data acquisition unit 41 and a learning unit 42.

[0036] The first data acquisition unit 41 acquires the molding conditions, which are the conditions for additive manufacturing by the additive manufacturing device 10, and evaluation values, which indicate the evaluation of the additive manufacturing by the additive manufacturing device 10. As a result, the first data acquisition unit 41 acquires training data that includes the corresponding molding conditions and evaluation values.

[0037] The molding conditions acquired by the first data acquisition unit 41 include at least one of the following: beam output, beam diameter, axis feed speed, material supply speed, shielding gas G supply speed, target temperature due to heating or cooling of the substrate 21, interlayer waiting time, work distance, thickness per layer of the 3D additive manufactured object, column pitch, material composition information, and dimensional information of the manufactured object 22. The axis feed speed is the speed at which the beam irradiation position on the workpiece moves. The interlayer waiting time is the waiting time from when one layer of the manufactured object 22 is formed until the next layer is formed. The work distance is the distance between the processing head of the additive manufacturing apparatus 10 and the workpiece. The column pitch is the interval between the beam scanning columns on the workpiece. The material composition information is information indicating the content of each of several elements contained in the material. The dimensional information of the manufactured object 22 is information indicating the dimensions of the manufactured object 22, such as length, width, height, diameter, and thickness.

[0038] The evaluation values ​​acquired by the first data acquisition unit 41 include at least one of the following: mechanical strength of the molded object 22, grain size in the internal structure of the molded object 22, crack length of the molded object 22, defect size of the molded object 22, residual stress of the molded object 22, deformation amount of the molded object 22, cooling rate during molding, interlayer temperature during molding, temperature gradient during molding, and solidification rate during molding. The deformation amount of the molded object 22 is the deformation amount of the molded object 22 when its shape changes due to some influence. The cooling rate during molding is the temperature change per unit time during the cooling process in which the molten pool 23 solidifies. The interlayer temperature during molding is the temperature of the workpiece from the formation of one layer to the formation of the next layer of the molded object 22. The temperature gradient during molding is the value obtained by dividing the temperature difference between any two points on the molded object 22 by the distance between those two points. The solidification rate during molding is the speed at which the solid-liquid interface, which is the interface between the molten pool 23 and the molded object 22, moves. The first data acquisition unit 41 acquires evaluation values ​​from, for example, the evaluation value acquisition device 17.

[0039] The learning unit 42 learns a prediction function and a predicted variance of the evaluation value for predicting the evaluation value corresponding to the molding conditions, based on training data that includes the corresponding molding conditions and evaluation values. The learning unit 42 generates a trained model for inferring the prediction function and predicted variance from the molding conditions and evaluation values ​​through learning based on the training data. Known methods such as Gaussian process regression can be used for training by the learning unit 42.

[0040] The learning unit 42 outputs a prediction function and a predicted variance by inputting the molding conditions and evaluation values ​​into the trained model. That is, the learning unit 42 outputs a prediction function and a predicted variance inferred from the input molding conditions and the input evaluation values. The learning device 31 outputs the prediction function and a predicted variance to the arithmetic unit 32.

[0041] In one example, the learning device 31 is provided with the additive manufacturing apparatus 10. The learning device 31 may also be an external device to the additive manufacturing apparatus 10. In this case, the learning device 31 is connected to the additive manufacturing apparatus 10 via a network. The learning device 31 may also be implemented by, for example, a cloud server. A cloud server is a server built in a cloud environment that includes computing resources provided on a cloud service platform.

[0042] Figure 4 is a diagram illustrating the overview of learning by the learning device 31 according to Embodiment 1. Figure 4 shows an example of the relationship between the molding conditions and the evaluation value corresponding to those molding conditions. Here, in order to simplify the explanation of this relationship, the molding conditions are assumed to consist of one parameter. In the graph shown in Figure 4, the horizontal axis, x, represents the molding conditions. In the graph shown in Figure 4, the vertical axis, f(x), represents the evaluation value.

[0043] The first data acquisition unit 41 acquires the molding conditions and the corresponding evaluation values ​​as initial data in advance. The method for acquiring the evaluation values ​​is not particularly limited, and the evaluation values ​​can be determined by any method. For example, the cooling rate under arbitrary molding conditions can be determined as an evaluation value by thermal analysis using the finite element method. Alternatively, the evaluation values ​​can be determined by actually measuring the temperature distribution during additive manufacturing by the additive manufacturing apparatus 10 using the evaluation value acquisition device 17.

[0044] The learning unit 42 outputs a prediction function and predicted variance of the evaluation value using methods such as Gaussian process regression, based on the acquired initial data. In the graph shown in Figure 4, the variance increases as the plot moves away from the initial data, which are the actual measured values.

[0045] Figure 5 shows an example of the configuration of the calculation device 32 according to Embodiment 1. The calculation device 32 comprises a second data acquisition unit 51 and a molding condition determination unit 52. The molding condition determination unit 52 comprises a calculation unit 53 and an estimation unit 54.

[0046] The second data acquisition unit 51 acquires the prediction function and prediction variance output by the learning device 31. The second data acquisition unit 51 outputs the prediction function and prediction variance to the molding condition determination unit 52. Based on the prediction function and prediction variance, the molding condition determination unit 52 determines the molding conditions necessary to obtain the target evaluation value, which is the target of the evaluation value.

[0047] The calculation unit 53 outputs an acquisition function that derives the molding conditions from the prediction function and prediction variance output from the second data acquisition unit 51. The acquisition function is a function that determines the molding conditions to obtain the target evaluation value based on the prediction function and prediction variance. The function to be used as the acquisition function is selected by the user of the additive manufacturing apparatus 10. The user can select any function as the acquisition function, taking into account the type of evaluation value for which optimization is desired or the characteristics of the function.

[0048] Known functions such as UCB (Upper Confidence Bound), LCB (Lower Confidence Bound), EI (Expected Improvement), and PI (Probability of Improvement) can be applied to the acquisition function. PI makes it easy to select candidates near the current maximum value, but it has the disadvantage of being prone to getting stuck in local optima. LCB is the opposite of UCB and is used when you want to find the minimum value of the evaluation. The calculation unit 53 outputs the acquisition function to the estimation unit 54.

[0049] The estimation unit 54 estimates the molding conditions for obtaining the target evaluation value based on the acquisition function and outputs the estimated molding conditions. The arithmetic unit 32 outputs the molding conditions for obtaining the target evaluation value.

[0050] In one example, the computing unit 32 is provided in the additive manufacturing apparatus 10. The computing unit 32 may also be an external device to the additive manufacturing apparatus 10. In this case, the computing unit 32 is connected to the additive manufacturing apparatus 10 via a network.

[0051] Figure 6 is a diagram illustrating the outline of the process by which the calculation device 32 according to Embodiment 1 determines the molding conditions. Figure 6 shows an example of the relationship between the molding conditions and the evaluation value corresponding to those molding conditions. Here, in order to simplify the explanation of this relationship, the molding conditions are assumed to consist of one parameter. In the graph shown in Figure 6, the horizontal axis, x, represents the molding conditions. In the graph shown in Figure 6, the vertical axis, f(x), represents the evaluation value.

[0052] Here, the acquisition function output by the arithmetic unit 53 is assumed to be a UCB. The UCB(x) corresponding to the molding condition x is expressed by the following equation (1). UCB(x) = μ(x) + kσ(x) ... (1)

[0053] In equation (1), μ(x) represents the predicted value of the evaluation value at the molding condition x. k represents an arbitrary constant. σ(x) is the variance of the predicted value of the evaluation value at the molding condition x. In UCB, the variance is added to the predicted value, and the molding condition at which UCB is maximized is determined as the molding condition for obtaining the target evaluation value. Figure 6 shows the same initial data, prediction function, and predicted variance as in Figure 4, as well as an example of UCB and an example of a candidate point at which UCB is maximized.

[0054] By changing the value of k in equation (1), we can change the weighting used to adjust whether we search for candidates near the current maximum value or explore the unknown portion with a large variance. Therefore, we can determine the value of k depending on whether we want to search for candidates near the current maximum value or explore the unknown portion with a large variance.

[0055] If candidate points are obtained without considering variance, it may be possible to fall into a local minimum and be unable to derive a more optimal solution. According to Embodiment 1, the computing device 32 can search for a global optimal solution without falling into a local minimum.

[0056] Next, the processing procedures performed by the molding planning support system 30 and the additive manufacturing apparatus 10 will be described. Figure 7 is a flowchart showing an example of the processing procedures performed by the molding planning support system 30 and the additive manufacturing apparatus 10 according to Embodiment 1.

[0057] In step S1, the first data acquisition unit 41 of the learning device 31 acquires the molding conditions and evaluation values. In step S2, the learning unit 42 of the learning device 31 performs a learning process based on the training data, which is a combination of molding conditions and evaluation values ​​acquired in step S1. The learning unit 42 learns the prediction function and prediction variance of the evaluation values.

[0058] The second data acquisition unit 51 of the arithmetic unit 32 acquires the prediction function and prediction variance output from the learning device 31. In step S3, the arithmetic unit 53 of the arithmetic unit 32 performs calculations to obtain the acquisition function based on the prediction function and prediction variance output from the second data acquisition unit 51. Based on the prediction function and prediction variance, the arithmetic unit 53 outputs an acquisition function that derives the molding conditions from the prediction function and prediction variance.

[0059] In step S4, the estimation unit 54 of the arithmetic unit 32 determines the molding conditions for obtaining the target evaluation value based on the acquisition function output from the arithmetic unit 53.

[0060] Subsequently, in step S5, the additive manufacturing apparatus 10 acquires evaluation values ​​based on the manufacturing conditions determined in step S4. The additive manufacturing apparatus 10 acquires evaluation values ​​for additive manufacturing under the determined manufacturing conditions. The method for acquiring evaluation values ​​is arbitrary. The additive manufacturing apparatus 10 may, for example, acquire evaluation values ​​by CAE (Computer Aided Engineering) analysis. The additive manufacturing apparatus 10 may also acquire evaluation values ​​by measurement during manufacturing or by evaluation of test pieces taken from the 3D additive manufactured object.

[0061] In step S6, the additive manufacturing apparatus 10 determines whether the evaluation value obtained in step S5 satisfies the target. If it is determined that the evaluation value satisfies the target (step S6, Yes), the manufacturing planning support system 30 and the additive manufacturing apparatus 10 terminate the process according to the procedure shown in Figure 7.

[0062] On the other hand, if the evaluation value is determined not to satisfy the target (step S6, No), the molding planning support system 30 returns to step S1. The first data acquisition unit 41 of the learning device 31 receives the molding conditions determined in step S4 and the evaluation value acquired in step S5 as new learning data. The process from step S1 to step S6 is repeated until the evaluation value acquired in step S5 satisfies the target.

[0063] Next, the hardware configuration for realizing the learning device 31 and the arithmetic device 32 will be described. Each of the learning device 31 and the arithmetic device 32 is realized by a processing circuit. The processing circuit is, for example, a circuit in which a processor executes software.

[0064] When the processing circuit is implemented by software, the processing circuit is, for example, the control circuit 60 shown in Figure 8. Figure 8 is a diagram showing an example configuration of the control circuit 60 according to Embodiment 1. The control circuit 60 comprises an input unit 61, a processor 62, a memory 63, and an output unit 64. The input unit 61 is an interface circuit that receives data input from outside the control circuit 60 and provides it to the processor 62. The output unit 64 is an interface circuit that sends data from the processor 62 or the memory 63 to the outside of the control circuit 60.

[0065] The learning device 31 is implemented by software, firmware, or a combination of software and firmware. The software or firmware is written as a program and stored in memory 63. The control circuit 60 implements the functions of the learning device 31 or the arithmetic unit 32 by having the processor 62 read and execute the program stored in memory 63. In other words, the control circuit 60 has memory 63 for storing the program that will result in the execution of each process of the learning device 31 and the arithmetic unit 32. This program can also be said to cause the computer system to execute the procedures and methods for each process of the learning device 31 and the arithmetic unit 32. Memory 63 is also used as temporary memory when the processor 62 executes various processes.

[0066] The learning unit 42 of the learning device 31 and the calculation unit 53 and estimation unit 54 of the arithmetic unit 32 are realized by using a processor 62 and memory 63. The processor 62 is a CPU (Central Processing Unit). The processor 62 may also be a central processing unit, processing unit, arithmetic unit, microprocessor, microcomputer, processor, or DSP (Digital Signal Processor). The memory 63 may be non-volatile or volatile semiconductor memory such as RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory), EEPROM (Registered Trademark) (Electrically Erasable Programmable Read Only Memory), magnetic disks, flexible disks, optical disks, compact disks, minidiscs, or DVDs (Digital Versatile Discs). The first data acquisition unit 41 of the learning device 31 and the second data acquisition unit 51 of the arithmetic unit 32 are realized by using an input unit 61.

[0067] The functions of the learning device 31 and the arithmetic device 32 may be implemented by dedicated hardware circuits. These dedicated hardware circuits include processing circuits. The processing circuits may be single circuits, composite circuits, programmed processors, parallel programmed processors, ASICs (Application Specific Integrated Circuits), FPGAs (Field Programmable Gate Arrays), or combinations thereof. Furthermore, the learning device 31 and the arithmetic device 32 may each be implemented by combining the control circuit 60 with hardware circuits.

[0068] In one example, the learning device 31 and the arithmetic unit 32 are distributed across separate hardware configurations. The learning device 31 and the arithmetic unit 32 may also be integrated into a single hardware configuration.

[0069] The program according to Embodiment 1 may be provided on a recording medium such as a CD (Compact Disc)-ROM or DVD-ROM. The program according to Embodiment 1 may be provided by being stored on a computer connected to a network such as the Internet and downloaded via the Internet or other network. The program according to Embodiment 1 may be provided or distributed via a network such as the Internet.

[0070] According to Embodiment 1, the learning device 31 takes the molding conditions, which are the conditions for additive manufacturing, and the evaluation value, which indicates the evaluation of additive manufacturing, as input information, and outputs a prediction function and a prediction variance of the evaluation value corresponding to the molding conditions. The calculation device 32 determines the molding conditions to obtain the target evaluation value based on the prediction function and the prediction variance. Compared to conventional adjustment methods that are dependent on individual expertise, the molding planning support system 30 makes it possible to efficiently obtain molding conditions that satisfy the target evaluation value with fewer trials. Furthermore, the molding planning support system 30 can reduce variability in prediction accuracy and variability in learning time. In additive manufacturing, where molding conditions consisting of many parameters are used, the molding planning support system 30 makes it easy to adjust the molding conditions to obtain the target evaluation value even if the operator does not have advanced expertise. As a result, the molding planning support system 30 has the effect of easily optimizing the molding conditions.

[0071] The configurations shown in the embodiments described above are examples of the content of this disclosure. The configurations of the embodiments can be combined with other known technologies. Some parts of the configurations of the embodiments can be omitted or modified without departing from the spirit of this disclosure. [Explanation of symbols]

[0072] 10 Additive manufacturing device, 11 Stage, 12 Beam irradiation device, 13 Gas injection device, 14 Wire supply device, 15 Head drive device, 16 Laser oscillator, 17 Evaluation value acquisition device, 18 Temperature control device, 19 Control device, 21 Substrate, 22 Manufactured object, 23 Weld pool, 30 Manufacturing planning support system, 31 Learning device, 32 Calculation device, 41 First data acquisition unit, 42 Learning unit, 51 Second data acquisition unit, 52 Manufacturing condition determination unit, 53 Calculation unit, 54 Estimation unit, 60 Control circuit, 61 Input unit, 62 Processor, 63 Memory, 64 Output unit, G Shielding gas, L Laser beam, W Wire.

Claims

1. A manufacturing planning support system that assists in formulating a plan for additive manufacturing using an additive manufacturing device, A first data acquisition unit that acquires the molding conditions, which are the conditions for the additive manufacturing process, and an evaluation value that indicates the evaluation of the additive manufacturing process. A learning unit learns a prediction function and a predicted variance of the evaluation value for predicting the evaluation value corresponding to the molding condition, based on training data including the molding conditions and evaluation value associated with each other, and outputs the prediction function and the predicted variance inferred from the input molding conditions and the input evaluation value. A learning device equipped with, A second data acquisition unit acquires the prediction function and the prediction variance output by the learning device, A molding condition determination unit determines the molding conditions for obtaining the target evaluation value, which is the target of the evaluation value, based on the prediction function and the prediction variance. A computing device having A design planning support system characterized by the following features.

2. The molding condition determination unit is, A calculation unit that outputs an acquisition function for deriving the molding conditions from the prediction function and the prediction variance, The system includes an estimation unit that estimates the molding conditions for obtaining the target evaluation value based on the acquisition function and outputs the estimated molding conditions. The molding planning support system according to feature 1.

3. The molding conditions include at least one of the following: beam output and beam diameter of the beam used to melt the material; the speed at which the beam irradiation position on the workpiece moves; the material supply speed; the supply speed of the shielding gas supplied to the irradiation position; the target temperature of the substrate on which the molded object is formed by heating or cooling; the interlayer waiting time during molding; the distance between the processing head of the additive manufacturing apparatus and the workpiece; the thickness of each layer of the additively manufactured object; the spacing of the beam scanning trains on the workpiece; the composition information of the material; and the dimensional information of the molded object. The molding planning support system according to claim 1 or 2.

4. The evaluation value includes at least one of the following: the mechanical strength of the molded object, the grain size of the crystals in the internal structure of the molded object, the crack length of the molded object, the defect size of the molded object, the residual stress of the molded object, the amount of deformation of the molded object, the cooling rate during molding, the interlayer temperature during molding, the temperature gradient during molding, and the solidification rate during molding. The molding planning support system according to claim 1 or 2.

5. A data acquisition unit that acquires the molding conditions, which are the conditions for additive manufacturing using an additive manufacturing device, and an evaluation value that indicates the evaluation of the additive manufacturing, The system includes a learning unit that learns a prediction function and a predicted variance of the evaluation value for predicting the evaluation value corresponding to the molding condition, based on training data including the molding conditions and evaluation value associated with each other, and outputs the prediction function and the predicted variance inferred from the input molding conditions and the input evaluation value. A learning device characterized by the following features.

6. A data acquisition unit that acquires a prediction function for predicting an evaluation value that indicates the evaluation of additive manufacturing by an additive manufacturing device and corresponds to the manufacturing conditions which are conditions for the additive manufacturing, and the predicted variance of the evaluation value, The system includes a molding condition determination unit that determines the molding conditions for obtaining a target evaluation value, which is the target of the evaluation value, based on the prediction function and the prediction variance. A computing device characterized by the following features.

7. A method for supporting the formulation of a manufacturing plan for additive manufacturing using an additive manufacturing device, A step of obtaining the printing conditions, which are the conditions for the additive manufacturing process, and an evaluation value that indicates the evaluation of the additive manufacturing process. A step of learning a prediction function for predicting the evaluation value corresponding to the molding conditions and the predicted variance of the evaluation value based on training data including the corresponding molding conditions and evaluation values, A step of determining the prediction function and the prediction variance inferred from the input molding conditions and the input evaluation value, The step includes determining the molding conditions for obtaining a target evaluation value, which is the target of the evaluation value, based on the prediction function and the prediction variance. A method for supporting design planning, characterized by the features described above.

8. In the computer system, A step of obtaining the printing conditions, which are the conditions for additive manufacturing using an additive manufacturing device, and an evaluation value that indicates the evaluation of the additive manufacturing, A step of learning a prediction function for predicting the evaluation value corresponding to the molding conditions and the predicted variance of the evaluation value based on training data including the corresponding molding conditions and evaluation values, A step of determining the prediction function and the prediction variance inferred from the input molding conditions and the input evaluation value, The process involves determining the molding conditions for obtaining the target evaluation value, which is the target of the evaluation value, based on the prediction function and the prediction variance. A program characterized by the following features.

Images