Machine learning based fast parameter development for additive manufacturing and related methods

Abstract

An additive manufacturing system may include an additive manufacturing apparatus and a process chart generation system. The process map generation system may generate an operational process map in response to a mathematical simulation of the additive manufacturing device, cause the additive manufacturing device to print a plurality of test structures in response to the operational process map, detect one or more defects in each of the plurality of test structures, automatically update the operational process map in response to the defects, and generate a plurality of test structures. And modifying one or more operating parameters of the additive manufacturing apparatus in response to the updated operating process map.

Description

[0001] Priority declaration

[0002] This application claims the benefit of the filing date of U.S. patent application serial number 18 / 157,221, filed on January 20, 2023, entitled “MACHINE LEARNING BASED RAPIDPARAMETER DEVELOPMENT FOR ADDITIVE MANUFACTURING AND RELATED METHODS,” and claims the benefit of Republic of India application serial number 202211052764, filed on September 15, 2022, the disclosures of each of which are hereby incorporated by reference in their entirety. Technical Field

[0003] Embodiments of the present disclosure generally relate to methods and systems for developing operating parameters of additive manufacturing equipment. Background Art

[0004] Three-dimensional additive manufacturing (or "printing") is increasingly used in a variety of industries due to its ability to create complex three-dimensional structures from a variety of materials. For example, additive manufacturing equipment is able to produce structures formed from a variety of plastics and metals based on computer-generated three-dimensional images. Various methods have been developed to perform additive manufacturing, each requiring unique operating parameters depending on the method used. However, the development of these operating parameters requires a lot of trial and error, and conventional methods are slow and inefficient. Summary of the invention

[0005] Some embodiments of the present disclosure include an additive manufacturing system. The additive manufacturing system may include an additive manufacturing device and a process map generation system. The process map generation system may include: at least one processor; a memory device that can store data representing multiple mathematical simulations of the operation of the additive manufacturing device; and at least one non-transitory computer-readable storage medium, the non-transitory computer-readable storage medium storing instructions thereon, which, when executed by the at least one processor, cause the process map generation system to: obtain an operating process map that defines operating parameters of the additive manufacturing device; cause the additive manufacturing device to print multiple test structures in response to the operating process map, wherein each of the multiple test structures is printed based on at least one different operating parameter relative to other test structures in the multiple test structures; detect one or more defects in each of the multiple test structures; automatically update the operating process map in response to the one or more defects; and modify one or more operating parameters of the additive manufacturing device in response to the operating process map.

[0006] Additional embodiments of the present disclosure include a method for determining operating parameters of an additive manufacturing device. The method may include generating an operating process map based on a plurality of pre-computed mathematical simulations of the operation of the additive manufacturing device; printing a plurality of test structures via the additive manufacturing device, each of the plurality of test structures being printed using at least one different operating parameter relative to other test structures in the plurality of test structures; detecting one or more defects in each of the plurality of test structures; automatically updating the operating process map in response to the one or more defects; and modifying one or more operating parameters of the additive manufacturing device in response to the updated operating process map.

[0007] Further embodiments of the present disclosure may include a non-transitory computer-readable medium storing instructions thereon that, when executed by at least one processor, cause the at least one processor to perform steps. The steps may include: generating an operational process map based on a plurality of pre-computed mathematical simulations of the operation of an additive manufacturing apparatus; printing a plurality of test structures via the additive manufacturing apparatus, each of the plurality of test structures being printed using at least one different operational parameter relative to other of the plurality of test structures; obtaining one or more computed tomography (CT) images of each of the plurality of test structures via a CT imager; extracting one or more features from the one or more CT images; classifying one or more defects represented in the CT images into one or more defect types based at least in part on the one or more features; automatically updating the operational process map in response to the one or more defect types; and modifying one or more operational parameters of the additive manufacturing apparatus in response to the operational process map. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] While the present disclosure concludes with claims that particularly point out and distinctly claim certain examples, the various features and advantages of examples within the scope of the present disclosure may be more readily ascertained from the following description when read in conjunction with the accompanying drawings, wherein:

[0009] Figure 1 is a schematic diagram of an exemplary additive manufacturing system according to one or more embodiments of the present disclosure;

[0010] Figure 2 is a perspective view of a diagram showing a non-limiting example of operating parameters of an additive manufacturing apparatus according to one or more embodiments of the present disclosure;

[0011] Figure 3 is a flow chart illustrating the operation of an additive manufacturing system performed by a processor executing instructions stored on a computer-readable storage medium, according to one or more embodiments of the present disclosure;

[0012] Figure 4 An exemplary operational process diagram according to one or more embodiments of the present disclosure is shown;

[0013] Figure 5 An additive manufacturing system including a computed tomography (CT) imager is shown according to one or more embodiments of the present disclosure;

[0014] Figure 6 is a flow chart illustrating the operation of an additive manufacturing system performed by a processor executing instructions stored on a computer-readable medium, according to one or more embodiments of the present disclosure;

[0015] Figure 7 is a graphical representation of defects that have been detected and classified via one or more operations of an additive manufacturing system as described herein;

[0016] Figure 8 are block diagrams of circuits that, in some examples, may be used to implement various functions, operations, acts, processes and / or methods disclosed herein. DETAILED DESCRIPTION

[0017] Three-dimensional additive manufacturing (AM), also known as "printing", involves spreading layers of particulate material and then bonding selected portions of the particulate layers together, such as by laser powder bed fusion, activating a binder material coating the particles, or spraying an adhesive binder fluid onto the particulate layer. This sequence is repeated for additional layers of particles until the desired structure has been built. The material that makes up the particulate layer is often referred to as the "build material" or "build material powder", and the coating or sprayed fluid is often referred to as the "binder" or in some cases an "activator". For example, laser powder bed fusion involves using a laser to repeatedly melt layers of powder (usually metallic) layer by layer to form a geometric structure. In order to strengthen and / or densify the part, post-processing of the three-dimensional printed part is often required.

[0018] Various methods can be used to provide each new layer of powder for 3D printing. For example, some 3D printers have a powder supply platform that contains powder supported on a vertical indexable platform and uses counter-rotating rollers or a recoater to transfer a predetermined amount of powder from the top of the powder supply platform to the top of the build platform. This method is typical for laser powder bed fusion. Some other 3D printers utilize a traveling dispenser nozzle to dispense each new layer of powder and a traveling blade to smooth the newly deposited layer.

[0019] The process of AM can be difficult because there may be hundreds of variables to consider, which, if not properly calibrated, may cause defects in the article being manufactured, such as keyhole phenomena, vertical or horizontal poor fusion (LOF), balling, surface closed pores, etc. As a specific example, in an AM machine configured to perform laser bed fusion, keyhole phenomena may occur when the laser moves too slowly over the powder material and / or if the laser directs too much heat to the powder. This may result in a keyhole-shaped cavity at the bottom of the molten pool created by the laser, which may form holes or bubbles at the bottom of the keyhole. These holes or bubbles may then be trapped in the solidifying molten pool, thereby compromising the structural integrity of the structure being manufactured. In addition to the speed and intensity of the laser, other parameters of the laser bed fusion AM machine may cause keyhole phenomena or other defects, such as hatch spacing (e.g., the spacing between each pass of the laser over the powder material), powder layer thickness, area of ​​the laser, type of material used for the powder, etc. In addition, these parameters often must be changed or adapted when using new materials, different AM processes, different machines, different powders, or even different parts of the print itself. Furthermore, the large number of parameters that interact and influence each other during the manufacturing process can make it difficult to determine the root cause, making optimizing parameters to avoid future defects slow and expensive.

[0020] To calculate the operating parameters of an AM device or process, conventional methods include building an operating process map to help select parameter values ​​for the AM device or process. Conventional methods typically limit operating process maps to a small number of parameters, and may require manually performing single-track tests and large experimental design (DOE) prints to evaluate each parameter type individually. This conventional process may also involve a person performing a manual defect type analysis on a computed tomography (CT) scan of a printed test structure, and then changing the parameters of the process map of the AM device and performing additional prints to see if the defects are corrected. This conventional process is slow and expensive because it may take months to build a process map, which will still be limited to a small number of parameters. Therefore, conventional process map development for AM equipment is slow, expensive, and may not take into account a large number of parameters that may contribute to defects in the printed structure.

[0021] According to the present disclosure, an additive manufacturing system may use machine learning for automatic process map generation. For example, an additive manufacturing system may automatically generate and / or update an operating process map for an additive manufacturing device, which may enable faster development of the operating process map and allow for a more comprehensive number of parameters to be considered without taking into account the labor and time intensive limitations of conventional methods. In addition, the additive manufacturing system may allow for automatic classification of defects detected in multiple printed test structures, which may allow for more robust accurate defect detection and enable faster determination of the root cause of the defect to improve parameter development of the operating process map.

[0022] Figure 1 is a schematic diagram of an additive manufacturing system 100 according to one or more embodiments of the present disclosure. In some embodiments, the additive manufacturing system 100 may include a controller 140 and an additive manufacturing device 110. The additive manufacturing device may include a build assembly 142, a heat applicator assembly 144, and a powder supply assembly 146. Each of the build assembly 142, the heat applicator assembly 144, and the powder supply assembly 146 may be capable of being operably coupled to the controller 140. The build assembly 142 may include a build platform 130 and a build plate 131, which is disposed on an upper surface of the build platform 130 and is used to support a build powder 124 and an article 120 to be constructed (e.g., an additively manufactured article or a "printed construction"). In some embodiments, one or more fasteners removably secure the build plate 131 to the build platform 130.

[0023] In some embodiments, build platform 130 may include a build platform piston 132 that translates in a vertical direction (e.g., a vertical Z-axis) during formation of article 120 as a result of the printing process. For example, build platform 130 may include any conventional build platform known in the art.

[0024] Powder supply assembly 146 may include powder supply platform 129, powder supply 122, and material applicator 118. In some embodiments, powder supply platform 129 may include powder supply piston 128 that translates in a vertical direction (e.g., vertical Z axis) to translate powder supply platform 129 in a vertical direction. In addition, powder supply 122 may be disposed on an upper surface of powder supply platform 129. For example, powder supply platform 129 may be configured to translate upward in a vertical direction during the process of printing an article (e.g., article 120), and material applicator 118 (e.g., powder roller, powder blade) may move powder supply 122 from the top of powder supply platform 129 and may spread a layer of powder supply 122 on build platform 130. For example, powder supply assembly 146 may include any conventional powder supply assembly known in the art.

[0025] The heat applicator assembly 144 may include a laser 112 and a scanner 116. The laser 112 may be configured to emit an energy beam 114, which may be configured to melt a powder (e.g., a metal powder) so that the particles of the powder are melted together. In some embodiments, the scanner 116 may include one or more mirrors configured to redirect the energy beam 114 so that the energy beam 114 translates on the surface of the powder disposed on the build platform 130. For example, the heat applicator assembly 144 may include any conventional heat application system known in the art. In some embodiments, the heat applicator assembly may be configured to apply heat to the powder (e.g., a powder supply 122 disposed on the build platform 130) without using a scanner 116. For example, the heat applicator assembly 144 may be configured to guide the laser 112 and thereby guide the energy beam 114 to scan on the powder disposed on the build platform 130 without relying on any intermediate scanner or mirror.

[0026] After the heat applicator assembly 144 applies heat to the powder layer, another layer of the powder supply 122 can be spread over the previous layer, and the laser 112 can be directed again to scan the energy beam 114 over the new powder layer. Thus, this process of spreading layers and melting portions of layers according to the geometry of the article can be repeated to form layers of the article, and ultimately the article itself.

[0027] The controller 140 may include a process map generation system. The process map generation system may include a processor, a memory, a storage device, an I / O interface, and a communication interface, which may be communicatively connected via a communication infrastructure. In one or more embodiments, the processor includes hardware for executing instructions such as those constituting a computer program. The memory may be used to store data, metadata, and programs for execution by the processor. The storage device includes a storage device for storing data or instructions. The I / O interface allows a user to provide input to the additive manufacturing system 100, receive output from it, and otherwise transmit data to it and receive data from it. The communication interface may include hardware, software, or both. In any case, the communication interface may provide one or more interfaces for communication (such as, for example, packet-based communication) between the additive manufacturing system 100 and one or more other computing devices or networks.

[0028] In operation, the controller 140 can slice the three-dimensional model into layers to create a two-dimensional image of each layer through a conventional process. Subsequently, the laser 112 and the scanner 116 can direct the energy beam 114 to a thin layer (e.g., a 0.1 mm thick layer of powder) of the powder supply 122 that has been set on the building platform 130. Excess powder 126 can be set on the powder overflow platform 135, and the powder overflow platform 135 can be configured to translate in a downward vertical direction (e.g., along the vertical Z axis) through the powder overflow piston 134 as needed. The energy beam 114 can heat a portion of the powder supply 122 so that the heated portion is melted together into a first layer according to the two-dimensional image of the first layer. Then, the building platform 130 can be incrementally lowered (e.g., lowered by the same amount as the thickness of the powder supply 122 layer), and the process is repeated until all (or at least a portion) of the three-dimensional model is created.

[0029] Although discussed in terms of a particular type of additive manufacturing apparatus 110 (e.g., a laser powder bed fusion apparatus), the additive manufacturing system 100 may include any additive manufacturing apparatus. For example, the additive manufacturing system 100 may include additive manufacturing apparatus for performing one or more of the following: stereolithography (SLA), sol-gel or liquid dispensing methods, binder jetting, inkjet 3D printing, direct metal deposition, microplasma powder deposition, direct laser sintering, selective laser sintering, electron beam melting, electron beam freeform fabrication, fused deposition modeling, or other additive manufacturing processes.

[0030] Figure 2 is a perspective view of a diagram 150 showing operating parameters of an additive manufacturing device (e.g., additive manufacturing device 110) according to one or more embodiments. For example, the operating parameters of the additive manufacturing device may include at least one parameter selected from laser power, laser scanning speed, hatch spacing, powder layer thickness, build direction, laser spot size, scan vector, powder bed composition, or melt pool size. In addition, what operating parameters are used for the additive manufacturing device may change based on what type of process the additive manufacturing device is performing. For example, an additive manufacturing device configured to perform binder jetting may not include operating parameters related to the laser, but may include operating parameters related to the binder jetting process, such as the binder material, the volume of material deposited over time, the size of the binder material applicator nozzle, etc.

[0031] Figure 3302 is a flowchart illustrating the operation of the additive manufacturing system 100 performed by a processor executing instructions stored on a computer-readable storage medium. At operation 302, the additive manufacturing system 100 generates or otherwise obtains an operation process diagram in response to a plurality of mathematical simulations, wherein the operation process diagram defines one or more operating parameters of an additive manufacturing device. The plurality of mathematical simulations may include a physics-based simulation of the operation of an additive manufacturing device (e.g., the additive manufacturing device 110). For example, the plurality of simulations may simulate the operation of the additive manufacturing device, wherein each of the plurality of simulations changes the operating parameters of the additive manufacturing device relative to other simulations in the plurality of simulations. In other words, the plurality of simulations may consider and change one or more operating parameters of each simulation to simulate the operation of the additive manufacturing device. In addition, each of the plurality of simulations may predict whether the operating parameters of the corresponding simulation will produce defects in an article printed using those operating parameters.

[0032] In addition, multiple simulations can be used to generate an initial operation process map that defines n-dimensional values, where n is the number of operating parameters. The operation process map can define the relationship between two or more operating parameters. In addition, the operation process map can also define an optimal process area or "sweet spot" based at least in part on each operating parameter of the operation process map, and the optimal process area defines a range of operating parameter values ​​configured to reduce defects in the manufactured product (e.g., 3D printed structure). In some embodiments, the optimal process area may define a range of operating parameter values, which is configured to reduce defects in the manufactured product according to one or more operating parameter values. For example, the operation process map can define an optimal process area based at least in part on each operating parameter of the operation process map, so that the optimal process area defines an area or point on the operation process map where defects are unlikely to occur, taking into account all operating parameters of the operation process map. In some embodiments, the operation process map can be generated before the operation of the additive manufacturing system 100. For example, one or more general operation process diagrams may be generated externally to or by the additive manufacturing system 100, wherein the general operation process diagrams may be generated based on experimental data (e.g., multiple simulations) related to the operation of the additive manufacturing device (e.g., the additive manufacturing device 110). The additive manufacturing system 100 may then obtain (e.g., select from the one or more general operation process diagrams) an initial operation process diagram to initialize the operating parameters of the additive manufacturing device 110. For example, the operation process diagram may be generated on an external device (e.g., an external computing device), and the additive manufacturing device may be configured to communicate with the external device (e.g., via a wired or wireless connection) so that the additive manufacturing device can access and obtain the operation process diagram from the external device.

[0033] In some embodiments, a plurality of mathematical simulations and / or operation process diagrams may be stored on a memory device. In some embodiments, the memory device may be away from the additive manufacturing system. In other words, the memory device may not be integrated into the additive manufacturing system 100. For example, the memory device may be an external hard drive connected to the computing device by a cable (e.g., USB, micro USB, serial line, etc.) or a wireless connection (e.g., Bluetooth, virtual local area network (VLAN), etc.). In some embodiments, the memory device may be a removable storage device configured to be connected to a processor, such as a flash drive, a compact disk (CD), a digital versatile disk (DVD), a floppy disk, etc. In some embodiments, the memory device may be an integral part of the controller 140. In some embodiments, the memory device may include a database, which may be configured to store a plurality of simulations. For example, a plurality of simulations may be stored in a format accessible by a program in the controller 140.

[0034] At operation 304, the additive manufacturing system 100 causes the additive manufacturing device to print a plurality of test structures in response to the operational process map, wherein each of the plurality of test structures is printed based on at least one different operational parameter relative to other of the plurality of test structures. For example, in some embodiments, a plurality of test structures may be printed based on operational parameters that are known to be close to an operational parameter value threshold for a defined optimal process zone for each operational parameter. In this manner, the additive manufacturing system 100 may print one or more test structures with the goal of testing an operational parameter at a value that may present a defect.

[0035] At operation 306, the additive manufacturing system detects one or more defects in each of the plurality of test structures. For example, the additive manufacturing system may scan each of the plurality of test structures using computed tomography (CT), and may develop a machine learning model configured to identify and classify the one or more defects in each structure (with respect to Figure 5 and Figure 6 discussed in more detail).

[0036] At operation 308, the additive manufacturing system 100 automatically updates the operational process map in response to the one or more defects. For example, for each of the plurality of test structures, the detected defects of the test structure may be examined along with the operating parameters associated with the test structure. The defects are then identified and evaluated for root cause analysis. Thereafter, the operational process map is updated based on the location and / or type of defects detected for a given set of operating parameters. The root cause analysis may allow for an insightful and accurate assessment of the operating parameters that need to be changed to reduce the detected defects, thereby allowing for targeted and informed changes to the process map and thereby improving the accuracy of the operational process map in determining the operating parameters that may reduce defects in manufactured articles.

[0037] At operation 310, the additive manufacturing system 100 modifies one or more operating parameters of the additive manufacturing equipment (e.g., the manufacturing equipment 110) in response to the updated operating process map. For example, in some embodiments, the additive manufacturing system 100 can update the additive manufacturing equipment 110 so that the operating parameters found on the operating process map are at or near the defined optimal process zone. In addition, to further improve the operating process map, operations 304, 306, 308, and 310 can be repeated until the desired optimal process zone within the operating process map is defined.

[0038] In some embodiments, the additive manufacturing system 100 can dynamically modify one or more operating parameters of the additive manufacturing device during the printing process in response to the operating process map. For example, the optimal process zone can define different operating parameter values ​​during the printing process. Specifically, the operating parameters at one portion of the print (e.g., the edge of the printed structure) can use different operating parameters at other portions of the print (e.g., the center of the printed structure). Therefore, when the location of the printing process changes, the operating parameters can change to remain within the optimal process zone of the operating process map.

[0039] Figure 4 An exemplary operational process map 350 defining two operating parameters (scan speed and laser power) in connection with an AM process using a particular metal alloy (alloy 718) according to one or more embodiments is shown. The operational process map can define regions of the process map (e.g., regions 352, 354, and 356) where values ​​of the operational parameters of the operational process map may produce defects in a printed structure. For example, the operational process map 350 can define a map region 352 that can define operational parameter values ​​for a scan speed variable and a laser power variable that may produce a keyhole phenomenon defect in a printed structure using those operational parameters. For example, as Figure 4As shown in , an additive manufacturing apparatus using operating parameters including a scan speed variable from about 0 m / s to about 0.5 m / s and a laser power variable from about 150 W to about 400 W may increase the likelihood of causing a keyhole phenomenon defect in a printed structure manufactured using those operating parameters. Operational process graph 350 may also define a graph region 354 that may define operating parameter values ​​for the scan speed variable and the laser power variable that may produce a balling defect in a printed structure using those parameters. For example, Figure 4 As shown in , an additive manufacturing apparatus using operating parameters including a scan speed variable from about 0.8 m / s to about 2.0 m / s and a laser power variable from about 300 W to about 400 W may increase the likelihood of causing balling defects in printed structures manufactured using those operating parameters. In addition, the operating process graph 350 may define a graph region 356 that may define operating parameter values ​​for the scan speed variable and the laser power variable that may produce particle poor fusion (e.g., vertical poor fusion or horizontal poor fusion) defects in printed structures manufactured using those operating parameters. For example, Figure 4 As shown in , an additive manufacturing apparatus using operating parameters including a scan speed variable from about 0 m / s to about 2 m / s and a laser power variable from about 0 W to about 325 W may increase the likelihood of causing poor fusion defects in a printed structure manufactured using those operating parameters.

[0040] In addition, the process map may also define a "sweet spot" (e.g., an optimal operating zone 358) that defines an area of ​​the process map in which values ​​of operating parameters are unlikely to cause defects in a printed structure using those operating parameters. Although two operating parameters are shown mapped, the operating process map may include any number of operating parameters and, therefore, any number of dimensions. In addition, although three types of defects (e.g., keyhole phenomenon, balling, or poor fusion) are shown mapped, the operating process map may define any number of defects as areas of the process map. For example, the process map may also include a map area that defines an area of ​​the process map that indicates operating parameter values ​​that are likely to produce surface-enclosed pore defects in a printed structure using those operating parameters. For example, surface-enclosed pores (also known as subsurface or near-surface pores) typically appear around the contours (boundaries) and cores of a printed structure, where the laser slows down and turns near the contours. Therefore, additional variables such as hatch spacing can be mapped as related to laser speed to create a map area that defines the following area: in this area, values ​​of hatch speed and laser speed are likely to cause surface-enclosed pore defects in a printed structure using those parameters.

[0041] Furthermore, each operating parameter included in the operational process map can affect or change the region defined by the operational process map, such as, for example, the optimal operating region 358. For example, including additional operating parameters (e.g., hatch spacing) in the operational process map 350 can cause the region defined by the operational process map 350 (e.g., the optimal operating region 358) to change. Furthermore, including additional operating parameters can cause the operational process map 350 to map additional defect regions (e.g., a map region defining operating parameter values ​​that are likely to produce surface-enclosed porosity defects).

[0042] Figure 5 An additive manufacturing system 100 including a computed tomography (CT) imager 400 is shown according to one or more embodiments of the present disclosure. In some embodiments, the CT imager 400 can be configured to obtain one or more CT images of an object (eg, a 3D printed structure).

[0043] Figure 6 is a flow chart illustrating the operation of the additive manufacturing system 100 performed by a processor executing instructions stored on a computer readable medium. Figure 5 and Figure 6 Both, at operation 602, the additive manufacturing system 100 may obtain one or more CT images of a plurality of test structures printed by an additive manufacturing device (e.g., additive manufacturing device 110) via a CT imager (e.g., CT imager 400). In some embodiments, the one or more CT images may represent one or more 2D cross-sectional slices of the test structure being imaged. For example, in some embodiments, the CT imager 400 may be separate from the additive manufacturing device 110, and an operator may place a structure printed by the additive manufacturing device 110 within the CT imager and generate one or more CT images of the printed structure via the CT imager 400. In other embodiments, the CT imager may be integrated with the additive manufacturing device 110 so that the additive manufacturing system 100 may automatically obtain one or more CT images from the structure printed by the additive manufacturing device 110 after the printing process is completed.

[0044] In some embodiments, the additive manufacturing system can perform one or more image processing techniques to enhance features of one or more CT images. For example, the additive manufacturing system 100 can perform at least one technique selected from edge detection, histogram equalization, noise reduction, edge enhancement, image sharpening, signal enhancement, or signal attenuation. As a specific example, the additive manufacturing system 100 can apply any conventionally known convolution matrix to at least a portion of one or more CT images to enhance features of the image, such as possible defects represented in the image.

[0045] At operation 604, the additive manufacturing system 100 may extract one or more features from one or more CT images. For example, the additive manufacturing system 100 may extract features using any conventionally known feature extraction techniques, such as convolution, rectified linear unit transform, pooling, etc. At operation 606, the additive manufacturing system 100 may detect one or more defects in each of the multiple test structures in response to the one or more features. For example, the detection may include detecting the location of the defect within the structure as the CT imaging object. At operation 608, the additive manufacturing system 100 may classify each of the one or more defects into a defect type (e.g., keyhole phenomenon, balling, poor fusion, etc.). In some embodiments, operations 606 and 608 may be performed using a machine learning model. For example, the additive manufacturing system may train a classification model (e.g., a machine learning model) that is configured to identify defect types and / or defect locations based at least in part on features represented in a CT image (e.g., a CT image obtained via a CT imager 400). The classification model may be trained using decision tree learning, regression tree, boosted tree, gradient boosted tree, multi-layer perception, one-to-many, gradient boosted tree, k-nearest neighbor association rule learning, neural network, deep learning, pattern recognition, or any other type of machine learning. In some embodiments, the machine learning model may be trained using at least 5000 data points in the form of holes found in the 3D printed structure, wherein the height and width of the holes may be greater than or equal to 100 nm. As a specific example, one iteration of the unsupervised training process may include printing a test structure using different combinations of varying parameters to create one or more defect types for a given material used in the printing process. The test structure may then be scanned by a CT imaging device (e.g., CT imager 400) to create one or more CT images of the test structure. A machine learning method (e.g., an unsupervised machine learning method such as clustering or manifold learning) may then be used to form a category of defect types based on features extracted using the machine learning method. The operator may then manually inspect the test structure and / or the CT image of the test structure to confirm the accuracy of at least one defect classified by the machine learning method. The machine learning model may then be generated or updated in response to the detected defects and / or the operator's accuracy findings. This process can be repeated any number of times until the desired classification model is obtained.

[0046] As another example, the classification model can be implemented through a supervised training process. For example, one iteration of the supervised training process can include printing a test structure using different combinations of varying parameters to create one or more defect types for a given material used in the printing process using a generated operational process map. The test structure can then be scanned by a CT imaging device (e.g., CT imager 400) to produce one or more CT images of the test structure. The test structure and / or the CT image can then be manually inspected to identify defects and assign the identified defects to categories. The identified categories can then be used to generate or train a classification model based on the identified categories and features present in the CT images. For example, a machine learning model can be trained based on the labeled categories by performing a feature selection process on the features identified in the CT images to identify distinguishing characteristics of different defect types. This process can be repeated any number of times until the desired classification model is obtained.

[0047] In some embodiments, the classification model may be generated before generating the operational process map. In addition, the classification model may be continuously trained during operation of the additive manufacturing system 100 with or without manual feedback.

[0048] At operation 610, the additive manufacturing system 100 may determine one or more root causes of at least one of the one or more defects. In some embodiments, the additive manufacturing system 100 may determine one or more root causes of the defect based on the location and / or type of the detected defect. For example, the detection of a keyhole phenomenon type defect may indicate that the laser scanning speed is too slow and / or the laser power is too high. In some embodiments, the additive manufacturing system may generate one or more recommendations via a user interface display of the additive manufacturing system to change one or more operating parameters of the additive manufacturing device (e.g., the additive manufacturing device 110) to resolve the defect. The operator can then take corrective action based on the recommendation. In some embodiments, the additive manufacturing system 100 may automatically update the operating process map (e.g., the operating process map of the additive manufacturing device 110) in response to the classified defect type and / or one or more root causes. For example, when one or more keyhole phenomenon defects are detected, the additive manufacturing system 100 may automatically update the operating process map to define an optimal process area, which defines operating parameters with faster laser scanning speeds and / or lower laser powers. Thereafter, the additive manufacturing system 100 may modify one or more operating parameters of an additive manufacturing device (eg, manufacturing device 110 ) in response to the updated operational process map.

[0049] Figure 7 650 is a diagram of exemplary defects that have been detected and classified via one or more operations (e.g., operation 600) of the additive manufacturing system 100 described herein. Figure 7As shown, the additive manufacturing system 100 can detect defects such as keyhole phenomenon, vertical poor fusion (V-LOF) and horizontal poor fusion (H-LOF) within the CT image, and classify the detected defects into defect types based on detected features that generally indicate the defect type using a machine learning model. By doing so, defects can be quickly identified and classified according to defect type, which can make it easier to evaluate the root cause of the defect compared to conventional techniques that may require manual identification of the defect type in order to make desired changes to the operating parameters of the additive manufacturing equipment.

[0050] Reference together Figures 1 to 7 , developing operating parameters of the additive manufacturing system 100 by automatically updating the operating process map of the additive manufacturing device 110 can be advantageous over conventional methods of developing operating parameters of the additive manufacturing device. For example, the methods described herein can enable greatly improved efficiency in developing operating parameters of the additive manufacturing device, while also allowing for the development of a more comprehensive and therefore more accurate operating process map. For example, the methods described herein allow for automatic updating of the operating process map of the additive manufacturing device based on detected defects, which can allow for greater efficiency in developing operating parameters of the additive manufacturing device.

[0051] In addition, when developing test structures for operating process map development, the methods described herein can allow for more insightful and targeted experiments. For example, the methods described herein can enable machine learning methods to detect and identify defects within 3D printed structures. This can allow automatic updating of operating process maps and provide accelerated and more accurate root cause analysis, which can allow for more targeted and insightful operating parameter adjustments for each 3D printing iteration, thereby reducing the required number of 3D printing iterations. In addition, the machine learning-based classification described in the methods herein can allow for increased scalability to include different materials and additive manufacturing processes and the ability to detect additional additive manufacturing defect types.

[0052] Those of ordinary skill in the art will appreciate that the functional elements (eg, functions, operations, actions, processes, and / or methods) of the examples disclosed herein may be implemented in any suitable hardware, software, firmware, or combination thereof. Figure 8 Non-limiting examples of specific implementations of the functional elements disclosed herein are shown.In some examples, some or all of the functional elements disclosed herein may be performed by hardware specifically configured to perform the functional elements.

[0053] Figure 8700 is a block diagram of a circuit 700 that can be used to implement various functions, operations, actions, processes and / or methods disclosed herein in some examples. The circuit 700 includes one or more processors 702 (sometimes referred to herein as "processors 702") operably connected to one or more data storage devices (sometimes referred to herein as "storage devices 704"). The storage device 704 includes machine executable code 706 stored thereon, and the processor 702 includes logic circuits 708. The machine executable code 706 includes information describing functional elements that can be implemented (e.g., executed) by the logic circuits 708. The logic circuits 708 are suitable for implementing (e.g., executing) the functional elements described by the machine executable code 706. When executing the functional elements described by the machine executable code 706, the circuit 700 should be considered to be dedicated hardware configured to perform the functional elements disclosed herein. In some examples, the processor 702 can execute the functional elements described by the machine executable code 706 sequentially, simultaneously (e.g., on one or more different hardware platforms), or in one or more parallel process flows.

[0054] When implemented by the logic circuit 708 of the processor 702, the machine executable code 706 is configured to adapt the processor 702 to perform the operations of the examples disclosed herein. For example, the machine executable code 706 may adapt the processor 702 to perform Figure 3 Operation 300 and / or Figure 6 As another example, the machine executable code 706 may cause the processor 702 to perform at least a portion or all of the operations 600. Figure 1 As a specific non-limiting example, the machine executable code 706 may cause the processor 702 to generate or otherwise obtain an operation process map in response to a plurality of mathematical simulations stored on a non-transitory storage device. As another specific non-limiting example, the machine executable code 706 may cause the processor 702 to detect one or more defects in each of a plurality of test structures, automatically update the operation process map in response to the one or more defects detected, and modify one or more operating parameters of the additive manufacturing device in response to the operation process map. As yet another specific non-limiting example, the machine executable code 706 may cause the processor 702 to train a machine learning model based on one or more input CT images. As yet another specific non-limiting example, the machine executable code 706 may cause the processor 702 to receive one or more CT images, extract one or more features from the one or more CT images, use the machine learning development model to detect one or more defects in each of the plurality of test structures, use the machine learning development model to classify the detected defects, and determine one or more root causes of at least one of the one or more defects.

[0055] The processor 702 may include a general-purpose processor, a special-purpose processor, a central processing unit (CPU), a microcontroller, a programmable logic controller (PLC), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, other programmable devices, or any combination thereof designed to perform the functions disclosed herein. A general-purpose computer including a processor is considered a special-purpose computer, and a general-purpose computer executes functional elements corresponding to the machine executable code 706 (e.g., software code, firmware code, hardware description) associated with the examples of the present disclosure. It should be noted that the general-purpose processor (also referred to as a host processor or simply a host herein) may be a microprocessor, but in an alternative, the processor 702 may include any conventional processor, controller, microcontroller, or state machine. The processor 702 may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in combination with a DSP core, or any other such configuration.

[0056] In some examples, the storage device 704 includes a volatile data storage device (e.g., a random access memory (RAM)), a non-volatile data storage device (e.g., a flash memory, a hard disk drive, a solid state drive, an erasable programmable read-only memory (EPROM), etc.). In some examples, the processor 702 and the storage device 704 can be implemented into a single device (e.g., a semiconductor device product, a system on a chip (SOC), etc.). In some examples, the processor 702 and the storage device 704 can be implemented into separate devices.

[0057] In some examples, machine executable code 706 may include computer readable instructions (e.g., software code, firmware code). By way of non-limiting example, the computer readable instructions may be stored by storage device 704, directly accessed by processor 702, and executed by processor 702 using at least logic circuitry 708. Also by way of non-limiting example, the computer readable instructions may be stored on storage device 704, transferred to a memory device (not shown) for execution, and executed by processor 702 using at least logic circuitry 708. Thus, in some examples, logic circuitry 708 includes electrically configurable logic circuitry 708.

[0058] In some examples, machine executable code 706 may describe hardware (e.g., circuits) to be implemented in logic circuit 708 to perform a functional element. The hardware may be described at any of a variety of abstraction levels, from low-level transistor layouts to high-level description languages. At a high level of abstraction, a hardware description language (HDL) may be used, such as an IEEE standard hardware description language (HDL). By way of non-limiting example, a VERILOG TM 、SYSTEMVERILOG TM or Very Large Scale Integration (VLSI) Hardware Description Language (VHDL TM ).

[0059] The HDL description can be converted into a description of any of many other abstraction levels as needed. As a non-limiting example, the high-level description can be converted into a logic-level description, such as a register transfer language (RTL), a gate-level (GL) description, a layout-level description, or a mask-level description. As a non-limiting example, a micro-operation to be performed by a hardware logic circuit (e.g., without limitation, a gate, a flip-flop, a register) of the logic circuit 708 can be described in RTL, and then the micro-operation can be converted into a GL description by a synthesis tool, and the GL description can be converted into a layout-level description of the physical layout of an integrated circuit corresponding to a programmable logic device, discrete gate or transistor logic, discrete hardware components, or a combination thereof by a placement and routing tool. Therefore, in some examples, the machine executable code 706 may include HDL, RTL, GL description, mask-level description, other hardware descriptions, or any combination thereof.

[0060] In examples where the machine executable code 706 includes a hardware description (at any level of abstraction), a system (not shown, but including the storage device 704) may implement the hardware description described by the machine executable code 706. By way of non-limiting example, the processor 702 may include a programmable logic device (e.g., an FPGA or a PLC), and the logic circuit 708 may be electrically controlled to implement circuits corresponding to the hardware description into the logic circuit 708. Also by way of non-limiting example, the logic circuit 708 may include hard-wired logic manufactured by a manufacturing system (not shown, but including the storage device 704) according to the hardware description of the machine executable code 706.

[0061] Regardless of whether the machine executable code 706 includes computer readable instructions or a hardware description, the logic circuit 708 is adapted to perform the functional elements described by the machine executable code 706 when implementing the functional elements of the machine executable code 706. It should be noted that although the hardware description may not directly describe the functional elements, the hardware description indirectly describes the functional elements that the hardware elements described by the hardware description are capable of performing.

[0062] The terms used in this disclosure and especially in the appended claims (e.g., the bodies of the appended claims) are generally intended to be “open” terms (e.g., the term “including” should be interpreted as “including but not limited to,” the term “having” should be interpreted as “having at least,” the term “comprising” should be interpreted as “including but not limited to,” etc.).

[0063] Additionally, if a specific number of claim recitations is intended to be introduced, such intent will be explicitly recited in the claim, and in the absence of such recitation, such intent is absent. For example, to aid understanding, the following appended claims may contain the use of the introductory phrases "at least one" and "one or more" to introduce claim recitations. However, the use of such phrases should not be construed as implying that the introduction of a claim recitation by the indefinite article "a" or "an" limits any particular claim containing such introduced claim recitation to an example containing only one such recitation, even when the same claim includes the introductory phrases "one or more" or "at least one" and an indefinite article such as "a" or "an" (e.g., "a" or "an" should be interpreted as meaning "at least one" or "one or more"); the same is true for the use of definite articles used to introduce claim recitations.

[0064] In addition, even if a specific number of an introduced claim statement is explicitly stated, one skilled in the art will recognize that such a statement should be interpreted to mean at least the stated number (e.g., an unmodified statement of "two statements" without other modifiers means at least two statements or two or more statements). Moreover, in those cases where a convention like "at least one of A, B, and C, etc." or "one or more of A, B, and C, etc." is used, generally such constructions are intended to include A alone, B alone, C alone, A and B together, A and C together, B and C together, or A, B, and C together, etc.

[0065] In addition, any disjunctive word or phrase presenting two or more alternative terms, whether in the specification, claims or drawings, should be understood to contemplate the possibility of including one of the terms, any one of the terms, or both of the terms. For example, the phrase "A or B" should be understood to include the possibility of "A" or "B" or "A and B".

[0066] Although the present disclosure has been described herein with respect to certain illustrated examples, those of ordinary skill in the art will recognize and appreciate that the present invention is not so limited. Rather, many additions, deletions, and modifications may be made to the examples illustrated and described without departing from the scope of the present invention as claimed below and its legal equivalents. In addition, features from one example may be combined with features of another example while still being encompassed within the scope of the present invention as contemplated by the inventor.

Claims

1. An additive manufacturing system, comprising: Additive manufacturing equipment; A process diagram generation system, the process diagram generation system comprising: at least one processor; at least one non-transitory computer-readable storage medium storing thereon instructions that, when executed by the at least one processor, cause the process map generation system to: Obtaining an operational process diagram defining operating parameters of the additive manufacturing apparatus; causing the additive manufacturing device to print a plurality of test structures in response to the operational process map, each test structure in the plurality of test structures being printed based on at least one different operational parameter relative to other test structures in the plurality of test structures; detecting one or more defects in each of the plurality of test structures; automatically updating the operational process map in response to the one or more deficiencies; and One or more operating parameters of the additive manufacturing apparatus are modified in response to the operational process map.

2. The additive manufacturing system of claim 1, wherein the one or more defects include at least one defect selected from the group consisting of a keyhole phenomenon, vertical poor fusion, horizontal poor fusion, balling, or surface closed pores.

3. The additive manufacturing system of claim 1 , wherein the operating parameters include at least one parameter selected from laser power, laser scanning speed, hatch spacing, powder layer thickness, build direction, laser spot size, scan vector, powder bed composition, or melt pool size.

4. The additive manufacturing system of claim 1 , wherein the operational process map defines an optimal process region based at least in part on each operational parameter of the operational process map, the optimal process region defining a range of operational parameter values ​​configured to reduce defects in a structure printed by the additive manufacturing device.

5. The additive manufacturing system of claim 1 , further comprising a computed tomography (CT) imager, wherein the instructions stored on the at least one computer readable storage medium, when executed by the processor, cause the process map generation system to: obtaining, via the CT imager, one or more CT images of each of the plurality of test structures; extracting one or more features from the one or more CT images; detecting one or more defects in each of the plurality of test structures in response to the one or more features; as well as Each defect of the one or more defects is classified into a defect type.

6. The additive manufacturing system of claim 5, wherein the instructions stored on the at least one computer-readable storage medium, when executed by the processor, cause the process map generation system to: determining one or more root causes of at least one of the one or more defects; and The operational process map is automatically updated in response to the classified defect type and the one or more root causes.

7. The additive manufacturing system of claim 5, wherein the instructions stored on the at least one computer-readable storage medium, when executed by the processor, cause the process map generation system to: A classification model is trained, the classification model being configured to identify defect types and / or defect locations based at least in part on features represented in the CT images.

8. The additive manufacturing system according to claim 1, further comprising: a memory device storing data representing experimental data related to a process of the additive manufacturing apparatus; wherein the instructions stored on the at least one computer-readable storage medium, when executed by the processor, cause the process map generation system to: An operational process map is generated in response to the experimental data.

9. The additive manufacturing system of claim 1, wherein the plurality of mathematical simulations are physics-based simulations of the operation of the additive manufacturing device.

10. A method for determining operating parameters of an additive manufacturing device, comprising: generating an operational process map based on a plurality of pre-computed mathematical simulations of the operation of the additive manufacturing device; printing, via the additive manufacturing device, a plurality of test structures, each test structure in the plurality of test structures being printed using at least one different operating parameter relative to other test structures in the plurality of test structures; detecting one or more defects in each of the plurality of test structures; automatically updating the operational process map in response to the one or more deficiencies; as well as One or more operating parameters of the additive manufacturing apparatus are modified in response to the updated operational process map.

11. The method of claim 10, wherein detecting one or more defects in each of the plurality of test structures comprises: obtaining, via a computed tomography (CT) imager, one or more CT images of each of the plurality of test structures; extracting one or more features from the one or more CT images; detecting one or more defects in each of the plurality of test structures in response to the one or more features; as well as Each defect of the one or more defects is classified into a defect type.

12. The method of claim 11, further comprising performing one or more image processing techniques on the one or more CT images.

13. The method of claim 12, further comprising selecting the one or more image processing techniques to include at least one technique selected from edge detection, histogram equalization, noise reduction, edge enhancement, image sharpening, signal enhancement, or signal attenuation.

14. The method of claim 11, further comprising training a classification model configured to identify defect types and / or defect locations based at least in part on features represented in the CT images.

15. The method of claim 11, further comprising automatically updating the operational process map in response to the classified defect type.

16. The method of claim 10, further comprising detecting the one or more defects from keyhole phenomenon, vertical poor fusion, horizontal poor fusion, balling, or surface closed porosity.

17. The method of claim 10, wherein modifying the one or more operating parameters comprises modifying at least one parameter selected from laser power, laser scanning speed, hatch spacing, powder layer thickness, build direction, laser spot size, scanning vector, powder bed composition, or melt pool size.

18. The method of claim 10, further comprising configuring the operational process map to define an optimal process region based at least in part on each operational parameter of the operational process map, the optimal process region defining a range of operational parameter values ​​configured to reduce defects in a structure printed by the additive manufacturing device.

19. A non-transitory computer readable medium storing thereon instructions which, when executed by at least one processor, cause the at least one processor to perform steps comprising: generating an operational process map based on a plurality of pre-computed mathematical simulations of the operation of the additive manufacturing device; printing, via the additive manufacturing device, a plurality of test structures, each test structure in the plurality of test structures being printed using at least one different operating parameter relative to other test structures in the plurality of test structures; obtaining, via a computed tomography (CT) imager, one or more CT images of each of the plurality of test structures; extracting one or more features from the one or more CT images; classifying one or more defects represented in the CT image into one or more defect types based at least in part on the one or more features; automatically updating the operational process map in response to the one or more defect types; and One or more operating parameters of the additive manufacturing apparatus are modified in response to the operational process map.

20. The non-transitory computer readable medium of claim 19, further comprising dynamically modifying one or more operating parameters of the additive manufacturing device during a printing process in response to the operational process map.

Images