Additive manufacturing condition search device and additive manufacturing condition search method

Abstract

To improve the efficiency of searching for an optimum solution of an additive manufacturing condition for an additive manufacturing apparatus.SOLUTION: An additive manufacturing condition searching apparatus 2 includes:a defect database storing materials, shape information, additive manufacturing conditions, monitoring information for a forming process and defect information in association with each other; a first machine learning unit 47 which outputs additive manufacturing conditions dependent on material information and apparatus information, and outputs new additive manufacturing conditions from combinations of a plurality of additive manufacturing conditions and defect information; an identification unit 43 which makes an additive manufacturing apparatus perform forming with the additive manufacturing conditions to acquire monitoring information for the forming process, and acquires shape information and defect information by examining a molded article; a second machine learning unit 48 in which a model which has learned the defect database as teacher data estimates defect information for the molded article from the monitoring information and stores the defect information in the defect database; and a determination unit 44 which determines whether the defect information for the molded article achieves an evaluation target value or not.SELECTED DRAWING: Figure 12

Description

Technical Field

[0001] The present invention relates to an additive manufacturing condition exploration device and an additive manufacturing condition exploration method.

Background Art

[0002] It is known that additive manufacturing (layered manufacturing) includes, for example, a powder bed fusion method, a directed energy deposition method, and the like. In the powder bed fusion method, additive manufacturing is performed by irradiating a flatly spread powder with a light beam (such as a laser beam or an electron beam). The powder bed fusion method includes SLM (Selective Laser Melting) and EBM (Electron Beam Melting). The directed energy deposition method performs additive manufacturing by controlling the position of a head that irradiates a light beam and discharges a powder material. The directed energy deposition method includes LMD (Laser Metal Deposition) and DMP (Direct Metal Deposition).

[0003] On the other hand, for these additive manufacturing methods, it is necessary to set appropriate additive manufacturing conditions (recipes) according to the material. In particular, in the powder bed fusion method, there are many types of parameters to be controlled, and a lot of labor is required to derive appropriate additive manufacturing conditions.

[0004] By the way, with the recent improvement in the processing speed of computers, artificial intelligence has been rapidly developing. For example, Patent Document 1 describes generating laser processing condition data by machine learning. These describe determining the quality of shaping by monitoring the state during shaping.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

SUMMARY OF THE INVENTION

PROBLEMS TO BE SOLVED BY THE INVENTION

[0006] In powder bed fusion, since there are many factors of additive manufacturing conditions, it is not easy to determine the optimal factors. For the condition of filling the inside of the molded object alone, there are factors such as heat source output, scanning speed, scanning line interval, scanning line length, layer thickness, etc. In order to obtain the optimal conditions, it is necessary to manufacture and evaluate a large number of molded objects. Therefore, the construction of the process window has problems that require enormous costs and time, and that the process window with individual differences is constructed by the person conducting the experiment.

[0007] On the other hand, a method is used in which machine learning is used to match control factors with molding results, and regression analysis is performed for optimization. However, even when using this method, since the control factors of the additive manufacturing conditions and the settable range are wide, if the conditions are set comprehensively, the number of evaluations increases, and many conditions in which the manufacturing result (evaluation score) deteriorates are included. If many of the manufacturing results are poor, it takes time for the analysis to converge until the appropriate conditions are found.

[0008] In powder bed fusion, since evaluation samples of multiple conditions are manufactured at once, if there are many samples with poor manufacturing results, it may have an adverse effect on the samples with good conditions. For example, in samples with poor results, it may come into contact with a tool for pushing powder called a squeegee during powder spreading, or generate relatively large spatter and scatter into the manufacturing area with good conditions. There is also a problem that it takes time for a person to assign according to the control factors for which the initial learning conditions can be set.

[0009] Therefore, an object of the present invention is to improve the efficiency of searching for the optimal solution of the additive manufacturing conditions of an additive manufacturing apparatus.

MEANS FOR SOLVING THE PROBLEM

[0010] To solve the above problems, the additive manufacturing condition exploration apparatus of the present invention includes a defect database that stores materials, shape information, additive manufacturing conditions, monitoring information during manufacturing, and defect information in association with each other, a first machine learning unit that outputs additive manufacturing conditions according to material information and apparatus information, and outputs new additive manufacturing conditions from combinations of a plurality of additive manufacturing conditions and defect information, a specific unit that causes an additive manufacturing apparatus to perform manufacturing according to the additive manufacturing conditions and acquires monitoring information during manufacturing Do and a second machine learning unit that estimates defects of the manufactured object from the monitoring information and stores the defects in the defect database, and a determination unit that determines whether or not the defect information of the manufactured object has achieved an evaluation target value by from the monitoring information determine the defect information of the shaped object from, and until the of the manufactured object shaping result score and stores the defects in the defect database, and a determination unit that determines whether or not the defect information of the manufactured object has achieved an evaluation target value the determination unit determines that the defect information of the shaped object has achieved the evaluation target value, output a new additional manufacturing condition from a combination of a plurality of additional manufacturing conditions and defect information by the first machine learning unit, cause the shaping to be performed by the additional manufacturing apparatus under the additional manufacturing condition by the specifying unit, acquire monitoring information during shaping, and a model learned using the defect database as teacher data determines the defect information of the shaped object from the monitoring information, estimates the shaping result score of the shaped object, and stores it in the defect database, and repeat a series of operations It is characterized by this

[0011] The additive manufacturing condition exploration method of the present invention The first machine learning unit includes a step of outputting additive manufacturing conditions according to material information and apparatus information, or outputting new additive manufacturing conditions from combinations of a plurality of additive manufacturing conditions and defect information, The specifying unit a step of causing an additive manufacturing apparatus to perform manufacturing according to the additive manufacturing conditions and acquiring monitoring information during manufacturing Do and a step of The second machine learning unit learning a model using, as teacher data, a defect database of combinations of monitoring information and defect information during manufacturing, by estimating defect information of the manufactured object from the monitoring information, and storing the defect information in the defect database, The determination unit and a step of determining whether or not the defect information of the manufactured object has achieved an evaluation target value and until the determination unit determines that the defect information of the shaped object has achieved the evaluation target value, the first machine learning unit outputs a new additional manufacturing condition from a combination of a plurality of additional manufacturing conditions and defect information; the specifying unit causes the additional manufacturing apparatus to perform shaping under the additional manufacturing condition and acquires monitoring information during shaping; and the second machine learning unit estimates the defect information of the shaped object from the monitoring information by a model learned using a combination of the monitoring information during shaping and the defect database as teacher data, and stores it in the defect database, and repeat these steps It is characterized by this Other means will be described in the mode for carrying out the invention

Effects of the Invention

[0012] According to the present invention, it is possible to improve the efficiency of searching for the optimal solution of the additive manufacturing conditions of the additive manufacturing apparatus

Brief Description of the Drawings

[0013]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22

[0014] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0015] Additive manufacturing equipment FIG. 1 is a configuration diagram showing an example of an additive manufacturing device 5. An additive manufacturing apparatus 5 to which the additive manufacturing condition exploration device and additive manufacturing condition exploration method of this embodiment can be applied will be described with reference to FIG. 1. The additive manufacturing apparatus 5 is a powder bed fusion type. The additive manufacturing apparatus 5 irradiates a light beam onto metal powder laid in layers to heat it, causing the metal powder to melt and solidify. The additive manufacturing apparatus 5 then manufactures a molded object by repeatedly forming a powder bed and irradiating it with a light beam.

[0016] The additive manufacturing apparatus 5 includes a chamber 510, a gas supply unit 511, an exhaust mechanism 512, a material supply unit 514, an additive manufacturing unit 515, a recovery unit 516, a recoater 513, a light beam source 501, and a control unit 530.

[0017] Chamber 510 houses all of the components of additive manufacturing apparatus 5 except for light beam source 501 and exhaust mechanism 512. Chamber 510 has a transparent window 502 fitted with protective glass. This transparent window 502 allows the light beam emitted from light beam source 501 located outside chamber 510 to pass through and reach a powder bed placed on stage 518 of additive manufacturing unit 515 inside chamber 510.

[0018] The additive manufacturing apparatus 5 is also equipped with a temperature sensor 56, a pressure sensor 57, an oxygen sensor 58, and the like. The temperature sensor 56 includes a contact-type temperature sensor such as a thermocouple that measures the temperature of the stage 518, and a non-contact-type temperature sensor such as an infrared radiation thermometer that measures the temperature of the powder bed formed on the stage 518.

[0019] The pressure sensor 57 measures the pressure of the reduced pressure environment in the chamber 510. The oxygen sensor 58 measures the amount of oxygen (oxygen concentration) in the reduced pressure environment in the chamber 510. Although not shown, the chamber 510 may also have a camera that takes images of the powder bed formed on the stage 518 of the additive manufacturing unit 515, for example.

[0020] The gas supply unit 511 is connected to the chamber 510 and supplies an inert gas to the inside of the chamber 510. The gas supply unit 511 includes, for example, a gas supply source and a control valve, which are not shown. The gas supply source is configured as a high-pressure tank filled with an inert gas. The control valve is controlled by the control unit 530 and controls the flow rate of the inert gas supplied from the gas supply source to the chamber 510. For example, nitrogen or argon can be used as the inert gas.

[0021] The exhaust mechanism 512 is configured by a vacuum pump and is connected to the chamber 510 via a vacuum piping. The exhaust mechanism 512 is controlled by the control unit 530, and may exhaust gas from the chamber 510 to create a vacuum pressure inside the chamber 510 that is lower than atmospheric pressure, thereby creating a reduced pressure environment inside the chamber 510.

[0022] The material supply unit 514 is recessed and has an opening at the top, with an open top. The material supply unit 514 has a stage 517 that can be moved up and down on which the material powder is placed and supplied. The stage 517 forms the bottom wall of the material supply unit 514. The stage 517 is provided so that it can be raised and lowered at a predetermined pitch by an appropriate lifting mechanism. The lifting mechanism of the stage 517 is connected to and controlled by the control unit 530. The material supply unit 514 may also be of a type that drops the material powder instead of being of a lifting type.

[0023] Examples of material powders used in additive manufacturing of shaped objects include powders of metal materials such as hot work tool steel, copper, titanium alloys, nickel alloys, aluminum alloys, cobalt-chromium alloys, and stainless steel; powders of resin materials such as polyamide; and powders of ceramics.

[0024] Similar to the material supply unit 514 described above, the additive manufacturing unit 515 is recessed and capable of accommodating material powder, with an open top and an opening at the top end. The additive manufacturing unit 515 has a stage 518 for spreading the material powder to form a powder bed. The stage 518 forms the bottom wall of the additive manufacturing unit 515. The material powder supplied from the material supply unit 514 and a model to be manufactured by additive manufacturing are placed on the stage 518.

[0025] The opening of the additive manufacturing unit 515 and the opening of the material supply unit 514 are approximately equal in vertical height and aligned approximately horizontally. The additive manufacturing stage 518, like the material supply stage 517 described above, is provided so as to be able to be raised and lowered at a predetermined pitch by an appropriate lifting mechanism. The stage 518 may also be equipped with a preheating mechanism including a heater that preheats the stage 518. The lifting mechanism and preheating mechanism of the stage 518 are connected to and controlled by, for example, the control unit 530.

[0026] The recovery unit 516 is provided in a concave shape capable of accommodating the material powder, similar to the aforementioned material supply unit 514 for example, with an open top and an opening at the upper end. In the example shown in FIG. 1, the bottom wall of the recovery unit 516 is fixed to the lower end, but may be constituted by a liftable stage, similar to the material supply unit 514 and the additive manufacturing unit 515.

[0027] The opening of the recovery unit 516 and the opening of the additive manufacturing unit 515 are approximately equal in vertical height and are arranged approximately horizontally. The recovery unit 516 accommodates and recovers, for example, the excess material powder supplied from the material supply unit 514 to the additive manufacturing unit 515 by the recoater 513.

[0028] The recoater 513 forms a powder bed on the stage 518 by transporting the material powder supplied from the material supply unit 514 onto the stage 518 of the additive manufacturing unit 515 and spreading it evenly while leveling. The recoater 513 is provided with a moving mechanism. The moving mechanism is, for example, a linear motor, and moves the recoater 513 along the generally horizontal traveling direction from the material supply unit 514 towards the additive manufacturing unit 515. The light beam source 501 can use a laser light source that generates a light beam with an output of about several W to several kW. The light beam source 501 of the additive manufacturing apparatus 5 of the present embodiment is a single-mode fiber laser, that is, a laser light source that generates a laser with an energy intensity having a Gaussian distribution. Further, the light beam source 501 includes a galvanometer scanner for scanning the light beam on the powder bed.

[0029] Here, the light beam includes a laser beam and an electron beam, and also includes various beams capable of melting metal powder. Also, various lasers such as a near-infrared wavelength laser, a CO2 laser (far-infrared laser), and a semiconductor laser can be applied to the laser beam, and are appropriately determined according to the type of the target metal powder.

[0030] The control unit 530 is configured with a microcontroller and firmware. The control unit 530 includes a processing unit such as a CPU, a storage device such as a RAM (Random Access Memory) and a ROM (Read Only Memory), and an input / output unit that exchanges signals with programs and data stored in the storage device and with each unit of the additive manufacturing apparatus 5. The control unit 530 controls the gas supply unit 511, the exhaust mechanism 512, the material supply unit 514, the additive manufacturing unit 515, and the light beam source 501 by executing the programs stored in the storage device using the processing unit. In addition, the detection results of the temperature sensor 56, the pressure sensor 57, and the oxygen sensor 58, as well as the output of the camera, are input to the control unit 530.

[0031] Standard sample for exploring additive manufacturing conditions in powder bed fusion Next, we describe a standard sample used to explore additive manufacturing conditions for powder bed fusion.

[0032] Fig. 2 is a schematic diagram showing an example of the shape of the standard sample 1. Fig. 3 is a development view showing an example of the shape of the standard sample 1. Fig. 4 is a cross-sectional view showing an example of the shape of the standard sample 1.

[0033] The standard sample 1 has an overall hexahedral block shape, and the three surfaces, namely the bottom surface 16, top surface 12, and back surface 14, are smooth. Rectangular holes are drilled on the left side surface 13 and right side surface 15. The front surface 11 has parallelogram and circular hole shapes. In other words, the front surface 11 is a surface where hole shapes consisting of straight lines and curves are integrated.

[0034] As shown in Figure 4, the standard sample 1 includes two or more independent regions in the DD and GG cross sections cut from any one layer in the center of the stacking direction. These two or more regions are divided into a small region cut at a predetermined width from the outer edge of the standard sample 1 and a large region consisting of the rest of the sample.

[0035] For example, consider the case where a process window for "in-skin" conditions, in which a simple block shape fills the printing area, is derived and the additive manufacturing equipment 5 prints a detailed shape within that range. In areas with small areas, heat tends to accumulate, resulting in an over-melting state and causing deformation such as a bulging shape. When the additive manufacturing equipment 5 spreads the next layer of powder, the printed object may come into contact with the squeegee, causing the operation to stop or destroy the printed object.

[0036] By searching for additive manufacturing conditions using a standard sample 1 with the shape described above, it is possible to derive conditions that can be used in advance for objects with different build areas, especially those that include small areas.

[0037] Fig. 5 is a diagram showing three regions of the cross section of the standard sample 1. Fig. 6 is a schematic diagram of the front surface 11 of the standard sample 1. Fig. 7 is a schematic diagram of the left side surface 13 of the standard sample 1. The front surface 11 in Figure 6 has an upper end region 111, an upper side region 117 of the large parallelogram hole, and a lower side region 116. The front surface 11 has an upper side region 118 of the small parallelogram hole and a lower side region 119. The front surface 11 has an upper side region 112 of the large circular hole and a lower side region 113. The front surface 11 has an upper side region 114 of the small circular hole and a lower side region 115.

[0038] Areas 111, 116, 119, 113, and 115 are "up-skin" areas that form the top surface in the height direction of the object. Areas 117, 118, 112, and 114 are "down-skin" areas that form overhangs. The remaining areas are "in-skin" areas that fill the object and are the base areas for forming the object.

[0039] The left side surface 13 in Fig. 7 has an upper end region 131, an upper side region 132 of the rectangular hole, and a lower side region 133. Regions 131 and 133 are "up-skins" that form the outermost surface in the build height direction. Region 132 is a "down-skin" that forms an overhang.

[0040] "Down-skin," which forms an overhang, compares any slice data with the slice data one layer (or multiple layers) before it when a three-dimensional shape is sliced for each layer thickness, and finds that the slice data one layer (or multiple layers) before it does not have a printing area. Therefore, "Down-skin" sets different conditions than "In-skin," which fills in a printing area when a printing area exists in any slice data.

[0041] When a beam is irradiated onto powder, if there is nothing to bond with the molten powder or if heat cannot be dissipated quickly through thermal conduction, the molten powder will shrink into a spherical shape and form a relatively large spherical mass on top of the powder. This phenomenon is called "balling." When balling occurs, not only does it deteriorate the surface condition of the overhanging area, but the balled mass is also carried away when the powder is laid, leaving it empty. For this reason, conditions with reduced energy are selected for the "down-skin" that forms the overhang.

[0042] "Up-skin," which forms the top surface in the build height direction, compares any slice data with the slice data one layer (or multiple layers) later when a three-dimensional shape is sliced for each layer thickness, and finds that the slice data one layer (or multiple layers) later does not have a build area. Therefore, "Up-skin" sets different conditions for filling the build area when a build area exists in any slice data.

[0043] Conventionally, after setting the process window for filling the printing area using a simple block shape, the "Down-skin" conditions, which form the overhang for a tilted-shape sample, and the "Up-skin" conditions, which form the top surface in the printing height direction, are selected. However, the conditions for processing these three types of areas each affect the printing process. As a result, their effects change in areas with different printing areas. Therefore, by evaluating a shape such as Standard Sample 1 where each processing condition is effective, it is possible to approach the appropriate additive manufacturing conditions, and by utilizing machine learning, it is possible to arrive at the appropriate additive manufacturing conditions more quickly. In other words, the efficiency and / or accuracy of searching for the optimal additive manufacturing conditions for additive manufacturing equipment can be improved.

[0044] In standard sample 1, the areas for evaluating shape reproducibility are concentrated on the front surface 11. As a result, when inspecting standard sample 1, it is sufficient to acquire images or displacement data of the front surface 11. When measuring surface roughness and internal defect rate over a large area, even if shape reproducibility deteriorates (the shape becomes distorted and destroyed), the impact on surface roughness and internal defect rate is small. Therefore, each evaluation item can be measured accurately and easily. Furthermore, it is possible to change the difficulty of manufacturing by changing the angle and width of the parallelogram and the diameter of the circle.

[0045] Additive manufacturing conditions and measurement of printing results for standard samples In this embodiment, an example will be described in which several tens of control factors are changed to create a data set of several tens of additive manufacturing conditions, and a standard sample 1 is additively manufactured by an additive manufacturing device 5.

[0046] The control factors here are scale correction in the x, y, and z directions, the offset amount from the contour line in the CAD data for irradiating the contour line, the power and scanning speed of the light beam irradiating the contour line, the offset amount from the contour line irradiation position in the "in-skin" that fills the printing area, the scan line spacing, the scan pattern, the scan line length, the offset amount between scan patterns, the light beam power and scanning speed, the offset amount from the contour line irradiation position in the "down-skin" that forms the overhang, the scan line spacing, the scan pattern, the scan line length, the offset amount between scan patterns, the light beam power and scanning speed, and the offset amount from the contour line irradiation position in the "up-skin" that forms the outermost surface in the printing height direction, whether or not to perform double irradiation, the scan line spacing, the light beam power and scanning speed. However, the above control factors are merely examples and are not intended to limit the technical scope of the present invention to these alone.

[0047] 8 to 10 are diagrams showing an example of a method for measuring the additively manufactured standard sample 1. Figure 8 shows the measurement locations of the total height Z, width X and width Y of the additively manufactured standard sample 1 as a result of its shape. This measurement method measures the roughness of the left side surface 13, the right side surface 15 (not shown), and the top surface 12. Here, the roughness of each surface is expressed as the arithmetic mean roughness Ra, the maximum height Ry, and the ten-point mean roughness Rz.

[0048] As shown in Figure 9, this measurement method measures the average dimensional error of the width dimensions px1 to px4 of the larger parallelogram punched holes and the average dimensional error of the height dimension py1. This measurement method measures the average dimensional error of the width dimensions px5 to px8 of the smaller parallelogram punched holes and the average dimensional error of the height dimension py2. The average dimensional error of the width dimension cx2 of the larger circular punched hole and the average dimensional error of the height dimension cy2 are measured. The average dimensional error of the width dimension cx1 of the smaller circular punched hole and the average dimensional error of the height dimension cy1 are measured. The above measurements, along with the degree of damage to the parallelogram and the circle, are visually judged, and the degree of damage is rated on a scale from 0 to 3, with 0 representing a high degree of shape reproduction accuracy and 3 representing a poor shape such as destruction.

[0049] Figure 10 shows defects when standard sample 1 is cut at a cross section. The cross section is observed in this way, and the cross-sectional defect rate is measured.

[0050] <<Additive manufacturing conditions and measurement example of standard sample printing results>> In addition to the sample evaluation method described above, the additive manufacturing condition search device 2 of this embodiment compares monitoring information, which extracts the intensity of a specific wavelength when a laser is irradiated onto powder during additive manufacturing, with the X-ray CT results of the molded object to obtain a correlation between the intensity of the specific wavelength and defects. The additive manufacturing condition search device 2 then uses a database containing the correlation between the intensity of the specific wavelength and defects to determine defects in the molded object using a second machine learning unit, and measures the result as a defect rate. Furthermore, the additive manufacturing condition search device 2 of this embodiment performs shape measurement during printing using the intensity data of the specific wavelength and image data from an optical camera, and represents this three-dimensionally to obtain the printing result of the standard sample 1.

[0051] FIG. 11 is a block diagram showing an example of the hardware configuration of the additive manufacturing condition searching device 2. The additive manufacturing condition search device 2 searches for input parameter values that will be a solution from the search area. The additive manufacturing condition search device 2 has a processor 21, a memory unit 22, an input device 23, an output device 24, and a communication unit 25. The processor 21, the memory unit 22, the input device 23, the output device 24, and the communication unit 25 are connected by a bus 26. The processor 21 controls the additive manufacturing condition search device 2. The memory unit 22 serves as the working area of the processor 21.

[0052] The storage unit 22 is a non-temporary or temporary recording medium that stores various programs and data. Examples of the storage unit 22 include a ROM, a RAM, an HDD (Hard Disk Drive), and a flash memory.

[0053] The input device 23 inputs data. Examples of the input device 23 include a keyboard, a mouse, a touch panel, a numeric keypad, and a scanner. The output device 24 outputs data. Examples of the output device 24 include a display and a printer. The communication unit 25 connects to a network and transmits and receives data. Examples of the communication unit 25 include a network interface.

[0054] 《Functional Configuration Example of the Additive Manufacturing Condition Search Device 2》 FIG. 12 is a block diagram showing a functional configuration example of the search device. The additive manufacturing condition search device 2 includes a first machine learning unit 47, an input unit 41, a generation unit 42, a specification unit 43, a second machine learning unit 48, a determination unit 44, a setting unit 45, and an output unit 46. The first machine learning unit 47, the input unit 41, the generation unit 42, the specification unit 43, the determination unit 44, the setting unit 45, and the output unit 46 are realized by the processor 21 executing a program stored in the storage unit 22. This additive manufacturing condition search device 2 implements an additive manufacturing condition search method.

[0055] The first machine learning unit 47 includes a control factor information unit 471, an input unit 472, a higher-level item 473, a lower-level item 474, arithmetic units 475 and 478, a recipe database 476, and a material type input unit 477. The first machine learning unit 47 outputs the initially learned conditions automatically assigned to the input unit 41. This first machine learning unit 47 outputs additive manufacturing conditions according to the material information and the device information.

[0056] The control factor information unit 471 is a storage unit that stores the control factors of the additive manufacturing device 5. Note that the control factor information unit 471 may accept the selection of the control factors of the additive manufacturing device 5 and the input of the setting range of these control factors. The input unit 472 assigns the upper-level item 473, the lower-level item 474, and the setting order to the control factors of the additive manufacturing apparatus 5 input from the control factor information unit 471, and outputs them to the calculation unit 475.

[0057] The recipe database 476 accumulates by associating the material type, material physical properties, past shaping conditions for each material, and manufacturing results. The manufacturing results include monitoring information and defect information. When the calculation unit 478 acquires the monitoring information from the recipe database 476 and acquires the attribute information of each material type from the material type input unit 477, it calculates the setting range of the energy density and outputs it to the calculation unit 475. The material type input unit 477 accepts the input of the material type and its material physical properties by user operation. The material type input unit 477 is a storage unit in which the material type and its material physical properties are stored, and the calculation unit 478 may acquire the material type and its material physical properties from the material type input unit 477.

[0058] Based on the control factors of the additive manufacturing apparatus 5 and the setting range of the energy density, the calculation unit 475 calculates the additive manufacturing conditions and outputs them to the input unit 41. Note that the first machine learning unit 47 outputs an initial recipe, which is an automatically assigned initial learning condition.

[0059] The input unit 41 receives the additive manufacturing conditions set for the additive manufacturing apparatus 5 from the first machine learning unit 47, and accepts the input of the evaluation target value and the condition reference value by user operation. The additive manufacturing conditions set for the additive manufacturing apparatus 5 are the input parameters described above.

[0060] The input parameters specifically include the scale correction in the xyz directions, the offset amount from the contour line on the CAD data for irradiating the contour line, the output and scanning speed of the light beam for irradiating the contour line, the offset amount from the contour line irradiation position in "In-skin" for filling the shaping area, the scanning line interval, the scanning pattern, the scanning line length, the offset amount between the scanning patterns, the output and scanning speed of the light beam, the offset amount from the contour line irradiation position in "Down-skin" for forming an overhang, the scanning line interval, the scanning pattern, the scanning line length, the offset amount between the scanning patterns, the output and scanning speed of the light beam, the offset amount from the contour line irradiation position in "Up-skin" for forming the outermost surface in the shaping height direction, the presence or absence of double irradiation, the scanning line interval, and the output and scanning speed of the light beam.

[0061] FIG. 13 is an explanatory diagram showing the scanning line interval among the input parameters. The powder bed 82 shows the uppermost layer, and its thickness is σz. The powder bed 81 is a layer below the uppermost layer and was laid in the past. The scanning line 83 is the part irradiated in the previous scan. The scanning line 85 is the part where irradiation is planned in this scan. And the beam spot 84 is the part being irradiated currently. The interval between the scanning line 83 and the scanning line 85 is δy.

[0062] FIG. 14 is an explanatory diagram of the scanning line interval, the scanning pattern, the scanning line length, and the offset amount between the scanning patterns. The contour line irradiation area 92 is the irradiation area forming the contour of the shaped object, and a plurality of scanning lines 931 to 939, scanning lines 941 to 949, etc. are drawn so as to fill its interior. Thereby, a molten pool can be formed so as to fill the interior of the shaped object.

[0063] FIG. 15 is an explanatory diagram of the offset amount from the contour line irradiation. The contour line 91 is the contour of the shaped object on the CAD data. The contour line irradiation area 92 is irradiated inside by a predetermined offset amount from this contour line 91. Thereby, considering the size of the molten pool by the light beam, a contour with less error can be formed.

[0064] Continuing the description, returning to Fig. 12, the input unit 41 receives an input of an evaluation target value, which is a target value of a modeling result additively manufactured by the additive manufacturing device 5. The actual measured values of the modeling results additively manufactured by the additive manufacturing apparatus 5 are the output parameters described above. The output parameters include the actual measured values of the modeling results of the standard sample 1 additively manufactured by the additive manufacturing apparatus 5 and the actual measured values related to the device status of the additive manufacturing apparatus 5.

[0065] The input unit 41 also accepts input of a search area defined by the range of additive manufacturing conditions for the standard sample 1, which are input parameters, and the range of actual measured values of the modeling results for the standard sample 1, which are output parameters, as well as reference values of the additive manufacturing conditions for this search area. The search area is the area in which the values of the input parameters are searched for, and more specifically, the input range that can be set as a control factor for the additive manufacturing conditions. The search area is defined by the control range of the input parameters and the target range of the output parameters of the additive manufacturing device 5. The reference values of the additive manufacturing conditions are the reference values of the input parameters, and are values of the input parameters obtained in the past.

[0066] The generation unit 42 generates a prediction model that indicates the relationship between the additive manufacturing conditions and the actual measured values of the modeling results, based on a combination of the setting values of the additive manufacturing conditions within the search area and the actual measured values of the modeling results when these setting values are provided to the additive manufacturing device 5. The setting values of the additive manufacturing conditions are values of input parameters prepared as learning data. The actual measured values of the modeling results are the actual measured values of the modeling results when the additive manufacturing device 5 models the standard sample 1.

[0067] A prediction model is a function that indicates the relationship between input parameters and output parameters. The generation unit 42 generates a prediction model that indicates the relationship between the set values of conditions within the search region and the actual measured values of outputs by using regression analysis that can handle multiple inputs and multiple outputs, such as neural networks and support vector machines, or statistical analysis, such as correlation analysis, principal component analysis, and multiple regression analysis.

[0068] The identification unit 43 identifies an area where the predicted value exists from the prediction model by providing the evaluation target value input by the input unit 41 to the prediction model generated by the generation unit 42. The identification unit 43 further sets this predicted value in the additive manufacturing apparatus 5 and performs a demonstration experiment in which a standard sample 1 is formed, acquires monitoring information during the formation, and obtains the results as actual measurement values. Here, the actual measurement values are shape information and defect information acquired by the identification unit 43 by inspecting the formed object. The judgment unit 44 judges whether the actual measurement values resulting from the demonstration experiment, i.e., the defect information of the formed object, achieve the evaluation target values.

[0069] The second machine learning unit 48 estimates a modeling result score, which is a defect determination result for the modeled standard sample 1, from monitoring information acquired in a demonstration experiment in which the predicted value is set in the additive manufacturing apparatus 5 and the standard sample 1 is modeled. Here, the modeling result score is an actual measurement value that is the result of the demonstration experiment, and is also defect information for the modeled object.

[0070] If the judgment unit 44 determines that the actual measured value, which is the result of a demonstration experiment of the predicted value, does not achieve the evaluation target value, the setting unit 45 adds the combination of this predicted value and the actual measured value to the combination of the setting value of the additive manufacturing conditions and the modeling result, and causes the generation unit 42 to update the prediction model.

[0071] When the determination unit 44 determines that the actual measurement value, which is the result of the demonstration experiment of the predicted value, achieves the evaluation target value, the output unit 46 outputs this predicted value. The output unit 46 may display the predicted value that achieves the evaluation target value on a display, which is an example of the output device 24, or may transmit the predicted value that achieves the evaluation target value to an external device via the communication unit 25, or may store it in the memory unit 22 or the recipe database 476. This predicted value is a setting value for the additive manufacturing conditions.

[0072] FIG. 16 is a block diagram showing an example of the functional configuration of the first machine learning unit 47. The first machine learning unit 47 includes a recipe database 476, a data processing unit 62, an algorithm selection unit 63, and a parameter set construction unit 64. The recipe database 476 stores a material type energy density range database 611, a selected parameter / setting range database 612, and a parameter set / result database 613. The first machine learning unit 47 outputs additive manufacturing conditions according to material information and equipment information as an initial learning recipe. The first machine learning unit 47 then outputs new additive manufacturing conditions as a recommended recipe from a combination of multiple additive manufacturing conditions and defect information. The recipe database 476 stores material types, material properties, additive manufacturing conditions for each material used in the past, and manufacturing results.

[0073] The material type energy density range database 611 stores the material type, material properties, and the characteristics of the object when a specified energy density is applied to this material. The selected parameter / setting range database 612 stores the parameters and setting ranges selected by the user. The parameter set / result database 613 stores the additive manufacturing conditions for each material that have been implemented in the past, as well as the manufacturing results. The manufacturing results refer to defect information for the object additively manufactured under these additive manufacturing conditions.

[0074] When material property data 71, recipe 72, parameter setting range 73, and initial learning recipe derivation number 74 are input to this first machine learning unit 47, various machine learning is performed based on a recipe database 476, and then a recipe 75, which is an additive manufacturing condition, is calculated. The recipe 75 includes heat source output, scanning speed, scanning line spacing, and layer thickness, which are control factors for filling the interior of the object. The first machine learning unit 47 calculates the setting range of energy density from the control factors for filling the interior of the object, and assigns additive manufacturing conditions for initial learning according to the setting range of energy density. The first machine learning unit 47 accepts the selection of these control factors and the input of the setting range of the control factors, and assigns upper and lower items and setting orders to these control factors.

[0075] The material property data 71 is a combination of material types and material properties. The recipe 72 is a group of parameters indicating additive manufacturing conditions, such as laser power and operating speed in In-Skin, Down-Skin, or contour.

[0076] The algorithm selection unit 63 constructs calculation rules such as parameter determination rules, energy density calculations, automatic assignment rules, pass / fail determination rules, etc., according to the items to be set, and provides them to the data processing unit 62. When the data processing unit 62 obtains a parameter set by performing calculations according to the calculation rules given by the algorithm selection unit 63, it outputs the parameter set to the parameter set construction unit 64. The parameter set construction unit 64 derives a recipe 75, which is an initial learning recipe according to the required unit, from the parameter set calculated by the data processing unit 62.

[0077] Figure 17 is a diagram showing the physical property values of materials. The material property data 71 stores the combination of each material name and physical property data such as thermal conductivity and absorption rate.

[0078] Figure 18 is a graph showing the relationship between the energy density and the defect value of each material. The horizontal axis of the graph indicates the energy density. The vertical axis indicates the density of the additive manufactured product. The higher the density of the additive manufactured product, the lower the defect rate. For each of the materials A to C, it is necessary to set the additive manufacturing conditions at an appropriate energy density so that the density of the additive manufactured product exceeds a predetermined value.

[0079] Figure 19 is a block diagram showing a functional configuration example of the second machine learning unit 48. The second machine learning unit 48 includes a defect database 66, a data processing unit 67, an algorithm selection unit 68, and a defect determination unit 69. The defect database 66 stores a monitoring information database 661, a defect determination result database 662, and a recipe / defect rate database 663. The second machine learning unit 48 is a part that calculates a modeling result score 77 from the monitoring information 76 acquired from the additive manufacturing apparatus 5 during modeling. As a result, it is possible to obtain the score of the modeling result without manually measuring the modeled object such as the standard sample 1. The monitoring information 76 includes the luminance, temperature, wavelength, optical image, etc. of the laser irradiation unit during modeling. In the second machine learning unit 48, a model learned using the defect database as teacher data estimates the defect information of the modeled object from the monitoring information during modeling. Then, the second machine learning unit 48 stores the defect information of this modeled object, as well as the material, shape information, additive manufacturing conditions, and monitoring information during modeling, in this defect database.

[0080] The monitoring information database 661 stores the past monitoring information during modeling. The defect determination result database 662 stores the defect determination results obtained by manual work or the like of the standard sample 1 modeled when the monitoring information was acquired. The recipe / defect rate database 663 stores the correspondence between each recipe (additive manufacturing conditions) and the defect determination rate. That is, the defect database 66 accumulates the material, shape information, additive manufacturing conditions, monitoring information during modeling, and defect information in association with each other.

[0081] The data processing unit 67 learns the past monitoring information in the monitoring information database 661 and the past defect determination results in the defect determination result database 662 as teacher data, and creates a model that predicts the defect determination result when the monitoring information is input. The algorithm selection unit 68 selects an algorithm for creating a correlation map between the monitoring information and the defect determination result, and provides it to the data processing unit 67.

[0082] The defect determination unit 69 determines defects from the monitoring information 76 using the model generated by the data processing unit 67, and derives a modeling result score 77.

[0083] FIG. 20 is a flowchart showing the search process performed by the additive manufacturing condition search device 2. First, the first machine learning unit 47 acquires the material type and material properties (thermal property data) via the material type input unit 477 (step S30). Then, the first machine learning unit 47 acquires the type of parameters, their settable ranges, and the number of recipes via the input unit 472 (step S31). Here, a recipe refers to additive manufacturing conditions according to material information and equipment information.

[0084] The first machine learning unit 47, by the calculation unit 475, creates an initial learning recipe for the new modeling conditions (Step S32). The additive manufacturing apparatus 5 additively manufactures a model using this recipe while acquiring monitoring information during additive manufacturing (step S33). Then, the second machine learning unit 48 estimates defects in the model from the monitoring information (step S34). Note that, in parallel with step S34, an inspector may inspect the model manufactured by the additive manufacturing apparatus 5 and acquire shape information and defect information of the model. Then, the determination unit 44 associates the material, recipe, monitoring information during additive manufacturing, and defect determination results used when additively manufacturing this shaped object, and stores them in the recipe database 476 (step S35).

[0085] In step S36, the determination unit 44 determines whether the score of the modeling result has reached the evaluation target value. If the score of the modeling result has reached the evaluation target value (Yes), the determination unit 44 ends the processing in Fig. 20. If the score of the modeling result has not reached the evaluation target value (No), the determination unit 44 proceeds to step S37. In step S37, the first machine learning unit 47 performs a regression analysis of the modeling result scores and parameters to derive a new recommended recipe, and then returns to step S33. Then, the additive manufacturing condition searching device 2 repeats the series of processes from steps S33 to S36 based on this recommended recipe. This allows the first machine learning unit 47 to modify the recommended recipe until it reaches the evaluation target value.

[0086] 《Exploring additive manufacturing conditions》 The search for additive manufacturing conditions is realized by conducting demonstration experiments based on the prediction model to search for the optimal solution that meets the goal. For this reason, the additive manufacturing condition search device 2 adds the modeling results, which are the results of the demonstration experiments, to the learning data to update the prediction model, and repeats this process until the goal is met. The additive manufacturing condition search device 2 can further gradually update the goal toward the final goal, thereby efficiently searching for the optimal solution.

[0087] In additive manufacturing, the additive manufacturing condition searching device 2 creates a dataset of the predetermined number of additive manufacturing conditions described above. Then, the additive manufacturing condition searching device 2 causes the additive manufacturing device 5 to additively manufacture the standard sample 1, and generates a prediction model using the demonstration experiment results (modeling results) as learning data.

[0088] The additive manufacturing condition search device 2 generates a prediction model from learning data that includes the results of the demonstration experiment, and calculates the prediction results using the prediction model. By repeating the above process until the demonstration experiment results (modeling results) using the prediction results calculated above as additive manufacturing conditions meet the target, it is possible to efficiently search for the optimal solution.

[0089] FIG. 21 is a flowchart of the additive manufacturing condition search process in the additive manufacturing device 5. The additive manufacturing condition searching device 2 receives input of a target value for the modeling result of the standard sample 1 additively manufactured by the additive manufacturing device 5, and search settings (step S11). The search settings are, for example, an allowable value for the difference or deviation between the search result and the target value.

[0090] Next, the additive manufacturing condition search device 2 receives input of a base solution and input of information related to that solution via the input unit 41 (step S12). Specifically, the additive manufacturing condition search device 2 receives input parameters of the data set of several tens of additive manufacturing conditions described above, and output parameters when the input parameters are used. The additive manufacturing condition search device 2 also receives input of the optimal solution (input parameter values) before the start of the search, the output parameters when the optimal solution is used, target values of the output parameters before the start of the search, and a model function that explains the relationship between the input parameters and the output parameters.

[0091] The additive manufacturing condition searching device 2 generates, via the generation unit 42, a prediction model for predicting input parameters that are solutions that satisfy the target values of the modeling results of the standard sample 1 (step S13). Specifically, the additive manufacturing condition searching device 2 generates, as a prediction model, a function that indicates the relationship between input and output data of the additive manufacturing device 5, using data (e.g., initial data) stored in the storage unit 22. The input and output data is a combination of input data and output data, where the input data is the value of the input parameter given to the additive manufacturing device 5, and the output data is the actual measured value obtained from the modeling results of the standard sample 1 additively manufactured by the additive manufacturing device 5.

[0092] As a method for analyzing the relationship between input and output data, regression analysis that can handle multiple inputs and multiple outputs, such as neural networks, support vector regression, and regression using kernel methods, can be used. Also, statistical analysis such as correlation analysis, principal component analysis, and multiple regression analysis can be used.

[0093] Next, the additive manufacturing condition search device 2 uses the generated predictive model to predict the parameters of the additive manufacturing conditions that will result in the desired solution or a modeling result that is close to the desired solution, and outputs and saves the predicted result (step S14).

[0094] In order to search for the optimal solution in a single prediction, it is necessary to obtain and analyze data that covers the entire setting range of the parameters that can be set under the additive manufacturing conditions. However, as described above, as the number of parameters increases, the number of parameter combinations becomes extremely large. Therefore, searching the entire area requires an extremely long search time and is extremely difficult to implement.

[0095] In order to efficiently search for a solution while avoiding these problems, (a) acquisition of data for creating a prediction model, (b) creation of a prediction model, (c) acquisition of prediction results, (d) conduct an experimental verification of the prediction results, and furthermore, by (a2) adding the experimental verification results to the database for model creation, prediction and verification can be repeated.

[0096] The acquisition of data for creating a prediction model corresponds to the process of step S12. The generation of a prediction model corresponds to the process of step S13. The acquisition of prediction results corresponds to the process of step S14. The experimental verification of the prediction results corresponds to the process of step S15. Adding the experimental verification results to the database for model creation corresponds to the process of step S16.

[0097] Specifically, the additive manufacturing condition search device 2 uses the prediction conditions as search conditions and conducts an experimental verification by the additive manufacturing device 5 (step S15). Then, the additive manufacturing condition search device 2 acquires the input / output data of the additive manufacturing device 5 under each search condition as the experimental verification results, that is, the search results.

[0098] The additive manufacturing condition search device 2 stores the acquired search results in the recipe / defect rate database 663 (step S16). That is, the additive manufacturing condition search device 2 stores, as search results, the input / output data that is a pair of the value of the additive manufacturing conditions, which is the input parameter used in the experimental verification, and the value of the shaping result of the standard sample 1 additively manufactured by the additive manufacturing device 5 using the value of this input parameter. Here, the value of the shaping result of the standard sample 1 is the defect information of the standard sample 1.

[0099] Next, the additive manufacturing condition exploration device 2 identifies the optimal solution from the acquired input / output data (step S17), and stores the identified optimal solution in the storage unit 22. After that, the additive manufacturing condition exploration device 2 determines whether the final goal has been achieved (step S18). If the final goal has been achieved (step S18: Yes), the additive manufacturing condition exploration device 2 ends the process of FIG. 21. On the other hand, if the final goal has not been achieved (step S18: No), the additive manufacturing condition exploration device 2 proceeds to step S19 to update the goal, and then returns to step S12 to update the learning data.

[0100] Specifically, in step S18, the additive manufacturing condition exploration device 2 determines that the final goal has been achieved when the output parameter corresponding to the updated optimal solution is equal to the final target value or the difference from the final target value is within the allowable range (step S18: Yes). On the other hand, the additive manufacturing condition exploration device 2 determines that the goal has not been achieved when the output parameter corresponding to the updated optimal solution is not equal to the final target value or the difference from the final target value is not within the allowable range (step S18: No), and proceeds to step S20.

[0101] In step S20, the additive manufacturing condition exploration device 2 updates the target value, the difference between the exploration result and the target value, or the allowable value of the deviation, and returns to the process of step S12. When proceeding with the processes from step S13 to S18, if the final goal is given from the beginning, or if a very small value is given as the difference between the exploration result and the target value, or the allowable value of the deviation, the difficulty of finding the optimal solution increases, and there is a possibility that the additive manufacturing condition exploration device 2 cannot find the solution. To avoid this, the additive manufacturing condition exploration device 2 may give a goal different from the final goal at the initial stage of the exploration. When the current goal has been achieved and the final goal has not been satisfied (step S18: No), in step S20, by gradually approaching the target value to the final target value, the possibility of finding a solution that achieves the final goal can be increased.

[0102] In addition, if the additive manufacturing condition search device 2 sets a large value as the difference between the search result and the target value, or the tolerance for deviation, as the current target, and the current target is achieved but the final target is not met (step S18: No), the target value can be gradually brought closer to the final target value, thereby increasing the possibility of finding a solution that achieves the final target.

[0103] A method for gradually updating from the initial target to the final target is to prepare multiple target values between the initial target and the final target, set the initial target as the initial current target, and update the current target value to a target value approaching the final target each time the current target is achieved. Alternatively, an initial target value may be set as the initial current target, and multiple target values may be prepared and used that gradually approach the final target at a predetermined rate.

[0104] Second Embodiment In the additive manufacturing condition search of the first embodiment, by using a standard sample, it is possible to evaluate the number of modeling conditions that can be arranged in the modelable area in a single demonstration experiment.

[0105] On the other hand, when dealing with parameters related to the manufacturing environment, it is necessary to assign manufacturing conditions to each environment. Parameters related to the manufacturing environment include, for example, layer thickness, preheating temperature, manufacturing environment pressure, powder particle size, etc. In other words, in additive manufacturing, since the parameters related to the manufacturing environment cannot be changed in a single demonstration experiment, it is necessary to change the dataset of additive manufacturing conditions every time the parameters related to the manufacturing environment are changed.

[0106] When changing parameters related to the printing environment, it is necessary to clarify the purpose. For example, it is necessary to clarify whether the purpose of changing parameters related to the printing environment is to optimize the above printing conditions and parameters related to the printing environment, or to determine parameters related to the printing environment. For example, the inventor's verification results revealed that when additive manufacturing conditions are searched for using a dataset of additive manufacturing conditions including layer thickness, the additive manufacturing condition search device 2 derives conditions with a thinner layer thickness as a predicted value.

[0107] Therefore, when dealing with lamination thickness, it is preferable for the engineer to determine the required lamination thickness in advance and optimize it as a fixed condition using the process in Figure 21. However, if there is a certain deviation between the solution and the target value even after searching for a solution for a certain development period, the lamination thickness can be changed to a smaller value to find the optimal solution, and the optimal solution for the lamination thickness for that metal material can be found.

[0108] FIG. 22 is a flowchart of the additive manufacturing condition search process in the additive manufacturing condition search device 2 of the second embodiment. First, the input unit 41 receives input of environmental conditions such as layer thickness (step S10), and proceeds to a process similar to step S11 in Fig. 21. Parameters related to the modeling environment may be obtained in advance by another method and then input to the additive manufacturing condition searching device 2 of the second embodiment.

[0109] The processing from steps S11 to S17 is the same as that in FIG. In step S18, the additive manufacturing condition search device 2 determines that the final target has been achieved if the output parameter corresponding to the updated optimal solution is equal to the final target value or the difference from the final target value is within an acceptable range (step S18: Yes).

[0110] On the other hand, if the output parameter corresponding to the updated optimal solution is equal to the final target value or the difference from the final target value is not within the allowable range, the additive manufacturing condition search device 2 determines that the target has not been achieved (step S18: No) and proceeds to step S19.

[0111] In step S19, the additive manufacturing condition search device 2 determines whether or not the specified development time has been reached. If the additive manufacturing condition search device 2 has reached the specified development time (Yes), it proceeds to step S21, changes the layer thickness, and then returns to step S12. If the additive manufacturing condition search device 2 has not reached the specified development time (No), it proceeds to step S20, updates the target in the same manner as in FIG. 21, and then returns to step S12.

[0112] (Modification example) The present invention is not limited to the above-described embodiments, and includes various modification examples. For example, the above-described embodiments have been described in detail for easy understanding of the present invention, and are not necessarily limited to those having all the configurations described. It is possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Further, it is possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.

[0113] Each of the above configurations, functions, processing units, processing means, etc. may be realized by hardware such as an integrated circuit, for example, for a part or all of them. Each of the above configurations, functions, etc. may be realized by software by a processor interpreting and executing a program for realizing each function. Information such as a program, table, file, etc. for realizing each function can be placed in a recording device such as a memory, hard disk, SSD (Solid State Drive), or a recording medium such as a flash memory card, DVD (Digital Versatile Disk).

[0114] In each embodiment, the control lines and information lines show those considered necessary for explanation, and do not necessarily show all the control lines and information lines on the product. In reality, it may be considered that almost all the components are interconnected. As modification examples of the present invention, for example, there are the following (a) to (c).

[0115] (a) The present invention is not limited to an additive manufacturing apparatus using a powder bed fusion method, and may be applied to a directed energy deposition method or other additive manufacturing apparatuses. (b) The standard sample only needs to have at least three smooth surfaces. (c) The standard sample is not limited to a cubic shape and may be a rectangular parallelepiped.

Description of Reference Numerals

[0116] 1 Standard sample 11 Front surface 111~119 Regions 12 Upper surface 13 Left side surface 131~133 Regions 14 Rear surface 15 Right side surface 16 Bottom surface 2 Additive manufacturing condition exploration apparatus 21 Processor 22 Memory unit 23 Input device 24 Output device 25 Communication unit 26 Bus 41 Input unit 42 Generation unit 43 Identification unit 44 Judgment unit 45 Setting unit 46 Output unit 47 First machine learning unit 471 Control factor information unit 472 Input unit 473 Upper item 474 Lower item 475 Arithmetic unit 476 Recipe database 477 Material type input unit 478 Arithmetic unit 48 Second machine learning unit 5 Additive manufacturing apparatus 501 Light beam source 502 Transmission window 510 Chamber 511 Gas supply unit 512 Exhaust mechanism 513 Recoater 514 Material supply section 515 Additive manufacturing section 516 Recovery section 517 Stage 518 Stage 530 Control section 56 Temperature sensor 57 Pressure sensor 58 Oxygen sensor 611 Material type energy density range database 612 Selection parameter / setting range database 613 Parameter set / result database 62 Data processing section 63 Algorithm selection section 64 Parameter set construction section 66 Defect database 661 Monitoring information database 662 Defect determination result database 663 Recipe / defect rate database 67 Data processing section 68 Algorithm selection section 69 Defect determination section 71 Material physical property data 72 Recipe 73 Parameter setting range 74 Initial learning recipe derivation number 75 Recipe 81 Powder bed 82 Powder bed 83,85 Scanning line 84 Beam spot 91 Contour line 92 Contour line irradiation area 931~939 Scanning line 941~949 Scanning line

Claims

1. A defect database that associates and stores materials, shape information, additive manufacturing conditions, monitoring information during manufacturing, and defect information; A first machine learning unit that outputs additive manufacturing conditions according to material information and device information, and outputs new additive manufacturing conditions from combinations of a plurality of additive manufacturing conditions and defect information; A specifying unit that causes an additive manufacturing apparatus to perform manufacturing according to the additive manufacturing conditions and acquires monitoring information during manufacturing; A second machine learning unit that determines defect information of a manufactured object from the monitoring information by a model trained using the defect database as teacher data, estimates a manufacturing result score of the manufactured object, and stores the score in the defect database; A determination unit that determines whether or not the defect information of the manufactured object achieves an evaluation target value; comprising until the determination unit determines that the defect information of the manufactured object achieves the evaluation target value, the first machine learning unit outputs new additive manufacturing conditions from combinations of a plurality of additive manufacturing conditions and defect information, the specifying unit causes the additive manufacturing apparatus to perform manufacturing according to the additive manufacturing conditions and acquires monitoring information during manufacturing, and a model trained using the defect database as teacher data determines defect information of the manufactured object from the monitoring information, estimates a manufacturing result score of the manufactured object, and stores the score in the defect database, and a series of operations are repeated; An additive manufacturing condition search apparatus characterized by the above.

2. If the determination unit determines that the defect information of the manufactured object does not achieve the evaluation target value, the first machine learning unit performs regression analysis of the manufacturing result score and parameters to output new additive manufacturing conditions. The additive manufacturing condition search apparatus according to claim 1, characterized by the above.

3. An input unit that accepts a manufacturing result of a standard sample manufactured by an additive manufacturing apparatus, additive manufacturing conditions corresponding thereto, an evaluation target value of the standard sample, and a search region defined by ranges of the additive manufacturing conditions and the manufacturing result; The additive manufacturing condition search apparatus according to claim 1, characterized by comprising the above.

4. A generation unit that generates a prediction model showing the relationship between additive manufacturing conditions and manufacturing results based on set values of the additive manufacturing conditions within the search region and manufacturing results when the set values of the additive manufacturing conditions are set in the additive manufacturing apparatus; The specific part calculates a predicted value from the prediction model by providing the evaluation target value received by the input part to the prediction model, and acquires, as the measured value, the result of the proof test in which the predicted value is set in the additive manufacturing apparatus. The additive manufacturing condition search apparatus according to claim 3, characterized in that.

5. An output unit that outputs the predicted value as a set value of the additive manufacturing conditions when the evaluation target value is achieved, The additive manufacturing condition search apparatus according to claim 4, characterized in that it comprises.

6. When the measured value does not achieve the evaluation target value, a setting unit that causes the generation unit to update the prediction model by adding the combination of the predicted value and the measured value to the combination of the set value of the additive manufacturing conditions and the shaping result, The additive manufacturing condition search apparatus according to claim 4, characterized in that it comprises.

7. The standard sample is a hexahedron with three or more smooth surfaces, Filling in the shaping area, which is three types of areas set by the additive manufacturing conditions, The area forming the overhang, The area forming the outermost surface in the shaping height direction, Are involved, and has one surface in which the punched hole shapes composed of straight lines and curves are aggregated, The additive manufacturing condition search apparatus according to claim 3, characterized in that.

8. The slice data of the standard sample includes two or more independent regions in any one layer at the central part in the stacking direction, It comprises a small region cut by a predetermined width from the outer edge of the standard sample and a large region composed of other parts, The additive manufacturing condition search apparatus according to claim 3, characterized in that.

9. It comprises a recipe database that stores material types, material physical properties, additive manufacturing conditions and manufacturing results for each material implemented in the past, When the material type and material physical properties for exploring the additive manufacturing conditions are input, the first machine learning unit calculates the setting range of the energy density from the heat source output, scanning speed, scanning line interval, and layer thickness, which are control factors for filling the inside of the shaped object, based on the recipe database. The additive manufacturing condition search apparatus according to claim 2, characterized in that.

10. The first machine learning unit receives an input of the setting range of the energy density, The additive manufacturing condition search apparatus according to claim 9, characterized in that.

11. The first machine learning unit assigns additive manufacturing conditions for initial learning according to the setting range of the energy density, The additive manufacturing condition search apparatus according to claim 9, characterized in that.

12. The first machine learning unit receives the selection of the control factor and the input of the setting range of the control factor. The additive manufacturing condition search device according to claim 9, characterized in that.

13. The first machine learning unit assigns a higher-level item, a lower-level item, and a setting order to the control factor. The additive manufacturing condition search device according to claim 12, characterized in that.

14. A step in which the first machine learning unit outputs additive manufacturing conditions according to material information and device information, or outputs new additive manufacturing conditions from a combination of a plurality of additive manufacturing conditions and defect information; A step in which the specifying unit causes the additive manufacturing apparatus to perform shaping according to the additive manufacturing conditions and acquires monitoring information during shaping; A step in which the second machine learning unit estimates defect information of the shaped object from the monitoring information by using a model learned with a defect database of a combination of monitoring information during shaping and defect information as teacher data, and stores the defect information in the defect database; A step in which the determination unit determines whether or not the defect information of the shaped object achieves an evaluation target value; is executed, Until the determination unit determines that the defect information of the shaped object achieves the evaluation target value, a step in which the first machine learning unit outputs new additive manufacturing conditions from a combination of a plurality of additive manufacturing conditions and defect information; A step in which the specifying unit causes the additive manufacturing apparatus to perform shaping according to the additive manufacturing conditions and acquires monitoring information during shaping; The step in which the second machine learning unit estimates defect information of the shaped object from the monitoring information by using a model learned with a defect database of a combination of monitoring information during shaping and defect information as teacher data, and stores the defect information in the defect database is repeated. An additive manufacturing condition search method, characterized in that.

Images