Three-dimensional modeling method, program, and structure

Abstract

[Problem] To suppress deformation occurring during modeling in a method for performing three-dimensional modeling by laminating thin layers that are formed by melting and solidifying a powder. [Solution] In a powder bed type three-dimensional modeling device, a powder of metal is irradiated with a laser, the powder is melted and solidified to form a thin layer, and this thin layer is laminated to perform three-dimensional modeling. When so doing, the energy amount applied by the laser irradiation is changed, thereby mixing and forming high-density layers L21 having a high modeling density and porous layers L22 having a low modeling density. The high-density layers and the porous layers can be formed at various thicknesses (test pieces P2 through P4), and the high-density layers and the porous layers may be alternately formed (test pieces P6 through P9).　By mixing and forming the high-density layers and the porous layers in this manner, it is possible to suppress distortion without reducing the mechanical strength of a modeled article.

Description

Three-dimensional modeling method, program, and structure

[0001] The present invention relates to a method for three-dimensionally manufacturing by laminating thin layers formed by melting and solidifying powder, and in particular to a technique for alleviating residual stress and suppressing deformation that occurs during manufacturing.

[0002] One known additive manufacturing technique for creating three-dimensional shapes by stacking thin layers of a predetermined thickness is to melt and solidify powder, such as metal powder, to create three-dimensional shapes. Various methods are known for this purpose, including powder bed fusion (PBF), binder jetting (BJT), and directed energy deposition (DED).

[0003] In such three-dimensional modeling, it is known that residual stresses caused by the effects of heating during melting and cooling during solidification can lead to delamination, cracks, deformation or breakage of the model, etc. To avoid or mitigate these problems, various techniques have been attempted, such as dividing the model into small cells and irradiating them with a laser, using a base plate in PBF that has a linear expansion coefficient close to that of the powder, and re-irradiating the modeled layer with a laser.

[0004] Furthermore, Patent Document 1 discloses a technique for reducing thermal deformation and preventing warpage by embedding separately prepared block components inside a molded object during the three-dimensional modeling process. Patent Document 2 discloses a technique for forming a boundary within a molded object, which is not directly aimed at alleviating residual stress caused by heat, but which forms a solid boundary layer and a porous layer therein by modulating a laser module in a technique for producing a composite material that combines metal, plastic, and ceramic materials by melting and solidifying powder.

[0005] JP 2000-190086 A JP 2022-33816 A

[0006] However, the problem of residual stress in three-dimensional modeling has not been sufficiently solved. In particular, no technology has been established that can achieve a good balance between the requirements for suppressing distortion and other issues associated with residual stress, the strength of the model, and the modeling speed. The present invention has been made in light of this problem.

[0007] The present invention can be a three-dimensional modeling method for forming a three-dimensional object, comprising: (a) a step of defining the shape of each layer to form the object from thin layers of a predetermined thickness; (b) a step of applying a predetermined amount of energy to powder to melt it and solidify it to form the thin layer according to the shape; (c) a step of repeating step (b) and stacking the next thin layer on top of the formed thin layer; and (d) a step of controlling the amount of energy for each step (b) so that two or more layer regions with different amounts of energy are mixed in the direction of stacking.

[0008] If the amount of energy used to form the thin layers is reduced, the molding density of the molded object will decrease, resulting in a porous shape. Conversely, if the amount of energy is increased, a solid structure will result. In this way, the amount of energy affects the density of the formed thin layers. A porous shape tends to have a lower mechanical strength, but relatively small residual stress. Conversely, the more solid the structure, the greater the mechanical strength and the greater the residual stress. In the present invention, by controlling the amount of energy used for each thin layer, it is possible to stack layers with different molding densities, i.e., layers with different mechanical strengths and residual stresses, and to suppress distortion of the molded object due to residual stress.

[0009] In the present invention, when energy is applied by laser irradiation or the like, controlling the amount of energy can involve controlling the irradiation intensity, irradiation speed, laser irradiation area, etc., either individually or in combination. "Controlling the amount of energy for each thin layer" refers to controlling a reference energy amount for each thin layer. It is also possible to further control the amount of energy depending on the location within a single thin layer. Furthermore, "for each thin layer" does not necessarily mean changing the amount of energy each time a thin layer is formed; multiple adjacent thin layers may be formed with the same amount of energy. A layer region refers to a region consisting of one or more thin layers formed with the same amount of energy, i.e., thin layers that are considered to have equivalent mechanical strength and residual stress. For example, if the amount of energy is changed every time five thin layers are formed, five thin layers would form one layer region. The present invention is applicable to various three-dimensional fabrication methods that use powder melting and solidification, such as powder bed fusion (PBF), binder jetting (BJT), and directed energy deposition (DED).

[0010] In the three-dimensional modeling method of the present invention, the step (d) may include forming the plurality of thin layers with the first amount of energy, and then switching to the second amount of energy.

[0011] As explained above, the energy amount may be changed each time a thin layer is formed. Alternatively, as in the above embodiment, the energy amount may be changed after forming multiple thin layers. According to the above embodiment, multiple thin layers formed with the same energy amount form one layer region. In this embodiment, since it is not necessary to change the energy amount for each thin layer, the modeling time can be shortened. Furthermore, since layer regions can be formed with a certain thickness, the impact of variations in the formation state of each thin layer on the mechanical strength and residual stress of the entire model can be suppressed. In the above embodiment, the number of thin layers formed with the first energy amount can be determined arbitrarily. Furthermore, the number of thin layers formed with the second energy amount may be one or multiple. The number of layers formed with the first energy amount does not need to be the same as the number of layers formed with the second energy amount. In the above embodiment, modeling may be performed using three or more different energy amounts.

[0012] In the three-dimensional modeling method of the present invention, the step (d) may switch the amount of energy at least twice during the modeling process.

[0013] According to the above aspect, three layer regions with different energy amounts are laminated. This makes it possible to form a variety of layer regions with different properties, such as mechanical strength and residual stress, and to flexibly adjust the properties of the entire shaped object. In the above aspect, the three layer regions may each have a different energy amount. Alternatively, a layer region with a first energy amount, a layer region with a second energy amount, and a layer region with a first energy amount may be laminated.

[0014] In the three-dimensional modeling method of the present invention, step (d) may control the amount of energy so that first layer regions consisting of the thin layers formed by a first amount of energy and second layer regions consisting of the thin layers formed by a second amount of energy are alternately stacked.

[0015] In the above embodiment, any number of first and second layer regions may be alternately stacked, and the thickness of each layer region may be constant or may vary.

[0016] The three-dimensional modeling method of the present invention may further include: (e) setting the amount of energy to be applied to each thin layer according to the mechanical strength required for the model; and step (d) may control the amount of energy according to the setting result of step (e).

[0017] In the present invention, layer regions with different build densities, i.e., layer regions with different properties such as mechanical strength and residual stress, are mixed together to form the object. If the amount of energy is reduced and the number of layer regions with low mechanical strength is increased, the mechanical strength of the entire object will decrease. On the other hand, if the amount of energy is increased and the number of layer regions with high mechanical strength is increased, the mechanical strength of the entire object will also increase. In this way, the amount of energy affects the mechanical strength of the object. According to the above aspect, the amount of energy can be determined according to the mechanical strength required of the object, making it possible to form the object so that the entire object meets the required mechanical strength.

[0018] The three-dimensional modeling method of the present invention may further include: (f) setting the amount of energy to be applied to each thin layer according to the range of distortion allowable for the model; and step (d) may control the amount of energy according to the setting result of step (e).

[0019] In the present invention, increasing the number of layer regions with low density and low residual stress can suppress distortion of the entire molded object. On the other hand, increasing the number of layer regions with high density and high residual stress increases distortion of the entire molded object. In this way, the amount of energy affects the density of the thin layers, which in turn affects distortion due to residual stress of the molded object. According to the above aspect, the amount of energy can be determined according to the range of distortion allowable for the molded object, making it possible to mold the object so that the distortion of the entire object satisfies the required value.

[0020] Generally, layer regions with high mechanical strength tend to have high residual stress, while layer regions with low mechanical strength tend to have low residual stress. Therefore, if the amount of energy is determined based on the requirements for both the mechanical strength and residual stress of the object, it is possible to create a shape that satisfies both requirements.

[0021] The three-dimensional modeling method of the present invention is preferably applied to a three-dimensional modeling method in which the step (b) is performed by stacking the thin layers using metal powder.

[0022] The present invention is applicable not only to metal powders but also to various other molding methods that involve melting and solidifying powders. However, the issue of residual stress caused by melting is particularly important in molding methods that use metal powders. Therefore, the present invention is particularly useful for three-dimensional molding that uses metal powders.

[0023] The various features of the present invention described above do not necessarily need to be provided in their entirety, and some of them may be omitted or combined as appropriate. Furthermore, the present invention can be configured in various forms other than the above-described three-dimensional modeling method.

[0024] For example, the present invention may be a three-dimensional printing apparatus for forming a three-dimensional object, comprising: an input unit that reads shape data for each layer to form the object with thin layers of a predetermined thickness; a thin layer forming unit that applies a predetermined amount of energy to powder to melt it and solidify it to form the thin layer in accordance with the shape data; a lamination unit that moves the thin layer forming unit and the already formed thin layer relatively to each other so that the next thin layer is laminated on top of the formed thin layer; and a control unit that controls the amount of energy in the thin layer forming unit so that two or more layer regions with different energy amounts are mixed in the lamination direction.

[0025] The 3D modeling apparatus of the present invention can form thin layers by controlling the amount of energy, which makes it possible to realize the 3D modeling method described in the present invention.The various features described in the 3D modeling method can also be applied to the 3D modeling apparatus of the present invention.

[0026] The present invention may also be a program for causing a computer that controls a three-dimensional printing device that forms a three-dimensional object to execute the following steps: (a) receiving input of the shape of each layer to form the object from thin layers of a predetermined thickness; (b) controlling the powder to be melted by applying a predetermined amount of energy to the powder, and then solidifying the powder to form the thin layer according to the shape; (c) controlling the step (b) to repeat the step (b) and stack the next thin layer on top of the formed thin layer; and (d) controlling the amount of energy for each step (b) so that two or more layer regions with different amounts of energy are mixed in the stacking direction.

[0027] The present invention may also provide a shaped object manufactured by a three-dimensional modeling method comprising the steps of: (a) defining the shape of each layer to form a shaped object from thin layers of a predetermined thickness; (b) applying a predetermined amount of energy to powder to melt it and solidify it to form the thin layer according to the shape; (c) repeating step (b) to stack the next thin layer on the formed thin layer; and (d) controlling the amount of energy for each step (b) so that two or more layer regions with different amounts of energy are mixed in the direction of stacking.

[0028] The present invention may also be a profile setting method for setting an energy amount profile by a computer for a three-dimensional printing device that forms a three-dimensional object by applying a predetermined amount of energy to powder to melt it, solidify it, and stack the resulting thin layers, comprising the steps of: reading printing data that stores the shape of a previously printed object, a profile of the energy amount applied to form each thin layer in the printing process of the object, and one or both of the magnitude of distortion caused in the object and the mechanical strength required of the object; performing a regression analysis by machine learning based on the printing data, using one or both of the magnitude of distortion and the mechanical strength and the shape as explanatory variables and the profile as a target variable, to obtain a learning model; inputting the shape of the object to be printed, and the required values ​​for the allowable magnitude of distortion and mechanical strength that correspond to the learning model; and setting the energy amount profile for the printing process using the learning model based on the input.

[0029] This method uses so-called machine learning to set a profile of the amount of energy to be applied to each thin layer. By doing so, it is possible to easily set a profile that satisfies requirements such as mechanical strength and distortion by utilizing past modeling results. The required values ​​may be set for both the magnitude of distortion and mechanical strength, or for either one of them. In the above embodiment, constraints may be imposed, such as a condition that all layer regions are composed of a certain number of thin layers, or a condition that the amount of available energy must be selected from a predetermined number of options. This makes it possible to easily obtain a learning model using machine learning. Furthermore, the "shape of the model" used in learning may be represented by the thickness in the stacking direction. Distortion may also be given in terms of distortion in the stacking direction, distortion within the plane of the layers, etc. As such, various methods for assigning explanatory variables and target variables are possible.

[0030] FIG. 1 is an explanatory diagram schematically showing the configuration of a three-dimensional printing apparatus in an embodiment; FIG. 2 is an explanatory diagram showing a printing example in an embodiment; FIG. 3 is an explanatory diagram showing experimental results of tensile strength; FIG. 4 is an explanatory diagram showing experimental results of tensile strength, deformation amount, and printing speed; FIG. 5 is a flowchart of a printing process; and FIG. 6 is a flowchart of a learning model generation process and a profile setting process.

[0031] The present invention will be described in detail below with reference to an example of powder bed fusion molding. The present invention is applicable not only to this method, but also to various other molding methods that melt and solidify powder, such as BJT (Binder Jetting) and DED (Directed Energy Deposition).

[0032] A. Device Configuration: FIG. 1 is an explanatory diagram illustrating a schematic configuration of a three-dimensional modeling device according to an embodiment. The three-dimensional modeling device 10 according to the embodiment includes a modeling table 11 for forming a model and a recoater 13 that reciprocates above the modeling table 11 as indicated by arrow A. When the recoater 13 reciprocates during modeling, the surface of the modeling table 11 is covered with a substantially uniform thickness of metal powder used in modeling. The three-dimensional modeling device 10 includes a laser irradiation unit 15 as a mechanism for heating and melting the metal powder. The laser irradiation unit 15 includes a laser light source and a mechanism for moving the irradiated portion in accordance with the shape of the model. The metal powder melts at the portion irradiated with the laser by the laser irradiation unit 15, and then solidifies as it cools, forming a thin layer that constitutes part of the model. The three-dimensional modeling device 10 also includes a movement mechanism 12 that moves the modeling table 11 downward. After forming a thin layer, the modeling table 11 is moved downward by one layer, and the next layer can be stacked on top of the already formed thin layer by dispersing metal with the recoater 13 and melting and solidifying with the laser irradiation unit 15. The three-dimensional modeling device of the embodiment is capable of forming a three-dimensional object by repeatedly forming thin layers and stacking them in this way.

[0033] The three-dimensional modeling apparatus 10 of this embodiment switches the amount of energy supplied to the metal powder by laser irradiation when forming a thin layer in the process of achieving the above-mentioned modeling. Specifically, in the following embodiment, two energy amounts are used: irradiation with an energy amount for forming a thin layer in a high-density state where the metal can be considered solid (hereinafter, irradiation conditions based on such energy amount will be referred to as "high-density conditions"), and irradiation with an energy amount for forming a thin layer with a low modeling density (hereinafter, irradiation conditions based on such energy amount will be referred to as "porous conditions"). Although thin layers formed under porous conditions often end up in a porous state, this does not necessarily mean that they must be porous.

[0034] In this embodiment, the information specifying the amount of energy to be used for laser irradiation for each thin layer is called an energy amount profile. The profile can be set in various ways, such as by providing an energy amount for each layer, or by providing an energy amount for a predetermined number of thin layers. The energy amount can also be controlled in various ways according to the profile, such as by changing the scanning speed of the laser irradiation, by changing the intensity of the laser irradiation, by changing the spot area of ​​the laser irradiation, or by a combination of these. In this embodiment, the method of changing the scanning speed of the laser irradiation was adopted.

[0035] In order to realize the above-described modeling and energy amount control, the three-dimensional modeling apparatus 10 has an input unit 17 and a control unit 16. These can also be configured as hardware, but in this embodiment, they are configured as software by incorporating computer programs that realize the respective functions into a computer equipped with a CPU and memory.

[0036] The input unit 17 reads shape data of the object. Since the shape of each laminated thin layer is required during modeling, the shape data may be shape data for each layer that constitutes the object. Alternatively, data representing the three-dimensional shape of the object may be read, and the three-dimensional modeling device 10 may generate shape data for each thin layer. In addition to the shape data, the input unit 17 also inputs the required mechanical strength of the object and the required tolerance for distortion during modeling. The input unit 17 also inputs a profile of the energy amount of laser irradiation. The profile may be set and input by an operator, for example, or may be set using artificial intelligence.

[0037] The control unit 16 controls the operation of the three-dimensional modeling apparatus 10 and controls the laser irradiation by the laser irradiation unit 15, i.e., the amount of energy supplied, to form a model having a mixture of layers formed under high-density conditions and porous conditions. As described above, the control unit 16 can be configured as software by incorporating a program into a computer. For example, such a program may execute the following steps: (a) receiving input of the shape of each layer for forming a model with thin layers of a predetermined thickness; (b) controlling the powder to be melted by applying a predetermined amount of energy to the powder, and then solidifying the powder to form the thin layers according to the shape; (c) controlling the process to repeat step (b) and stack the next thin layer on top of the formed thin layer; and (d) controlling the amount of energy for each step (b) so that two or more layer regions with different amounts of energy are mixed in the stacking direction. Such a program may be pre-installed in the control unit 16 or distributed independently. For example, it may be stored on a storage medium and distributed, or provided via a communication line.

[0038] In this embodiment, a profile is set using artificial intelligence. The diagram also shows the configuration of a profile setting device 20 for setting a profile. While each component in the diagram can be provided as hardware, in this embodiment, it is configured as software by installing a program that realizes each function on a computer. While the diagram shows a single device, the profile setting device 20 may also be configured by connecting multiple computers, servers, etc. over a network to provide each function.

[0039] The modeling database 21 provided in the profile setting device 20 stores modeling data representing past modeling results. The content of the modeling data can be determined arbitrarily, and may include, for example, the shape of the modeled object, its mechanical strength, the magnitude of distortion during modeling, and the profile used for modeling. The shape of the modeled object may be a three-dimensional shape, or may be represented by a parameter such as the thickness in the stacking direction. Various indicators such as tensile strength and bending strength can be used for the mechanical strength. The magnitude of distortion may also be represented by parameters such as distortion in the stacking direction or distortion within a layer.

[0040] The machine learning unit 23 generates a learning model through machine learning regression using the modeling database 21. The purpose of machine learning is to create a learning model that provides a profile that can achieve required values, such as mechanical strength and distortion magnitude, of a future object to be modeled, when those values ​​are specified. Therefore, machine learning is performed using the shape, mechanical strength, distortion, etc. as explanatory variables and the profile as the objective variable. Machine learning can be performed using various algorithms. The generated learning model is stored in the profile setting device 20.

[0041] The required value input unit 22 inputs required values ​​for the object to be formed, such as mechanical strength and tolerance for distortion. The required values ​​preferably have a format that matches the format of the forming data used by the machine learning unit 23 for machine learning. The profile generation unit 24 generates a profile of the amount of energy to be supplied to the object using the learning model generated by the machine learning unit 23. Note that machine learning and profile generation do not need to be performed simultaneously. A learning model may be generated in advance by machine learning and saved. The profile generated in this manner is passed to the 3D printing device 10 and used for printing.

[0042] In order to easily achieve these in machine learning and profile generation, several constraints may be set. First, in this example, the amount of energy supplied is limited to two conditions: high density and porous. Furthermore, the amount of energy is changed not for each layer, but for a predetermined number of layers. Furthermore, only one of mechanical strength and strain may be used as the required value. This makes it possible to create a learning model relatively easily.

[0043] As described above, according to the three-dimensional modeling apparatus 10 of this embodiment, by controlling the amount of energy of the laser irradiation using the profile set by the profile setting device 20, it is possible to form a mixture of layers formed under high-density conditions (hereinafter referred to as "high-density layers") and layers formed under porous conditions (hereinafter referred to as "porous layers").

[0044] B. Experimental Results: Next, the inventors conducted experiments to examine the effects of mixing a high-density layer and a porous layer on the mechanical strength and distortion of a shaped object. The results will be described.

[0045] FIG. 2 is an explanatory diagram showing an example of molding in the example. It shows a schematic diagram of the layer structure of test pieces P1 to P9 used in the experiment. Hereinafter, a layer formed at one time by the three-dimensional printing device 10 will be referred to as a "thin layer," and a portion where multiple thin layers formed under the same conditions are stacked will be referred to as a "layer region." In this example, the energy amounts of the high-density layer and the porous layer were controlled by changing the scanning speed of the laser irradiation. Specifically, the scanning speed when forming the porous layer was twice the scanning speed when forming the high-density layer.

[0046] The test piece P1 is formed entirely of high-density layers. The bottom portion Lb0 is formed of 20 high-density layers. The upper portion Lu0 is formed of 10 high-density layers. The layer region L1 of the main body is formed of 100 high-density layers. The proportion of high-density layers in the main body is 100%.

[0047] The bottom portion Lb1 of specimen P2 is made up of 20 porous layers. The main body is composed of a layer region L21 formed of 75 high-density layers and a layer region L22 formed of 25 porous layers. The upper portion Lu0 is the same as specimen P1. The proportion of high-density layers in the main body is 75%.

[0048] Specimen P3 has the same bottom portion Lb1 and top portion Lu0 as specimen P2. The main body is composed of a layer region L31 formed of 50 high-density layers and a layer region L32 formed of 50 porous layers. The proportion of the high-density layers in the main body is 50%.

[0049] Specimen P4 has the same bottom portion Lb1 and top portion Lu0 as specimen P2. The main body is composed of a layer region L41 formed of 25 high-density layers and a layer region L42 formed of 75 porous layers. The proportion of the high-density layers in the main body is 25%.

[0050] The test piece P5 has a bottom portion Lb1 formed of 20 porous layers and an upper portion Lu0 formed of 10 porous layers. The main body is composed of a layer region L5 formed of 100 porous layers. The proportion of the high-density layer in the main body is 0%.

[0051] In specimen P6, the bottom portion Lb0 is formed of 20 high-density layers, and the top portion Lu0 is formed of 10 high-density layers. The main body is formed by alternately stacking a layer region L61 formed of 25 high-density layers, a layer region L62 formed of 50 porous layers, and a layer region L63 formed of 25 high-density layers. The proportion of high-density layers in the main body is 50%.

[0052] In specimen P7, the bottom portion Lb0 is formed of 20 high-density layers, and the top portion Lu0 is formed of 10 high-density layers. The main body is formed by alternating layers of 10 high-density layers and layers of porous layers, such as layer region L71 formed of 10 high-density layers and layer region L72 formed of 10 porous layers. The main body is 100 layers thick in total. The proportion of high-density layers in the main body is 50%.

[0053] In specimen P8, the bottom portion Lb0 is formed of 20 high-density layers, and the top portion Lu0 is formed of 10 high-density layers. The main body is formed by alternating high-density layer regions and porous layer regions every five layers, such as layer region L81 formed of five high-density layers and layer region L82 formed of five porous layers. The main body is 100 layers thick in total. The proportion of high-density layers in the main body is 50%.

[0054] In specimen P9, the bottom portion Lb0 is formed of 20 high-density layers, and the top portion Lu0 is formed of 10 high-density layers. The main body is formed by alternating high-density layer regions and porous layer regions every five layers, such as layer region L91 formed of two high-density layers and layer region L92 formed of two porous layers. The main body is 100 layers thick in total. The proportion of high-density layers in the main body is 50%.

[0055] As described above, in this example, the high-density layer and the porous layer can be formed in various ways. As with specimens P2 to P4, switching between high-density conditions and porous conditions may be performed only once when stacking the shaped object. Furthermore, as with specimens P6 and P7 to P9, switching between high-density conditions and porous conditions may be performed two or more times, or high-density layers and porous layers may be formed alternately. In this example, two types of energy amounts, high-density conditions and porous conditions, are selectively used. However, when three or more types of energy amounts are used, even more diverse switching is possible.

[0056] Figure 3 is an explanatory diagram showing the experimental results of tensile strength. It shows the relationship between the proportion of high-density layer and tensile strength. As shown in the figure, as the proportion of high-density layer increases, the tensile strength increases roughly linearly. However, when the proportion of high-density layer is around 50%, it can be seen that the tensile strength is affected by the thickness of the high-density layer and the porous layer.

[0057] Figure 4 is an explanatory diagram showing the experimental results for tensile strength, deformation, and build speed. Figure 4(a) shows the tensile strength experimental results. This is the same result as Figure 3, but in a different format. Looking at the results R1 for test specimens P1 to P5, we can see that the tensile strength gradually decreased. Test specimens P1 to P5 were created by gradually decreasing the proportion of the high-density layer under conditions in which the high-density and porous conditions were switched at most once. As confirmed in Figure 3, these test specimens confirm that the proportion of the high-density layer and the tensile strength change almost linearly. Next, we examine the results R2 for test specimens P3 and P6 to P9. These test specimens have a high-density layer proportion of approximately 50%, and the thicknesses of the high-density and porous layers vary. These results indicate that test specimen P9, which has the thinnest layers, achieved a tensile strength similar to that of test specimen P1. Considering that the layer thicknesses of test specimens P7 to P9 gradually decreased, it seems that a thinner layer thickness is preferable from the perspective of tensile strength. However, since the test piece P7, which has a thinner layer than the test piece P6, has a slightly lower tensile strength, it cannot be said with certainty that there is a linear relationship between the layer thickness and the tensile strength.

[0058] Figure 4(b) shows the experimental results for the deformation of the test specimens. The positive side represents downward convex deformation, and the negative side represents upward convex deformation. Looking at the results R3 for test specimens P6 to P9, it can be seen that even when the proportion of the high-density layer is 50%, the deformation amount is affected by the thickness of the layer region. For test specimens P6 to P8, the deformation amount decreases as the layer region becomes thinner. However, for test specimen P9, the deformation amount is large despite the thinnest layer region. Furthermore, while the results for test specimens P1 to P5 confirm that the proportion of the high-density layer affects the deformation amount, it is not possible to determine any correlation. The deformation amount is due to residual stresses caused by heating and cooling during the building process. However, in the case of the powder bed method used in this example, residual stresses are likely also affected by heat transfer to the building table 11, so a comprehensive analysis of various factors is necessary.

[0059] Figure 4(c) shows the experimental results of the building speed of the test pieces. This experiment was conducted only on test pieces P1 and P7-P9. These results show that when building a mixture of high-density and porous layers, as in test pieces P7-P9, the scanning speed of the porous layer is faster than that of the high-density layer, which has the advantage of enabling a faster building speed compared to when all layers are high-density (test piece P1). However, it was confirmed that the building speed was slightly reduced by thinning each layer, as in test pieces P7-P9.

[0060] Based on the experimental results shown above, it was confirmed that when performing three-dimensional printing, the tensile strength, deformation amount, and printing speed are all affected by changing the ratio of high-density to low-density layers and the thickness of each layer. In other words, by controlling the amount of energy, it is possible to create a model that appropriately meets the requirements for the mechanical strength, distortion tolerance, etc. required of the object.

[0061] C. Modeling Process: Next, the process of modeling performed by the 3D modeling device 10 of this embodiment will be described. FIG. 5 is a flowchart of the modeling process. This process is executed by the control unit of the 3D modeling device 10. When the process starts, the 3D modeling device 10 first reads shape data of the model (step S10). This shape data defines the shape of each thin layer when the model is constructed as a laminate of thin layers. If the shape data is three-dimensional data of the model, the 3D modeling device 10 needs to generate shape data of the thin layers based on the three-dimensional data.

[0062] Next, the 3D modeling device 10 sets an energy amount profile (step S11). In this embodiment, as will be described later, the profile set by the profile setting device 20 is read. Alternatively, the profile may be set manually by an operator of the 3D modeling device 10.

[0063] Once the above process is complete, the 3D modeling device 10 begins the modeling process. First, the first thin layer is formed (step S12). As explained with reference to FIG. 1, the thin layer is formed by irradiating a laser according to shape data to melt and solidify the metal powder. This laser irradiation is performed under either high-density or porous conditions according to the profile. If the modeling is not complete after forming a thin layer (step S13), the 3D modeling device 10 sets the energy amount for the next thin layer according to the profile (step S14). This process does not necessarily mean that the energy amount must be changed for each layer. If the next layer is also irradiated with the same energy amount, there is no need to change the energy amount. The 3D modeling device 10 repeatedly performs the above process until the modeling is complete (step S13).

[0064] In this embodiment, the energy profile is set by machine learning, and this method will be described below. Figure 6 is a flowchart of the learning model generation process and the profile setting process. These processes are executed by the profile setting device 20. However, as described in Figure 1, the profile setting device 20 does not need to be a single device, and may be configured by connecting multiple computers, servers, etc. via a network.

[0065] First, we will explain the learning model generation process executed by the profile setting device 20. When this process starts, the profile setting device 20 reads past shaping data from the shaping database (step S20). The shaping data may include, for example, the shape of the shaped object, its mechanical strength, the magnitude of distortion during shaping, and the profile used for shaping.

[0066] The profile setting device 20 then generates a learning model by machine learning regression (step S21) and saves the obtained learning model (step S22). This learning model is intended to provide a profile that realizes requirements such as mechanical strength and the magnitude of distortion during modeling. Various methods can be used as the algorithm for generating the learning model.

[0067] Next, the profile setting process will be described. This process is performed to determine the profile when a model is to be formed. When the process starts, the profile setting device 20 reads required values ​​(step S30). The required values ​​may be the mechanical strength or the tolerance range of distortion for the shape of the model. Next, the profile setting device 20 reads the learning model generated in the learning model generation process (step S31). Then, the learning model is used to set a profile of the amount of energy (step S32). The set profile is output to the 3D modeling device 10 (step S33).

[0068] In the above process, it is not necessary to generate a learning model when executing the profile setting process, but a learning model may be generated in advance. Also, a learning model that has been generated once may be continued to be used.

[0069] According to the present embodiment described above, when melting and solidifying metal powder to form a shape, by mixing layers with different energy amounts, i.e., high-density conditions and porous conditions, distortion due to residual stress can be suppressed. Furthermore, depending on the profile settings, it is possible to suppress distortion while satisfying mechanical strength requirements. Therefore, according to this embodiment, it is possible to improve the accuracy of three-dimensional modeling and enhance its convenience.

[0070] In this embodiment, a profile can be set by machine learning, which allows an appropriate profile to be set so as to satisfy required values ​​for mechanical strength, distortion, and the like.

[0071] It is not necessary to satisfy all of the various features described in this embodiment, and some of them may be omitted or combined as appropriate. Furthermore, the present invention is not limited to the embodiments, and various modifications can be made. For example, the amount of energy is not limited to the two conditions of high density and porousness as in the embodiments, and more diverse conditions may be set. The present invention is not limited to the powder bed method as in the embodiments, and can be applied to various molding methods that melt and solidify powder.

[0072] INDUSTRIAL APPLICABILITY The present invention can be used to suppress deformation that occurs during three-dimensional modeling in a method of laminating thin layers formed by melting and solidifying powder.

[0073] REFERENCE SIGNS LIST 10 Three-dimensional modeling device 11 Modeling table 12 Moving mechanism 13 Recoater 15 Laser irradiation unit 16 Control unit 17 Input unit 20 Profile setting device 21 Modeling database 22 Required value input unit 23 Machine learning unit 24 Profile generation unit

Claims

1. A three-dimensional modeling method for forming a three-dimensional object, comprising: (a) a step of defining the shape of each layer to constitute the object with thin layers of a predetermined thickness; (b) a step of applying a predetermined amount of energy to powder to melt it and solidify it to form the thin layers according to the shape; (c) a step of repeating step (b) and stacking the next thin layer on top of the formed thin layer; (d) a step of controlling the amount of energy for each step (b) so that two or more layer regions with different amounts of energy are mixed in the stacking direction; and (f) a step of setting the amount of energy to be applied to each thin layer according to the range of distortion allowable for the object, wherein step (d) controls the amount of energy according to the setting result of step (f).

2. A three-dimensional modeling method according to claim 1, wherein step (d) forms the plurality of thin layers with a first amount of energy, and then switches to a second amount of energy.

3. A three-dimensional modeling method according to claim 1, wherein said step (d) switches said amount of energy at least twice during said modeling process.

4. A three-dimensional modeling method according to claim 1, wherein step (d) controls the amount of energy so that first layer regions made of thin layers formed by a first amount of energy and second layer regions made of thin layers formed by a second amount of energy are alternately stacked.

5. A three-dimensional modeling method according to claim 1, further comprising: (e) setting the amount of energy to be applied to each thin layer in accordance with the mechanical strength required for the model; and in step (d), controlling the amount of energy in accordance with the setting result of step (e).

6. A three-dimensional modeling method according to claim 1, wherein said step (b) comprises stacking said thin layers with metal powder.

7. A three-dimensional printing device for forming a three-dimensional object, comprising: an input unit that reads shape data for each layer to form the object with thin layers of a predetermined thickness; a thin layer forming unit that applies a predetermined amount of energy to powder to melt it and solidify it to form the thin layer in accordance with the shape data; a lamination unit that moves the thin layer forming unit and the already formed thin layer relatively to each other so that the next thin layer is laminated on top of the formed thin layer; and a control unit that controls the amount of energy in the thin layer forming unit so that two or more layer regions with different amounts of energy are mixed in the lamination direction, wherein the control unit sets the amount of energy to be applied to each thin layer in accordance with the range of distortion allowable for the object, and controls the amount of energy in accordance with the setting result.

8. A computer program for controlling a three-dimensional printing device that forms a three-dimensional object, the program executing the following steps: (a) receiving input of the shape of each layer to form the object from thin layers of a predetermined thickness; (b) controlling the powder to be melted by applying a predetermined amount of energy to the powder, and then solidifying the powder to form the thin layer according to the shape; (c) controlling the process to repeat step (b) and stack the next thin layer on top of the formed thin layer; (d) controlling the amount of energy for each step (b) so that two or more layer regions with different amounts of energy are mixed in the stacking direction; and (f) setting the amount of energy to be applied to each thin layer according to the range of distortion allowable for the object, and in step (d), controlling the amount of energy according to the setting result of step (f).

Images