Systems and methods for measuring radiant heat energy during an additive manufacturing operation

Abstract

The present disclosure describes systems and methods for measuring radiant thermal energy during an additive manufacturing operation, and describes various methods and apparatuses for characterizing an additive manufacturing process. A method for characterizing an additive manufacturing process can include generating a plurality of scans of an energy source across a build plane; measuring, using an optical sensor, an amount of energy radiated from the build plane during each of the plurality of scans; determining an area of the build plane traversed during the plurality of scans; determining, based on the amount of energy radiated and the area of the build plane traversed by the plurality of scans, a thermal energy density for the area of the build plane traversed by the plurality of scans; mapping the thermal energy density to one or more locations in the build plane; determining that the thermal energy density is characterized by a density outside of a range of density values; and thereafter, adjusting a subsequent scan of the energy source across or near the one or more locations in the build plane.

Description

[0001] This application is a divisional application of Chinese Patent Application No. 2018800641011, filed on August 1, 2018, entitled "System and Method for Measuring Radiant Thermal Energy During Additive Manufacturing Operation".

[0002] Cross-references to related applications

[0003] This application claims priority to U.S. Provisional Patent Application No. 62 / 540,016, filed August 1, 2017; U.S. Provisional Patent Application No. 62 / 633,487, filed February 21, 2018; and U.S. Provisional Patent Application No. 62 / 643,457, filed March 15, 2018, each entitled “System and Method for Measuring Energy Input During Additive Manufacturing Operations,” the disclosures of which are incorporated herein by reference in their entirety for all purposes. Background Technology

[0004] Additive manufacturing, or sequential assembly or construction, of parts takes many forms through a combination of adding materials and applying energy, and currently exists in many specific implementations and embodiments. Additive manufacturing can be performed using any of a variety of different processes involving the virtual formation of three-dimensional parts of any shape. What these various processes have in common is the layer-by-layer sintering, solidification, or melting of liquid, powdered, or granular raw materials using ultraviolet light, high-power lasers, or electron beams, respectively. Unfortunately, established processes for determining the quality of parts manufactured in this manner are limited. Conventional quality assurance testing typically involves post-processing measurements of the mechanical, geometric, or metallurgical properties of the parts, which often results in the parts being destroyed. While destructive testing is a recognized method for verifying part quality, it cannot be applied to production parts for obvious reasons, as it allows for close inspection of various internal features of the part. Therefore, there is a need for a non-destructive and accurate method for verifying the mechanical, geometric, and metallurgical properties of production parts manufactured through additive manufacturing. Summary of the Invention

[0005] The described embodiments relate to additive manufacturing, which involves the use of an energy source in the form of a moving area that takes strong thermal energy. These processes are widely known as welding processes if this thermal energy causes the physical melting of the added material. In welding processes, material added incrementally and sequentially is melted by an energy source in a manner similar to fusion welding. Exemplary welding processes suitable for use with the described embodiments include processes using a scanning energy source with a powder bed and wire-fed processes using electric arcs, lasers, and electron beams as energy sources.

[0006] When the added material is in the form of powder layers, after each incremental layer of powder material is sequentially added to the part being constructed, a scanning energy source melts the incrementally added powder through the welding area of ​​the powder layer, thereby forming a moving molten zone, hereinafter referred to as a melt pool. Once solidified, these become part of the previously sequentially added and melted and solidified layers located beneath the new layer to form the part being constructed. Because additive machining processes can be lengthy and involve any number of melt pools, at least slight variations in the size and temperature of the melt pools are unavoidable when they are used to solidify the part. The embodiments described herein reduce or minimize discontinuities caused by variations in the size and temperature of the melt pools. It should be noted that, due to the high travel rate of the heating elements and the complex patterns required to form three-dimensional structures, additive manufacturing processes can be driven by one or more processors associated with computer numerical control (CNC).

[0007] The general purpose of the described embodiments is to apply optical sensing techniques, such as quality inference, process control, or both, to additive manufacturing processes. Optical sensors can be used to track the evolution of physical phenomena in a process by tracking the evolution of variables associated with those physical phenomena. In this document, optics may include a portion of the electromagnetic spectrum, which includes near-infrared (IR), visible, and near-ultraviolet (UV). Typically, the wavelengths of the spectrum are considered to be from 380 nm to 780 nm. However, the wavelengths of near-UV and IR can extend as low as 1 nm and as high as 3000 nm, respectively. Sensor readings collected from optical sensors can be used to determine in-process quality metrics (IPQM). One such IPQM is thermal energy density (TED), which helps characterize the energy applied to different regions of a part.

[0008] TED is a metric sensitive to user-defined laser powder bed fusion process parameters (e.g., laser power, laser speed, hatch spacing, etc.). This metric can then be used for analysis comparing the IPQM with a baseline dataset. The IPQM can be calculated for each scan and displayed graphically or in 3D using a point cloud. Furthermore, comparison of the IPQM with a baseline dataset indicating manufacturing defects can be used to generate control signals for process parameters. In some embodiments, where detailed thermal analysis is required, the thermal energy density can be determined for discrete portions of each scan. In some embodiments, thermal energy data from multiple scans can be divided into discrete grid regions, allowing each grid region to reflect the total amount of energy received at each grid region for a layer or a predetermined number of layers.

[0009] An additive manufacturing method is disclosed, comprising: generating an energy source across a plurality of scans of a build plane; measuring the amount of energy radiated from the build plane during each scan of the plurality of scans using an optical sensor monitoring the build plane; determining the area of ​​the build plane traversed during the plurality of scans; determining a thermal energy density for the area of ​​the build plane traversed by the plurality of scans based on the amount of radiated energy and the area of ​​the build plane traversed by the plurality of scans; mapping the thermal energy density to one or more locations in the build plane; determining that the thermal energy density is characterized by a density outside a density value range; and thereafter, adjusting subsequent scans of the energy source across or near one or more locations in the build plane.

[0010] An additive manufacturing method is disclosed, comprising: generating an energy source for scanning across a build plane; measuring the amount of energy radiated from the powder bed during scanning using an optical sensor that monitors the powder bed; determining an area associated with the scan; determining a thermal density for the scanned area based on the amount of radiated energy and the scanned area; determining that the thermal density is characterized by a density outside a density value range; and subsequently adjusting subsequent scans of the energy source across the build plane.

[0011] An additive manufacturing method is disclosed, comprising: performing an additive manufacturing operation using an energy source; receiving sensor data associated with an optical diode during a scan of the energy source across a powder bed; receiving drive signal data indicating when the energy source is turned on; using the energy source drive signal data to identify the sensor data collected when the energy source is turned on; dividing the sensor data into multiple sample segments, each of the multiple sample segments corresponding to a portion of the scan; determining the thermal energy density for each of the multiple sample segments; and identifying one or more portions of a part most likely to contain manufacturing defects based on the thermal energy density of each of the multiple sample segments.

[0012] An additive manufacturing method is disclosed, comprising: generating an energy source across multiple scans of a build plane; determining a grid region comprising the multiple scans, wherein the grid region is characterized by a grid area; generating sensor readings during each scan of the multiple scans using an optical sensor; determining, using the sensor readings, the total amount of energy radiated from the build plane during the multiple scans; calculating the thermal density associated with the grid region based on the total amount of radiated energy and the grid area; determining that the thermal density associated with the grid region is characterized by a thermal density outside a range of thermal density values; and subsequently adjusting the output of the energy source.

[0013] An additive manufacturing method is disclosed, comprising: defining a portion of a build plane as a grid comprising a plurality of grid regions, each grid region having a grid region area; generating an energy source across a plurality of scans of the build plane; generating sensor readings using an optical sensor during each of the plurality of scans; mapping a portion of each of the plurality of sensor readings to a corresponding grid region within the plurality of grid regions for each of the plurality of scans; and for each grid region within the plurality of grid regions:

[0014] The sensor readings mapped to each grid region are summed; and the grid-based thermal density is calculated based on the summed sensor readings and the grid region area; the grid-based thermal density associated with one or more grid regions is characterized by a thermal density value outside the range of thermal density values; and thereafter, the output of the energy source is adjusted. Attached Figure Description

[0015] This disclosure will be more readily understood through the following detailed description taken in conjunction with the accompanying drawings, in which the same reference numerals denote the same structural elements, and in the drawings:

[0016] Figure 1A This is a schematic illustration of an optical sensing device used in an additive manufacturing system with an energy source, in which a laser beam is used in this specific example;

[0017] Figure 1B This is a schematic illustration of an optical sensing device used in an additive manufacturing system with an energy source, in which an electron beam is employed in this specific instance;

[0018] Figure 2 This shows a sample scan pattern used in the additive manufacturing process;

[0019] Figure 3 A flowchart illustrating a method for identifying the parts of a component most likely to contain manufacturing defects is shown.

[0020] Figures 4A to 4H Data associated with the step-by-step process is shown to use thermal energy density to identify the parts of a part most likely to contain manufacturing defects;

[0021] Figure 5 A flowchart is shown that describes in detail how to use scanlet data segregation to complete IPQM evaluation;

[0022] Figures 6A to 6F Data associated with the stepper process is shown to use thermal energy density to identify the parts of a part most likely to contain manufacturing defects; and

[0023] Figures 7A to 7C Test results comparing IPQM metrics with post-processing metallography are shown;

[0024] Figure 8 Alternative processes are shown in which data recorded by optical sensors (such as non-imaging photodetectors) can be processed to characterize additive manufacturing build processes.

[0025] Figures 9A to 9D A visual depiction indicating how multiple scans can contribute to the power introduced at a single grid region is shown;

[0026] Figure 10A An exemplary turbine blade suitable for use according to the described embodiments is shown;

[0027] Figure 10B An exemplary manufacturing configuration is shown in which 25 turbine blades can be manufactured simultaneously on top of the build plane 1006;

[0028] Figures 10C to 10D It shows Figure 10B The different cross-sectional views of the different layers of the configuration depicted in the document;

[0029] Figures 11A to 11BCross-sectional views of the base of two different turbine blades are shown;

[0030] Figure 11C The image shown illustrates the differences in surface consistency between two different bases;

[0031] Figure 12 The illustration shows the thermal energy density for parts associated with multiple different builds;

[0032] Figures 13 to 14B The illustration shows an example of how thermal energy density can be used to control the operation of parts using in-situ measurements;

[0033] Figure 14C Another power density plot is shown, highlighting the various physical effects caused by energy settings that extend far beyond the process window;

[0034] Figure 14D This demonstrates how the size and shape of the melting pool can be varied according to laser power and scanning speed settings;

[0035] Figures 15A to 15F The illustration shows how meshes can be dynamically created to characterize and control additive manufacturing operations;

[0036] Figure 16 An exemplary control loop 1600 for establishing and maintaining feedback control of additive manufacturing operations is shown;

[0037] Figure 17A The standard distribution of powder across the building plate is shown;

[0038] Figure 17B This demonstrates how the thickness of the resulting powder layer can vary when the recoater arm recovers insufficient powder and this insufficient powder spreads across the build plate.

[0039] Figure 17C A black-and-white photograph of the build plate is shown, in which a small supply of powder results in only partial coverage of the nine workpieces arranged on the build plate; and

[0040] Figure 17D This demonstrates how the detected thermal energy density differs fundamentally when the energy source scans across all nine workpieces using the same input parameters. Detailed Implementation

[0041] Figure 1AAn embodiment of an additive manufacturing system is shown that uses one or more optical sensing devices to determine thermal energy density. Thermal energy density is sensitive to variations in process parameters such as energy source power, energy source speed, and scanning spacing. Figure 1A The additive manufacturing system uses a laser 100 as an energy source. The laser 100 emits a laser beam 101, which passes through a partially reflective mirror 102 and enters a scanning and focusing system 103. The scanning and focusing system 103 then projects the beam onto a small area 104 on a work platform 105. In some embodiments, the work platform is a powder bed. Light energy 106 is emitted from the small area 104 due to the high material temperature.

[0042] In some embodiments, the scanning and focusing system 103 may be configured to collect some of the light energy 106 emitted from the beam interaction region 104. A partial reflector 102 may reflect the light energy 106, as depicted by the optical signal 107. The optical signal 107 may be interrogated by a plurality of on-axis optical sensors 109, each receiving a portion of the optical signal 107 via a series of other partial reflectors 108. It should be noted that in some embodiments, the additive manufacturing system may include only one on-axis optical sensor 109 with a total internal reflection mirror 108.

[0043] It should be noted that the collected optical signal 107 may not have the same spectral content as the light energy 106 emitted from the beam interaction region 104 because the signal 107 suffers some attenuation after passing through multiple optical elements (such as partial reflectors 102, scanning and focusing systems 103, and a series of other partial reflectors 108). These optical elements each have their own transmission and absorption characteristics, resulting in variations in the amount of attenuation, thus limiting certain portions of the spectrum of energy radiated from the beam interaction region 104. The data generated by the coaxial optical sensor 109 can correspond to the amount of energy applied to the working platform.

[0044] Examples of coaxial optical sensors 109 include, but are not limited to, light-to-electrical signal transducers (such as pyrometers and photodiodes) (i.e., photodetectors). Optical sensors may also include spectrometers and low- or high-speed cameras operating in the visible, ultraviolet, or infrared spectrum. The coaxial optical sensor 109 is located in a reference frame that moves with the laser beam; that is, as the laser beam 101 scans across the work platform 105, the coaxial optical sensor 109 sees all areas touched by the laser beam and is able to collect optical signals 107 from all areas of the touched work platform 105. Since the path of the light energy 106 collected by the scanning and focusing system 103 is almost parallel to the laser beam, the sensor 109 can be considered a coaxial sensor.

[0045] In some embodiments, the additive manufacturing system may include an off-axis sensor 110, which is positioned in a fixed reference frame relative to the laser beam 101. These off-axis sensors 110 will have a given field of view 111, which may be very narrow or may comprise the entire work platform 105. Examples of such sensors may include, but are not limited to, pyrometers, photodiodes, spectrometers, high-speed or low-speed cameras operating in the visible, ultraviolet, or infrared spectrum, etc. An off-axis sensor 110 misaligned with the energy source is considered an off-axis sensor. Off-axis sensors can also be sensors that combine a range of physical measurement modes, such as laser-ultrasound sensors, which can actively excite or “ping” a deposit using a laser beam, and then use a laser interferometer to measure the resulting ultrasonic waves or the structure’s “approximate transient high-frequency resonant response” to measure or predict the mechanical properties or mechanical integrity of the deposit while it is being built. Laser-ultrasound sensor / interferometer systems can be used to measure the elastic properties of materials, which can provide a deep understanding of, for example, the porosity of the material and other material properties. In addition, laser ultrasonic / sensor interferometer systems can be used to measure the formation of defects that cause material vibration.

[0046] In addition, contact sensors 113 may be present on the mechanical device, the recoating arm 112 for spreading powder. These sensors may be accelerometers, vibration sensors, etc. Finally, other types of sensors 114 may be present. These sensors may include contact sensors such as thermocouples that measure macroscopic thermal fields, or may include acoustic emission sensors that can detect cracking and other metallurgical phenomena occurring in the deposit during the deposition process. These contact sensors can be used to characterize the operation of the recoating arm 112 during the powder addition process. Data collected by the coaxial optical sensor 109 and the off-axis sensor 110 can be used to detect process parameters associated with the recoating arm 112. Therefore, the system can detect and address non-uniformities on the surface of the spread powder. Rough surfaces resulting from variations during the powder spreading process can be characterized by the contact sensors 113 to anticipate potential problem areas or non-uniformities in the resulting part.

[0047] In some embodiments, the laser beam 101 can melt peaks in the powder spreading process, resulting in subsequent powder layers having corresponding peaks. At certain points, peaks may contact the recoating arm 112, potentially damaging it and causing further powder spreading inhomogeneity. Therefore, embodiments of the invention can detect inhomogeneities in the spread powder before they cause inhomogeneities in the build area on the work platform 105. Many variations, modifications, and alternatives will be recognized by those skilled in the art.

[0048] In some embodiments, the coaxial optical sensor 109, off-axis sensor 110, contact sensor 113, and other sensors 114 may be configured to generate in-process raw sensor data. In other embodiments, the coaxial optical sensor 109, off-axis optical sensor 110, contact sensor 113, and other sensors 114 may be configured to process data and generate reduced-order sensor data.

[0049] In some embodiments, a computer 116, including a processor 118, a computer-readable medium 120, and an I / O interface 122, is provided and coupled to appropriate system components of the additive manufacturing system to collect data from various sensors. The data received by the computer 116 may include raw sensor data and / or downgraded sensor data from the process. The processor 118 can use the raw sensor data and / or downgraded sensor data from the process to determine power and control information for the laser 100, including coordinates relative to the work platform 105. In other embodiments, the computer 116, including the processor 118, the computer-readable medium 120, and the I / O interface 122, can provide control over various system components. The computer 116 can send, receive, and monitor control information associated with the laser 100, the work platform 105, and the recoating arm 112 to control and adjust appropriate process parameters for each component.

[0050] Processor 118 can be used to perform calculations using data received from various sensors to generate quality metrics in the process. In some embodiments, data generated by coaxial optical sensors 109 and / or off-axis sensors 110 can be used to determine thermal density during the build process. The processor can receive control information associated with the movement of an energy source across the build plane. The processor can then use the control information to associate data from one or more coaxial optical sensors 109 and / or off-axis sensors 110 with corresponding locations. The associated data can then be combined to calculate thermal density. In some embodiments, processor 118 can use thermal density and / or other metrics to generate control signals for process parameters (e.g., laser power, laser speed, scan spacing, and other process parameters) in response to other metrics or thermal densities exceeding the desired range. In this way, problems that could otherwise damage the produced parts can be mitigated. In embodiments where multiple parts are produced at once, timely correction of process parameters in response to metrics exceeding the desired range can prevent adjacent parts from receiving too much or too little energy from the energy source.

[0051] In some embodiments, I / O interface 122 may be configured to transmit collected data to a remote location. I / O interface 122 may also be configured to receive data from a remote location. The received data may include benchmark datasets, historical data, post-processing inspection data, and classifier data. A remote computing system may use the data transmitted by the additive manufacturing system to calculate quality metrics in the process. The remote computing system may transmit information to I / O interface 122 in response to a specific quality metric in the process.

[0052] In the case of an electron beam system Figure 1BPossible configurations and sensor arrangements are illustrated. An electron beam gun 150 generates an electron beam 151, which is focused by an electromagnetic focusing system 152 and then deflected by an electromagnetic deflection system 153, thereby forming a precisely focused and oriented electron beam 154. The electron beam 154 forms a heat beam-material interaction region 155 on the workpiece 156. Light energy 158 radiated from the workpiece 156 can be collected by a series of optical sensors 159, each optical sensor 159 having its own corresponding field of view 160. The light energy 158 can be further locally isolated to the interaction region 155 or can surround the entire workpiece 156. Furthermore, the optical sensors 159 can have their own tracking and scanning system that can follow the electron beam 154 as it moves across the workpiece 156.

[0053] Regardless of whether sensor 159 has optical tracking, sensor 159 may include a pyrometer, photodiode, spectrometer, and low-speed or high-speed camera operating in the visible, UV, or IR spectral regions. Sensor 159 may also be a sensor combining a range of physical measurement modes, such as a laser ultrasonic sensor, which can actively excite or “bombard” the deposit using a laser beam, and then use a laser interferometer to measure the resulting ultrasonic waves or the structure’s “approximate transient high-frequency resonant response” to measure or anticipate the mechanical properties or mechanical integrity of the deposit during its construction. Additionally, contact sensors 113 may be present on the recoating arm. These sensors may be accelerometers, vibration sensors, etc. Finally, other types of sensors 114 may be present. These sensors may include contact sensors such as thermocouples that measure macroscopic thermal fields, or may include acoustic emission sensors that can detect cracking and other metallurgical phenomena occurring in the deposit during its construction. In some embodiments, one or more thermocouples may be used to calibrate the temperature data collected by sensor 159. It should be noted that, in combination with… Figure 1A and Figure 1B The described sensor can be used in the described manner to characterize the performance of any additive manufacturing process involving sequential material accumulation.

[0054] Figure 2Possible scanning patterns for scanning an energy source across a powder bed are illustrated. In 200, a region of the workpiece is processed by scanning an energy source along alternating long path lengths in a certain direction. In this embodiment, a scan interval 204 is shown between a first scan 206 and a second scan 208. In 202, the region of the workpiece is divided into smaller checkerboard grids 214, which can be scanned sequentially from left to right and from top to bottom by a first scan 210 and a second scan 212, respectively. In other embodiments, the scanning order for the individual checkerboard grids can be randomized. Many scanning patterns can be utilized in conjunction with the additive manufacturing processes disclosed herein. Many variations, modifications, and alternatives will be recognized by those skilled in the art.

[0055] Figure 3 A flowchart illustrating an exemplary process 300 is shown, which uses data generated by an additive manufacturing system to determine thermal density and identify portions of a part most likely to contain manufacturing defects. The thermal density can be determined using data generated by coaxial optical sensor 109 and off-axis optical sensor 110, either alone or in combination. At 302, a raw photodiode data trace is received. This raw photodiode data trace can be generated using, for example, voltage data generated by a sensor in response to detected emitted thermal energy. At 304, the raw photodiode trace is identified relative to a specific scan. i The corresponding part. In some embodiments, the individual photodiode data trajectory can be separated from the remaining readings of the sensor readings by referencing energy source drive signal data (drive signals responsible for manipulating and actuating the energy source). At 306, the scan is determined under the original photodiode data trajectory. i The area (in the following text, pdon) i In some embodiments, pdon i This represents the integrated photodiode voltage. In some embodiments, pdon i Indicates in scan i The average reading of the photodiode during the period. At 308, identification is performed with scan. i The associated part p. The part identified at 308 may also have an associated area A of the part. p This can be achieved by making pdon i These two values ​​are determined by correlating them with the energy source location data described above. The process can calculate the total scan count at 310. At 312, the scan count can be determined. i The associated length L i Equation (1) can be used to calculate L. i , where x1 i y1i and x2 i y2 i Indicates the use of scan i The corresponding start and end positions.

[0056]

[0057] At point 314, the total length Lsum of all scans used to produce the part can be determined. p The length L of each scan associated with the part can be determined. i Summation is used to determine Lsum on the part. p At point 316, the scanned area A can be determined. i A can be calculated using equation (2). i :

[0058]

[0059] At position 316, it can be determined that the value used for i is... th Scanned proportional thermal density (TED) i TED i This is an example of a set of descending process characteristics. The TED is calculated using raw photodiode data. From this raw sensor data, the TED calculation extracts descending process characteristics from the raw sensor data. i Sensitive to all user-defined laser powder bed fusion process parameters (e.g., laser power, laser speed, scanning spacing, etc.). TED can be calculated using equation (3). i :

[0060]

[0061] For the purposes of this discussion, "descending order" refers to one or more of the following: data compression, i.e., less data in a feature compared to the original data; data reduction, i.e., performing a systematic analysis of the original data to derive process metrics or other figures of merit; data aggregation, i.e., clustering data into discrete groups and smaller sets of variables, representing clusters that are the opposite of the original data itself; data transformation, i.e., using transformation laws or algorithms to mathematically manipulate the data to map the original data linearly or nonlinearly to another variable space of lower dimensionality; or any other related such technique that will have the net effect of reduced data density, reduced data dimensionality, reduced data size, transformation of data to another reduced space, or all of these effects simultaneously.

[0062] TED iIt can be used for analysis during comparison of process quality metrics (IPQM) with benchmark datasets. IPQM can be calculated for each scan. At 318, the IPQM quality benchmark dataset and the calculated TED can be compared. i In regions where the difference between the calculated TED of a part and the benchmark dataset exceeds a threshold, these regions can be identified as potentially containing one or more defects and / or where further processing can be performed in near real-time to improve any defects caused by variations in TED relative to the benchmark dataset. In some embodiments, a classifier can be used to identify defect-prone portions of the part. The classifier is capable of grouping results nominally or nominally and can be represented graphically and / or through text-based media. The classifier can use a variety of classification methods, including but not limited to: statistical classification using both univariate and multivariate methods; heuristic-based classifiers; expert system-based classifiers; lookup table-based classifiers; classifiers based solely on upper or lower control limits; classifiers that work in conjunction with one or more statistical distributions, which can establish nominal and nominal thresholds based on confidence intervals and / or consideration of degrees of freedom; or any other classification scheme, whether implied or explicit, capable of distinguishing whether a set of feature data is nominal or nominal. For the purposes of this discussion, "nominal" will mean a set of process results within a predefined specification (which result in the post-process measurement properties of the part thus manufactured falling within a range of values ​​considered acceptable) or any other quantitative, semi-quantitative, objective, or subjective method used to confirm an "acceptable" part. Further description of the classification of IPQM is provided in US Patent Application No. 15 / 282,822, filed September 30, 2016, the disclosure of which is incorporated herein by reference in its entirety for all purposes.

[0063] What will be understood is... Figure 3 The specific steps illustrated in the diagram provide a concrete method for collecting data and determining thermal energy density according to an embodiment of the present invention. Other sequences of steps may also be performed according to alternative embodiments. Furthermore, Figure 3 The steps illustrated in the diagram may include multiple sub-steps, which can be performed in various orders suitable for individual steps. Furthermore, depending on the specific application, additional steps may be added, or existing steps may be removed. Those skilled in the art will recognize many variations, modifications, and alternatives.

[0064] Figures 4A to 4H The illustration shows the steps used in process 300 to determine the TED and identify any part of the part that may contain defects. Figure 4ACorresponding to step 302, the raw photodiode signal 402 for a given scan length is shown. The x-axis 450 indicates time in seconds, and the y-axis 460 indicates the photodiode voltage. In some embodiments, optical measurements may be performed alternatively or additionally by a pyrometer. Signal 402 is the raw voltage of the photodiode. The rise 404 and fall 406 of the photodiode signal 402, as well as the scattering and variation 408 of the signal during the laser-on time, are clearly visible. Data is collected at a given number of samples per second. The variation 408 in the photodiode signal 402 may be caused by changes in the powder melting in the powder bed. For example, one of the minor troughs in the photodiode signal 402 may be caused by the energy absorbed by larger particles in the particle bed as they transition from a solid to a liquid state. Typically, the number of data points in a given segment of the photodiode signal between rise and fall events may be related to the scan duration and sampling rate.

[0065] Figure 4B The photodiode signal 402 and the laser drive signal 410 are shown. These signal data can be generated using an energy source to drive the signal. Figure 4B The laser drive signal 410 depicted in the diagram, in this case, is either a laser drive signal or a command signal that instructs the laser to be turned on or off for a specific scan length. A photodiode signal 402 is superimposed on the laser drive signal 410. Rising 421 and falling 414 of the laser drive signal 410 correspond to rising 404 and falling 406 of the photodiode signal 402. Figure 4B The data illustrated in the diagram can be used at step 304 to identify the portion of the original photodiode signal 402 corresponding to the scan. In some embodiments, the laser drive signal 410 is ~0V when the laser is off and ~5V when the laser is on. Step 304 can be accomplished by isolating all data associated with the photodiode signal when the laser drive signal 410 is above a certain threshold (e.g., 4.5V) and by analyzing and excluding all data when the laser is below that threshold.

[0066] Figure 4C An embodiment of step 306 is shown, which includes determining the area 416 under the original photodiode signal 402. In some embodiments, equation (4) can be used to calculate the area under the curve:

[0067]

[0068] The voltage of the integrated photodiode 418 can be used to determine the voltage for TED. i Calculated pdon i .

[0069] Figure 4DThe position of scan 420 relative to the part and the total scan count 424 are shown. Two values ​​can be used to determine the TED corresponding to the scan position on the part. Figure 4E The rendered area of ​​interest is shown. Figure 4E Several other parts 428 and embedded pieces 430 are also shown. Figure 4E All parts are depicted as being located on powder bed 432.

[0070] Figure 4F The diagram shows a trajectory associated with a portion of the photodiode data and laser drive signal data corresponding to four scans. This trajectory can be used with the remaining photodiode data to determine the total sample count 434. The total sample count can be used to calculate the total scan length LSum on the part. p The total sample count is determined by summing the laser on-time periods 436. In some embodiments, the total scan length can be determined using the sum of the laser on-time periods and the average speed of scanning the energy source during the laser on-time periods.

[0071] After collecting the scan data, the TED for each layer can be calculated using the TED associated with each laser scan, and then displayed in graph 440, as shown. Figure 4G As shown in the figure. Graph 440 illustrates the TED values ​​located within the nominal region 442 and the non-nominal region 444. The TED regions are divided by a reference threshold 438. In this way, layers in the part that may contain defects can be easily identified. Further analysis can then be performed on the set of layers with non-nominal TED values.

[0072] Figure 4H This illustrates how point cloud 446 can be used to display TED values ​​for each scan in three dimensions. Point cloud 446 illustrates the location of TED values ​​from nominal region 442 and non-nominal region 444 in three-dimensional space by displaying non-nominal values ​​as different colors or intensities than nominal values. Non-nominal values ​​indicate the parts of the part most likely to contain manufacturing defects, such as voids formed by keyholes or cavities due to lack of fusion. In some embodiments, the system may generate and transmit control signals based on TED that will change one or more process parameters.

[0073] Figure 5A flowchart illustrating an exemplary process 500 is shown. Exemplary process 500 uses data generated by an additive manufacturing system to determine thermal density and identify portions of a part most likely to contain manufacturing defects. Thermal density can be determined using data generated by coaxial optical sensor 109 and off-axis optical sensor 110, either alone or in combination. At 502, photodiode time-series data can be collected. The photodiode time-series data can be generated using, for example, voltage data associated with the sensors. At 504, laser drive time-series data is collected. The laser drive time-series data can be associated with additive manufacturing process parameters such as laser power, laser speed, scan spacing, xy position, etc. The process at 506 slices the photodiode time-series data by discarding portions of the photodiode time-series data corresponding to portions of the laser drive time-series data indicating a laser-off state. In some embodiments, the laser drive signal is ~0V when the laser is off and ~5V when the laser is on. The process at 506 can isolate all data when the laser drive signal 410 is above a certain threshold (e.g., 4.5V) and exclude all data when the laser is below that threshold through analysis. In some embodiments, photodiode signals that periodically drop to ~0.2V can be included in the sample sequence data because these sample sequence data represent the time when the laser is just turned on and is heating the material.

[0074] The process at 506 may output only the laser-on photodiode data 508. The process at 510 may use the laser-on photodiode data to convert time-series data into sample sequence data. The process at 510 segments the laser-on photodiode data into “N” sample segments. Using 20 sample segments is intended to provide an example of one embodiment of the invention. Any number of sample segments may be used with varying degrees of accuracy / resolution. In some embodiments, this group of sample segments may be referred to as subscan 520, as multiple subscans 520 are typically required to constitute a single scan. The process at 512 may count the number of samples 516. The process at 514 may render the area of ​​the laser-processed part. In some embodiments, the laser-processed part area 518 may be determined using the number of pixels associated with the laser-processed part on the display. In other embodiments, the area may be calculated using data associated with process parameters and the number of scans. At 522, the process normalizes the subscan data using the total sample count, the laser-processed part area 518, and the subscan data 520. In the illustrated embodiment, sub-scan metric data 524 is the thermal energy density of the portion of the part associated with each sub-scan. In some embodiments, scan data may also be subdivided by scan type. For example, additive manufacturing machines may utilize scans with different characteristics. In particular, contour scans or scans designed to finish the outer surface of a part may have much higher power than scans designed to sinter the internal regions of a part. Therefore, isolating data by scan type can also yield more consistent results. In some embodiments, scan type identification may be based on scan intensity, scan duration, and / or scan location. In some embodiments, scan type can be identified by associating detected scans with scans specified by a scan plan (which is associated with the part being built).

[0075] Next, process 528 receives baseline scan metric data and thermal density, and outputs an IPQM quality assessment 530. The IPQM quality assessment 530 can be used to identify the parts of a component most likely to contain manufacturing defects. Process 528 may include a classifier as discussed earlier in this specification. In addition to the methods and systems described above, process 528 may also use Mahalanobis distance to compare candidate data (e.g., sub-scan metric data 524) with baseline sub-scan metric data. In some embodiments, the baseline sub-scan metric data may be used to calculate the Mahalanobis distance for each sub-scan. While regarding… Figure 5 The disclosed embodiments have discussed the use of lasers as an energy source, but it will be apparent to those skilled in the art that many modifications and variations are possible in light of the above teachings, for example, an electron beam or other suitable energy source may be used instead of a laser.

[0076] What should be understood is that Figure 5 The specific steps shown provide a specific method for determining thermal density and identifying the parts most likely to contain manufacturing defects according to another embodiment of the invention. According to alternative embodiments, other sequences of steps may also be performed. For example, alternative embodiments of the invention may perform the steps outlined above in a different order. Furthermore, in Figure 5 The steps shown may include multiple sub-steps, which may be performed in various orders suitable for individual steps. Furthermore, other steps may be added or removed depending on the specific application. Many variations, modifications, and alternatives will be recognized by those skilled in the art.

[0077] Figure 6A The photodiode time series data 602 is shown. This can be seen from Figure 1 and... Figure 2 Various coaxial or off-axis sensors shown collect time-series data from photodiodes. The x-axis 604 indicates time in seconds, and the y-axis 606 indicates the voltage generated by the sensors. The voltage generated by the sensors is associated with energy emitted from the building plane that can strike one or more sensors. Sample 606 is illustrated on the trajectory of the photodiode time-series data 602. Figure 4B The process at 506 is described, in which photodiode data is correlated with laser drive signals.

[0078] Figure 6B The laser-activated photodiode data is shown. The x-axis represents the number of samples 608, and the y-axis 610 represents the voltage from the raw sensor data. The voltage drop 612 is included in the analysis because, although the voltage is generally low, the laser is still actively contributing to heating the material.

[0079] Figure 6C The sample sequence data of the laser-activated photodiode discussed in step 510 is shown. The 20 sample segments 620 can have any size. Twenty samples correspond to a laser travel distance of ~400 μm and a laser travel speed of 1000 mm / s. The noise of the XY signal itself is approximately ~150 μm. In some embodiments, with fewer than 20 sample segments, for example, 2 sample segments, the measured distance and noise will be in a ratio such that the location of the point cannot be reliably determined. In some embodiments, a limit of 50 samples can be used for spatial resolution less than 1 mm. Therefore, for 50 kHz data with a laser travel speed of 1000 mm / s, the number of samples used to segment the data should be in the range of 20 ≤ N ≤ 50. It will be apparent to those skilled in the art that many modifications and variations are possible in light of the above teachings.

[0080] Figure 6DCorresponding to process 522, an embodiment is illustrated in which the average value 618 for each sub-scan is determined. In some embodiments, the inputs to process 522 include total sample count 516, laser-processed area 518, and sub-scan line data 520. Using these inputs, the average value can be used to determine the area under the curve (AUC), as shown in Equation (5):

[0081] AUC = V(average) * N(sample) Equation (5)

[0082] Where V is the average voltage determined for each sub-scan, and N is the number of samples. Figure 6D Since the width of the data is fixed, the average voltage of the 20 sample segments is equal to the integral of the signal.

[0083] Figure 6E The laser processing area A is shown for a single sub-scan 622 and for all scans 624. i In addition to the area, the scan length L can also be calculated. i And the L on the entire part i The sum of LSum p L can be calculated using equation (6). i :

[0084]

[0085] The x and y coordinates for the start and end of a scan can be provided, or the x and y coordinates for the start and end of a scan can be determined based on measurements from one or more direct sensors.

[0086] Figure 6F A rendering area 626 of the laser processing section associated with a layer in the construction plane is shown. In some embodiments, once pdon is determined... i Area A of the part p Length of scan L i and total length LSum p TED can then be calculated using equation (7):

[0087]

[0088] TED is sensitive to all user-defined laser powder bed fusion process parameters (e.g., laser power, laser speed, scan spacing, etc.). TED values ​​can be used for analysis comparing IPQM with benchmark datasets. The resulting IPQM can be determined for each laser scan and displayed graphically or in 3D using point clouds. Figure 4G An exemplary graphic is shown. Figure 4H An exemplary point cloud is shown.

[0089] Figures 7A to 7CThe figure shows post-processing porosity measurements and corresponding normalized in-process TED measurements. This illustrates that in-process TED measurements can be an accurate IPQM predictor of porosity and other manufacturing defects. Figure 7A The figure shows a comparison of TED metric data with a benchmark dataset. The graph shows the value of each photodiode in the IPQM metric, both individually and in combination. Coaxial photodiode data 702 can be obtained from a sensor aligned with the energy source. Off-axis photodiode data 704 can be collected by a sensor not aligned with the energy source. The combination 706 of coaxial and off-axis photodiode data produces the highest sensitivity to variations in process parameters. The x-axis 708 shows the build plane layers of the part; the y-axis 710 shows the Mahalanobis distance between the calculated TED and the benchmark metric.

[0090] Mahalanobis distance can be used to standardize TED data. Mahalanobis distance indicates how many standard deviations each TED measurement differs from the nominal distribution of TED measurements. In this case, Mahalanobis distance indicates how many standard deviations each TED measurement differs from the collected average TED measurements when constructing control layers 526-600. Figure 7A The charts below also show how TED varies with total energy density (GED) and porosity. In particular, for this set of tests, TED can be configured to predict the porosity of a part without destructive testing.

[0091] In some embodiments, the performance of the additive manufacturing apparatus can be further validated by comparing quantitative metallographic features (e.g., the size and shape of pores or intermetallic particles) and / or mechanical property characteristics (e.g., strength, toughness, or fatigue) of the formed metal parts while performing test runs. Typically, the presence of unmelted metal powder particles in the test part indicates insufficient applied energy, while test parts receiving too much energy are prone to developing internal cavities, both of which compromise the integrity of the formed part. Porosity 714 can represent these defects.

[0092] In some embodiments, for generating Figure 7AThe nominal value is taken from previous tests. In some embodiments, since calculations are not required during additive manufacturing operations, the nominal value may also be taken from subsequent tests. For example, when attempting to compare the performance of two additive manufacturing apparatuses, the nominal value can be identified using the first run test of the additive manufacturing apparatus. The performance of the second additive manufacturing apparatus can then be compared to the nominal value defined by the first additive manufacturing apparatus. In some embodiments, if the performance of the two additive manufacturing apparatuses is within a predetermined threshold of five standard deviations, the two machines can be expected to have comparable performance. In some embodiments, the predetermined threshold may be a 95% statistical confidence level derived from an inverse chi-squared distribution. This type of testing method can also be used to identify performance that changes over time. For example, after calibrating the machine, the results of a test pattern can be recorded. After the apparatus has performed a certain number of manufacturing operations, the additive manufacturing apparatus can be run again. The initial test pattern performed immediately after calibration can be used as a benchmark to identify any changes in the performance of the additive manufacturing apparatus over time. In some embodiments, the settings of the additive manufacturing apparatus can be adjusted to restore the additive manufacturing apparatus to its subsequently calibrated performance.

[0093] Figure 7B The post-processing metallographic image of a part constructed using an additive manufacturing process is shown. Figure 7B Part 718 and its corresponding cross-section 720 are shown. Segments 1 to 11 correspond to... Figure 7A Segments 1 to 11 are shown in the part. Variations in process parameters and the resulting changes in porosity can be seen in cross-sectional view 720. Specifically, segments 2 and 3 have the highest porosities, at 3.38% and 1.62%, respectively. Higher porosities are also observed in the cross-section by increasing the number of defect markers 722 in the sample part.

[0094] Figure 7C IPQM results with corresponding cross-sections determined during metallographic analysis are shown. Each cross-section includes values ​​in J / mm². 2Energy density 724 and porosity 714 are expressed in units. Samples with the largest number of defect markers 722 correspond to TED measurements with the largest normalized distance from the reference. This figure illustrates that a low standard distance predicts higher density and lower porosity in the metallography, while a high standard Mahalanobis distance is highly associated with high porosity and poor metallography. For example, a low power setting used to generate the layer around layer 200 results in high porosity 714 and a large number of defect markers 722. In contrast, using a middle of the road settings on or around layer 600 results in no identifiable defect markers 722 and the lowest recorded porosity value is 0.06%.

[0095] Figure 8 An alternative process is illustrated in which data recorded by optical sensors (such as non-imaging photodetectors) is processed to characterize the additive manufacturing build-up process. At 802, raw sensor data is received, which may include build-plane intensity data and an energy source drive signal correlated together. At 804, by comparing the drive signal and the build-plane intensity data, individual scans are identified and positioned within the build-plane. Typically, the energy source drive signal will provide at least a start position and an end position from which the area spanned by the scan extension can be determined. At 806, the raw sensor data associated with the intensity or power of each scan can be categorized into corresponding X and Y grid regions. In some embodiments, the raw intensity or power data can be converted into energy units by correlating the dwell time of each scan in a particular grid region. In some embodiments, each grid region may represent a pixel of an optical sensor monitoring the build-plane. It should be noted that different coordinate systems, such as polar coordinates, can be used to store grid coordinates, and the storage of coordinates should not be limited to Cartesian coordinates. In some embodiments, different scan types can be categorized separately so that analysis can be performed only on specific scan types. For example, if those scan types are most likely to contain unwanted variations, the operator may want to focus on contour scans. At 808, the energy input at each grid region can be summed, so that Equation (8) can be used to determine the total amount of energy received at each grid region.

[0096]

[0097] The summation can be performed before a new powder layer is added to the build plane, or alternatively, the summation can be delayed until a predetermined number of powder layers have been deposited. For example, the summation can be performed only after five or ten different powder layers have been deposited and fused during the additive manufacturing process. In some embodiments, the sintered powder layers can increase the thickness of the part by about 40 micrometers; however, this thickness will vary depending on the type of powder being used and the thickness of the powder layers.

[0098] At 810, determine the standard deviation of the detected samples associated with each grid region. This helps identify grid regions with small or large variations in power readings. Variations in the standard deviation can indicate sensor performance problems and / or the following: missing one or more scans or power levels of one or more scans far exceeding the nominal operating parameters. The standard deviation can be determined using equation (9).

[0099]

[0100] At 812, the total energy density received at each grid region can be determined by dividing the power reading by the total area of ​​the grid regions. In some embodiments, the grid regions may have a square geometry with a length of approximately 250 micrometers. The energy density for each grid region can be determined using equation (10).

[0101]

[0102] At point 814, when more than one part is being constructed, different mesh regions can be associated with different parts. In some embodiments, the system may include stored part boundaries that can be used to quickly associate each mesh region and its associated energy density with its corresponding part using the coordinates of the mesh region and the boundary associated with each part.

[0103] At 816, the area of ​​each layer of the part can be determined. In cases where a layer contains voids or helps define internal cavities, a large portion of that layer may not receive any energy. Therefore, the affected area can be calculated by summing only the grid regions identified as receiving some amount of energy from the energy source. At 818, the energy density of that layer for the part can be determined by summing the total power received by the grid regions within a portion of the layer associated with the part and dividing by the affected area. Equations (11) and (12) can be used to calculate the area and energy density.

[0104]

[0105]

[0106] At 820, the energy densities of each layer are summed together to obtain a metric indicating the total energy received by the component. The overall energy density of the component can then be compared to the energy densities of other similar components on the build plane. At 822, the total energy from each component is summed. This allows for high-level comparisons between different builds. Build comparisons help identify system differences, such as powder variations and differences in total power output. Finally, at 824, the summed energy value can be compared with other layers, components, or build planes to determine the quality of those other layers, components, or build planes.

[0107] What should be understood is that Figure 8 The specific steps illustrated in the diagram provide a specific method for characterizing an additive manufacturing build process according to another embodiment of the present invention. According to alternative embodiments, steps in other sequences may also be performed. For example, alternative embodiments of the present invention may perform the steps outlined above in a different order. Furthermore, in Figure 8 The steps illustrated in the diagram may include multiple sub-steps, which can be performed in various orders suitable for individual steps. Furthermore, depending on the specific application, other steps may be added or removed. Many variations, modifications, and alternatives will be recognized by those skilled in the art.

[0108] Figures 9A to 9D A visual depiction is shown indicating how multiple scans can contribute to the power introduced at a single grid region. Figure 9A A grid pattern consisting of multiple grid regions 902 is depicted, which are distributed across a portion of a part constructed by an additive manufacturing system. Figure 9A A first energy scan pattern 904 extending diagonally across the grid region 902 is also depicted. The first energy scan pattern 904 can be applied by a laser or other strong thermal energy source that scans across the grid 902. Figure 9B The diagram illustrates how energy introduced onto a part is represented in each grid region 902 by a single grayscale color indicating the amount of energy received, where darker gray shades correspond to larger amounts of energy. It should be noted that in some embodiments, the size of the grid region 902 may be reduced to obtain higher resolution data. Alternatively, the size of the grid region 902 may be increased to reduce memory and processing power usage.

[0109] Figure 9C A second energy scan pattern 906 is shown, which overlaps at least a portion of the energy scan of the first energy scan pattern. (See attached diagram) Figure 8As discussed in the text, when the first and second energy scan patterns overlap, the grid areas are shown in a darker shade to illustrate how the energy from the two scans increases the amount of energy received on the overlapping scan pattern. Clearly, the method is not limited to two overlapping scans and may include many other additional scans that will be added together to fully represent the energy received at each grid area.

[0110] Figure 10A An exemplary turbine blade 1000 suitable for use according to the described embodiments is shown. The turbine blade 1000 includes multiple different surfaces and includes many different features that require many different types of complex scans to produce. In particular, the turbine blade 1000 includes a hollow blade portion 1002 and a tapered base 1004. Figure 10B An exemplary manufacturing configuration is shown in which 25 turbine blades 1000 can be manufactured simultaneously on top of the build plane 1006.

[0111] Figures 10C to 10D It shows Figure 10B The configuration depicted in the diagram has different cross-sectional views of different layers with a grid-based TED visualization layer. Figure 10C Layer 14 of turbine blade 1000 is shown. A TED-based visualization layer illustrates how the lower end of a selected portion of the base 1004 can define multiple voids in turbine blades 1000-1, 1000-2, and 1000-3. These voids, which would otherwise be completely hidden within the turbine blade, are clearly visible in this grid-based TED visualization because the energy density data is correlated with discrete grid regions. Figure 10D This illustrates how the upper end of the base 1004 can also define multiple hidden cavities within the turbine blades 1000-1, 1000-2, and 1000-3, which can be seen from... Figure 10D The mesh-based TED visualization layer depicted in the paper clearly identifies the multiple hidden holes.

[0112] Figures 11A to 11B Cross-sectional views of the base 1004-1 and base 1004-2 of two different turbine blades 1000 are shown. Figure 11A A base 1004-1 manufactured using nominal manufacturing settings is shown. The outer surfaces 1102 and 1104 of the base 1004-1 receive substantially more energy than the interior 1106 of the base 1004-1. The increased energy input to the outer surfaces provides a more uniform hardened surface, resulting in an annealing effect along the outer surfaces 1102 and 1104. This additional energy can be introduced using a contour scan with higher energy targeting the outer surfaces 1102 and 1104. Figure 11BBase 1004-2 is shown, which is manufactured with the same manufacturing settings as 1004-2, except that the contour scan is omitted. Although all scans used in the manufacturing operation of base 1004-2 are also included in the manufacturing of base 1004-1, the summation of the energy density inputs for all scans covering each grid region allows the operator to clearly see the differences between base 1004-1 and 1004-2.

[0113] Figure 11C The differences in surface consistency between base 1004-1 and base 1004-2 are shown. Clearly, omitting the contour scan for base 1004-2 has a substantial impact on the overall outer surface quality. In terms of consistency, the outer surface of base 1004-1 is smoother and has fewer pores. The annealing effect on base 1004-1 should also make it substantially stronger than that of base 1004-2.

[0114] Figure 12 The figure illustrates the thermal energy density for parts associated with multiple different builds. Builds A through G each include thermal emission data (represented by discrete circle 1202) for approximately 50 different parts (built during the same additive manufacturing operation). This figure shows how thermal emission data can be used to track differences between different builds. For example, batches A, B, and C all have similar TED distributions; however, builds D, E, and F, while still within tolerance, have consistently smaller thermal emission data. In some cases, these types of variations can be attributed to variations in powder batches. In this way, thermal emission data can be used to track systematic errors that may negatively impact overall output quality. It should be noted that while this graph depicts the average TED value based on grid-based TED, a similar graph can be constructed for scan-based TED methods.

[0115] Figures 13 to 14B The illustration shows an example of how thermal energy density can be used to control the operation of parts using in-situ measurements. Figure 13This illustrates how parts can be produced using different combinations of energy source power and scanning speed, and then subjected to destructive analysis to determine the resulting part density in grams per cubic centimeter, as depicted. In this experiment, a part with a part density of 4.37 g / cc was produced using manufacturer-recommended scanning and laser power settings, associated with part density 1302. Based on the resulting density reading, the location of dashed line 1304 can be determined. Line 1304 indicates the resulting reduction in energy input causing insufficiently heated portions of the powder to fuse together, resulting in a part density drop below a threshold density level. Part density can also be reduced when excessive energy is added to the system, resulting in keyhole-like pores within the part due to the powder being vaporized rather than melted during the manufacturing process. Dashed line 1306 can be determined experimentally based on density data, and in this example, it is identified by a part density as low as 4.33 g / cc. Line 1308 represents the optimal energy density profile along which the energy density and part geometry remain substantially constant. Density tests demonstrate how the average density of parts created using settings distributed along line 1308 remains relatively consistent.

[0116] Figures 14A to 14B The diagram shows a thermal density profile superimposed on a portion of the power density map, illustrating how thermal density measurements collected during additive manufacturing operations vary depending on the settings used by the energy source. Figures 14A to 14BAs depicted, darker shading indicates higher thermal density, and lighter shading indicates lower thermal density. Control limits for a specific part can be determined through part density testing and thermal density profiling. In this case, control limits indicated by ellipse 1402 have been determined, allowing power and scan speed parameters to vary by up to 3σ along line 1308 and up to 1σ along a line perpendicular to line 1308 as settings for producing part 1302 change. In some embodiments, allowable variation in power and speed allows for in-process adjustments to maintain the desired thermal density during production runs. Ellipse 1404 indicates the entire process window that can accommodate further deviations beyond the control limits. In some embodiments, this process window can be used to identify deviations that would still allow for validation of the resulting part using only in-process data. It should be noted that while the depicted control and process windows are shown as ellipses, other process windows are possible in shape and size, and may be more complex, depending on factors such as part geometry and material eccentricity. During manufacturing operations, thermal density can be determined in situ using readings from optical sensors that measure thermal radiation. If the thermal density falls outside the expected range while laser power and scanning speed are kept within the depicted control limits indicated by ellipse 1402, portions of the part with abnormal thermal density readings may be flagged as potentially defective.

[0117] In some embodiments, a process window can be incorporated into a modeling and simulation program that models one or more optical sensors that collect sensor readings for determining thermal density. Once the modeling and simulation system iterates to a first approximation of the instruction set for the workpiece, the expected thermal density can be output to an additive manufacturing machine for further testing. When the additive manufacturing machine includes optical sensors and computing devices configured to measure thermal density, it can use the modeled thermal density data for further testing and verification. The comparison between the modeled thermal density and the measured thermal density can be used to confirm the close match between the execution of the instruction set and the expected in-situ results.

[0118] Figure 14CAnother power density plot is shown, highlighting various physical effects resulting from the energy source setting being far beyond process window 1404. For example, the power density plot shows that adding a large amount of laser power at a low scan rate results in keyhole formation within the workpiece. Keyhole formation occurs because the powder material partially evaporates due to receiving too much energy. Furthermore, the combination of low power and high scan rate can prevent the powder from fusing together. Finally, the combination of high power level and high scan rate results in molten metal agglomeration during the build operation. It should be noted that changing the thickness of the deposited power layer can cause a shift in the line separating the conduction mode zone from the keyhole formation zone and the lack of fusion zone. For example, increasing the thickness of the powder layer has the effect of increasing the slope of the line separating the keyhole and lack of fusion regions from the conduction mode, because thicker layers generally require more energy to undergo liquefaction.

[0119] Figure 14C The diagram also illustrates how power and scan speed settings corresponding to the conduction mode region generally do not lead to any of the aforementioned serious defects; however, by maintaining laser settings corresponding to values ​​within process window 1404, the grain structure and / or density of the resulting part can be optimized. Another benefit of keeping energy source settings within process window 1404 is that these settings should keep thermal density readings within a narrow range of values. Any thermal density value falling outside the predetermined range during manufacturing operations can indicate a problem in the manufacturing process. These problems can be addressed by moving the settings closer to the central region of the process window and / or by updating the process window for subsequent parts. In some embodiments, the manufacturer may be able to identify that the failure is caused by defective powder that should not be considered for inclusion in subsequent manufacturing processes or some other infrequent aberration. It should be understood that the depicted process window may not be the same for all parts of the part and can vary considerably depending on what area or even a specific layer of the part is being processed at a particular time. In some embodiments, the size and / or shape of process window 1404 may also vary depending on other factors such as scan spacing, scan length, and scan direction.

[0120] Figure 14D This illustrates how the size and shape of the molten pool can vary depending on the laser power and scanning speed settings. Exemplary molten pools 1406 to 1418 present exemplary molten pool sizes and shapes for various laser power and speed settings. Generally, it can be seen that a larger molten pool results from a higher laser power and a lower scanning speed. However, for this particular configuration, the size of the molten pool depends more on the laser power than on the speed.

[0121] Figures 15A to 15F The illustration shows how meshes can be dynamically created to characterize and control additive manufacturing operations. Figure 15A A top view of a cylindrical workpiece 1502 located on a portion of a build plane 1504 is shown. The workpiece 1502 is shown as being in its state during an additive manufacturing operation. Figure 15B The diagram illustrates how a cylindrical workpiece 1502 is divided into multiple tracks 1506 along which an energy source can melt powder distributed on the upper surface of the cylindrical workpiece 1502. In some embodiments, the energy source can alternately change the orientation 1506 as depicted, while in other embodiments, the energy source can always move along one direction. Furthermore, the orientation of the tracks 1506 can vary between layers to further facilitate orientation randomization for scanning the workpiece 1502.

[0122] Figure 15C An exemplary scanning pattern for an energy source is shown when the energy source forms a portion of workpiece 1502. As indicated by arrow 1508, the exemplary energy source moves diagonally across the workpiece 1502. A single scan 1510 of the energy source may be oriented in a direction perpendicular to the direction of movement of the energy source along track 1506 and extend completely across track 1506. The energy source may be briefly switched off between consecutive individual scans 1510. In some embodiments, the duty cycle of the energy source may be approximately 90% as it traverses each of the tracks 1506. By employing this type of scanning strategy, the energy source can cover the width of track 1506 as it traverses workpiece 1502. In some embodiments, the strip 1510 may have a width of approximately 5 mm. This can significantly reduce the number of tracks required to form workpiece 1502, since in some embodiments, the width of the molten pool produced by the energy source may be on the order of approximately 80 micrometers.

[0123] Figures 15D to 15EThis diagram illustrates how grid regions 1512 are dynamically generated along each track 1506, and how the size of the grid regions 1512 is set to fit the width of each individual scan 1510. The system can predict the precise location of subsequent scans by referencing an energy source drive signal leading to the energy source along the path. In some embodiments, the width of the grid 1512 may match the length of the individual scan 1512 or be within 10% or 20% of the length of the individual scan 1512. Again, the scan length of the individual scan 1512 can be anticipated by referencing the energy source drive signal. In some embodiments, the shape of the grid regions 1512 may be square or rectangular. As the energy source continues along track 1506, the thermal density can be determined for each of the grid regions 1512. In some embodiments, the thermal density reading within grid region 1512-1 can be used to adjust the output of the energy source within the next grid region (in this case, grid region 1512-2). For example, if the heat density readings generated by individual scans 1510 within grid region 1512-1 are substantially higher than expected, the power output of the energy source can be reduced, the rate at which the energy source is scanned across each individual scan 1510 can be increased, and / or the interval between each individual scan 1510 within grid region 1512-2 can be increased. These adjustments can be made as part of a closed-loop control system. Although only five individual scans 1510 are shown in each region, this is for illustrative purposes only, and the actual number of individual scans within grid region 1512 can be substantially higher. For example, in the case where the melting zone generated by the energy source is approximately 80 micrometers wide, approximately 60 individual scans 1510 may be required to ensure that all powder within the 5mm square grid region 1512 falls into the melting zone.

[0124] Figure 15FThe edge region of workpiece 1502 is shown once the energy source has completed traversing the pattern of track 1506. In some embodiments, the energy source may continue to apply energy to workpiece 1502 after most of the powder has been melted and re-solidified. For example, contour scan 1514 may be performed along the outer periphery 1516 of workpiece 1502 to apply a surface finish to workpiece 1502. It should be understood that contour scan 1514, as depicted, is substantially shorter than individual scan 1510. For this purpose, grid region 1518 may be substantially narrower than grid region 1512. It should also be noted that the shape of grid region 1518 is not purely rectangular, as in this case, the shape of grid region 1518 follows the contour of the outer periphery of workpiece 1502. Another situation that may lead to differences in scan length could be that the workpiece includes walls of varying thickness. Walls of varying thickness can cause scan lengths to vary within a single grid region. In such cases, the area of ​​each grid region can be kept consistent by increasing the length of the grid region while narrowing the width to accommodate variations in the length of each scan.

[0125] Figure 16 An example of closed-loop control is illustrated, showing a feedback control loop 1600 for establishing and maintaining feedback control of additive manufacturing operations. At block 1602, a reference thermal density for the next grid region the energy source will traverse is input into the control loop. This reference thermal density reading can be confirmed from modeling and simulation programs and / or from previously run test / test runs. In some embodiments, this reference thermal density data can be adjusted by an energy density deviation block 1604, which includes energy density readings recorded during previous layers for various grid regions. The energy density deviation block 1604 can include adjustments to the reference energy density block in case the previous layer received too much or too little energy. For example, if an optical sensor reading indicates that the thermal density in a region of the workpiece is below the nominal value, the energy density deviation value can be increased for the reference energy density value for grid regions overlapping with the grid region with the below-nominal thermal density reading. In this way, the energy source can melt other powders that were completely melted during the previous one or more layers.

[0126] Figure 16The diagram also illustrates how inputs from blocks 1602 and 1604 collaboratively create the energy density control signal received by controller 1606. Controller 1606 is configured to receive the energy density control signal and generate heat source input parameters configured to generate the desired thermal energy density within the current grid region. Input parameters may include power, scan speed, scan spacing, scan direction, and scan duration. Energy source 1608 then receives the input parameters and employs arbitrary variations of the input parameters for the current grid region. Once the optical sensor has measured the scan of energy source 1608 constituting the current grid region, at block 1610, the thermal energy density for the current grid region is calculated and compared with the energy density control signal. If the two values ​​are the same, the energy density control signal is not changed based on the optical sensor data. However, if the two values ​​are different, the difference is added to or subtracted from the energy density control signal for the scan performed in the next grid region.

[0127] In some embodiments, the mesh regions for the current layer and all previous layers can be dynamically generated mesh regions, oriented according to the path and scan length / width of the scan performed by the energy source. In this type of configuration, both the reference energy density and the energy density deviation can be based on the dynamically generated mesh regions. In other embodiments, the mesh regions for the current layer can be dynamically generated, while the energy density deviation data 1604 can be based on energy density readings associated with a static mesh region defined before the start of the additive manufacturing operation, resulting in the static mesh region remaining fixed throughout the part and unchanged in position, size, or shape. When a Cartesian mesh system is required, the mesh can be uniformly shaped and spaced, but it can also take the form of mesh regions constituting a polar coordinate mesh system. In other embodiments, the mesh regions for the current layer can be statically generated before performing the build operation, and the energy density deviation data can also be statically generated and shared with the same mesh used for the current layer.

[0128] In some embodiments, control loop 1600 may be used to use thermal emission density instead of thermal energy density. Thermal emission density may refer to other factors besides thermal energy density. For example, thermal emission density may be a weighted average of multiple characteristics, including thermal energy density and one or more other characteristics such as peak temperature, minimum temperature, heating rate, cooling rate, average temperature, standard deviation from the average temperature, and rate of change of the average temperature over time. In other embodiments, one or more other characteristics may be used to verify that the scans constituting each of the grid regions are reaching the desired temperature, heating rate, or cooling rate. In such embodiments, the verification characteristic may serve as a flag indicating that input parameters for the energy source may need to be adjusted within a defined control window to achieve the desired temperature, heating rate, or cooling rate. For example, if the peak temperature within the grid region is too low, power may be increased and / or the scan rate reduced. While the aforementioned control loop 1600 has been discussed in relation to various types of grid TED, it should be noted that those skilled in the art will also understand that scan TED metrics can also be used in similar loop configurations.

[0129] TED Analysis on Small-Scale Supply of Recoating Arm

[0130] Figure 17A The standard distribution of powder 1702 across building plate 1704 is shown. This depiction shows how the powder spreads uniformly without any variation in height. Conversely, Figure 17B This illustrates how the thickness of the powder layer 1702 can vary significantly when insufficient powder 1702 is recovered by the recoating arm. Once the recoating arm begins to deplete the powder 1702, the thickness of the powder layer gradually decreases until a portion of the build plate 1704 is completely devoid of powder 1702. Because this error can have a rather negative impact on the overall quality of the part, early detection of this phenomenon is crucial for accurate defect detection.

[0131] Figure 17C A black-and-white photograph of the build plate is shown, in which a shortage of powder 1702 resulted in only partial coverage of the nine workpieces arranged on the build plate. Specifically, the three workpieces on the right side of the photograph are completely covered, while the three workpieces on the left side are almost completely uncovered.

[0132] Figure 17DThis illustrates how the detected thermal density differs substantially when the energy source scans across all nine workpieces using the same input parameters. Region 1706 produces a substantially higher thermal density reading because the powder has a substantially higher emissivity and the thermal conductivity of the building plate or cured powder material is greater than the effective thermal conductivity of the powder. The higher thermal conductivity reduces the amount of energy available for radiation back to the optical sensor, thus reducing the detected thermal radiation. Furthermore, the lower emissivity of the material itself also reduces the amount of radiated thermal energy.

[0133] Thermal density vs. total energy density

[0134] The power supplied by the energy source to the workpiece causes the material used to manufacture the workpiece to melt, but during additive manufacturing processes, this power can also be dissipated through several other heat and mass transfer processes. The following formulas describe the different processes by which the emitted power can be absorbed when the energy source is a laser sweeping across a powder bed:

[0135] P 总激光功率 =P 在激光处的光学损耗 +P 腔室气体的吸收 +P 反射 +P 颗粒和羽流相互作用 +P 维持熔化池所需的功率 +P 传导损耗 +P 辐射损耗 +P 对流损耗 +P 蒸发损耗 Equation (13)

[0136] Optical loss at the laser refers to power loss due to defects in the optical system responsible for transmitting and focusing the laser on the build plane. These defects cause absorption and reflection losses of the emitted laser within the optical system. Absorption by the chamber gas refers to power loss due to the absorption of a small portion of the laser power by the gas within the build chamber of the additive manufacturing system. The impact of this power loss will depend on the gas's absorptivity at the laser wavelength. Reflection loss refers to power loss due to light escaping the laser optics without being absorbed by the powder bed. Particle and plume interaction refers to the interaction between the laser and the ejected plume and / or particles during the deposition process. While power losses due to these effects can be mitigated by protecting the positive circulation of gas through the build chamber, a small reduction in power is generally not entirely avoidable. Power required to maintain the melt pool refers to a portion of the laser power absorbed by the working material to melt the powder and ultimately overheat it to any temperature the melt pool eventually reaches. Conduction loss refers to a portion of the power absorbed through conduction to the solidified metal beneath the powder and the powder bed itself. In this way, the powder bed and solidified material constituting the part conduct heat away from the melt pool. The conduction of this thermal energy is the primary form of energy loss from the molten pool. Radiation loss refers to a portion of the laser power emitted by the molten pool and the surrounding material, which is hot enough to emit thermal radiation. Convection loss refers to the loss caused by the transfer of thermal energy to the gas circulating through the constructed chamber. Finally, evaporation loss refers to a small portion of the powder material that evaporates under laser radiation. The latent heat of vaporization is very large, thus having a powerful cooling effect on the molten pool, and as the total laser beam power increases, the latent heat of vaporization can become a non-negligible source of energy loss.

[0137] Thermal energy density (TED) is a measure based on optical light, which is the result of the radiation of light from a heated region, the light being transmitted back through optical devices, the light being collected by a detector, and the light being converted into an electrical signal. The Stefan-Boltzmann equation shown in equation (14) below gives the formula controlling blackbody radiation at all possible frequencies:

[0138]

[0139] The variables from equation (14) are shown in Table (1) below.

[0140]

[0141]

[0142] Table (1)

[0143] Before the radiated light is collected by the sensor and before it causes the voltage used to calculate the TED metric, there are other interfering factors affecting the radiated light. These are summarized in equation (15):

[0144] V TED所使用的电压 ={P 辐射 –P V观察因子 –P 以辐射的波长的光学损耗 –P 传感器损耗因子}*(Sensor scaling factor) Equation (15)

[0145] The various terms from equation (15) are explained in Table (2) below.

[0146]

[0147] Table (2)

[0148] In additive manufacturing, the quality factor typically used is the global energy density (GED). This is a parameter that combines various process inputs, as shown in Equation (16) below:

[0149] GED = (electron beam power) / {(travel speed) * (scanning interval)} Equation (16)

[0150] We note that the unit of energy per area of ​​the GED is: (Joules / sec) / {(cm / sec)*(cm)} = Joules / cm 2 However, it should be noted that although GED may have the same units as TED, GED and TED are generally not equivalent. For example, TED is derived by dividing the radiant power from a high-temperature region by the area, while GED is a measurement of the input power. As described herein, TED relates to the response or process output, while GED relates to the process input. Therefore, the inventors consider TED and GED to be different measurements. In some embodiments, the area used to determine TED is different from the area of ​​the melt pool. Therefore, some embodiments do not have a direct correlation between TED and the melt pool area. Advantageously, TED is sensitive to a wide variety of factors that directly affect the additive manufacturing process.

[0151] While the embodiments described herein have used data generated by optical sensors to determine thermal energy density, the embodiments described herein can be implemented using data generated by sensors that measure the performance of other physical variables in a process. Sensors that measure the performance of physical variables in a process include, for example, force and vibration sensors, contact heat sensors, non-contact heat sensors, ultrasonic sensors, and eddy current sensors. It will be apparent to those skilled in the art that many modifications and variations are possible in light of the above teachings.

[0152] The various aspects, embodiments, implementations, or features of the described embodiments may be used alone or in any combination. The various aspects of the described embodiments may be implemented by software, hardware, or a combination of hardware and software. The described embodiments may also be implemented as computer-readable code on a computer-readable medium for controlling manufacturing operations, or as computer-readable code on a computer-readable medium for controlling a production line. The computer-readable medium is any data storage device capable of storing data that can subsequently be read by a computer system. Examples of computer-readable media include read-only memory, random access memory, CD-ROM, HDD, DVD, magnetic tape, and optical data storage devices. The computer-readable medium may also be distributed across a network-coupled computer system, thereby storing and executing the computer-readable code in a distributed manner.

[0153] For purposes of explanation, the foregoing description uses specific terminology to provide a thorough understanding of the described embodiments. However, it will be apparent to those skilled in the art that the specific details are not required to practice the described embodiments. Therefore, the foregoing description of specific embodiments is presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the described embodiments to the precise forms disclosed. It will be apparent to those skilled in the art that many modifications and variations are possible in light of the foregoing teachings.

Claims

1. A method of additive manufacturing, comprising: depositing a layer of build material on a build plane; dividing at least a portion of the build plane into a plurality of build areas, each build area having a respective area; producing a plurality of scans across one of the plurality of build areas using an energy source to fuse the layer of build material within the build area; detecting energy emitted from the build plane as the energy source fuses the layer of build material using a sensor; and determining a heat energy density of a build area of the plurality of build areas from the energy detected by the sensor during the fusing divided by the respective area of the build area. The energy emitted from the build area is determined from data produced by a photodiode.

2. The additive manufacturing method of claim 1, wherein, The build material comprises a metal powder.

3. The additive manufacturing method of claim 1, wherein, The energy source comprises a laser.

4. The additive manufacturing method of claim 1, wherein, The determined heat energy density is compared to a threshold value, and wherein the build area is identified as potentially having a defect when the determined heat energy density exceeds the threshold value.

5. The additive manufacturing method of claim 1, wherein, 6. A method of additive manufacturing, comprising: depositing a layer of build material on a build plane; dividing at least a portion of the build plane into a plurality of build areas, each build area having a respective area; fusing the build material within each of the plurality of build areas using an energy source; detecting energy emitted as the energy source fuses the layer of build material using a sensor; and determining a heat energy density of each of the plurality of build areas from the energy detected by the sensor during the fusing divided by the respective area of the build area. The energy emitted from each respective build area is derived from data received from a photodiode. The layer of build material comprises a metal powder.

7. The additive manufacturing method of claim 6, wherein, The fusing of the build material comprises melting the build material with an energy source.

8. The additive manufacturing method of claim 6, wherein, The determined heat energy density of each of the plurality of build areas is compared to a threshold value.

9. The additive manufacturing method of claim 6, wherein, The deposition of the layer is performed by a recoater arm that spreads the layer of powder.

10. The additive manufacturing method of claim 6, wherein, 12. A system for additive manufacturing, comprising:

11. The additive manufacturing method of claim 6, wherein, a build material disposed across a build plane; an energy source arranged to fuse at least a portion of the build material; a sensor arranged to detect energy emitted from the build plane; and a processor configured to: divide at least a portion of the build plane into a plurality of build areas; receive data from the sensor as the build material is fused by the energy source; detect energy emitted as the energy source fuses the at least a portion of the build material using a sensor; determine an area of each of the plurality of build areas; and calculate a heat energy density of each of the plurality of build areas from the data produced by the sensor and the area of each of the plurality of build areas. The energy source fuses the build material in each of the plurality of build areas by producing a plurality of scans in each of the plurality of build areas, wherein each of the plurality of scans comprises a turn on of the energy source, a movement of the energy source, and a turn off of the energy source. ​ 13. The additive manufacturing system of claim 12, wherein, ​ 14. The additive manufacturing system of claim 13, wherein, The processor is further configured to sum data generated by the sensor for each scan of the plurality of scans within each build area of the plurality of build areas.

15. The additive manufacturing system of claim 12, wherein, The build material includes a metal powder.

16. The additive manufacturing system of claim 12, wherein, The energy source includes a laser.

17. The additive manufacturing system of claim 12, wherein, The processor is further configured to compare the calculated heat energy density of each build area of the plurality of build areas to a threshold value.

Images