Patent Application
20210299754
The present invention relates to a process window generation technique in laminating and shaping.
In the above technical field, non-patent literatures 1 and 2 disclose a technique of obtaining optimum conditions of laminating and shaping in the following way. For example, the surface morphology of objects shaped under various conditions is observed, and the objects are classified into three states, that is, a state in which the surface is relatively flat, but unmolten powder remains (porous), a state in which powder is completely molten, and the surface is flat (even), and a state in which unevenness on the surface is large (uneven). Next, boundaries are set between the surface states, thereby defining the condition range of the even state as a process window. Also, the density of each shaped object is measured using a method such as the Archimedes method. X-ray CT, or quantitative evaluation of defects by analysis of a cross-sectional tissue observation image of the shaped object, and a condition for the highest density is obtained as an optimum condition.
Non-patent literature 1: H. E. Helmer, C. Korner, and R. F. Singer: “Additive manufacturing of nickel-based superalloy Inconel 718 by selective electron beam melting: Processing window and microstructure”, Journal of Materials Research, Vol. 29, No. 17 (2014), p. 1987-1996.
However, in the technique described in the above literatures, since the number of types of parameters (a heat source output, a scan speed, a scan line interval, a scan line length, a powder layer thickness, a shaping temperature, and the like) to be changed in shaping is large, and the ranges to vary are wide (for example, in a case of a commercially available 3-kW electron beam laminating and shaping apparatus, it is possible to vary the current within the range of 0 to 50 mA and the scan speed within the range of 0 to 8,000 m/s), it is necessary to perform evaluation by manufacturing many shaped objects to obtain the optimum condition even when experimental design is used. For this reason, to construct a process window, an enormous cost and time are needed, and different process windows are constructed depending on individuals who conduct experiments.
The present invention enables to provide a technique of solving the above-described problem.
One example aspect of the present invention provides an information processing apparatus comprising:
a parameter generator that generates sets of at least two parameters for controlling laminating and shaping such that the sets are scattered in a process window;
a sample shaping instructor that instructs a laminating and shaping unit to perform laminating and shaping of samples using the scattered sets of the at least two parameters:
a sample quality evaluator that acquires an image of the laminated and shaped samples and evaluates qualities of the laminated and shaped samples;
a boundary determiner that determines, in a process map generated by mapping evaluation results of the samples, a boundary of the evaluation results by machine learning; and
a process window generator that repeats parameter generation processing by the parameter generator, sample shaping instruction processing by the sample shaping instructor, quality evaluation processing by the sample quality evaluator, and boundary determination processing by the boundary determiner using a boundary zoon including the determined boundary as a new process window, and generates a process window separated by a finally determined boundary as a process window that guarantees quality of laminating and shaping.
Another example aspect of the present invention provides a laminating and shaping apparatus comprising:
the above-described information processing apparatus;
a laminating and shaping unit that, before shaping of a laminating and shaping target object, laminates and shapes samples based on a laminating and shaping instruction of the samples by the sample shaping instructor, and then laminates and shapes a laminating and shaping target object based on a laminating and shaping instruction of the laminating and shaping target object by the laminating and shaping instructor; and
an image acquirer that acquires images of the laminated and shaped samples.
Still other example aspect of the present invention provides an information processing method comprising:
generating sets of at least two parameters for controlling laminating and shaping such that the sets are scattered in a process window;
instructing a laminating and shaping unit to perform laminating and shaping of samples using the scattered sets of the at least two parameters;
acquiring images of the laminated and shaped samples and evaluating qualities of the samples;
determining, in a process map generated by mapping evaluation results of the samples, a boundary of the evaluation results by machine learning; and
repeating the generating, the instructing, the acquiring and evaluating, and the determining while using a boundary zoon including the determined boundary as a new process window, and generating a process window separated by a finally determined boundary as a process window that guarantees quality of laminating and shaping.
Still other example aspect of the present invention provides an information processing program for causing a computer to execute a method comprising:
generating sets of at least two parameters for controlling laminating and shaping such that the sets are scattered in a process window;
instructing a laminating and shaping unit to perform laminating and shaping of samples using the scattered sets of the at least two parameters;
acquiring images of the laminated and shaped samples and evaluating qualities of the samples;
determining, in a process map generated by mapping evaluation results of the samples, a boundary of the evaluation results by machine learning; and
repeating the generating, the instructing, the acquiring and evaluating, and the determining while using a boundary region including the determined boundary as a new process window, and generating a process window separated by a finally determined boundary as a process window that guarantees quality of laminating and shaping.
Still other example aspect of the present invention provides a process window generation method comprising:
performing laminating and shaping of samples using sets of at least two parameters for controlling laminating and shaping, which are scattered in a process window;
determining, in a process map generated by mapping evaluation results obtained by evaluating the laminated and shaped samples, a boundary of the evaluation results by machine learning; and
repeating the performing laminating and shaping and the determining while using a boundary region including the determined boundary as a new process window, and generating a process window separated by a finally determined boundary as a process window that guarantees quality of laminating and shaping.
According to the present invention, it is possible to construct a general-purpose process window while saving cost and time.
Example embodiments of the present invention will now be described in detail with reference to the drawings. It should be noted that the relative arrangement of the components, the numerical expressions and numerical values set forth in these example embodiments do not limit the scope of the present invention unless it is specifically stated otherwise.
An information processing apparatus 100 according to the first example embodiment of the present invention will be described with reference to
As shown in
According to this example embodiment, since the process window that guarantees the quality of laminating and shaping is generated by automatically repeating parameter generation processing, sample shaping instruction processing, quality evaluation processing, and boundary determination processing, it is possible to construct a general-purpose process window while saving cost and time.
An information processing apparatus according to the second example embodiment of the present invention will be described next. The information processing apparatus according to this example embodiment shapes a sample by setting parameter sets scattered in a process window, and evaluates the quality of the shaped sample. The process window is updated from the initial process window while obtaining a boundary (decision function) by machine learning from the evaluation result, and a process window that guarantees the quality of laminating and shaping is generated and presented to a user.
<<Laminating and Shaping System Including Information Processing Apparatus>>
The arrangement and operation of a laminating and shaping system including the information processing apparatus according to this example embodiment will be described below with reference to
(Processing Outline)
In this example embodiment, an automatic process map construction method is provided. Note that in “surface morphology monitoring” to be described later, unevenness on surfaces is observed, and the surfaces are mainly classified into non-flat surfaces (surfaces on which unmolten powder is observed or surfaces with large unevenness) and flat surfaces.
(Step S201): In the initial process window, ranges to vary process parameters (a beam output, a beam diameter, a scan speed, a line offset (scan line interval), a scan path length, a powder layer thickness, and the like) are determined, and data points (parameter sets) are set such that they are evenly distributed in the initial process window. Note that to always shape a high-quality sample and a low-quality sample, a process window including a range from the maximum value to the minimum value of each parameter settable by the laminating and shaping unit is used as the initial process window.
(Step S202): A powder layer (about 3 to 5 mm) is formed on the base plate of the laminating and shaping unit, and shaping (the height of a shaped object: about 10 to 15 mm) is performed on the powder layer under the set conditions. Note that in one sample laminating and shaping, a predetermined number of samples are laminated and shaped in the laminating and shaping unit using sets of different parameters.
(Step S203): The surface morphology of the laminated and shaped samples is monitored, and the surfaces are classified into flat surfaces and non-flat surfaces. If there is no flat surface, the surfaces are further classified into surfaces on which unmolten powder remains because of energy shortage and surfaces with large surface unevenness because of excessive energy. Alternatively, even if a flat surface exists, non-flat surfaces are classified into surfaces on which unmolten powder remains because of energy shortage and surfaces with large surface unevenness because of excessive energy. That is, an instruction is issued such that each laminated and shaped sample is captured by an image capturer, and the quality of the sample is evaluated based on the image of the sample captured by the image capturer, and evaluation value is assigned.
(Step S204): A process window is generated based on the qualities of the samples by machine learning. A boundary between the flat surfaces and the non-flat surfaces is obtained by the machine learning.
(Step S205): A new process window including the obtained boundary zoon is generated, and deficient data points are re-set to increase the accuracy of the boundary. At this time, the data points are set in a process window including the vicinity of the classification boundary (near a point where the decision function is “0”), thereby increasing the accuracy of the boundary.
(Step S206): During execution of steps S204 and S205, a powder layer (3 to 5 mm) is formed on the shaped samples, and then shaping (the height of new shaped samples: about 10 to 15 mm) is performed on the powder layer under the re-set conditions. That is, a powder layer generation instructor instructs generation of a predetermined powder layer that is not to be shaped before laminating and shaping the next samples.
(Step S207): The surface morphology of the next examined and shaped samples is monitored again, and the surfaces are classified again into flat surfaces and non-flat surfaces. If there is no flat surface, the surfaces are further classified into surfaces on which unmolten powder remains because of energy shortage and surfaces with large surface unevenness because of excessive energy. Alternatively, even if a flat surface exists, non-flat surfaces are classified into surfaces on which unmolten powder remains because of energy shortage and surfaces with large surface unevenness because of excessive energy.
(Step S208): A new boundary is obtained by machine learning, and the process window is updated to a new process window.
(Step S209): Steps S205 to S208 are repeated. Note that if process window does not change anymore, or if the height from the base plate to the uppermost surface of the samples exceeds a value set in advance (a maximum shaping height possible in the apparatus), the processing is ended. That is, the repeat is ended if the new boundary cannot be determined anymore, or if the repeat has been done a predetermined number of times (cycles). Note that the procedure is not limited to that shown in
(System Procedure)
The process window generation method in the laminating and shaping system includes sample shaping processing (step S301), boundary determination processing (step S303), and process window generation processing (step S305).
Sample shaping processing (step S301) corresponds to steps S201 and S202 in
Boundary determination processing (step S303) corresponds to steps S203 and S204 in
Process window generation processing (step S305) corresponds to steps S205 to S209 in
(System Arrangement)
The laminating and shaping system 400 includes the information processing apparatus 410 according to this example embodiment, a laminating and shaping apparatus 420, and a shaped sample image capturer 430. Note that the shaped sample image capturer 430 may be either incorporated in the laminating and shaping apparatus 420 or installed separately.
The information processing apparatus 410 includes a parameter adjuster 411 according to this example embodiment, which is configured to calibrate (adjust) parameters while shaping a sample by a laminating and shaping unit 422, and a laminating and shaping data provider 412 configured to shape a laminating and shaping target object by the laminating and shaping unit 422.
The laminating and shaping apparatus 420 includes a shaping controller 421 that controls the laminating and shaping unit 422 to shape a sample or a laminating and shaping target object in accordance with an instruction from the information processing apparatus 410, and the laminating and shaping unit 422 that shapes a sample or a laminating and shaping target object under the control of the shaping controller 421.
The shaped sample image capturer 430 captures the surface of a shaped sample. Note that the shaped sample image capturer 430 may be, for example, an X-ray camera capable of capturing an internal structure without processing such as cutting of a shaped sample.
(Laminating and Shaping Unit)
A laminating and shaping unit 441 shown in
<<Functional Arrangement of Information Processing Apparatus>>
The information processing apparatus 410 includes a communication controller 501, an input/output interface 502, a display unit 521, an operation unit 522, and the parameter adjuster 411 shown in
The communication controller 501 controls communication between the shaping controller 421 of the laminating and shaping apparatus 420 and the shaped sample image capturer 430. Note that the communication controller 501 may control communication with another external apparatus. The shaped sample image capturer 430 may be connected to the input/output interface 502 without interposing the communication controller 501.
The input/output interface 502 interfaces with input/output devices such as the display unit 521 and the operation unit 522. As described above, the shaped sample image capturer 430 may be connected.
The parameter adjuster 411 includes a scattered parameter set generator 511, a process window generator 512, and a database 513. The parameter adjuster 411 also includes a sample image acquirer 514, a sample quality evaluator 515, a process map generator 516, and a boundary determiner (machine learning) 517. The parameter adjuster 411 further includes an end determiner 518, and a process window output unit 519.
The scattered parameter set generator 511 generates scattered parameter sets in a process window generated by the process window generator 512 such that a boundary can be determined from the evaluation results of samples. The process window generator 512 initially uses an initial process window 531 held in the database 513, and in subsequent repeat, generates a region including a boundary determined by the boundary determiner 517 as a new process window. The database 513 holds the initial process window 531, an updating process window 532, and an adjusted process window 533. Note that although a detailed description thereof will be omitted, the database 513 also holds data to be used by the functional components of the parameter adjuster 411.
The sample image acquirer 514 acquires an image of a shaped sample captured by the shaped sample image capturer 430 via the communication controller 501 or the input/output interface 502. The sample quality evaluator 515 evaluates the quality of a sample, using a sample quality evaluation table 710 to be described later, based on the image of the shaped sample acquired by the sample image acquirer 514. For example, a sample shaping size, jags around a sample, unevenness on a sample surface, powder remaining on a sample surface, and the like are used as the criteria of evaluation. Note that the sample quality evaluator 515 includes an image capturing instructor that instructs to capture a laminated and shaped sample by an image capturer, and an evaluation value assigner that evaluates the quality of a sample based on the image of the sample captured by the image capturer, and assigns an evaluation value. The process map generator 516 generates a process map by mapping the evaluation results of the sample quality evaluator 515 in a process window. The boundary determiner (machine learning) 517 obtains the decision function of the evaluation result using machine learning from the process map generated by the process map generator 516 and determines a boundary. Note that in this example embodiment, a support vector machine is used as machine learning. However, a boundary may be determined by logistic regression, a k-nearest neighbor algorithm, a decision tree, random forest, a neural network, or naive bayes definition, and the same effect can be obtained.
The end determiner 518 determines the end of repeat of sample shaping→observation→evaluation→process window construction. The determination criteria include a condition that a process window does not change anymore and a condition that the height from the base plate to the uppermost surface exceeds a value set in advance (a maximum shaping height possible in the apparatus). That is, the repeat is ended if the boundary cannot be determined anymore, or if the repeat has been done a predetermined number of times. If the end determiner 518 determines to end the repeat, the process window output unit 519 functions as a presenter that presents a currently generated process window on the display unit 521 and supports parameter setting by the user. Note that the support can be made either by directly presenting the currently generated process window or by presenting the process window such that parameters can be set within a range in the process window where quality can be guaranteed.
(Process Window)
A process window set value 610 shown in
A process window set value 630 shown in
Note that
(Sample Quality Evaluation Table)
A sample quality evaluation method 720 according to this example embodiment will be described first with reference to
In excessive energy 721 shown in
In insufficient energy 723 shown in
Optimal energy 722 exists between the sample shape of high energy and the sample shape of low energy in which a fine shape is formed, and the surface become flat. In this example embodiment, boundary determination in a sample evaluation result is repeated.
The sample quality evaluation table 710 shown in
The evaluation index table 725 shown in
(Process Map)
A table 810 of sample quality evaluation results used to generate the process map 820 is shown. For each shaped sample, the table 810 stores an evaluation result 813 in association with a scan speed 811 and a current 812, and the evaluation results 813 are mapped.
A table 840 of sample quality evaluation results used to generate the process map 850 is shown. For each shaped sample, the table 840 stores an evaluation result 843 in association with a current 841 and a speed function (SF) 842, and the evaluation results 843 are mapped.
Note that in this example embodiment, machine learning is performed based on the process map 850 shown in
(Machine Learning: Support Vector Machine)
Here, process windows 961 and 962 shown in
If an evaluation result exists even in a boundary margin by the support vector machine, not a hard margin 970 but a soft margin 980 that considers the evaluation result in the boundary margin as well is employed, as shown in
Note that for more information about processing by the support vector machine, see [non-patent literature 3]. Also, in addition to the support vector machine, machine learning for deciding a boundary includes logistic regression, a k-nearest neighbor algorithm, a decision tree, random forest, a neural network, and naive bayes definition. These can also be used, and the same effect can be obtained.
(Presentation of Process Window)
In a presentation example 1010 shown in
In a presentation example 1020 shown in
Using the input or set parameter sets, a laminating and shaping instructor instructs the shaping controller 421 of the laminating and shaping apparatus 420 to laminate and shape a laminating and shaping target object. Note that the process window presentation examples are not limited to those in
<<Hardware Arrangement of Information Processing Apparatus>>
In
ARAM (Random Access Memory) 1140 is a random access memory used by the CPU 1110 as a work area for temporary storage. In the RAM 1140, an area to store data necessary for implementation of this example embodiment is allocated. The process windows 620 and 640 are data representing the process windows shown in
A storage 1150 stores databases, various kinds of parameters, and following data and programs which are necessary for implementation of this example embodiment and are used by the CPU 1110. A boundary determination machine learning algorithm 1151 is a machine learning algorithm that determines a boundary from a sample quality evaluation result of a process map. In this example embodiment, a support vector machine is used. The sample quality evaluation table 710 is a table shown in
The storage 1150 stores the following programs. An information processing program 1153 is a program configured to control the entire information processing apparatus 410. A parameter adjustment module 1154 is a module configured to implement the parameter adjuster 411. The parameter adjustment module 1154 includes a process window generation module, a scattered parameter set generation module, a sample image acquisition module, a sample quality evaluation module, a process map generation module, and a boundary determination module in correspondence with the functional components shown in
The display unit 521 and the operation unit 522 are connected as input/output devices to the input/output interface 502. In addition, the shaped sample image capturer 430 may be connected to the input/output interface 502.
Note that programs and data associated with general-purpose functions and other implementable functions of the information processing apparatus 410 are not shown in the RAM 1140 and the storage 1150 in
<<Processing Procedure of Information Processing Apparatus>>
In step S1201, the information processing apparatus 410 sets parameter sets scattered in a process window. In step S1203, the information processing apparatus 410 instructs the laminating and shaping apparatus 420 to laminate and shape samples using the parameter sets. In step S1205, the information processing apparatus 410 acquires the images of the laminated and shaped samples from the shaped sample image capturer 430. In step S1207, the information processing apparatus 410 executes quality evaluation of the samples based on the acquired images. In step S1209, the information processing apparatus 410 generates a process map by mapping the quality evaluations of the samples in the process window.
In step S1211, the information processing apparatus 410 determines a boundary in the process map by machine learning. In step S1213, as the end condition, the information processing apparatus 410 determines whether, for example, boundary accuracy is sufficient. If the end condition is not satisfied, in step S1215, the information processing apparatus 410 repeats the processing from step S1201 using a boundary zoon including the boundary as a new process window.
If the end condition is satisfied, in step S1217, the information processing apparatus 410 sets a process window separated by a finally determined boundary as a process window that guarantees the quality of laminating and shaping. In step S1219, the information processing apparatus 410 presents the process window that guarantees the quality of laminating and shaping from the display unit 521 to the user.
According to this example embodiment, since a process window that guarantees the quality of laminating and shaping is generated by automatically repeating parameter generation processing, laminating and shaping instruction processing, quality evaluation processing, and boundary determination processing, it is possible to construct a general-purpose process window while saving cost and time.
That is, in this example embodiment, the procedures of (1) forming a powder layer of a certain height, (2) shaping to a certain height. (3) observing a shaped object surface using a monitoring device implemented in the shaping apparatus, (4) classifying various shaping conditions by machine learning based on the surface morphology, and (5) deciding deficient data points and adding conditions are repeated, and the shaped object is not extracted from the apparatus for evaluation of the density or the like. Additionally, at the same time as the completion of a shaping test, process window construction and determination of process conditions under which shaping defects are less likely to occur are completed. Hence, since the processing can automatically be executed without the need of manual operations in almost all processes, cost and time needed for process window construction can largely be reduced.
An example executed according to the above-described example embodiment and a result thereof will be described below. Note that a cycle of process window construction will be described below for the sake of simplicity. The cycle is repeated similarly, thereby generating a process window of appropriate parameter conditions.
(Process Conditions)
A sample shaped in this example is metal powder of a CoCrMo alloy (ASTM F75) for medical use. The metal powder was shaped into 11 samples of different parameter sets, each of which had a diameter of 5 mm and a height of S mm, by Arcam A2X1310 (available from Arcam).
The surface image of each sample was acquired, and quality determination was executed based on feature amounts, thereby generating a process map. A boundary of quality in the process map was determined using a support vector machine. A region including the determined boundary was generated as a new process window, and 11 samples were further shaped using different parameter sets.
In this way, six layers with powder layers sandwiched between them were laminated for the 11 samples, and a region (process window) of an appropriate parameter set was determined.
In this example, as indicated by a temperature history 1330 at the time of shaping, three hours were needed for sample shaping, and about seven hours were needed for overall processing. However, processing such as cutting of the shaped samples was not necessary, the series of processes were automatically performed, and only the sample shaping time by the laminating and shaping apparatus was needed in fact.
As for the sample image, the process map, and the machine learning result, a cycle of repeat will be described below, and a description of the overall procedure will be omitted.
(Sample Image)
The sample image 1400 shows the surface images of the 11 samples simultaneously shaped in parallel. In these, a sample marked with ◯ is a sample determined as normal, and a sample marked with x is a sample determined as defective.
For example, the first sample of a set of (SF 15, 30.6 mA) and the fourth sample of a set of (SF 25, 43.9 mA) were determined as defective because of a large shaping size (8 mm>5 mm) and unevenness on the periphery. The cause is assumed to be high energy. The second sample of a set of (SF 45, 42.3 mA) and the third sample of a set of (SF 55, 14.5 mA) were determined as defective because of a residue of metal powder recognized on the sample surfaces. The cause is assumed to be low energy.
As described above, it can be seen that the 11 samples included both high energy and low energy, and were shaped as parameter sets capable of deciding a boundary in a process map.
(Process Map)
In
(Machine Learning Result)
(Optimum Conditions)
In a normal are of parameter sets separated by the boundary obtained in
As a result, the sample 1620 less distorted than the samples determined as normal in
As described above, according to this example, it is possible to present a process window including a parameter set close to optimum conditions to the user while saving cost and time.
According to this example, since a general-purpose process window is constructed while saving cost and time, the sample could be laminated and shaped by setting parameters that were optimum conditions.
An information processing apparatus according to the third example embodiment of the present invention will be described next. The information processing apparatus according to this example embodiment is different from the above-described second example embodiment in that a sample can be laminated and shaped while changing a variety of parameters of laminating and shaping. The rest of the components and operations is the same as in the second example embodiment. Hence, the same reference numerals denote the same components and operations, and a detailed description thereof will be omitted.
<<Laminating and Shaping System>>
Referring to
(Process Map)
The process map table 1840 stores an evaluation result 1849 in association with the dimensions of a current 1841, a speed function 1842, a scan line interval 1843, a scan line length 1844, a powder layer thickness 1845 a shaping temperature 1846, a beam diameter 1847, and other parameters 1848. A process map is generated based on the process map table 1840, and an n-dimensional boundary is determined by machine learning.
According to this example embodiment, it is possible to generate a process window of appropriate conditions while changing not only the current value and the scan speed but also the scan line interval, the scan line length, the powder layer thickness, the shaping temperature, the beam diameter, and the like. Note that the variety of parameters of laminating and shaping are not limited to the parameters described in this example embodiment.
An information processing apparatus according to the fourth example embodiment of the present invention will be described next. The information processing apparatus according to this example embodiment is different from the above-described second and third example embodiments in that after parameter calibration (adjustment) according to this example embodiment, an appropriate parameter set is set automatically without interposing a user setting operation, and a laminating and shaping target object is shaped. The rest of the components and operations is the same as in the second example embodiment. Hence, the same reference numerals denote the same components and operations, and a detailed description thereof will be omitted.
<<Functional Arrangement of Information Processing Apparatus>>
A parameter adjuster 2011 of the information processing apparatus 1910 further includes an appropriate parameter set generator 1920 that automatically sets an appropriate parameter set. The appropriate parameter set generator 1920 selects an appropriate parameter set without any influence of external disturbance (far apart from the boundary) based on the output from a process window output unit 519 and a process window in which the region of an appropriate parameter set is clear, and starts laminating and shaping.
According to this example embodiment, it is possible to set appropriate parameters without user setting and automatically shape a laminating and shaping target object of guaranteed quality. Note that the user may set the conditions of parameter sets in advance depending on a laminating and shaping target object and a material and the information processing apparatus may automatically set an appropriate parameter set in consideration of the conditions.
An information processing apparatus according to the fifth example embodiment of the present invention will be described next. The information processing apparatus according to this example embodiment is different from the above-described second to fourth example embodiments in that parameter set adjustment is performed in parallel even during shaping of a laminating and shaping target object. The rest of the components and operations is the same as in the second example embodiment. Hence, the same reference numerals denote the same components and operations, and a detailed description thereof will be omitted.
<<Functional Arrangement of Information Processing Apparatus>>
A parameter adjuster 2011 of the information processing apparatus 2010 further includes an adjusted parameter set generator 2021 that performs parameter set adjustment in parallel even during shaping of a laminating and shaping target object. The adjusted parameter set generator 2021 acquires an adjusted process window from a process window generator 512, and generates adjusted parameter sets in the process window, thereby further improving a laminating and shaping target object under shaping. Note that, for example, if a parameter set is close to a boundary, the adjusted parameter set generator 2021 performs adjustment such that the parameter set is apart from the boundary. This may be done based on comparison between a threshold and the distance of the parameter set from the boundary. As for the timing of parameter set adjustment, the adjustment may be performed for each layer, for each predetermined number of layers, or for each shaping of a laminating and shaping target object.
According to this example embodiment, even during shaping of a laminating and shaping target object, it is possible to adjust parameters to appropriate parameters and shape a laminating and shaping target object of guaranteed quality.
In this example embodiment, a support vector machine is used as machine learning. However, even if a boundary is determined by logistic regression, a k-nearest neighbor algorithm, a decision tree, random forest, a neural network, or naive bayes definition, the same effect can be obtained.
Additionally, this example embodiment is not limited to an application to metal laminating and shaping that performs laminating and shaping using metal powder.
Also, in this example embodiment, a powder bed method by electron beam melting has mainly been described. However, the same effect can also be obtained by an application to a laser melting method. Furthermore, the same effect can be obtained by an application to a powder deposition method.
While the invention has been particularly shown and described with reference to example embodiments thereof, the invention is not limited to these example embodiments. It will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention as defined by the claims. A system or apparatus including any combination of the individual features included in the respective example embodiments may be incorporated in the scope of the present invention.
The present invention is applicable to a system including a plurality of devices or a single apparatus. The present invention is also applicable even when an information processing program for implementing the functions of example embodiments is supplied to the system or apparatus directly or from a remote site. Hence, the present invention also incorporates the program installed in a computer to implement the functions of the present invention by the computer, a medium storing the program, and a WWW (World Wide Web) server that causes a user to download the program. Especially, the present invention incorporates at least a non-transitory computer readable medium storing a program that causes a computer to execute processing steps included in the above-described example embodiments.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/JP2018/031389 | 8/24/2018 | WO | 00 |
