Forming state estimation system and forming state estimation method

Abstract

Extract more detailed information about the quality of the additive manufacturing object while avoiding the complexity of the additive manufacturing device. [Solution] According to the present invention, there is provided a system for estimating the molding state of a layered object, the system comprising an image acquisition unit and an analysis unit, wherein the layered object is manufactured by repeating a material layer formation process in which material powder is supplied onto a molding area to form a material layer, and a solidified layer formation process in which one or more laser beams are irradiated onto the material layer to form a solidified layer, the image acquisition unit acquires images in real time of spatter generated around each molten pool formed by the irradiation of the one or more laser beams, the analysis unit extracts feature values ​​related to the spatter from the images and calculates coordinate values ​​indicating the position of the molten pool, and by inputting the feature values ​​into a trained model, the system estimates local parameters representing the molding state of the solidified layer and outputs the local parameters in a form associated with the coordinate values.

Description

[Technical Field]

[0001] The present invention relates to a system and method applicable to estimation of the printing state of a three-dimensional object. [Background technology]

[0002] Rapid prototyping, typified by three-dimensional additive manufacturing, has been attracting attention in manufacturing sites, and in recent years, rapid manufacturing, which applies rapid prototyping techniques to obtain final products, has also been attracting increasing attention. For example, Patent Document 1 listed below discloses an additive manufacturing device for laser additive manufacturing (LAM).

[0003] The additive manufacturing device described in Patent Document 1 acquires images of spatters generated by irradiating a material layer with laser light at an appropriate sampling rate, and analyzes these images to estimate a "virtual porosity" as a parameter indicating the molding state of the additive manufacturing object. The technology described in Patent Document 1 monitors this "virtual porosity" to determine whether the solidified layer is formed normally, and corrects the irradiation conditions of the laser light according to the determination result. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2022-121427 Summary of the Invention [Problem to be solved by the invention]

[0005] However, although the technology described in Patent Document 1 makes it possible to detect undesirable fluctuations in laser irradiation conditions during additive manufacturing, it is not necessarily suitable for evaluating whether the completed additively manufactured object has been manufactured as intended. It would be beneficial to be able to obtain more detailed information about the quality of the additively manufactured object without destroying the finished product, for example. [Means for solving the problem]

[0006] According to the present invention, the following inventions are provided. [1] A system for estimating the modeling state of an additively manufactured object, comprising an image acquisition unit and an analysis unit, wherein the additively manufactured object is manufactured by repeating a material layer formation process in which a material layer is formed by supplying material powder onto a modeling area, and a solidified layer formation process in which a solidified layer is formed by irradiating the material layer with one or more laser beams, wherein the image acquisition unit acquires images in real time of spatter generated around each molten pool formed by the irradiation of the one or more laser beams, and the analysis unit extracts feature values ​​related to the spatter from the images and calculates coordinate values ​​indicating the position of the molten pool, and estimates local parameters representing the modeling state of the solidified layer by inputting the feature values ​​into a trained model, and outputs the local parameters in a form associated with the coordinate values. [2] The system according to [1], wherein the local parameters are data representing the porosity of each part of the layered object, obtained for each coordinate value. [3] A system according to [1] or [2], wherein the analysis unit generates a three-dimensional image of the molding state of the additively molded object, represented by point cloud data, which is a collection of sets consisting of the coordinate values ​​and the local parameters. [4] A system according to any one of [1] to [3], further comprising a control unit, wherein the control unit controls the operation of a recoater head that supplies the material powder onto the building area, the coordinate values ​​including a Z coordinate value in the height direction of the layered object in addition to planar coordinate values ​​within the building area, and the analysis unit determines the cumulative number of layers of the solidified layer by detecting an image including the recoater head from a series of images acquired by the image acquisition unit, and calculates the Z coordinate value. [5] A system according to any one of [1] to [3], further comprising a control unit, wherein the coordinate values ​​include a Z coordinate value relating to the height direction of the layered object in addition to planar coordinate values ​​within the modeling area, and the analysis unit determines the cumulative number of layers of the solidified layers by analyzing a log of commands from the control unit, and calculates the Z coordinate value. [6] A system according to [4] or [5], wherein the control unit changes the irradiation conditions of the laser beam according to the estimated value of the local parameter. [7] A system described in any one of [1] to [6], wherein the one or more laser beams include a plurality of laser beams, and the analysis unit extracts features for each molten pool from the image and calculates coordinate values. [8] A method for estimating the modeling state of an additively manufactured object, comprising an image acquisition process and an analysis process, wherein the additively manufactured object is manufactured by repeating a material layer formation process in which material powder is supplied onto a modeling area to form a material layer, and a solidified layer formation process in which one or more laser beams are irradiated onto the material layer to form a solidified layer, wherein the image acquisition process acquires images of spatter generated around each molten pool formed by the irradiation of the one or more laser beams in real time, and the analysis process estimates local parameters representing the modeling state of the solidified layer by inputting feature values ​​related to the spatter into a trained model, and outputs the local parameters in a form associated with coordinate values ​​indicating the position of the molten pool. [9] The method according to [8], wherein the analyzing step includes a step of extracting the feature amount from the image and a step of calculating the coordinate value from the image. [Effects of the Invention]

[0007] In an embodiment of the present invention, not only are local parameters that represent the build state of the solidified layer estimated from the spatter image, but the coordinate values ​​of the molten pool are also calculated from the spatter image. This embodiment of the present invention makes it possible to obtain estimated local parameters in a form that is linked to the three-dimensional shape of the additively manufactured object, allowing, for example, a designer of the additively manufactured object to visually grasp the quality of the finished product. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 illustrates a schematic diagram of an exemplary build state estimation system in accordance with an embodiment of the present invention. [Figure 2] FIG. 2 is a perspective view of the appearance of a material layer forming mechanism 130. [Figure 3] FIG. 2 is a schematic top perspective view of a recoater head 136 of the material layer forming mechanism 130. [Figure 4] 10 is a schematic bottom perspective view of a recoater head 136 of the material layer forming mechanism 130. FIG. [Figure 5] 10 is a schematic diagram illustrating the arrangement of material powder in a modeling region R. FIG. [Figure 6] FIG. 2 is a diagram schematically illustrating an example of the configuration of a laser irradiation mechanism 140. [Figure 7] 2 is a schematic diagram illustrating an example of an image of spatter S acquired by an image acquisition unit 110. FIG. [Figure 8] FIG. 1 is an exemplary functional block diagram of a build state estimation system 1000. [Figure 9] FIG. 10 is a schematic diagram showing an example of expression of porosity by the modeling state estimating system 1000. [Figure 10] FIG. 10 is a schematic diagram showing another example in which the porosity is three-dimensionally visualized in association with the shape of an additive manufacturing object. [Figure 11] 10 is a flowchart illustrating an exemplary build state estimation method according to another embodiment of the present invention. [Figure 12] FIG. 12 is a diagram showing a step that can be included in step S5 shown in FIG. [Figure 13] FIG. 12 is a diagram showing a step that can be included in step S9 shown in FIG. [Figure 14] FIG. 10 is a diagram illustrating a schematic of an exemplary build state estimation system according to another embodiment of the present invention. [Figure 15] 10 is a schematic diagram illustrating another example of the image of the spatter S acquired by the image acquisition unit 110. FIG. [Figure 16]FIG. 10 is a functional block diagram of a modeling state estimation system according to yet another embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0009] As will be described in detail below with reference to the drawings, in a typical embodiment of the present invention, a three-dimensional object is obtained by a method similar to the additive manufacturing method described in Patent Document 1. More specifically, the additive manufacturing method produces the additive manufacturing object by repeating a material layer formation process and a solidified layer formation process. In the material layer formation process, a material powder is supplied onto a predetermined modeling area to form a material layer. In the solidified layer formation process, one or more laser beams are irradiated onto the material layer to form a solidified layer. That is, the overall shape of the additive manufacturing object is completed by sequentially forming multiple solidified layers, each having a predetermined thickness.

[0010] According to the technology described in Patent Document 1, by acquiring an image of spatter that occurs around the irradiation spot during manufacturing (hereinafter referred to as a "spatter image"), it is possible to obtain the "virtual porosity" of the layered object without destroying the layered object. Note that spatter images can be acquired multiple times during manufacturing.

[0011] Although the technology described in Patent Document 1 makes it possible to grasp the transition of "virtual porosity" during the manufacturing process of an additively manufactured object, these "virtual porosity" values ​​are only given for each acquired sputter image and are not obtained in a form corresponding to the three-dimensional shape of the additively manufactured object. Therefore, simply grasping the "virtual porosity" value may not be sufficient to accurately grasp the quality of the finished product. For example, if a low-density structure is intentionally created inside the additively manufactured object, a relatively high "virtual porosity" value may be estimated locally. Alternatively, conversely, a relatively low "virtual porosity" value may be estimated. In such cases, simply monitoring the transition of "virtual porosity" may not accurately evaluate the quality of the finished product.

[0012] The quality of a completed additive manufacturing object can be evaluated, for example, by cross-sectional observation based on the ratio of voids to a certain area. However, cross-sectional observation is a destructive test of a specific cross-section, and it is practically impossible to observe every cross-section of an additive manufacturing object. While the adoption of non-destructive testing using ultrasound or X-rays may be worth considering, preparing high-precision inspection equipment and inspecting the finished product every time an additive manufacturing object is completed simply for the sake of repeated manufacturing is also difficult to say practical from the perspective of cost and effort.

[0013] The present inventors have considered the above-mentioned circumstances and conducted extensive research, leading to the completion of the present invention. As will be described later, typical embodiments of the present invention make it possible to estimate, for example, the local porosity of the spot irradiated by the laser beam and its surrounding area, and to calculate the coordinate values ​​of the molten pool, from sputter images taken during processing. Embodiments of the present invention make it possible to extract more detailed information regarding the quality of the completed additively-shaped object from the sputter images. According to embodiments of the present invention, for example, it becomes easier to grasp the quality of the completed additively-shaped object at a glance, and the pass / fail of the additively-shaped object can be more effectively determined.

[0014] The following describes embodiments of the present invention. The various features shown in the following embodiments can be combined with each other. Furthermore, each feature can be an independent invention.

[0015] 1. Exemplary Embodiments of the Printing State Estimation System FIG. 1 illustrates an exemplary build state estimation system according to an embodiment of the present invention. The build state estimation system 1000 illustrated in FIG. 1 includes an additive manufacturing apparatus 100 and an external computing device 200. For ease of explanation, FIG. 1 illustrates three arrows indicating mutually orthogonal X-axis, Y-axis, and Z-axis. In this figure, the Z-axis is parallel to the vertical direction. Arrows indicating these X-axis, Y-axis, and Z-axis may also be illustrated in other figures following FIG. 1.

[0016] The additive manufacturing apparatus 100 of the manufacturing state estimation system 1000 has a manufacturing area R where an additively manufactured object is formed by irradiating a material layer with a laser, and further has an image acquisition unit 110. The image acquisition unit 110 includes one or more cameras, each capable of capturing images of the manufacturing area R. In the configuration illustrated in FIG. 1 , the image acquisition unit 110 is a digital still camera or a digital video camera disposed above the manufacturing area R.

[0017] The external computing device 200 of the modeling state estimation system 1000 is, for example, a personal computer that is communicatively connected to the additive manufacturing apparatus 100 via a wired or wireless connection. The external computing device 200 has one or more processors and one or more memories. As schematically shown in FIG. 1 , the external computing device 200 includes an analysis unit 210 and a display unit 220 such as a liquid crystal panel. The analysis unit 210 is configured by the processor and memory of the external computing device 200, and analyzes images acquired by the image acquisition unit 110 of the additive manufacturing apparatus 100.

[0018] As will be described in detail later, in a typical embodiment of the present invention, the image acquisition unit 110 of the additive manufacturing apparatus 100 acquires an image of spatter generated by laser irradiation of a material layer formed in the manufacturing region R. The analysis unit 210 of the external computing device 200 extracts certain features from the image acquired by the image acquisition unit 110. The analysis unit 210 also calculates coordinate values ​​indicating the position of a molten pool formed by laser irradiation from the spatter image. The analysis unit 210 estimates parameters (e.g., porosity) that represent the manufacturing state of the solidified layer using a trained model, and outputs the estimated parameters in association with the coordinate values ​​indicating the position of the molten pool. Specific examples of parameter output will be described later.

[0019] (1.1 Additive Manufacturing Apparatus 100) In the following, we will first explain an example configuration of the additive manufacturing apparatus 100, particularly the additive manufacturing apparatus 100, of the manufacturing state estimation system 1000. In general, the additive manufacturing apparatus 100 can have the same configuration as the additive manufacturing apparatus described in Patent Document 1. For reference, the entire disclosure of JP 2022-121427 A is incorporated herein by reference. Here, we will only provide an overview of the specific configuration of the additive manufacturing apparatus 100, avoiding excessively detailed explanations.

[0020] 1, the additive manufacturing apparatus 100 includes a chamber 120 in addition to the image acquisition unit 110 described above. The additive manufacturing apparatus 100 also includes a material layer formation mechanism 130, a laser irradiation mechanism 140, and a control unit 150. The control unit 150 includes a controller that controls the operations of the material layer formation mechanism 130 and the laser irradiation mechanism 140. As will be described later, the additive manufacturing apparatus 100 may further include a temperature sensor that monitors the temperature of the molten pool.

[0021] 1.1.1 Chamber 120 The chamber 120 has a structure that covers the manufacturing region R and may have, for example, a door on the front surface thereof for accessing the manufacturing space 120v inside the chamber 120. When performing additive manufacturing, an inert gas of a predetermined concentration is introduced into the manufacturing space 120v from an inert gas supply device (not shown in FIG. 1) through an inlet 20c provided in the chamber 120.

[0022] The inert gas can be any gas that does not substantially react with the material layer and / or solidified layer formed in the building region R. The inert gas is selected appropriately depending on the material powder used in additive manufacturing. Typical examples of inert gases are nitrogen gas, argon gas, and helium gas. By filling the chamber 120 with an inert gas, the oxygen concentration in the building space 120v can be kept sufficiently low. For example, in additive manufacturing of metals, maintaining a low oxygen concentration in the building space 120v contributes to suppressing deterioration of the material powder that constitutes the material layer and to stable irradiation of the laser beam onto the material layer.

[0023] The inert gas introduced into the chamber 120 is collected through the exhaust port 20d. The exhausted gas is sent to a fume collector (not shown), where fumes in the gas are removed by the fume collector and the gas is returned to the chamber 120. That is, the inert gas can be circulated between the chamber 120 and the fume collector. Examples of fume collectors are a dry electrostatic precipitator and a filter-type dust collector.

[0024] (1.1.2 Material layer formation mechanism 130) 2 shows an example of the external appearance of the material layer forming mechanism 130 taken out of the additive manufacturing apparatus 100. The material layer forming mechanism 130 includes a base 132 and a recoater head 136 that is movable above the base 132.

[0025] The base 132 has a drive mechanism 32 for the recoater head 136. In the configuration illustrated in Fig. 2, the drive mechanism 32 includes two guide rails 32L, each extending along the X-axis in the figure, and an actuator 32A such as a servo motor. Here, the two guide rails 32L, each extending parallel to the X-axis, are spaced apart along the Y-axis. The recoater head 136 is supported by these guide rails 32L and is reciprocatable along the X-axis by the actuator 32A.

[0026] As shown in FIG. 2, the printing region R is located between these guide rails 32L. In other words, the recoater head 136 can be installed in the material layer formation mechanism 130 so as to straddle the printing region R. Here, the shape of the printing region R in plan view as seen in the positive direction of the Z axis is rectangular, and one side of this rectangle is parallel to the X axis. Of course, the printing region R is not limited to a rectangular shape. The printing region R may have a circular, oval, or other shape in plan view.

[0027] 3 and 4 show the recoater head 136 of the material layer forming mechanism 130. The recoater head 136 includes a main body 36 having a rectangular parallelepiped shape extending along the Y axis. As shown in FIG. 3, the main body 36 has a reservoir 36R that opens upward. The reservoir 36R temporarily stores material powder for forming the material layer.

[0028] 4, a slit 36S that communicates with the reservoir 36R is provided on the bottom surface 136b of the recoater head 136. The material powder contained in the reservoir 36R can be supplied to the modeling region R via the slit 36S by moving the recoater head 136. As shown in the figure, a blade 35 and a blade 37 for leveling the material powder applied to the modeling region R can be provided on the front surface 136f and the rear surface 136r of the recoater head 136, respectively.

[0029] Referring again to Fig. 1, as schematically shown in Fig. 1, the base 132 of the material layer forming mechanism 130 includes retaining walls 34 that extend downward relative to a plane on which the recoater head 136 is disposed. The retaining walls 34 are disposed to surround the build region R, and in this case, the base 132 includes four retaining walls 34 corresponding to the build region R having a rectangular shape.

[0030] The base 132 of the material layer formation mechanism 130 further includes a modeling table 5 and an actuator 7. The modeling table 5 is disposed in a cylindrical space (which may also be called a "shaft") defined by the holding wall 34, and is movable up and down along the Z axis in predetermined steps by the actuator 7. The movement amount of the modeling table 5 is, for example, in the range of 20 μm to 200 μm, and preferably in the range of 30 μm to 70 μm. Here, the movement amount of the modeling table 5 per step is 50 μm. The upper surface 5a of the modeling table 5 corresponds to the above-mentioned modeling region R.

[0031] After the modeling table 5 is lowered in a predetermined number of steps, the recoater head 136 is moved along the X-axis while supplying material powder from the recoater head 136 into the space created by the lowering of the modeling table 5, thereby forming a material layer having a predetermined thickness on the modeling region R. Typically, a base plate 6 that is removable from the modeling table 5 is placed on the upper surface 5a, and a model is formed on the base plate 6.

[0032] 5 shows the formation of a material layer in the printing region R. For example, when the recoater head 136 is moved along the X-axis from the position shown in FIG. 1 to the position shown in FIG. 5, material powder is supplied from the recoater head 136 to the space created by the lowering of the printing table 5. By moving the recoater head 136, a material layer 10 having a thickness extending from the upper surface 5a of the printing table 5 to the bottom surface 136b of the recoater head 136 can be formed in the printing region R, as schematically shown in FIG. 5. The actuator 7 for moving the printing table 5 up and down and the actuator 32A (see FIG. 2) for moving the recoater head 136 are driven under the control of the control unit 150.

[0033] (1.1.3 Laser irradiation mechanism 140) After the material layer 10 is formed in the manufacturing region R, laser irradiation is performed on the material layer 10. In the example shown in Fig. 5, a laser irradiation mechanism 140 is installed above the chamber 120, and the material layer 10 is irradiated from above with a laser beam B emitted from the laser irradiation mechanism 140. Note that Fig. 5 shows an example in which one laser beam B is irradiated onto the material layer 10, but as will be described later, if the additive manufacturing apparatus has multiple laser irradiation mechanisms, multiple laser beams may be irradiated onto the material layer 10 simultaneously.

[0034] 5, the laser beam B from the laser irradiation mechanism 140 is irradiated onto the material layer 10 through a window 22 provided in the upper part of the chamber 120. Examples of materials for the window 22 include quartz glass, borosilicate glass, or crystals of germanium, silicon, zinc selenide, potassium bromide, or the like, and may be selected appropriately depending on the laser light source. If a fiber laser or YAG laser is used as the laser light source, for example, a quartz glass plate may be used for the window 22.

[0035] As shown in FIG. 5, a fume diffusion unit 24 for preventing fumes from adhering to the window 22 may be provided inside the chamber 120. The fume diffusion unit 24 has a shape that covers the window 22 from below. In this example, the fume diffusion unit 24 includes, for example, a cylindrical housing 24H and a diffusion member 24D having a large number of pores 2. The diffusion member 24D has, for example, a cylindrical shape like the housing 24H, and, as shown schematically in FIG. 5, is disposed in a space 24v formed by the ceiling of the chamber 120 and the housing 24H, thereby separating the space 24v into two spaces. However, these spaces are connected to each other by the pores 2 in the diffusion member 24D.

[0036] During additive manufacturing, clean inert gas is supplied from the inert gas supply device (not shown) to the outer space 2e of the two spaces partitioned by the diffusion member 24D. The inert gas introduced into space 2e passes through the pores 2 of the diffusion member 24D and flows into space 2c surrounded by the diffusion member 24D. The clean inert gas introduced into space 2c through the pores 2 is discharged into the manufacturing space 120v from opening 24d provided in a portion of the housing 24H located below the window 22.

[0037] As shown in Fig. 5, the lower surface 22b of the window 22 is exposed to the space 2c. By filling the space 2c with clean gas and discharging the gas toward the shaping space 120v, the intrusion of fumes into the space 2c can be reduced. Preventing fumes from entering the space 2c significantly reduces the adhesion of fumes to the window 22.

[0038] By irradiating the material layer 10 with the laser beam B through the window 22 of the chamber 120 and the opening 24d of the fume diffusion unit 24, it is possible to melt or sinter a portion of the material powder that constitutes the material layer 10. By cooling after irradiation with the laser beam B, a solidified layer is formed from the melted or sintered material powder. In other words, by irradiating with the laser beam B, it is possible to selectively change a portion of the material powder layer into a solidified layer.

[0039] Fig. 6 shows an example of the configuration of the laser irradiation mechanism 140. In the configuration shown in Fig. 6, the laser irradiation mechanism 140 includes a laser oscillator 143 and a galvano unit 144 as a scanning optical system. The operations of the laser oscillator 143 and the galvano unit 144 are controlled by a laser control unit described below.

[0040] There are no particular limitations on the laser light source that can obtain a laser output that can melt or sinter material powder, and it is possible to use such a laser light source as the laser oscillator 143. Examples of the laser light source include a fiber laser, a CO2 laser, and a YAG laser.

[0041] In the example shown in FIG. 6, the galvanometer unit 144 includes a collimator 44, a focus control unit 46, and a galvanometer 48. The collimator 44 has a collimator lens 44L therein and shapes the laser beam emitted from the laser oscillator 143 into a parallel beam. The focus control unit 46 has, for example, a movable lens 46L and a condenser lens 46M therein and adjusts the beam diameter of the parallel beam from the collimator 44. The movable lens 46L of the focus control unit 46 can be moved along the optical axis of the beam by an actuator (not shown). By adjusting the position of the movable lens 46L, the focal position of the laser beam B irradiated onto the material layer 10 can be adjusted. Note that the number of lenses and the shape of each lens shown in FIG. 6 are merely exemplary and are not intended to limit the actual configuration.

[0042] The galvanometer 48 includes a first mirror 48A and a second mirror 48B, each of which is connected to an actuator (not shown) and thereby can rotate independently. The galvanometer 48 steers the beam that has passed through the focus control unit 46 under the control of the laser control unit. The laser irradiation mechanism 140 uses the steering of the galvanometer 48 to two-dimensionally scan the laser beam B over the material layer 10, thereby selectively melting or sintering the material powder in the portion of the material layer 10 that is irradiated with the laser beam B.

[0043] (1.2 External computing device 200) Next, attention will be focused on the external computing device 200 of the printing state estimation system 1000. As described with reference to FIG. 1, the external computing device 200 includes an analysis unit 210. In a typical embodiment of the present invention, the analysis unit 210 roughly performs the following three functions.

[0044] The first is to acquire a sputter image obtained by the image acquisition unit 110 of the additive manufacturing apparatus 100 and extract feature quantities related to the sputter from the sputter image. The second is to calculate coordinate values ​​indicating the position of the molten pool formed by irradiation with the laser beam. The third is to output local parameters indicating the molding state of the solidified layer in association with the coordinate values ​​indicating the position of the molten pool. As will be explained in detail later, in an embodiment of the present invention, the analysis unit 210 estimates the local porosity as a local parameter by inputting the feature quantities extracted from the sputter image into a trained model.

[0045] (1.2.1 Extraction of spatter-related features) As shown schematically in FIG. 6 , when the material layer 10 formed in the build region R by the recoater head 136 is irradiated with a laser beam B, some of the material powder constituting the material layer 10 at the irradiated location melts to form a molten pool P, and spatter S occurs around the molten pool P. In this specification, spatter S refers to particles that fly off from the molten pool P or its periphery during laser irradiation. The actual substance of the spatter S may include molten metal particles and unmelted material powder that fly off from the molten pool P.

[0046] The image acquisition unit 110 of the layered manufacturing apparatus 100 captures images of the manufacturing region R at predetermined intervals (i.e., at a predetermined frame rate), for example, and transmits image data of the spatters S to the external computing device 200. The frame rate for capturing images is, for example, in the range of 5 fps to 40 fps, and preferably in the range of 10 fps to 30 fps. Here, the frame rate is set to 23 fps.

[0047] Typically, the size of the field of view (FOV) of the image acquisition unit 110 is set so that the entire printing region R is included in the image plane. In this embodiment, the image acquisition unit 110 acquires images of the spatter S in real time during additive manufacturing. The acquisition of images of the spatter S is not limited to regular intervals, and may be performed at irregular intervals.

[0048] The analysis unit 210, which receives the image data from the image acquisition unit 110, extracts feature quantities related to the sputter S through image analysis. The feature quantities related to the sputter S include, for example, one or more of the following: Of the sputter particles captured in the image, the number of sputter particles (hereinafter sometimes referred to as "sputter particles") that are of interest for feature quantity extraction may be one, or two or more. Brightness value at the center of the sputtered particle Brightness value of the red component (R component) at the center of the sputter particle Brightness value of the green component (G component) at the center of the sputtered particle Ratio of the brightness value of the G component to the brightness value of the R component (G / R) Distance between the center of the sputter particle and the molten pool Distance in the X direction between the center of the sputter particle and the molten pool Distance in the Y direction between the center of the sputter particle and the molten pool The ratio of the length of the major axis to the minor axis when the outline of a sputtered particle is approximated as an ellipse The ratio of the length of the major axis to the minor axis when the outer shape of a sputtered particle is approximated as a rectangle (elongation rate) Hydraulic diameter when the outer shape of a sputtered particle is considered to be the cross section of a non-circular pipe Circumference of sputtered particle L Area of ​​sputtered particle A Total number of sputtered particles The total number of sputtered particles sorted based on some characteristic

[0049] FIG. 7 shows an example of an image of spatter S acquired by the image acquisition unit 110. FIG. 7 shows particles scattering from the center of the molten pool P toward the upper right of the figure. As can be seen from FIG. 7, the "image of spatter S" in this specification refers to an image that includes not only an image of spatter particles but also an image of the molten pool P. Typically, the molten pool P is identified as a nearly circular region that includes the brightest part in the image and is brighter than a certain threshold. Spatter particles can also be identified as regions in the image that include pixels with relatively high brightness. Here, the spatter particle is represented by the largest area among multiple relatively bright regions.

[0050] In Figure 7, the distance indicated by the double-headed arrow d corresponds to the aforementioned "distance between the center of the sputter particle and the weld pool." The center of the sputter particle can be defined as the geometric center of the sputter particle's outline, and the position of the weld pool can also be represented by the geometric center of the weld pool's outline. In Figure 7, the distance dx indicated by the double-headed arrow parallel to the X-axis corresponds to the aforementioned "distance in the X direction between the sputter particle center and the weld pool." Similarly, the distance dy indicated by the double-headed arrow parallel to the Y-axis corresponds to the aforementioned "distance in the Y direction between the sputter particle center and the weld pool." Note that in Figure 7, the outline of the weld pool P is depicted as a circle, and the outline of the sputter particle is depicted as an oval or ellipse. However, please note that this is merely for convenience of explanation and is not intended to represent the actual shape of the weld pool P or the sputter particle.

[0051] The hydraulic diameter D of a sputtered particle can be calculated as D = (4A / L) using the above-mentioned area A and perimeter L. Examples of the "total number of sputtered particles sorted based on some characteristic quantity" include the number of sputtered particles whose brightness at the center is equal to or greater than a predetermined value, the number of sputtered particles whose major axis / minor axis length ratio when the outer shape is approximated as an ellipse is less than a predetermined value, and the number of sputtered particles whose elongation rate is equal to or greater than a predetermined value.

[0052] (1.2.2 Estimation using a trained model) Fig. 8 is an exemplary functional block diagram of the modeling state estimation system 1000. In the configuration illustrated in Fig. 8, the control unit 150 of the additive manufacturing apparatus 100 includes a numerical control unit 50, a gas system control unit 128, a recoater control unit 138, a laser control unit 148, and a table control unit 158.

[0053] The gas system control unit 128, recoater control unit 138, laser control unit 148, and table control unit 158 ​​are connected to the numerical control unit 50 and function as controllers that control the operation of each mechanism in the layered manufacturing apparatus 100 based on commands from the numerical control unit 50. For example, the gas system control unit 128 controls the operation of the inert gas supply unit and the fume collector based on commands from the numerical control unit 50. The recoater control unit 138 drives the actuator 32A (see FIG. 2) based on commands from the numerical control unit 50, and controls the reciprocating movement of the recoater head 136.

[0054] In this example, the additive manufacturing apparatus 100 further includes a temperature sensor 112. The temperature sensor 112 is a radiation thermometer such as a pyrometer, and monitors the temperature of the molten pool P formed by laser irradiation of the material layer 10. The installation of the temperature sensor 112 is not essential to the embodiment of the present invention, but the output of the temperature sensor 112 may be used as an auxiliary means for estimating local parameters and / or calculating coordinate values ​​indicating the position of the molten pool.

[0055] Next, attention will be focused on the external calculation device 200. In the configuration illustrated in FIG. 8, the analysis unit 210 of the external calculation device 200 has a learning unit 12, a memory 14, and an image generation unit 16. Here, the learning unit 12 includes a storage unit 12M that holds the trained model LM, and a calculation unit 12C that executes input to the trained model LM and receives output from the trained model LM. The calculation unit 12C also has the function of updating a set of parameters in the training stage of the trained model LM.

[0056] The analysis unit 210 receives an image of the spatter S sent from the image acquisition unit 110 of the additive manufacturing apparatus 100, extracts a set of features from the image of the spatter S, and inputs the set of features into the trained model LM. The analysis unit 210 obtains local parameters related to the quality of the additive manufacturing object as output from the trained model LM. In this embodiment, an estimated value of porosity is used as an example of the local parameter. Note that, in this specification, "estimation" refers to the process of predicting or approximating an unknown quantity using the trained model LM.

[0057] Here, the porosity output from the trained model LM is related to the position of the molten pool P formed by irradiating the material layer 10 with the laser beam B when the image of the spatter S is acquired. In other words, based on the trained model LM, the analysis unit 210 obtains the porosity corresponding to the position of the molten pool P for each image of the spatter S. In this sense, in this embodiment, the porosity value estimated based on the trained model LM is "local."

[0058] The local parameters obtained by the analysis unit 210 based on the input of feature quantities to the trained model LM are not limited to the porosity described above and may be other quantities. For example, the analysis unit 210 may obtain one or more of the laser power, spot diameter, and laser power density of the laser beam B as output from the trained model LM when acquiring an image of the spatter S. As explained in Patent Document 1, the conditions of laser irradiation on the material layer surface may change from moment to moment due to the influence of fumes and other factors generated as the modeling process progresses. Therefore, these values ​​obtained from model inference using the trained model LM can also be considered local estimates of the irradiation location on the material layer 10.

[0059] In this way, instead of or in addition to the porosity, other local parameters may be obtained as estimated values ​​from the trained model LM. In addition to the above-mentioned parameters, the trained model LM may be used to estimate the dryness of the material powder constituting the material layer 10 and the thickness of the material layer 10 at the time of irradiation with the laser beam B (i.e., the distance from the surface of the material powder layer to the solidified layer below the material powder). The local parameters may be given, for example, as a single value related to the porosity, or may be given in the form of a set of estimated values ​​related to multiple attributes (e.g., a numeric vector having the porosity, laser power, and spot diameter as components).

[0060] The estimated values ​​obtained by using the learned model LM are temporarily stored in a memory 14 such as a RAM. In a typical embodiment, the memory 14 stores the estimated values ​​obtained for each image of the spatter S until the additive manufacturing is completed.

[0061] (1.2.3 Calculation of coordinate values ​​indicating the position of the molten pool P) In addition to extracting feature amounts from the image of the spatter S, the analysis unit 210 calculates coordinate values ​​indicating the position of the molten pool P formed by the irradiation of the laser beam B. Here, the coordinate values ​​calculated by the analysis unit 210 include not only the planar coordinate values ​​of the molten pool P within the printing region R, but also the Z coordinate value relating to the height direction of the layered object.

[0062] Calculation of coordinate values ​​indicating the position of the molten pool P is typically performed for each image of the spatter S. The calculated coordinate values, together with the estimated value of the porosity, are stored, for example, in memory 14 and retained until the additive manufacturing is completed. Below, the calculation of the planar coordinates (a set of X and Y coordinates) and the calculation of the Z coordinate will be explained separately.

[0063] As described above, in a typical embodiment of the present invention, the field of view of the image acquisition unit 110 is adjusted to include the entire build region R. Therefore, each of the series of images of the spatter S acquired by the image acquisition unit 110 includes an image of the molten pool P formed by the irradiation of the laser beam B. The image of the molten pool P is generally expressed as a region that is large in area and high in brightness compared to the spatter S, and has a nearly circular shape. Therefore, for example, by using an appropriate filter to extract an area that is likely to be the molten pool P from the image and determining the geometric center of that area, the position of the molten pool P can be represented by the coordinate values ​​of the geometric center.

[0064] In this embodiment, what is ultimately desired is a coordinate value in the world / object coordinate system. On the other hand, the position of the molten pool P in the image of the spatter S is expressed as a two-dimensional coordinate value in the image plane coordinate system. Therefore, in reality, camera calibration is performed using a known method before the actual additive manufacturing. By completing camera calibration in advance, it becomes possible to convert the plane coordinate values ​​relating to the position of the molten pool P in the image of the spatter S into real-world XY coordinate values, i.e., XY coordinate values ​​in the world coordinate system. This coordinate conversion may be performed, for example, by the analysis unit 210.

[0065] The calculation of the XY coordinate values ​​relating to the position of the molten pool P may be performed based on other methods. For example, the analysis unit 210 of the external calculation device 200 may obtain information relating to the XY coordinate values ​​of the molten pool P through the control unit 150 of the additive manufacturing apparatus 100.

[0066] 8, the control unit 150 includes a laser control unit 148. The laser control unit 148 is a controller that controls the operation of the galvano unit 144 (see FIG. 6). As will be described below, the analysis unit 210 may obtain information regarding the XY coordinate values ​​of the molten pool P from the laser control unit 148.

[0067] The position of the molten pool P formed in the material layer 10 is related to the steering of the laser beam B. Therefore, by obtaining a drive signal for the galvanometer unit 144 from the laser control unit 148, which is the controller of the galvanometer unit 144, information regarding the position of the molten pool P in the XY plane of the world coordinate system or the image plane coordinate system can be obtained. For example, an AD converter is connected to the laser control unit 148 to obtain a drive signal related to the steering of the laser beam B in the XY plane. More specifically, in addition to the drive signal related to the steering of the laser beam B, a drive signal related to the ON / OFF state of the laser and the analog output of the temperature sensor 112 are also obtained. By combining these drive signals with a signal carrying information regarding the surface temperature of the material layer 10, the XY coordinate values ​​regarding the position of the molten pool P can be calculated.

[0068] However, such a method requires the acquisition of signals from the laser control unit 148, which can complicate the configuration and processing of the entire system. In particular, if the laser control unit 148 and the components for estimating the local parameters are supplied by different manufacturers, a bus or the like must be provided between the galvano unit 144 and the laser control unit 148 for extracting signals, which can make the entire system expensive. In contrast, the method of numerically calculating coordinate values ​​from an image of the spatter S is simple and inexpensive.

[0069] In this embodiment, the analysis unit 210 calculates or acquires not only the XY coordinate values ​​but also the Z coordinate value of the molten pool P. The method of calculating and acquiring the Z coordinate value is not limited to a specific method, and various methods can be used.

[0070] As is well known, in additive manufacturing, a solidified layer is formed by irradiating a laser beam onto the material powder constituting the material layer, and then the modeling table is lowered in a predetermined number of steps, and a recoater head is moved to form a new material layer on the solidified layer. Then, a second solidified layer is formed on the solidified layer by irradiating the material layer on the solidified layer with a laser beam. That is, if the image acquisition unit 110 performs image capture at a regular interval, the recoater head 136 will appear in the image of the sputter S each time a material layer 10 is formed. The analysis unit 210 can count the number of times the modeling table 5 has been lowered by detecting the recoater head 136 in the image through image analysis or the like. The number of times the modeling table 5 has been lowered is stored, for example, in the memory 14 and is updated each time the modeling table 5 is detected to be lowering. In other words, the analysis unit 210 can determine the cumulative number of layers of solidified layers by detecting an image including the recoater head 136, and the analysis unit 210 can obtain the Z coordinate value of the molten pool P from the cumulative number of layers of solidified layers through image analysis.

[0071] Another example of a method for acquiring the Z-coordinate value is to acquire and analyze a log of commands from the numerical control unit 50. As shown in FIG. 8 , the control unit 150 of the additive manufacturing apparatus 100 includes a table control unit 158 ​​connected to the numerical control unit 50. The table control unit 158 ​​controls the operation of the actuator 7, which raises and lowers the modeling table 5, based on commands from the numerical control unit 50. The analysis unit 210 can count the number of times the modeling table 5 is lowered by, for example, acquiring a log of commands from the numerical control unit 50 to the table control unit 158. The number of times the modeling table 5 is lowered can be considered to be the same as the cumulative number of solidified layers. Therefore, the Z-coordinate value of the molten pool P can also be calculated by analyzing the log of commands from the numerical control unit 50 to determine the cumulative number of solidified layers. However, the method of calculating the Z-coordinate value by analyzing the image acquired by the image acquisition unit 110 is still advantageous in that the process of calculating the Z-coordinate value can be completed in a manner independent of the format of the data output from the numerical control unit 50.

[0072] (1.2.4 Presentation of local parameters in relation to coordinate values) The analysis unit 210 obtains the estimated values ​​of the local parameters using the trained model LM and the coordinate values ​​of the molten pool P, and visualizes them in a linked form rather than individually. As can be understood from the explanation so far, the porosity as a local parameter and the three-dimensional coordinate values ​​indicating the position of the molten pool P are obtained for each image of the spatter S. Below, we will explain an example of how to express the porosity when the porosity is estimated as a local parameter.

[0073] 8, the estimated porosity value and the three-dimensional coordinate values ​​of the molten pool P obtained for each image of the spatter S are stored in the memory 14 in a mutually associated form. The image generating unit 16 of the analysis unit 210 reads out the data stored in the memory 14 and generates a three-dimensional image based on the read data. The image generating unit 16 displays the three-dimensional image based on the data read out from the memory 14 on, for example, the display unit 220.

[0074] FIG. 9 shows an example of the representation of the porosity by the manufacturing state estimation system 1000. Here, the local parameters are data representing the porosity of each part of the additively manufactured object, obtained for each coordinate value of the molten pool P. In other words, in a typical embodiment, the porosity as a local parameter is given as a set with the three-dimensional coordinate values ​​of each part of the additively manufactured object. Therefore, the image generation unit 16 can three-dimensionally visualize the porosity in association with the shape of the additively manufactured object.

[0075] In the example shown in Fig. 9, the image generation unit 16 constructs a three-dimensional image, as shown in the bottom row, of the additively-modeled object having the shape shown in the top row, based on a set of porosity and three-dimensional coordinate values. In this example, the porosity of each part of the additively-modeled object is expressed using grayscale. In the three-dimensional image shown in the bottom row of Fig. 9, for example, parts drawn with relatively low brightness have small estimated porosity values. In other words, areas of the additively-modeled object with a relatively high material density are expressed in colors close to black.

[0076] It goes without saying that the representation of porosity is not limited to the brightness of the pixels that make up the three-dimensional image. For example, a three-dimensional image may be drawn using a collection of dots. By changing the shading (brightness), size, or color of each dot depending on the magnitude of the porosity, the porosity of each part of the additively manufactured object can be visually represented. The image generation unit 16 may be configured to switch between multiple representations and display them on the display unit 220.

[0077] As can be understood from the above description, the manufacturing state of a completed additively-molten object can be represented by point cloud data, which is a collection of sets consisting of coordinate values ​​of the molten pool P and local parameters. In this embodiment, the analysis unit 210 generates a 3D image of the manufacturing state of the additively-molten object based on this point cloud data and displays it on the display unit 220 of the external computing device 200 or the operation panel of the additively-molten object 100. This allows the designer of the additively-molten object to easily visually grasp the quality of the completed additively-molten object in association with the shape of each part of the additively-molten object. According to this embodiment, defects such as an unintentional increase in porosity or a low porosity in a part that was intentionally made to have a low density (e.g., volume density) are visualized. In response to this, the operator of the manufacturing state estimation system 1000 can appropriately feedback the laser irradiation conditions during manufacturing to the next manufacturing, and this embodiment of the present invention contributes to reducing the reject rate.

[0078] FIG. 10 shows another example of three-dimensional visualization of porosity in association with the shape of an additively-made object. In the example shown in FIG. 10, the three-dimensional shape of the additively-made object is represented by a set of voxels Vx, and the porosity of each part of the additively-made object is represented by the brightness of the voxels Vx. As such, three-dimensional images generated based on point cloud data are not limited to the example shown in FIG. 9. By formatting the data handled by the analysis unit 210 as point cloud data, it becomes possible to generate images in a variety of representations. It is even easy to visualize the distribution of porosity in a cross section of an additively-made object.

[0079] As described above, according to the embodiment of the present invention, the local parameters (here, porosity) can be linked to three-dimensional coordinate values ​​and presented to the operator of the modeling state estimation system 1000 or the designer of the additively manufactured object. By receiving the porosity distribution in the finished product in the form of, for example, a three-dimensional image, the designer of the additively manufactured object can visually and easily grasp whether the additively manufactured object has been completed with the expected quality. Furthermore, if an unintended value is included in the local parameters, the operator of the modeling state estimation system 1000 can appropriately change the laser irradiation conditions, etc., before proceeding with the next manufacturing. In other words, since feedback to the manufacturing conditions is facilitated, the defect rate of the additively manufactured object can be reduced, thereby realizing efficient additive manufacturing.

[0080] (1.2.5 Feedback to manufacturing conditions) According to the inventors' research, the greater the energy imparted to the material layer by the laser, the greater the amount of molten material powder scattered from the irradiation spot, and the farther it spread. However, if the laser does not impart sufficient energy to the material layer, the smaller the molten pool formed at the irradiation spot, and the material layer is less likely to melt and solidify deep within, resulting in voids in the solidified layer. Therefore, the porosity distribution within an AM object indirectly reflects the appropriateness of the laser irradiation conditions during AM. Obtaining an estimate of the local porosity makes it possible to determine whether the manufacturing parameters, such as laser power and spot diameter, were within appropriate ranges without destroying the AM object.

[0081] Alternatively, instead of or in addition to the estimated porosity, the analysis unit 210 may obtain estimated values ​​for one or more of the laser power, spot diameter, and laser power density of the laser beam B as local parameters from the trained model LM. The state of laser irradiation of the surface of the material layer 10 changes from moment to moment as the modeling progresses. For example, by presenting the estimated laser power in the form of a three-dimensional image as shown in FIG. 9 or 10, it becomes possible to directly and visually grasp the change in laser power over time during additive manufacturing.

[0082] The determination of whether the parameters during manufacturing were within an appropriate range may be made by the analysis unit 210 of the external computing device 200 or the control unit 150 of the additive manufacturing apparatus 100. In the example shown in FIG. 8 , the numerical control unit 50 constituting part of the control unit 150 includes a memory 52 and a determination unit 54. The memory 52 may store, in advance, threshold values ​​related to the irradiation conditions of the laser beam B (e.g., laser power). The determination unit 54, for example, acquires data related to the local parameters (e.g., laser power) obtained by the analysis unit 210 of the external computing device 200 from the learning unit 12 and compares the data with the threshold values ​​stored in the memory 52. ​​The determination unit 54 can determine whether the parameters related to manufacturing the additive manufacturing object, such as the laser power, were within a predetermined range by comparing the estimated values ​​of the local parameters with the threshold values ​​read from the memory 52.

[0083] When a determination result is obtained that a parameter (e.g., laser power) related to the production of an additive manufacturing object is outside a predetermined range, the determination unit 54 may update the setting value of the parameter to an appropriate value. In response to the parameter update by the determination unit 54, the numerical control unit 50 sends a command based on the updated value to the laser control unit 148 for the next additive manufacturing. That is, the control unit 150 of the additive manufacturing apparatus 100 may be configured to change the laser beam irradiation conditions according to the estimated values ​​of the local parameters. According to this embodiment, even if the laser irradiation conditions (laser power, spot diameter, laser power density, etc.) during actual additive manufacturing fall outside the appropriate range, the appropriate laser irradiation conditions can be immediately reflected in the next manufacturing, thereby efficiently reducing the defect rate of additive manufacturing objects.

[0084] The update of parameters related to the manufacturing of the layered object by the determination unit 54 is not limited to a change in the laser irradiation conditions. For example, a decrease in laser power on the surface of the material layer 10 may be caused by an increase in the fume concentration in the chamber 120, resulting in partial shielding of the laser beam B by the fumes. In this case, the numerical control unit 50 may send a command to the laser control unit 148 to compensate for the attenuation of laser power due to the fumes, or may send a command to the gas system control unit 128 to correct settings related to the operation of the fume collector (e.g., the fan rotation speed of the fume collector).

[0085] The magnitude of the porosity in an additively manufactured object depends not only on irradiation conditions, such as the actual laser power applied to the material layer 10, taking into account the effects of fumes, but also on the shape of the object. Even if the effects of laser attenuation due to fumes could be eliminated, the heat received by the material powder from the laser may vary depending on the shape desired after melting and solidifying the material powder. For example, even within the same additively manufactured object, the quality of the solidified layer may differ between sharply shaped portions and more bulky portions. In other words, the increase in porosity in an additively manufactured object is due to at least two factors: time-related factors and shape-related factors. It is generally difficult to determine whether the time-related factor or shape-related factor contributes more to the increase in (local) porosity solely from numerical monitoring of the virtual porosity.

[0086] In contrast, according to a typical embodiment of the present invention, a local parameter, such as porosity, can be presented to an operator and a designer in a form linked to the shape of the additively-shaped object. Because more detailed information about the quality of the additively-shaped object is presented, for example, in the form of a three-dimensional image, according to a typical embodiment of the present invention, it becomes possible to set more appropriate manufacturing conditions that take into account temporal factors and the shape of the additively-shaped object. Furthermore, by storing a local parameter (e.g., porosity) linked to the shape of the additively-shaped object as, for example, three-dimensional point cloud data, it is possible to expect improved traceability regarding the quality of the additively-shaped object.

[0087] 2. Exemplary Operation Flow of the Printing State Estimation System 1000 FIG. 11 is a flowchart showing an exemplary build-state estimation method according to another embodiment of the present invention. In this exemplary embodiment, the build-state estimation method using the build-state estimation system 1000 includes an image acquisition process and an analysis process. The image acquisition process is, generally, a process of acquiring an image of spatter S generated around the molten pool P. The analysis process is a process of estimating local parameters using machine learning. More specifically, the analysis process is a process of estimating local parameters (e.g., porosity) by inputting feature quantities related to the spatter S into the trained model LM and outputting the local parameters in association with coordinate values ​​indicating the position of the molten pool P.

[0088] (2.1 Table lowering process S1) Prior to the start of additive manufacturing, the manufacturing space 120v is filled with inert gas by introducing the inert gas into the chamber 120. Then, the manufacturing table 5 is lowered by a predetermined amount along the Z axis by driving the actuator 7 (see FIG. 1).

[0089] (2.2 Material layer formation step S2) Next, the actuator 32A is driven to move the recoater head 136 along the X-axis from one end to the other end of the build region R. At this time, material powder is supplied to the build region R from the reservoir 36R of the recoater head 136 through the slit 36S of the recoater head 136 (see FIG. 4). As the recoater head 136 moves, the material powder discharged from the recoater head 136 is leveled by the blades 35 and 37 of the recoater head 136, and a material layer 10 of a predetermined thickness is formed in the build region R (see FIG. 5). If necessary, the recoater head 136 may be reciprocated along the X-axis.

[0090] (2.3 Laser irradiation step S3) After the material layer 10 is formed, a desired portion of the material layer 10 is irradiated with a laser beam B emitted from a laser irradiation mechanism 140 (see FIG. 5). By scanning with the laser beam B, the material powder in the desired portion of the material layer 10 can be selectively sintered or melted.

[0091] (2.4 Step S4 of acquiring images of spatter S) In an embodiment of the present invention, in parallel with the irradiation of the laser beam B, an image of spatter S (see FIGS. 6 and 7) generated around the molten pool P formed by the irradiation of the laser beam B is acquired by the image acquisition unit 110. The acquisition of the image of the spatter S is performed in real time during the additive manufacturing process. The image data acquired by the image acquisition unit 110 is sent to the analysis unit 210 of the external computing device 200.

[0092] (2.5 Image analysis step S5) Upon receiving the image data of the spatters S, the analysis unit 210 extracts feature values ​​related to the spatters S from the image of the spatters S and estimates local parameters by inputting the feature values ​​into the trained model LM, as described above. The analysis unit 210 also calculates coordinate values ​​indicating the position of the molten pool P from the image of the spatters S. That is, the image analysis step S5 shown in FIG. 11 may include a feature extraction step S51, a local parameter estimation step S52, and a coordinate value calculation step S53, as illustrated in FIG. 12. Regardless of the order illustrated in FIG. 12, the coordinate value calculation step S53 may be performed before steps S51 and S52, or may be performed in parallel with the set of steps S51 and S52. Data related to the estimated values ​​obtained as output from the trained model LM and data related to the coordinate values ​​obtained using image analysis or the like are stored in the memory 14 (see FIG. 8) of the analysis unit 210 in association with each other, for example, in the form of a list (step S54 shown in FIG. 12).

[0093] (2.6 Step S6 of forming solidified layer) The material sintered or melted by laser irradiation is then cooled to form a solidified layer. By scanning the laser beam B to selectively sinter or melt a portion of the material powder that constitutes the material layer 10, a solidified layer having a desired shape can be obtained in a manner similar to laser line drawing.

[0094] (2.7 Step S7 of determining whether scanning is complete) The scanning of the laser beam B is performed based on a command from the numerical control unit 50 of the layer-by-layer manufacturing apparatus 100. When scanning for one layer is completed, the manufacturing table 5 is lowered, a material layer 10 is formed, and scanning of the material layer 10 with the laser beam B is performed again (steps S1 to S3 in FIG. 11). Between the formation of a new material layer 10 on the solidified layer and the formation of a second solidified layer by laser irradiation on the new material layer 10, the image acquisition unit 110 acquires an image of the spatter S again (step S4 in FIG. 11).

[0095] There is no particular limit to the number of times that the image acquisition unit 110 acquires images of the spatters S between the formation of one material layer 10 and the formation of the solidified layer. The number of times that images of the spatters S are acquired may be determined appropriately taking into consideration the resolution of the three-dimensional image that is ultimately to be obtained, the amount of computational resources available for analyzing the images, and other factors. Furthermore, it is not essential that the analysis of the images of the spatters S for a given material layer 10 (or solidified layer) be completed during the period from the formation of that material layer 10 to the next descent of the modeling table 5. It is also possible to store the images of the spatters S acquired by the image acquisition unit 110 in a memory 14 (see FIG. 8 ) or the like, and then perform step S5 of analyzing the images of the spatters S after the completion of additive manufacturing.

[0096] (2.8 Process S8 for determining whether the modeling is complete) The layered object is manufactured by repeating a material layer forming process by supplying material powder onto the manufacturing region R, and a solidified layer forming process by scanning with the laser beam B. When the number of times (n times) these processes are performed exceeds a predetermined number (e.g., N times), the layered object manufacturing device 100 ends the layered object manufacturing. The completion of the layered object manufacturing can be determined, for example, by counting the number of times the manufacturing table 5 is lowered, storing the count in the memory 52 of the numerical control unit 50, or the like, and comparing the count with a predetermined threshold value N.

[0097] (2.9 Step S9 of outputting three-dimensional images) As illustrated in Fig. 11, the method for estimating the printing state may additionally include a step S9 of outputting a three-dimensional image. As illustrated in Fig. 13, the step S9 of outputting the three-dimensional image may include a step S91 of reading estimated values ​​and coordinate values, a step S92 of generating a three-dimensional image, and a step S93 of displaying the three-dimensional image.

[0098] To output a three-dimensional image, the image generation unit 16 (see FIG. 8) of the analysis unit 210 reads out multiple sets of data stored in the memory 14, each containing estimated values ​​and coordinate values ​​(step S91 in FIG. 13). The entire set of data related to estimated values ​​of local parameters and coordinate values ​​of the molten pool P constitutes so-called point cloud data. Based on the read-out data, the image generation unit 16 generates a three-dimensional image, such as that shown in FIG. 9 or 10 (step S92 in FIG. 13). Because the data set includes information on the coordinate values ​​of the molten pool P, the generated three-dimensional image reflects the shape of the completed additive manufacturing object. The three-dimensional image generated by the image generation unit 16 is displayed, for example, on the screen of the display unit 220 of the external computing device 200 (step S93 in FIG. 13).

[0099] If necessary, the determination unit 54 may determine whether the local parameters are within a predetermined range, and the laser irradiation conditions may be updated or corrected according to the determination result. The method for estimating the modeling state may additionally include a step of changing the laser irradiation conditions.

[0100] 3. Other Embodiments of the Printing State Estimation System (3.1 Additive manufacturing using multiple laser beams) Fig. 14 shows an outline of an exemplary modeling state estimation system according to another embodiment of the present invention. The modeling state estimation system 1001 shown in Fig. 14 differs from the modeling state estimation system 1000 described with reference to Figs. 1 to 6 in that it includes an additive manufacturing apparatus 101 instead of the additive manufacturing apparatus 100. Note that in Fig. 14, the external calculation device 200 is not shown to avoid overly complicating the drawing.

[0101] 14, the additive manufacturing apparatus 101 has a laser irradiation mechanism 141 and a laser irradiation mechanism 142, each of which is disposed above the chamber 120. That is, in the embodiment described here, the additive manufacturing apparatus 101 includes two or more laser irradiation heads.

[0102] As schematically shown in FIG. 14 , in additive manufacturing using an additive manufacturing apparatus 101, two laser irradiation heads each independently irradiate a material layer 10 with a laser beam. In the example shown in FIG. 14 , a laser irradiation mechanism 141 scans the material layer 10 with a laser beam B1, and a laser irradiation mechanism 142 scans the material layer 10 with a laser beam B2. That is, in this example, two molten pools can be simultaneously formed at separate locations on the material layer 10. Note that in this example, the material layer 10 is irradiated with the laser beam B1 from the laser irradiation mechanism 141 and the laser beam B2 from the laser irradiation mechanism 142 via a common window 22 and a common fume diffusion unit 24. However, this configuration is not limited thereto, and a window and a fume diffusion unit that allow the laser beams to pass through may be provided for each laser irradiation mechanism.

[0103] Similar to the above-described embodiment, the image acquisition unit 110 photographs the build region R and acquires images of spatter generated around the molten pool formed by the irradiation of the laser beam. However, in this embodiment, the image acquired by the image acquisition unit 110 includes two images of the molten pool, since the material layer 10 is irradiated with two beams, laser beam B1 and laser beam B2. In other words, the image acquisition unit 110 acquires images of spatter generated around each of the molten pools formed by the irradiation of the two laser beams.

[0104] The analysis unit 210 of the external calculation device 200 extracts feature values ​​from the spatter image and calculates coordinate values, similar to the above-described embodiment. However, in this case, the analysis unit 210 extracts feature values ​​and calculates coordinate values ​​for each of the multiple molten pools formed on the material layer 10 from the image acquired by the image acquisition unit 110.

[0105] Figure 15 is a schematic diagram of an example of an image containing images of multiple molten pools. Figure 15 is an example of an image of spatter S obtained when four laser beams are simultaneously irradiated onto a material layer 10. In the example shown in Figure 15, four molten pools, namely, molten pool P1, molten pool P2, molten pool P3, and molten pool P4, are formed at different locations on the material layer 10, and spatter occurs in each of these molten pools.

[0106] In this example, when the image acquired by the image acquisition unit 110 includes images of multiple weld pools, the analysis unit 210, upon receiving the image data from the image acquisition unit 110, first performs the image analysis step S5 (see FIG. 11 ) by extracting the image into four regions, each containing a weld pool P1, P2, P3, and P4. For example, the analysis unit 210 extracts four regions, R1, R2, R3, and R4, from the image acquired by the image acquisition unit 110, as shown by the dashed rectangles in FIG. 15 . Region R1 includes an image of weld pool P1 and an image of surrounding spatter Sp1. Similarly, region R2 includes an image of weld pool P2 and an image of surrounding spatter Sp2. Region R3 includes an image of weld pool P3 and an image of surrounding spatter Sp3. Region R4 includes an image of weld pool P4 and an image of surrounding spatter Sp4.

[0107] The method for extracting multiple regions, each containing an image of a molten pool, from the image acquired by the image acquisition unit 110 is not limited to a specific method, and any appropriate method that can achieve the purpose may be used. For example, a region containing images of multiple particles that make up the spatter may be detected, and a high-intensity, approximately circular portion located near the geometric center of a figure that defines the shape of that region may be identified as the molten pool. Alternatively, multiple regions, each containing an image of a molten pool, may be extracted using machine learning.

[0108] In the feature extraction step S51 (see FIG. 12), the analysis unit 210 treats each of the cut-out regions as an image of spatter generated around the molten pool and extracts feature values ​​from each of regions R1 to R4. In the subsequent local parameter estimation step S52, the analysis unit 210 inputs the feature values ​​for each image corresponding to the cut-out region into the trained model LM to obtain estimates of the local parameters for each image. In the coordinate value calculation step S53, the analysis unit 210 calculates coordinate values ​​indicating the position of the molten pool for each image corresponding to the cut-out region. In the subsequent storage step S54, four sets of estimates and coordinate values ​​for each image corresponding to the cut-out region are stored in memory 14.

[0109] As shown in FIG. 15 , for example, when irradiating the material layer 10 with four laser beams, four independent layered objects can be manufactured in parallel. By extracting feature quantities from the images of sputters Sp1 to Sp4 and estimating local parameters, three-dimensional images such as those illustrated in FIGS. 9 and 10 can be obtained for each layered object. In other words, this embodiment makes it easy to understand the quality of each layered object. Of course, a single layered object can also be manufactured using multiple laser beams. In this case, the lead time required for manufacturing the layered object can be shortened, and the collection of local parameters for generating three-dimensional images such as those illustrated in FIGS. 9 and 10 can be accelerated.

[0110] As described above, the method of calculating the coordinates of the molten pool position by obtaining a drive signal for steering the laser beam from a laser control unit tends to make the entire system complex and expensive. The greater the number of laser irradiation mechanisms, the more complex the entire system becomes. It is also necessary to know in advance the number of laser irradiation mechanisms emitting laser beams when acquiring an image. In contrast, the method using image analysis, as in this embodiment, allows for numerical calculation of coordinate values ​​easily and inexpensively.

[0111] (3.2 Use of trained models via networks) Fig. 16 is a functional block diagram of a modeling state estimation system according to yet another embodiment of the present invention. Compared to the example described with reference to Fig. 8, in the example shown in Fig. 16, an external calculation device 201, instead of the external calculation device 200, is combined with the additive manufacturing apparatus 100, thereby constructing a modeling state estimation system. In the configuration illustrated in Fig. 16, the external calculation device 201 has an analysis unit 211, instead of the analysis unit 210. As schematically shown in Fig. 16, the analysis unit 211 of the external calculation device 201 does not have the above-mentioned learning unit 12.

[0112] Here, the learning unit 12, which includes a memory unit 12M in which the trained model LM is stored, is housed in a server 300 separate from the external computing device 201. The server 300 is configured to be able to communicate with the external computing device 201 via a network Wb such as the Internet.

[0113] The analysis unit 211 of the external computing device 201 extracts feature quantities related to the spatter S, and then sends the data related to the feature quantities to the server 300 via the network Wb. The server 300 inputs the data related to the feature quantities into the trained model LM based on, for example, an instruction from the analysis unit 211, and returns the estimated values ​​output from the trained model LM to the analysis unit 211 via the network Wb. The analysis unit 211 also calculates coordinate values ​​indicating the position of the molten pool P. The analysis unit 211 stores the estimated values ​​obtained from the trained model LM in correspondence with the coordinate values ​​of the molten pool P in, for example, the memory 14.

[0114] In this way, it is not necessary for the external computing device of the additive manufacturing state estimation system to have a trained model LM that returns the estimated values ​​of the local parameters as an output. The deployment of the trained model LM may be either a server-side model deployment as shown in FIG. 16 or a client-side model deployment as shown in FIG. 8. Alternatively, a hybrid model deployment may be adopted in which the trained model LM is deployed to both the external computing device of the additive manufacturing system and an external server.

[0115] <4. Other components of the printing state estimation system> Next, with reference to FIG. 8 again, other configurations of the printing state estimating system 1000 will be described in detail.

[0116] (4.1 CAM and CAD Equipment) 8 , the control unit 150 of the additive manufacturing apparatus 100 includes the numerical control unit 50. The numerical control unit 50 sends commands to controllers for operating each unit of the additive manufacturing apparatus 100, such as the gas system control unit 128, the recoater control unit 138, the laser control unit 148, and the table control unit 158, based on a machine-readable processing program that describes instructions for obtaining the shape of the additive manufacturing object. The processing program is prepared by a device separate from the additive manufacturing apparatus 100 and is stored in advance in the memory 52 of the numerical control unit 50 before the start of additive manufacturing.

[0117] Here, the above-mentioned processing program is prepared prior to additive manufacturing by a CAM device 500 installed outside the additive manufacturing apparatus 100, and is sent from the CAM device 500 to the control unit 150 of the additive manufacturing apparatus 100 via wired or wireless communication. The CAM device 500 imports a file containing data representing a three-dimensional model of the additive manufacturing object (hereinafter referred to as a "CAD model" for convenience), which has been prepared by the CAD device 400, and generates a processing program based on the CAD data.

[0118] The CAD device 400 and the CAM device 500 may be independent devices, or may be realized by a single computing device. For example, a machining program may be generated using a personal computer on which a CAD tool is installed in addition to CAM software. In such a configuration, the transfer of the CAD model from the CAD tool to the CAM software is completed within the computer.

[0119] 9 and 10, a CAD model may be used. For example, the estimated local parameters may be expressed in a form associated with the geometric shape of the additive manufacturing object by appropriately dividing the CAD model into meshes and assigning estimated values ​​of porosity to each divided region.

[0120] (4.2 Pre-trained model LM) In an embodiment of the present invention, model inference is performed using a machine-learning trained model LM that inputs sputter particle feature quantities and outputs local parameters such as porosity. The architecture of the trained model LM is not particularly limited. Here, a neural network model including an input layer to which sputter particle feature quantities are provided, an output layer that outputs local parameters, and one or more hidden layers (e.g., seven hidden layers) is applied to the trained model LM. Machine learning in an embodiment of the present invention is not limited to neural network-type techniques and may be performed based on various techniques that can learn using, for example, a large amount of training data (pairs of known input data and correct answer data). Note that the edges and nodes shown in FIG. 8 are merely examples for convenience of explanation and do not limit the actual architecture of the trained model LM.

[0121] For example, supervised learning can be applied to training a neural network model. Training data can be prepared by conducting preliminary test manufacturing, as described in Patent Document 1. More specifically, additive manufacturing is performed while acquiring images of spatters under different laser beam irradiation conditions. By observing the cross section of the additively manufactured object obtained in this process, a dataset can be obtained, consisting of the feature quantities of spatter particles extracted from the spatter images, the laser beam irradiation conditions, and the corresponding porosity measurements. The dataset obtained in this way can be used as training data to train a neural network model.

[0122] As described above, the heat that the material powder receives from the laser depends on the shape of the additive manufacturing object to be obtained. That is, even if the laser irradiation conditions are the same, the manner in which spatter scatters will differ between, for example, the center and the edges of the three-dimensional shape of the additive manufacturing object. Therefore, training data may be devised to enable highly accurate estimation according to the three-dimensional shape of the additive manufacturing object.

[0123] The memory unit 12M that stores the trained model LM, the memory 14 of the analysis unit 210, and the memory 52 of the above-mentioned numerical control unit 50 can use non-volatile storage devices such as RAM, as well as non-volatile storage devices such as magnetic disk drives and solid-state drives, depending on the purpose. The image generation unit 16, the calculation unit 12C of the learning unit 12, and the numerical control unit 50 of the additive manufacturing apparatus 100 can use processors such as CPUs or GPUs depending on the purpose.

[0124] (4.3 Other) The configuration of the modeling state estimation system is not limited to the examples shown in Fig. 8 and Fig. 16. The modeling state estimation system may be configured by combining either the additive manufacturing apparatus 100 or the additive manufacturing apparatus 101 with either the external computing device 200 or the external computing device 201.

[0125] Note that the image acquisition unit 110 of the additive manufacturing apparatus 100 and the additive manufacturing apparatus 101 is not essential for the embodiments of the present invention. The image acquisition unit 110 may be part of the external computing device 200 or the external computing device 201. For example, by using a method for calculating the coordinate values ​​of the molten pool from a spatter image, the image analysis step S5 and the three-dimensional image output step S9 shown in FIG. 11 can be completed on the external computing device side. This means that the additive manufacturing state estimation method according to the present invention can be applied by subsequently installing the image acquisition unit 110 and the external computing device 200 or the external computing device 201 in an existing additive manufacturing apparatus. In this sense, the present invention is also useful.

[0126] Various embodiments of the present invention have been described above, but these are presented as examples and are not intended to limit the scope of the invention. The novel embodiments may be embodied in various other forms, and various omissions, substitutions, and modifications may be made without departing from the spirit of the invention. Such embodiments and modifications are included within the scope and spirit of the invention, and are also included in the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0127] 2: Pore 2c, 2e: Space 5: Modeling table 5a:Top surface 6: Base plate 7: Actuator 10: Material layer 12: Learning Department 12C: Arithmetic section 12M: Storage section 14: Memory 16: Image generation unit 20c:Inlet 20d: Exhaust port 22: Window 22b: Bottom surface 24: Fume diffusion section 24D: Diffusion element 24H: Cabinet 24d:Aperture 24v: Space 32: Drive mechanism 32A: Actuator 32L: Guide rail 34: Retaining wall 35, 37: Blade 36: Main body 36R: Reservoir 36S: Slit 44: Collimator 44L: Collimator lens 46: Focus control unit 46L: Movable lens 46M: Condenser lens 48: Galvanometer 48A: 1st mirror 48B: 2nd mirror 50: Numerical control unit 52: Memory 54: Judgment section 100, 101: Additive manufacturing device 110: Image acquisition unit 112: Temperature sensor 120: Chamber 120v: sculptural space 128: Gas system control unit 130: Material layer formation mechanism 132: Bass 136: Recoater head 136b: bottom 136f:Front 136r: Rear 138: Recoater control unit 140-142: Laser irradiation mechanism 143: Laser oscillator 144: Galvano unit 148: Laser control unit 150: Control unit 158: Table control unit 200, 201: External calculation device 210, 211: Analysis Department 220: Display section 300: Server 400:CAD equipment 500:CAM device 1000, 1001: Modeling state estimation system B, B1, B2: Laser beams LM: Pre-trained model P, P1 to P4: molten pool R: Printing area R1~R4: Area S, Sp1 to Sp4: Sputter Vx: Voxel Wb:Network

Claims

1. A system for estimating a modeling state of a layered object, comprising: An image acquisition unit and an analysis unit are provided, the layered object is manufactured by repeating a material layer forming step of supplying material powder onto a modeling region to form a material layer, and a solidified layer forming step of irradiating the material layer with one or more laser beams to form a solidified layer; the image acquisition unit acquires in real time images of spatters generated around each molten pool formed by irradiation with the one or more laser beams; The analysis unit extracts features related to the spatter from the image and calculates coordinate values ​​indicating the position of the molten pool, and by inputting the features into a trained model, estimates local parameters representing the molding state of the solidified layer and outputs the local parameters in association with the coordinate values.

2. 10. The system of claim 1, The system, wherein the local parameters are data representing the porosity of each part of the additive manufacturing object, obtained for each of the coordinate values.

3. 10. The system of claim 1, The analysis unit generates a three-dimensional image of the modeling state of the additively manufactured object, the image being represented by point cloud data, which is a collection of sets made up of the coordinate values ​​and the local parameters.

4. 4. The system according to claim 1, further comprising: Further comprising a control unit, the control unit controls an operation of a recoater head that supplies the material powder onto the build region; the coordinate values ​​include, in addition to plane coordinate values ​​within the printing area, Z coordinate values ​​in the height direction of the layered object, The analysis unit determines the cumulative number of layers of the solidified layer by detecting an image including the recoater head from the series of images acquired by the image acquisition unit, and calculates a Z coordinate value.

5. 4. The system according to claim 1, further comprising: Further comprising a control unit, the coordinate values ​​include, in addition to plane coordinate values ​​within the printing area, Z coordinate values ​​in the height direction of the layered object, The analysis unit determines the cumulative number of layers of the solidified layers by analyzing a log of commands from the control unit and calculates a Z coordinate value.

6. 5. The system of claim 4, The control unit changes the irradiation conditions of the laser beam in accordance with the estimated value of the local parameter.

7. 6. The system of claim 5, The control unit changes the irradiation conditions of the laser beam in accordance with the estimated value of the local parameter.

8. 4. The system according to claim 1, further comprising: the one or more laser beams include a plurality of laser beams; The analysis unit extracts feature values ​​for each molten pool from the image and calculates coordinate values.

9. A method for estimating a modeling state of a layered object, comprising: An image acquisition step and an analysis step are provided, the layered object is manufactured by repeating a material layer forming step of supplying material powder onto a modeling region to form a material layer, and a solidified layer forming step of irradiating the material layer with one or more laser beams to form a solidified layer; In the image acquisition step, images of spatter generated around each molten pool formed by irradiation with the one or more laser beams are acquired in real time; In the analysis process, local parameters representing the molding state of the solidified layer are estimated by inputting feature quantities related to the sputtering into a trained model, and the local parameters are output in a form associated with coordinate values ​​indicating the position of the molten pool.

10. 10. The method of claim 9, The analyzing step extracting the feature amount from the image; calculating the coordinate values ​​from the image; A method comprising:

Images