JP7404833B2 - quality prediction system - Google Patents

Description

The present invention relates to a quality prediction system.

Patent Document 1 discloses a laser processing system including a machine learning device that learns laser processing condition data. The machine learning device learns laser processing condition data in association with state quantities and processing results of the laser processing system. Then, the machine learning device is configured to output laser processing condition data that provides an optimal processing result by referring to the learned laser processing condition data.

Japanese Patent Application Publication No. 2017-164801

In recent years, additive manufacturing has been carried out in which a shaped object is manufactured by irradiating material powder with a light beam (laser beam, electron beam, etc.). It is known that additive manufacturing includes, for example, a powder bed fusion method, a directed energy deposition method, and the like. Powder bed fusion bonding performs additive manufacturing by irradiating a flat layer of powder with a light beam. Powder bed melting methods include SLM (Selective Laser Melting), EBM (Electron Beam Melting), and the like. The directional energy deposition method performs additive manufacturing by controlling the position of a head that irradiates a light beam and discharges material powder. Directional energy deposition methods include LMD (Laser Metal Deposition), DMP (Direct Metal Printing), and the like. Even in such additive manufacturing, it is important to set the conditions for irradiating the powder with the light beam, as in the conventional laser processing system described above, in order to maintain the quality of the shaped product.

By the way, among the additive manufacturing described above, especially in the powder bed fusion bonding method, a light beam scans the material powder in the horizontal direction to melt and solidify the material powder to form (manufacture) the object. )do. In this case, the shrinkage that occurs when the material powder is melted and solidified affects the quality of the molded object, that is, the shape accuracy and the presence or absence of cracks. Here, the magnitude of the shrinkage that occurs when the material powder is melted and solidified depends, for example, on the volume of the material powder that is melted and solidified, and the larger the volume, the larger the shrinkage tends to be. In particular, when material powder is spread in layers with a constant layer thickness, the volume to be melted and solidified is determined by the area scanned by the light beam in the horizontal direction.

For this reason, although it is possible to meet the quality of the model by reducing the volume to be melted and solidified, in other words by reducing the area scanned by the light beam, the inclination of the model with respect to the base during printing (manufacturing) may be reduced. It is necessary to appropriately set the modeling posture to be expressed. However, this modeling posture needs to be set depending on the type of powder and the shape of the object, which requires experience and time for the operator and is extremely complicated.

SUMMARY OF THE INVENTION An object of the present invention is to provide a quality prediction system that is applied to additive manufacturing and can easily set a molding posture for ensuring the quality of a molded object.

The quality prediction system is applied to additive manufacturing equipment that manufactures objects by irradiating the material powder with a light beam and melting and solidifying the material powder. a learned model storage unit that stores a learned model generated by performing machine learning using as a training data set printing data including data and printing quality data representing the quality of the manufactured object; a quality prediction unit that predicts the quality of each model when changing the model orientation data as a parameter using the model and data during printing used when manufacturing a new model ; The posture data includes a first angle around the reference axis of the modeled object, a second angle around the first axis perpendicular to the reference axis, and a third angle around the second axis perpendicular to the reference axis and the first axis. It is expressed using

According to this, a trained model generated by performing machine learning using printing data including printing posture data and printing quality data as a training dataset, and printing data used when manufacturing a new object. The quality of the object can be predicted based on the In this case, by changing the modeling orientation data included in the modeling data as a parameter, it is possible to predict a plurality of qualities of the object depending on the modeling orientation. Here, the generated learning model is a model that defines at least the relationship between the shaping posture and the quality of the shaped object during additive manufacturing.

Thereby, it is possible to very easily set the modeling posture when producing a shaped object that satisfies the standard quality in a short time. Furthermore, by setting the printing posture using the trained model, it is possible to set the printing posture objectively, without being based on the operator's experience, for example.

FIG. 1 is a diagram showing the configuration of a quality prediction system. FIG. 1 is a diagram showing an additive manufacturing device. It is an enlarged front view showing a base, a modeled object, and a support member, and for explaining a modeling posture. It is an enlarged front view showing a base, a modeled object, and a support member, and for explaining a modeling posture. It is a figure which shows the example of a support member. It is a figure which shows the example of a support member. FIG. 2 is a block diagram showing the configuration of a machine learning section. FIG. 3 is a diagram for explaining modeling posture data. FIG. 3 is a diagram for explaining a learning phase of a first example. FIG. 3 is a diagram for explaining the inference phase of the first example. FIG. 7 is a diagram for explaining a learning phase of a second example. FIG. 7 is a diagram for explaining the inference phase of the second example. FIG. 7 is a diagram for explaining the first learning phase in the third example. FIG. 7 is a diagram for explaining a first-stage inference phase according to a third example. FIG. 7 is a diagram for explaining a second learning phase according to a third example. FIG. 7 is a diagram for explaining a second inference phase in a third example.

(1. Application target of quality prediction system)
The quality prediction system is applied to an additive manufacturing method for additively manufacturing 3D objects by employing a powder bed fusion bonding method or a directed energy deposition method. In this example, an example of an additive manufacturing apparatus using an SLM of the powder bed fusion bonding method will be described, but the additive manufacturing apparatus may also use an EMB of the powder bed fusion bonding method. Regarding the material powder, in this example, metal powder is exemplified, but resin powder may also be used.

(2. Configuration of quality prediction system 100)
The quality prediction system 100 includes one or more additive manufacturing devices 1, a 3D modeling model generation device 2, an inspection device 3, and a machine learning device 110, which will be described in detail later.

The 3D printing model generation device 2 generates a 3D printing model based on the required specifications so that the additive manufacturing device 1 can manufacture the object. The 3D printing model generation device 2 applies, for example, CAE (Computer Aided Engineering) to generate an optimal 3D printing model. In this example, the information included in the 3D printing model includes a 3D design shape (3D model data), a printing posture in the 3D design shape (printing posture data), and the like.

The inspection device 3 is a device for evaluating the modeling quality of the object manufactured by the additive manufacturing device 1. The inspection device 3 inspects the shape error (i.e., shape accuracy) of the manufactured object with respect to the reference shape of the object manufactured by the additive manufacturing device 1, the presence or absence of cracks in the object, and the like. The inspection device 3 outputs build quality data representing shape errors and build quality.

The machine learning device 110 performs machine learning using at least the modeling data when the additive manufacturing device 1 models (manufactures) a modeled object and the modeling quality data of the modeled (manufactured) object as a training data set. Here, the data at the time of modeling includes 3D model data that represents the three-dimensional shape of the object to be modeled and includes modeling posture data, and modeling condition data that represents the modeling conditions at the time of modeling. Further, the modeling condition data includes at least one of laser output, scanning speed, scanning pitch, irradiation spot diameter, lamination thickness, and type of metal powder.

As a result, the machine learning device 110 generates a learned model regarding the posture of the object at the time of manufacture and the quality of the manufactured object (shape accuracy of the object, presence or absence of cracks in the object, etc.). Then, the machine learning device 110 determines the quality of the newly additively manufactured object (shape accuracy, presence of cracks in the object, etc.) to predict. Here, the machine learning device 110 changes the modeling posture data included in the printing data as a parameter, and predicts the printing quality of the object corresponding to each of the changed parameters, that is, the plurality of printing postures.

Then, the machine learning device 110 evaluates a printing posture corresponding to a printing quality that satisfies a preset standard quality for the plurality of predicted printing qualities. Thereby, the additive manufacturing apparatus 1 manufactures a molded object that satisfies the standard quality by molding (manufacturing) the molded object using the evaluated molding posture.

Machine learning device 110 includes a first server 111 and a second server 112. However, although the first server 111 and the second server 112 will be described as separate devices, they can also be configured by the same device. Moreover, regarding the first server 111 and the second server 112, it is also possible to provide at least the first server 111 on the cloud.

The first server 111 functions as a learning phase in machine learning. The first server 111 generates a learned model by machine learning using the acquired training data set. The first server 111 is provided to be able to communicate with the plurality of additive manufacturing apparatuses 1, and acquires printing data used when each of the plurality of additive manufacturing apparatuses 1 forms a modeled object as part of a training data set. do. The data during printing includes, for example, 3D model data, printing condition data (at least one of the following: light beam output, scanning speed, irradiation spot diameter, lamination thickness, support type, and material powder type), and the reference axis direction of the printed object. , for example, includes modeling orientation data that represents the inclination with respect to the horizontal direction.

The first server 111 further includes a shape error (an error in the manufactured shape with respect to the 3D model data of the standard shape) of the model formed by each of the plurality of additive manufacturing apparatuses 1, and a model that indicates the presence or absence of cracks in the model. Obtain quality data as teacher data in the training dataset. Then, the first server 111 performs supervised learning to generate a learned model regarding the modeling state (specifically, the printing posture) of the object and the quality of the object. Note that although the machine learning in the first server 111 will be described using supervised learning as an example, it is also possible to apply other machine learning algorithms.

The first server 111 may acquire the modeling quality data measured by the inspection device 3 by an operator inputting it. Further, the first server 111 may directly acquire the modeling quality data measured by the inspection device 3 from the inspection device 3. Note that the modeling quality data is data linked to the corresponding modeled object.

In this way, in the quality prediction system 100, the first server 111 acquires the data at the time of modeling and the modeling quality data when each of the plurality of additive manufacturing apparatuses 1 models the object. Thereby, the first server 111 generates a learned model by machine learning using a large amount of printing data and printing quality data as a training data set. Therefore, the first server 111 can improve the learning accuracy of the trained model, and can improve the accuracy of the trained model.

The second server 112 functions as an inference phase in machine learning. The second server 112 acquires the trained model generated by the first server 111. Further, the second server 112 is provided to be able to communicate with each of the plurality of additive manufacturing apparatuses 1 . Then, the second server 112 uses the trained model generated by the first server 111, and uses as input data the data at the time of printing when each of the plurality of additive manufacturing devices 1 newly prints a model, Predict the quality of a newly created object. Furthermore, the second server 112 predicts the printing quality corresponding to each of the plurality of printing postures by changing the printing posture data included in the printing data as a parameter, and predicts the printing quality corresponding to the printing quality that satisfies the standard quality ( Printing posture data) is output as output data.

The printing quality of the object predicted by the second server 112 and the evaluated printing posture (printing posture data) can be sent to the additive manufacturing device 1 and used to adjust the printing conditions of the additive manufacturing device 1. . Further, the printing quality of the object predicted by the second server 112 and the evaluated printing posture (printing posture data) can also be transmitted to the 3D printing model generation device 2. As a result, if it is determined that the predicted printing quality of the printed object is poor, the 3D printing model generation device 2 changes the shape of the printed object according to the predicted printing quality, that is, the 3D model It is possible to perform processes such as modifying data and modeling posture data, and changing the type of support member that supports the object during additive manufacturing.

Here, although not shown, a quality prediction device that performs the same processing as the second server 112 may be arranged for each of the plurality of additive manufacturing devices 1. That is, the quality prediction device, like the second server 112, executes the inference phase in machine learning. Then, the quality prediction device determines the quality of the modeling of the object formed by the corresponding additive manufacturing device 1 based on the modeling data in the corresponding additive manufacturing device 1 and the learned model generated by the first server 111. etc. to predict.

Furthermore, the quality prediction system 100 may be configured by a single additive manufacturing device 1 and a machine learning device. The machine learning device is capable of executing a learning phase of machine learning corresponding to the first server 111, and is also capable of executing an inference phase of machine learning corresponding to the second server 112 or the quality prediction device.

(3. Configuration of additive manufacturing device 1)
The configuration of the additive manufacturing apparatus 1 will be explained with reference to the drawings. The additive manufacturing apparatus 1 is an apparatus that uses the SLM method and shapes (manufactures) a shaped object W by repeatedly irradiating a metal powder P, which is a material powder arranged in layers, with a light beam.

Here, the light beam includes various beams capable of melting the metal powder P, and can be exemplified by, for example, a laser beam. Various lasers can be used as the laser beam, such as a near-infrared wavelength laser, a CO ₂ laser (far-infrared laser), and a semiconductor laser, and is appropriately determined depending on the target metal powder P. Further, various metal materials can be used as the metal powder P, such as aluminum, copper, steel materials such as maraging steel and Inconel, stainless steel, and alloy tool steel.

As shown in FIG. 2, the additive manufacturing device 1 includes a chamber 10, a shaped object support device 20, a powder supply device 30, and a light beam irradiation device 40. The chamber 10 is configured such that the air inside can be replaced with an inert gas such as He (helium), N ₂ (nitrogen), Ar (argon), or the like. Note that the chamber 10 may have a configuration in which the pressure inside the chamber 10 can be reduced instead of replacing the interior with an inert gas.

The object support device 20 is provided inside the chamber 10 and is a part for modeling the object W. The object support device 20 includes a object container 21, a lifting table 22, and a base 23. The modeling container 21 has an opening on the upper side and an inner wall surface parallel to the vertical axis. The elevating table 22 is provided inside the modeling container 21 so as to be movable in the vertical direction along the inner wall surface. The base 23 is removably attached to the upper surface of the lifting table 22, and the upper surface of the base 23 becomes a part for modeling the object W.

The powder supply device 30 is provided inside the chamber 10 and adjacent to the shaped object support device 20 . The powder supply device 30 includes a powder storage container 31, a supply table 32, and a recoater 33. The powder storage container 31 has an opening on the upper side, and the height of the opening of the powder storage container 31 is set to be the same as the height of the opening of the modeling container 21. The powder storage container 31 has an inner wall surface parallel to the vertical axis. The supply table 32 is provided inside the powder storage container 31 so as to be movable in the vertical direction along the inner wall surface. In the powder storage container 31, the metal powder P is stored in the upper region of the supply table 32.

The recoater 33 is provided so as to be able to reciprocate over the entire area of the opening of the modeling container 21 and the opening of the powder storage container 31 along the upper surfaces of both openings. When the recoater 33 moves from right to left in FIG. 2, it transports the metal powder P protruding from the opening of the powder storage container 31 toward the modeling container 21 side. Further, the recoater 33 arranges the transported metal powder P in a layered manner and horizontally on the upper surface of the base 23.

The light beam irradiation device 40 irradiates the surface of the metal powder P horizontally arranged in layers on the upper surface of the base 23 by scanning the light beam in the horizontal direction (reference direction). The light beam irradiation device 40 changes the irradiation position and irradiation intensity based on 3D model data representing the shape of the object W by executing a preset program. By changing the irradiation position, a desired layered object W based on 3D model data can be formed. Furthermore, by changing the irradiation intensity, the bonding strength between the metal powders P can be changed.

(4. Method for manufacturing a shaped object W using the additive manufacturing device 1)
A method for manufacturing a shaped object W using the additive manufacturing apparatus 1 will be described with reference to FIG. 2. With the supply table 32 of the powder supply device 30 positioned downward, the metal powder P is stored in the powder storage container 31. The supply table 32 is raised so that a desired amount of metal powder P protrudes from the opening of the powder storage container 31. At the same time, in the object support device 20, the base 23 is attached to the upper surface of the elevating table 22, and the elevating table 22 is positioned so that the upper surface of the base 23 is located slightly below the opening of the modeling container 21. .

Subsequently, the recoater 33 is moved from the powder supply device 30 side toward the shaped object support device 20 side. As a result, the metal powder P in the powder supply device 30 moves to the upper surface of the base 23 and is arranged in a layer having the same thickness on the upper surface of the base 23.

Subsequently, the light beam irradiation device 40 irradiates the metal powder P arranged in layers on the upper surface of the base 23 with the light beam while scanning the metal powder P in the horizontal direction. The metal powder P at the irradiated position is melted and solidified to be integrated. By repeating the above operations, the support member S is formed on the upper surface of the base 23.

Subsequently, the metal powder P arranged in a layer on the upper surface of the support member S is irradiated with a light beam by the same operation as described above. By repeating this operation, the object W is formed (manufactured) on the upper surface of the support member S.

The shaped object W is joined to the upper surface of the support member S, and the support member S is joined to the upper surface of the base 23. Therefore, after the object W is formed, the base 23 with the object W and the support member S joined is removed from the lifting table 22. Then, the shaped article W and the support member S are separated from the base 23, and then the support member S is separated and removed from the shaped article W. In this way, the shaped object W is completed.

(5. About the shaped object W by additive manufacturing)
As described above, the shaped object W by additive manufacturing is formed by melting the metal powder P by being irradiated with a light beam, and then solidifying the molten metal powder P. By the way, when the metal powder P solidifies after being melted, that is, when heat is dissipated, the shaped object W contracts slightly. Therefore, a deforming force is generated in the shaped object W due to contraction. When such deformation force becomes large, distortion becomes large and the shape accuracy of the shaped object W deteriorates or cracks occur. Here, the contraction or deformation force of the shaped object W changes depending on the volume in which the metal powder P melts and solidifies, and the larger the volume in which the metal powder P melts and solidifies, the larger the contraction (deformation force) becomes.

Therefore, in order to reduce the volume to be melted and solidified, in other words, to reduce the area scanned by the light beam, in additive manufacturing, as shown in FIGS. 2 and 3, A modeling posture representing the inclination of the object W is set. When setting this modeling posture, it is necessary to take into consideration the volume to be melted and solidified, that is, the area to be scanned by the light beam, which changes depending on the shape of the object W, for example.

When the printing posture is appropriately set, as shown in FIG. The shape accuracy is improved and the occurrence of cracks is prevented. On the other hand, if the printing posture is inappropriately set, as shown in FIG. 4, the deformation force generated increases despite the setting of the printing posture, and for example, the support member S and the printed object W are separated. It deforms to such an extent that the shape accuracy deteriorates. For this reason, there is a risk that cracks may occur in the shaped object W, and the modeling posture needs to be appropriately set for each shaped object W.

(6. Regarding support member S)
The support member S is used on the base 23 to support the shaped article W while ensuring the shape accuracy of the shaped article W and preventing the occurrence of cracks. The support member S has multiple variations (types). The support member S can be provided in a columnar shape, for example, as shown in FIGS. 3 and 4. Further, the support member S can be provided in a mesh shape, as shown in an example in FIG. Furthermore, the support members S can be provided with different thicknesses, as shown in FIG. Note that for other shapes of the support member S, for example, the shape disclosed in detail in Japanese Unexamined Patent Publication No. 2019-81923 can be adopted.

(7. Configuration of machine learning device 110)
Next, the configuration of the machine learning device 110 (see FIG. 1) will be described with reference to FIG. 7. As shown in FIG. 7, the machine learning device 110 includes a learning processing device 210 capable of executing a learning phase, a quality prediction device 220 capable of executing an inference phase, and a database 230. Here, the learning processing device 210 corresponds to the first server 111 in the quality prediction system 100 described above. Furthermore, the quality prediction device 220 corresponds to the second server 112 in the quality prediction system 100 described above. Moreover, the database 230 stores and accumulates updatable data and modeling quality data obtained each time the additive manufacturing device 1 models (manufactures) a modeled object W.

The learning processing device 210 includes a learning data input section 211 , a training data set acquisition section 212 , a training data set storage section 213 , and a learned model generation section 214 . The learning data input unit 211 inputs modeling data linked to the corresponding shaped object W. The modeling data includes, for example, 3D model data, modeling condition data (at least one of light beam output, scanning speed, irradiation spot diameter, lamination thickness, type of support, and type of metal powder P), modeling posture data described below, and Support type data can be listed.

The training data set acquisition unit 212 acquires a training data set that uses the modeling data acquired by the learning data input unit 211 as an explanatory variable and the modeling quality data input from the inspection device 3 as an objective variable. The modeling quality data inputted from the inspection device 3 includes the shape accuracy of the object W and the presence or absence of cracks in the object W. The acquired training data set is stored in the training data set storage unit 213. Further, the training data set acquisition unit 212 can also acquire the modeling data and the modeling quality data stored in the database 230 as a training data set.

The trained model generation unit 214 performs machine learning based on the printing data and printing quality data stored in the training data set storage unit 213, using the linked printing data and printing quality data as a training data set. . As a result, the learned model generation unit 214 generates a learned model regarding at least the modeling posture (or/and support member S) of the object W during additive manufacturing and the modeling quality (shape accuracy and presence or absence of cracks). .

The quality prediction device 220 mainly includes a learned model storage section 221 , a modeling data acquisition section 222 , a quality prediction section 223 , an evaluation section 224 , and an output section 225 . The trained model storage unit 221 stores the trained model generated by the trained model generation unit 214. The data acquisition unit 222 acquires the data used when the additive manufacturing apparatus 1 newly shapes (manufactures) the object W.

The quality prediction unit 223 determines the shape accuracy and cracks of the newly modeled object W based on the modeling data acquired by the modeling data acquisition unit 222 and the learned model stored in the learned model storage unit 221. Predict the quality of printing, such as the presence or absence of The quality prediction unit 223 changes the modeling orientation data (or/and support type data) included in the modeling data as a parameter, and generates the changed modeling orientation data, i.e., the modeling orientation (or/and support type data, i.e., the support member S). The modeling quality of the object W is predicted according to each of the above.

Here, the modeling posture data includes a first angle around the reference axis when manufacturing the object W, a second angle around the first axis perpendicular to the reference axis, and a second angle around the first axis perpendicular to the reference axis and the first axis. It is a parameter (variable) expressed using one of the third angles around the two axes. In this example, as shown in FIG. 8, the reference axis for manufacturing the shaped object W is, for example, the X-axis, and the first angle is the X-axis rotation angle around the X-axis. Further, the first axis is a Y-axis perpendicular to the X-axis, and the second angle is a Y-axis rotation angle around the Y-axis. Furthermore, the first axis is the Z-axis perpendicular to the X-axis and the Y-axis, and the third angle is the Z-axis rotation angle around the Z-axis.

The evaluation unit 224 evaluates, for the modeling quality of each shaped object W predicted by the quality prediction unit 223, a printing posture corresponding to a printing quality that satisfies a preset standard quality. Then, the evaluation unit 224 outputs the evaluated modeling posture, that is, modeling posture data to the output unit 225.

The output unit 225 outputs the prediction result by the quality prediction unit 223 and/or the evaluation result by the evaluation unit 224 to the additive manufacturing device 1 and the 3D modeling model generation device 2. The output unit 225 outputs modeling data (including modeling posture data) to the additive manufacturing apparatus 1, for example, in accordance with the evaluation result by the evaluation unit 224. Thereby, the additive manufacturing apparatus 1 can model (manufacture) a shaped object W that satisfies the standard quality.

In addition, the output unit 225 outputs the prediction result of the quality prediction unit 223 to provide, for example, guidance by display on a display device (not shown), guidance by voice, guidance by indicator lights, etc. In this case, the output unit 225 may provide guidance to a display device, etc. provided in the quality prediction device 220, or may provide guidance to a display device, etc. provided in each of the plurality of additive manufacturing devices 1. You can also do it.

Further, the output unit 225 can also output prediction results and evaluation results to the 3D modeling model generation device 2. Thereby, the 3D modeling model generation device 2 can modify the 3D model data based on the prediction result, for example, to improve shape accuracy or to prevent cracks from occurring.

Furthermore, when the quality prediction unit 223 performs a quality determination, the output unit 225 outputs the quality prediction result obtained from the quality prediction unit 223 to the additive manufacturing apparatus 1 and performs the quality determination for the additive manufacturing apparatus 1. It is also possible to execute processing according to the prediction result. For example, if it is determined (predicted) to be defective in the predicted result of the pass/fail determination of the shape accuracy included in the modeling quality of the object W, the output unit 225 sends the additive manufacturing device 1 to the additive manufacturing apparatus 1 as being defective. It is possible to perform a disposal process or a sorting process for the shaped object W determined to be present.

In this way, in the learning processing device 210, the learned model generation unit 214 performs at least the printing posture and printing quality ( A trained model regarding shape accuracy and presence/absence of cracks is generated. Further, in the quality prediction device 220, the trained model storage unit 221 stores the trained model generated by the trained model generation unit 214.

In the quality prediction device 220, the quality prediction unit 223 extracts at least 3D model data and posture data (or/and support type data) from among the data during printing obtained when printing (manufacturing) a new object W. , and the learned model stored in the learned model storage unit 221, predicts the modeling quality of the (scheduled) object W to be newly formed (manufactured). Therefore, the machine learning device 110 can predict the modeling quality of the object W, that is, the shape accuracy and the presence or absence of cracks, with high accuracy. Hereinafter, a method for predicting modeling quality using the machine learning device 110 will be described using a specific example.

(8. Operation of machine learning device 110)
(8-1. First example)
A first example of the machine learning device 110 will be described using FIGS. 9A and 9B. The machine learning device 110 of the first example determines the printing posture when printing (manufacturing) the object W based on the modeling quality of the object W, that is, the shape accuracy and the presence or absence of cracks, predicted using the learned model. decide.

Therefore, in the first example, the learned model generation unit 214 of the learning processing device 210 generates printing data including 3D model data, printing condition data, and printing posture data as explanatory variables, as shown in FIG. 9A. , modeling quality data as objective variables, and machine learning is performed to generate a trained model.

Subsequently, in the quality prediction device 220, as shown in FIG. 9B, the data-at-printing acquisition unit 222 receives, as input data, 3D model data that is data at the time of printing regarding the object W to be newly modeled (manufactured) and modeling data. In addition to acquiring condition data, modeling posture data is acquired as a parameter. Then, the quality prediction unit 223 uses the 3D model data, the modeling condition data, the modeling posture data, and the trained model stored in the trained model storage unit 221 to newly create (manufacture) as output data. The modeling quality of the object W is predicted.

Here, the modeling quality of the object W includes the shape accuracy of the object W, more specifically, the shape error and the presence or absence of cracks in the object W. Note that the shape error in the shaped object W can be an error at a predetermined measurement position.

The quality prediction unit 223 changes at least one of the X-axis rotation angle, Y-axis rotation angle, and Z-axis rotation angle that constitute the modeling posture data that is a parameter, and calculates the printing quality of the object W in each case. Predict. Then, the quality prediction unit 223 outputs the prediction result linked to the modeling posture data to the evaluation unit 224.

The evaluation unit 224 compares the plurality of prediction results output from the quality prediction unit 223 with a reference quality, which is a preset standard regarding modeling quality, to determine whether each prediction result satisfies the standard. evaluate. Among the evaluated prediction results, the evaluation unit 224 uses the modeling posture data when the prediction result predicts that the shape error of the newly formed (manufactured) object W is small and no cracks will occur, that is, Determine the X-axis rotation angle, Y-axis rotation angle, and Z-axis rotation angle. The evaluation unit 224 then outputs the determined modeling posture data to the output unit 225.

Note that when the evaluation unit 224 determines the modeling orientation data, in addition to the predicted modeling quality satisfying the standard quality, for example, if the cycle time when manufacturing the object W while satisfying the standard quality is shortened. The modeling posture data can be determined. That is, the evaluation unit 224 can determine the printing posture data based on evaluation factors other than the printing quality.

The output unit 225 outputs the modeling posture data determined by the evaluation unit 224 to the 3D modeling model generation device 2. Upon acquiring the modeling orientation data output from the output unit 225, the 3D modeling model generation device 2 generates 3D model data of the object W to be newly modeled (manufactured) by reflecting the acquired modeling orientation data. . That is, the 3D modeling model generation device 2 rotates the 3D model around any one of the X axis, the Y axis, and the Z axis according to the acquired modeling posture data. Thereby, the 3D modeling model generation device 2 generates 3D model data of the object W to be newly modeled (manufactured) by the additive manufacturing device 1. Note that, in addition to the shape of the object W, an appropriately selected support member S is added to the 3D model data newly generated by the 3D modeling model generation device 2.

When the final 3D model data is generated by the 3D modeling model generation device 2, the 3D model data is outputted via the output unit 225, for example. Then, as described above, the additive manufacturing apparatus 1 forms (manufactures) a new object W in accordance with the printing data including the output 3D model data. Then, the shaped (manufactured) object W is inspected by the inspection device 3.

Here, as shown in FIG. 7, the data at the time of printing when the additive manufacturing device 1 prints (manufactures) a new object W and the inspection results by the inspection device 3, that is, the printing quality data, are linked to each other. The information is stored and accumulated in the database 230 in an updatable manner. As a result, for example, the training dataset acquisition unit 212 acquires printing data and printing quality data from the database 230 via the learning data input unit 211, and the learned model generation unit 214 uses these data as a training dataset. Machine learning can be performed. Therefore, the trained model generation unit 214 can generate a highly accurate trained model.

In addition, in the quality prediction by the quality prediction unit 223, if a shape error is large or a crack occurs due to changing the modeling orientation data, the output unit 225 detects a part of the modeled object W with a large shape error or a crack. The generated region can be output to, for example, the 3D modeling model generation device 2 (see the broken line in FIG. 7).

Thereby, the 3D modeling model generation device 2 generates new 3D model data by correcting the shape of the 3D model so that the shape error is small and no cracks occur. Then, the 3D printing model generation device 2 outputs the generated 3D model data, that is, 3D model data for printing (manufacturing) a new object W to the data acquisition unit 222 during printing, so that the quality prediction unit 223 , the modeling quality of the newly modeled (manufactured) object W is predicted as described above.

(8-2. Second example)
In the first example described above, the evaluation unit 224 determines the modeling posture data when the modeling quality of the object W is predicted to satisfy the reference quality. By the way, in additive manufacturing, by appropriately selecting and setting the support member S that supports the object W during modeling from among a plurality of types, the quality of the object W can be more reliably maintained. . Therefore, in the second example, when the modeling quality of the object W predicted using the learned model satisfies the standard quality as in the first example, the evaluation unit 224 determines the printing posture data and Determine S.

In the second example, the learning data input unit 211 acquires support type data as explanatory variables in addition to 3D model data, modeling condition data, and modeling posture data, as shown in FIG. 10A. The support type data is data for identifying each of the plurality of types of support members S as described above. The learned model generation unit 214 then generates a learned model by performing machine learning using the training data set including the support type data.

Next, in the quality prediction device 220, as shown in FIG. 10B, the data-at-printing acquisition unit 222 receives 3D model data and modeling condition data as input data regarding the object W to be newly modeled (manufactured). At the same time, modeling posture data and support type data are obtained as parameters. Here, the support type data is a parameter (variable) that identifies each of the plurality of types of support members S. Note that the modeling posture data is the same as in the first example described above. Then, the quality prediction unit 223 uses the 3D model data, printing condition data, printing posture data, support type data, and the learned model stored in the learned model storage unit 221 to newly print ( Predict the quality of the object W to be manufactured (manufactured).

At this time, the quality prediction unit 223 changes the modeling posture data, which is a parameter, in the same manner as in the first example described above, and sequentially changes the support type data, and predicts the printing quality of the object W in each case. Then, the quality prediction unit 223 outputs the prediction result linked to the modeling posture data and the support type data to the evaluation unit 224.

The evaluation unit 224 evaluates whether each prediction result satisfies the standard by comparing the plurality of prediction results output from the quality prediction unit 223 with the reference quality. Among the evaluated prediction results, the evaluation unit 224 uses the modeling posture data and the support when the prediction result is that the shape error of the newly formed (manufactured) object W is predicted to be small and no cracks will occur. The type data, that is, which one of the plurality of types of support members S is determined. The evaluation unit 224 then outputs the determined modeling posture data and support type data to the output unit 225.

The output unit 225 outputs the modeling posture data and support type data determined by the evaluation unit 224 to the 3D modeling model generation device 2. The 3D modeling model generation device 2 generates 3D model data of the object W to be newly modeled (manufactured) by reflecting the acquired modeling posture data and support type data. That is, the 3D modeling model generation device 2 rotates the 3D model around any one of the X axis, the Y axis, and the Z axis according to the acquired modeling posture data. The 3D modeling model generation device 2 includes a support member S that optimally supports the rotated 3D model, that is, a support member S determined by the evaluation unit 224, and the additive manufacturing device 1 newly creates a model ( 3D model data of the object W to be manufactured (manufactured) is generated.

When the final 3D model data is generated by the 3D printing model generation device 2, as described above, the additive manufacturing device 1 prints (manufactures) a new object W according to the printing data including the generated 3D model data. )do. At this time, the additive manufacturing apparatus 1 shapes (manufactures) the determined support member S together with the molded object W. Then, the shaped (manufactured) object W is inspected by the inspection device 3.

Here, in this second example, as shown in FIG. 7, the additive manufacturing apparatus 1 includes data at the time of printing (manufacturing) a new object W, that is, printing posture data and support type data. The modeling data and the modeling quality data are linked to each other, stored in the database 230 in an updatable manner, and accumulated. As a result, for example, the training data set acquisition unit 212 acquires printing data (printing posture data and support type data) and printing quality data from the database 230 via the learning data input unit 211, and the trained model generation unit 214 can perform machine learning using these data as a training data set. Therefore, in the second example, the trained model generation unit 214 can generate a highly accurate trained model especially when the support member S is used.

(8-3. Third example)
In the first example described above, the printing posture of the object W (3D model) is determined, and in the second example described above, the posture of the object W (3D model) is determined and the support member S is determined. . In the third example, as shown in FIGS. 11A and 11B, the printing posture of the object W is determined as a first step, and then, as shown in FIGS. 11C and 11D, as a second step, support is determined according to the printing posture. Determine member S.

That is, as in the first example, the learned model generation unit 214 of the third example generates 3D model data, printing condition data, and printing posture data as explanatory variables as shown in FIG. 11A in the first step. Machine learning is performed using the modeling data included and the modeling quality data as the objective variable to generate a learned model. Then, as in the first example, the quality prediction unit 223 of the third example, as a first step, inputs 3D model data, printing condition data, and printing posture data (parameters) as input data, as shown in FIG. 11B. and the trained model stored in the trained model storage unit 221, the modeling quality of the newly created (manufactured) object W is predicted as output data. Thereby, the evaluation unit 224 determines the modeling orientation data corresponding to the modeling orientation that satisfies the standard quality. Note that hereinafter, the modeling orientation data determined by the evaluation unit 224 will be referred to as "determined modeling orientation data."

Subsequently, as in the second example, the learned model generation unit 214 of the third example generates 3D model data as explanatory variables, modeling condition data, and determined modeling as a second step, as shown in FIG. 11C. A trained model is generated by performing machine learning using modeling data including posture data and support type data, and modeling quality data as an objective variable. Then, as in the second example, the quality prediction unit 223 of the third example, as shown in FIG. Using the support type data (parameters) and the trained model stored in the trained model storage unit 221, the modeling quality of the newly created (manufactured) object W is predicted as output data. Thereby, the evaluation unit 224 finally determines the modeling posture and the support member S.

(9. Effect)
According to the quality prediction system 100, the machine learning device 110 generates a learned model by performing machine learning using printing data including printing posture data (or/and support type data) and printing quality data as a training data set. can do. Here, the generated learning model becomes a model that defines the relationship between the modeling posture of the object W (or/and the support member S) and the quality of the object W during additive manufacturing. The quality prediction system 100 can predict the modeling quality of the object W based on the generated trained model and the data at the time of manufacturing a new object W. In this case, the quality prediction system 100 changes the printing posture data (or/and support type data) included in the printing data as a parameter, thereby adjusting the printing object W according to the printing posture (or/and support member S). It is possible to predict multiple printing qualities.

Thereby, it is possible to very easily set the shaping posture when manufacturing the shaped object W that satisfies the standard quality in a short time. Furthermore, by setting the printing posture using the trained model, it is possible to set the printing posture objectively, without being based on the operator's experience, for example. Furthermore, the support members S necessary for manufacturing the shaped object W meeting the standard quality can be set very easily and in a short time.

(10. Others)
In the above description, the material powder is metal powder P. However, as the material powder, for example, a resin material may be used. When using a resin material, instead of manufacturing a shaped object using an additive manufacturing device using a powder bed fusion bonding method as described above, for example, a resin material dissolved in a solvent may be ejected to form the object. An inkjet method for modeling (manufacturing) objects can also be used.

DESCRIPTION OF SYMBOLS 1... Additive manufacturing device, 2... 3D modeling model generation device, 3... Inspection device, 10... Chamber, 20... Modeled object support device, 21... Modeling container, 22... Elevating table, 23... Base, 30... Powder supply device , 31... Powder storage container, 32... Supply table, 33... Recoater, 40... Light beam irradiation device, 100... Quality prediction system, 110... Machine learning device, 111... First server, 112... Second server, 210... Learning Processing device, 211... Learning data input unit, 212... Training dataset acquisition unit, 213... Training dataset storage unit, 214... Learned model generation unit, 220... Quality prediction device, 221... Learned model storage unit, 222 ...data acquisition unit during modeling, 223...quality prediction unit, 224...evaluation unit, 225...output unit, 230...database, P...metal powder, S...support member, W...modeled object

Claims (12)

Applied to an additive manufacturing device that manufactures a shaped object by irradiating a material powder with a light beam and melting and solidifying the material powder,
Generated by performing machine learning using, as a training data set, modeling data including modeling posture data representing the modeling posture relative to the base when the object was manufactured, and modeling quality data representing the quality of the manufactured object. a trained model storage unit that stores the trained model;
Quality prediction for predicting the quality of each of the objects when the modeling posture data is changed as a parameter, using the learned model and the modeling data used when newly manufacturing the object. and ,
The modeling orientation data includes a first angle around a reference axis of the object, a second angle around a first axis perpendicular to the reference axis, and a second angle around a second axis perpendicular to the reference axis and the first axis. expressed using any of the third angles of
Quality prediction system. 2. The quality prediction system according to claim 1, further comprising: an evaluation unit that evaluates, with respect to the quality of each of the predicted objects, the modeling posture corresponding to the quality that satisfies a preset reference quality. The data at the time of modeling includes support type data that identifies a plurality of types of support members that support the modeled object on the base at the time of manufacturing the modeled object,
3. The quality prediction unit predicts the quality of each of the objects when changing the modeling posture data and the support type data included in the modeling data as parameters. Quality prediction system. The quality prediction according to claim 3, further comprising an evaluation unit that evaluates the modeling posture and the support member corresponding to the quality that satisfies a preset standard quality for the predicted quality of each of the printed objects. system. comprising an output unit that outputs the modeling data corresponding to the evaluation result by the evaluation unit,
The quality prediction system according to claim 2 or 4, wherein the additive manufacturing device newly manufactures the molded object by using the molding data output from the output unit. The output unit transmits the evaluation results to a 3D modeling model generation device that generates 3D model data representing the shape of the object and the shape of a support member that supports the object on the base during manufacture of the object. The quality prediction system according to claim 5, which outputs the corresponding modeling data. The quality prediction system according to claim 6, wherein the output unit outputs the modeling data corresponding to the prediction result by the quality prediction unit to the 3D modeling model generation device. The modeling data includes modeling condition data when manufacturing the modeled object,
The quality according to any one of claims 1 to 7, wherein the modeling condition data includes at least one of laser output, scanning speed, scanning pitch, irradiation spot diameter, lamination thickness, and type of the material powder. Prediction system. The quality includes either shape accuracy expressed using a shape error between the manufactured shape of the shaped object and a reference shape, and the presence or absence of cracks in the shaped object. The quality prediction system according to any one of the items. The quality prediction system includes:
The learned model is generated by performing machine learning using the modeling data including the modeling posture data and the modeling quality data as a training data set, and the generated trained model is stored in the learned model storage unit. The quality prediction system according to any one of claims 1 to 9 , comprising a trained model generation unit for storing. The trained model generation unit is
11. The learned model is updated by performing machine learning using the modeling data and the modeling quality data of the object newly manufactured by the additive manufacturing apparatus as the training data set. Quality prediction system described. Applied to an additive manufacturing device that manufactures a shaped object by irradiating a material powder with a light beam and melting and solidifying the material powder,
By performing machine learning using, as a training data set, modeling data including modeling posture data representing the modeling posture relative to the base when the object was manufactured, and modeling quality data representing the quality of the manufactured object. a trained model generation unit that generates a trained model;
Each of the objects when the modeling posture data included in the modeling data is changed as a parameter using the learned model and the modeling data used when newly manufacturing the modeling object. a quality prediction unit that predicts the quality of the
an evaluation unit that evaluates the modeling posture corresponding to the quality that satisfies a preset reference quality with respect to the quality of each of the predicted objects ;
The modeling orientation data includes a first angle around a reference axis of the object, a second angle around a first axis perpendicular to the reference axis, and a second angle around a second axis perpendicular to the reference axis and the first axis. expressed using any of the third angles of
Quality prediction system.

Images